Tuesday 19 March 2019
Katla Satyanarayana, Soton Uni: From Pigeon Post to Hologram Phone
Calls: The Future of Wireless Technology.
Wireless communication has evolved from pigeon-post to paging, voice
calls, text messages, video calls – and now Internet everywhere. It
has become the ubiquitous means of socializing, doing business and of
entertainment. There are around 5 billion mobile phones in use through
which we transmit around 60 terabytes of data every month. And yet,
this is just the beginning – the future is even more exciting as we
are moving from the internet-of-things to holographic video calls,
which can conjure up the image of a person right in the room when we
talk to them. However, one of the key issues of this technology,
whether we have the capacity to accommodate all these users at a high
quality-of-service. An obvious solution to circumvent this problem is
to increase the bandwidth used. But we only have a limited bandwidth.
In this talk, I shall shed light on how to address this problem.
I'm working on the next generation of mobile phones.
A hologram image, where I can participate in a conference in
Hydrabad and I can simultaneously do a talk here in
soton. This is the future of next generation mobile phones.
From pigeons being used as message carriers. These days a day doesn't
pass without the mobile phone.
A skeleton block diagram of what is going on in a MP.
If you open one up you will see each of these blocks.
Your voice is the message source, the input , encoder then modulator .
In the pigeon post , the pigeon is the carrier , message is enclosed i nan
envelope attach it to a pigeon and it takes it to a friend or family.
Your voice message is modulated on to a radio signal and out
on an antenna. Transmit over a channel which could be air
,copper or fibre, the medium for transmission.
If by air- its wirelss comms, via water its acoustic underwater comms,
copper its a landline, fibre its optical comms.
Dependent on its medium, is its name.
The transmitted message will be received by another antenna
of the second party. There its decoded and the message appears
in the earpiece of that MP.
Early comms were smoke signals where smoke carried the
message , to gather at a spot or whatever. Then pigeons then
telegraphic comms. Then landlines and then first gen MP.
Then text transfer on 2 ,3 and 4G MPs.
1G , for first generation, currently 4G and next year 5G.
1G was analogue signal , no coding involved, just modulated onto a carrier
and transmission via antenna. Voice only transmitted, analogue as the
way we speak.
2G was digital comms. The message is multiplied onto the carrier
, like a pigeon carrying its envelope. The message is carried in the
amplitude of the carrier wave. This is amplitude modulation
com, frequency stays the same and amplitude varies with the signal. It is susceptible to blockages including by the human body.
Also FM , frequency modulation where the amplitude stays the same and the frequency changes with the signal.
2G,3G,4G is digital , it is digitised or sampled . Instead of the complete
signal just one sample of the message , a discrete set of values rathe rthan
continuous. There are represented as bits , 0 or 1 as a sequence.
you can use the spectrum more efficiently and its more robust
against blockages and quality of service is high. The secret is inthe 2 blocks , not in G1 but is in all later G. The encoder and channel encoder
encrypt your info, so when decoded at the other end can
still work through a blockage. Datarate is higher 2G from 1G and
3G higher than 2G. The change is not so much the available
bandwidth but TDM time-division-multiplex of frequency division
multiplex. Say i want to talk to 2 groups i can't do it simultaneously.
I can apportion time slots to each group . I can talk or I
can hear , not both at the same time. With FDM there is a range
of spectrum alloting frequencies to each group. With this allocation
I'm losing the chance to transfer data , this was in 2G.
For 3 code , i send a unique code to 1 and another non-overlapping code to 2.
This is spread spectrum , as invented by the Hollywood actress
Hedy Lamarr in the 1940s. The are 2 flavours direct sequenced
spectrum and the other is frequency hopping core division multiplexing .
There is no overlapping and I can use the time allocations
to 1 and 2 simultaneously. Hence higher data rates.
In 4G there is orthography division multiplexing . Divide the entire
bandwidth into chunks of frequencies , called sub-carriers,
orthoganally . There are also other techniques that make 4G
faster than the other Gs.
We are now moving to 5G. An exponential growth in data use expected between 2017 and 2023, requiring 5G. Trying to enhance MP bandwidth .
Heavy data transfer like films presently require buffering and at
congested places like Waterloo Station sometimes you can't even get a dial tone.
The demand is too high . We also want low latency to connect
machines together and to have an internet of things.
Say a medical device monitoring heart-rate and pressure, sugar levels etc that can report via your MP to a GP, without any intervention by
yourself. Every device connected to every other device by the IoT.
Requires more bandwidth than availablr to 4G. This is what we aim
for , for 5G, to address some of these issues. Remotely operated
surgery, a future scenario, surgeon on the ground but patient
in a surgery in space somewhere, a robot doing the surgery.
For this you need ultra reliable and low latency comms,
someone's life at stake. Not just in space but surgery in
remote areas of the Earth where there is no doctor.
With 5G you could watch sports from the plyer's point of
view. You can stream high definition video to a high-speed
train, only by buffering currently.
A scenario of a disaster and an air ambulnce helicopter
dispatched to the scene. Energy consumption is of
critical importance. Another app would be remotely operated
agricultural drones disposing of pests or weeds, again
energy consumption is utmost importance. Have we opened Pandras'
box full of problems or exciting news. The potential problem
with wanting enhanced mobile broadband , is it actually possible.
Imagine 2 taps filling a bucket that has a hole in the bottom.
Can only access the water via that hole. If you add 2 more taps it will
just overflow. You cannot access the extra water, its spilled over the side.
So increase the size of the hole . but how much further can you increase.
The hole in wireless comms is called bandwidth BW, the water is called
information , the taps are MP, laptop , cooker , washing machine, etc.
For important comms like medical, it should not get lost in the
traffic. Its more important than pics of cats say. There is only
limited BW available. BW is basically yhr set of frequencies usable.
But we don't have unlimited frequencies. There are 5 magic
bullets around this.
Massive MIMO , millimetre waves, beamforming, full duplex and small cells.
MIMO is multiple input /multiple output. A MP usually has 2 antenna,
a laptop maybe 4 so multiple both input and output. Massive
is just scaling that up to more antennas, perhaps several hundred.
mm wavelengths, my terrain, my PhD area and including
beamforming. So I use a massive number of antenas at mm
wavelength, to enable beamforming. Small cells is the range where
the signals can cover. A base station communicates with your
MP contained within its cell reach or area. Its better to have
small cells than large . For small cell you can have multiple
base-stations , continuously talking to a number of base stations.
So reduce cell size from currently about 1km area to 500 or 250m
, the sevice quality is enhanced. Because I'm nearer to dedicating
each base station to 1 user . The individual base-stations then
talk to each other. Full duplex is talking to someone and
receiving from someone. Only politicians can do both at the
same time. So with MPs with 2 frequencies f1 and f2 I use
f1 for transmitting and f2 for receiving. They are orthogonal
so they don't interfere with each other. I can transmit without
causing interference to the received signal. I can listen
and decode without any error in the message. Full duplex is
transmitting on the same frequency at the same time.
I only need f1 or f2 , not both. So I'musing the available
frequencies more efficiently, nearly doubling the spectrum,
doubling the datarate.
mm waves . 4G MP work on about 2.4 GHz . Wavelength is the speed
of light devided by f. Typically the spacing of elements in an
antenna is half the wavelength. 2.4GHz means that spacing is of the
order 100mm . mm waves start from about 30GHz and above,
and the wavelength is of the order mms. The antenna size reduces
and you have more BW. BW for 4G is about 3GHz - 300MHz
about 2.7 GHz. Go up to 300 GHz and there is 270 GHz , 100 times the BW.
Lets say each f step is 1GHz then I have 270 frequencies available,
wheras before only 2 or 3. So I can transmit loads of data on these.
The spacing between the antennas is so small I can now accommodate a
few hundred antennas in a MP or laptop. More BW but more
susceptible to blockages and attenuation because of likes of trees and rain.
With low levels of received signal there is less success at decoding the
signal, not enough energy. If i lower my voice, you cant hear me,
the energy being received by ypu is less. I raise my voice level so
you can hear me. but then there is a wall and if you were the other
side you will not hear me. Windows, doors etc causes attenuation and
some surfaces reflect. Signal fades through a series of trees,
getting progressively weaker. 5G more susceptible to blockages
compared to 4G frequencies. Even user blockages, if I hold
a phone close to me and am between handset and base station.
For mm waves a human in the way can block , reflect and
refract those sorts of signals. I could raise the power level
but we have limited power and increase too much and they
become open microwave ovens and would cause cancer.
Increase microwave energy and it will ionise the cells in your
body, which is carcinogenic. The Tlecoms union has put a limit
on the amount of power we can use.
A way to increase power without using any extra source of power
is to do directional tranmission. In 4G there is omnidirectional
transmission , the signal energy is all-pervading,
spreading in all directions. With a suitable monitor of electromagnetic
radiation you can see my signal everywhere in range.
So why waste energy to the surrounding area, just transmit
directly between A and B . I would need to steer the transmit
signal into the desired direction. This is called beam-forming.
There are Yagi antennas , you see on rooftops for TV,
they are directional. Each element is an antenna, multiple such
strips , steer the beam in that direction. Twist the Yagi around ,
you may not receive a signal. Parabolic antennas are another such
For 5G we will have a few hundred antennas closely spaced , spacing
of the order mm. Beam-forming is only possible like this with a large
number of antennas. For a very narrow beam you need loads
of antennas, giving a laser-like signal . These are arranged
in a plane, 2D fashion, and have control over elevation
and azimuth , ie 3D control. 4 beams coming from the same array
can be directed to 4 different users in different directions ,but all on the same frequency and at the same time. There is no interference between the beams as they are so highly directional. Its not duplex , its only
transmitting. There isa bit of cross-over leakage but multiple
users can use the same channel, same frequency at the same time.
An example of 8 antennas with phse-shifting meaning the resultant
beam is steerable. The more phase-shifters you have the
better the resolution, the better pencil-beam i have.
Nothing to see in the way of steering, done electronically by the phase-shifters.
More than one array of such antennas , for diversity. For me speaking,
my voice is going in all directions , some has a direct line, some reflected
and a direct path blocked because someone moves in between, but you can still
here me because of my voice coming by less direct routes.
There isa diversity of signal paths , if you receive more than
one indirect signal path you can better understand the signal I'm
transmitting. With a total blockage, you still have access via another
path. It improves the fidelity of the signal, having multiple transmission
routes. Say 3 separate routes to 1 user, conveys more signal power ,
its more robust. Like fake-news going around, 4 people conveying the
story 3 people say the same and one says something else , you
could make a judgement based on the numbers. Massive
MIMO equals diversity.
Where are we at with this, I've heard of deployments in Korea.
Is it work in progress or established tech or
an embelishement on early tech that is being deployed?
The ITU has collected the companies involved like Samsung
together and are shceduling roll-out. By the end of 2020
there should be commercial deployment. There has been
preliminary trials by Qualcom in US and Samsung in Korea
and some others. No one is sure that 5G works.
mm waves is a big technical difficulty to handle. What we hear of 5G
is massive MIMO and the beam-forming but not mm-wave, 4.5G
at the moment. 4G does not do what 4G was asked to do,
whether 5G will do what 5G was supposed to do , when the
hype is around still.
5G handsets will have those multiple beam-forming
arrays built in. I'm assuming everyone's MP will have built-in,
updated regularly, a database of where all the base-stations are.
The x,y,z coordinates, so when you make a call , the initial hook-up
, that will be beam-formed in the direction of the nearest expected
Yes, instant access happens with 4G . For beam-forming to happen it
needs to know the angles of departure and arrival. If that is line-of-sight
then fine . But lets say there is multiple blockages, no direct line of sight.
Initial contact then would be omni-directional , learns the intervening environment,then it will do beam-forming.
Is the term phase-array the same as beam-forming?
In the early days of satellite TV there was things the public called
squarials , was that this beam-forming structure insdie?
yes. The beam forming ability has been around for ages, the orthography/
multiplexing knowledge was around in the late 1940s or so
but the technology only emerged mid 1970s when the electronic chips
became available to do it.
What do you reckon the future will be like?
I hope to make a holographic phone-call sometime soon.
I assume there is an infrastructure chenge required to go to 5G.
At what level is that required, satellites, the transmitters or can you use
much of what is already htere ?
For mm-waves its a completely different structure required bu tthe rest
of the infrastructure should be much the same as current deployment.
Can you see 5G being all pervasive or just augmenting 4G, having
5G where it warrants it?
Yes, where 5G is not available somewhere then fall back to 4G.
Like at the moment ,where 4G is not available , goes back to 3G.
Just a sheet of paper would block 300GHz ?
Depends on the type of paper. Metal reflects the signal, buildings
block the signal.
What about fog, would that block?
Yes. At 60GHz there is a total absorbtion band of water vapour.
So 60GHz cannot be used for telecoms. You cant use all
available frequencies in the mm wave region .
Only possible for indoor 60GHz wifi.
Tuesday 16 April 2019 , Adam Barton of Solent Uni . Title: The future
in Video Games. Video games are now a significant influence in today’s
global economy. This talk speculates about the future of the games
industry; the challenges it faces, the impact of new technologies and
what it means for those who wish to develop interactive media.
The state of the industry. The rise of the mobile, spiralling costs,
whichare insurmountable for the future of video-games vg.
The ris eof robots will hopefully solve some of the problems, tHe massive
skill shortage and I'll make some predictions at the end.
I wrote this presentation about a year ago and some of those predictions are
not so wild.
For the future , a tip someone gave me, look at whats not going to
change. If things are going to be consistent that helps you
form what the future might be. Some things we are happy to say is
performance will improve. Moore's law is beginning to taper out now .
Automation will increase . We know prices will
rise and dependent on leisure time and free space available but we're
pretty sure prices will rise. Leisure time will increase and probably
facing rising unemployment, long-term if not permanent.
Advertisers will advertise and want to advertise.
Nicholas de Ponte? wrote a book Be Digital in 1995 where he made
predictions about the economic digital future . He recently gave a talk
where he went through his book exploring what hew got wrong and
Precising him, there will be a growing number of consumers of the
digital domain and be a contracting number of creators.
If your a creator you're going to be financially secure , you will be able to
monetise your IP. THese people wil lbe supporting you with cash for what
you do. 20 years on from 1995 and it still holds true.
Cultural shifts we have to be aware of - the stigma of gaming . There seems to be a social construct where people don't like the term gamers or don't
like people who play games. Ideas of what a gamer and a non-gamer is.
There are now people who just watch people who play games.
Then how does marketing interfere with this. Data shows 50% of the
population are gaming, a staggering amount. The console market is a
big market, mobile gaming 1.1 billion and pc gaming just shy of 1 billion.
It make sa lot of money and is prevalent within media society.
Bu tif you survey the average man in hte street they don't
expound. Politically we are on the back foot , government not recognising
any of the significant contributions that gaming makes and gives the industry a
hard time about tax-breaks. There is a rise of non-gamers and a problem with
identifying them. A big proportion out there playing games
regularly but if you ask them , they say they are not a gamer.
You may catch someone playing solitaire on an ipad or similar whimsy
games on a pc , but firmly say they ar eno gamer.
So an aging group who are gamers , younger people coming thru
with a perception of not being gamers . Its not like there is a categorisation
of newspaper buyers or tv watchers. S ogaming is not necessarily thr
same construct. Also market size and gender split , the perception
is its very male oriented thing , it is roughly 50/50
but it depends how you split the market, whether console , pc
or mobile. We need to identify which grouping. A weird sector,
even to me , is people who watch people play games. On you-tube they
wil lsit and watch others. At which point my mind is blown.
They will even watch people playing a vg they themselves have .
I can stretch to understanding trying out, before buying
concept . Something weird and psychological going on.
I don't watch footbsl bu ta lot of people do. I understand that because they
might not be able to play football. But there is indication that it
doesn't matter thet you cannot play, you just like to watch football.
The gaming viewer market is huge but poorly understood.
We don't know why its popular and its rising in
popularity. Then marketting. Sony and No Man Sky? they saw a product built be a small team in
Guildford and they wanted to be first to market this . They put a lot
of money into marketting , getting ahead of the thing, they dragged the developement team along and ended up selling a game that was not
finished. This debacle made a big impact on gaming and its
marketting. A huge backlash, massive complaints , law suits.
Similar with Fallout1976? and EA are going thru similar with a game called
Anthem , where the hype is not supported by the developemnt.
Gaming is worth 130 billion globally , the markets are steadily
growing and expanding. Asia and S Pacific areas ar estil a big chunk
of that market. Looking by platform its roughly split console
and pc as everyone thinks that is the big market , but its only
half the market , the othe rhalf is in mobile.
The prediction for the future is the real market is in mobile.
Its set to grow for the next 10 to 15 years. Look at the trends
, its mobile that will be gobbling up the money and making the money.
Everything else is growing a bit but fairly static in percentage terms.
Then the predicted death of the console , which is a bit of a stunner.
Sony is announcing playstation5 , but there isa distinct downward trend.
Switch has kicked it up a biyt , but is switch a console or a
mobile. The rise of the PS4 but does not compensate for the market
loss as a whole. So maybe by 2030 , no mor econsoles.
The UK used to be one of the top vg producing countries , 2004/5
we had 3 out of 10 of the top developers in the world.
Now we are not in the top 10, probably in the top 30.
We've been overtaken by Japan and the USA. Canada introduced
tax-breaks and Ubi-soft moved to Canada , where all the money went.
We are making a healthy contribution to UK GDP .
We employ in the UK 20,000 people 50/50 split between
retail and developers. That contribution to GDP is about 1.5 billion
thats about 75,000 per individual contribution to GDP, quite
staggering. If we had more people, then a bigger contribution
to GDP. The problem is we don't have enough people.
Mobile trend is to increase.
By useage , making phone-calls is not even in this list.
Checking the weather is a common useage of them. Playing games
is in the middle of uses. We know people are very attached to their mobile data
, because we track all their data.
Mobile is set to expand, international markets set to expand.
The Middle East is brimming to be exploited by vg industry.
As it stands we don't know what games to make them , our games
don't appeal ot them, a cultural difference.
The pc console market is saturated , people have gone from
14 years old with a computer to 50 yearolds with computers
and no more room for expansion.
The markets are unsustainable because of developement costs.
In 1985 30,000 GBP would get a vg on the shelves. Then a logarithmic
graph . A broadening of the costs . Unless you are a billion dollar
company , how do you make a vg. another thing difficult to see
management of this in the future. If these companies continue either
they go bust or people cannot afford to make a vg. The industry
is not sure of the solutuion to this. Redlib2 rumoured to cost >.5 billion
to make and 7 years of developement . It made back that money in
about the first week. But you need substantial capital
to get in on that track to begin with. It never used to cost that
much to make a vg. They used to be produced in about 4 to 5
man-weeks and now man-years.
So need to sell more consoles, why buy a PS4 or 5 if its only
as good as a PS3 or even2. Some people argue the earlier ones were
mor eplayable. But they make money by selling new consoles.
You need to outshine your previous products, something better/
more glamorous or exciting. A lot of this comes down to
graphics , when you look at a vg advert, its how pretty it looks.
There is little commentary on the playability , even at the
bottom of the screen it says "This is not a playable section of the
game", its just a cut-scene . The marketeers know that is what sells the
game. The killer is that manufacturing costs ar etiny but developement
costs are high . To make your first beta version is many thousands of
man-hours. But once you've made it and want to make a copy
its miniscule costs. Here lies the problem. In comparison to say
car-making. The engineering in that car is expensive to make anothe r
car . When manufacturing costs are so cheap, marketing
takes over, outstripping developent costs often, the root problem.
So we must get cheaper . A Marxist theory of econmics , to get cheaper,
is to get rid of the humans, because they cost the money. The quality
must not go down , or the consoles would not sell. You need more
effective staff , better tools , automation , creativity and you need some
machine learning in there. So what do the robots have in
store for us. There is an interesting symbiotic relationship bertween the
vg industry and machine learning. In the 1990s not enough computing power
to do descent machine learning. At the same time GPU, Graphics Processing
Units , took off. A parallel processing much dedicated to do clever
graphics in a pc or console. AI scientists discovered that these GPU
were good for doing AI and machine learning, because they are a
parallel processing system. Comes into Single Instruction Multiple Data
, one instruction to stuff across a whole range of different day=ta.
And GPUs are god at this. Graphics cards got cheaper because so many
were used for this. now cheap, scientists could play around with thes e
GPUs. Now there are GPUs dedicated to just doing computing.
So we have AI working, the other big problem
facing the games industry is actually no one knows how to build a
vg. I've been in this industry for 30 years , I know, what many outside the
industry don;t know, we don't know how to build games.
Shocking how inefficient we are, how many games get canned
mid-developement , how we make a long series of bad choices
and the only way of fixing it, is to throw away and srart again.
There are some processes that don't work, get your ducks
in a row but it does not work for vg developeent.
We use Agile which is good for software developemnet ,
the build things that have a certain flexibility built into them
but talk to the big names in the industry and its a big failure
but they don't want it known publically , or their
share-holders may run away , when they realise we don't know what we're doing.
We had a success in the rise of rapid prototyping , get a team
to make some stuff, then get iteratively desgn and develop it.
Famously Valv are very keen on it , but they have not made a
game for at least 8 years. So what is going on there.
We've seen a return of Immersive Play , mainly Indie developers
going "I'm just going to make a game" and if I like it, I'll
put it out there and we'll see what happens. But if you're trying to
doa billion dollar game , thats very risky.
We've seen the integration of View? control and asset
tracking, again wholly ineffective systems , from a broad
The biggest one is asset tracking where you make a vg ,
then take the assets and chuck them away and make the next one.
This is insane, its not done in other industries. It seems to be prevalent
in vg design industry. We need people to come in and sort it out,
otherwise we will not be able to continue to make bigger and bigger
games. Consider engineering infrastructure where the building costs 1 billion,
you know people know what they're doing on that project.
With things like RedDeadRedemption the vast majority of the
team just hope it all comes together towards the end.
If games wen't so cost effective to manufacture , we'd all
be broke by now.
The Art Pipeline. Everything i na vg you see, is traditionally made
by an artist. Every chair, every table , every wall or character
somebody or a team built that. Very labour intensive .
so historically we had 1st second, third ,fourth generation.
Build models, textures and artwork that goes on models ,
shaders for how things look whether something looks like a piece
of metal or a piece of plastic. Then rigging which is putting a
mechanism inside a character so it can animate. Then animating
getting your character to do (e?)motions frame by frame. Traditionally this
was all human , all processes done by humans.
We are at a midway point, we can scan models , using photogrammetry
and another process using a laser scanner . We can use photos for
texturing . We have smart materials and smart systems
of visual programming languages which makes the shading easier
. We are still hand-rigging but Maxi-Mo ? and a couple of othe r
companies are slowly solving this . Currently a load of
motion capture, mo-cap. See any of the "making of" films and games,
people in black cat suits with whit ebobbles in a big room
with loads of cameras. Captures their movement and organises
it to a form of data that a computer can use.
We don't know what goes between scanning and synthesis but w e
hope for some systems that can automate this. We have something
called Substance from Algorithmic for simulating real physical
surfaces and generating textures and shading from them.
We've no idea what can happen about rigging and then with animation
we have asimilation systems , where you don't need to mocap
any more. You can set a program running and it will
work out what that character should be doing.
The end prediction is no more humans, we will be synthesising our
models , texturing using procedural texture generators ,
synthesising our shaders and machine-learning algorithms
will be able to do via classic teacher/training systems and it will
learn to rig.
Some examples of some clever stuff. They work out a patch grid .
Take a video and let the comouter look at the video and the
computer has computed where parts of the body are and can
then work out the animation . Can then return that as functional
data . So someone can capture me just walking around a room ,
no special rig or camera or other facilities and 20 minutes later
they have 3D animation in a computer.
Repeated but with crowds.
Using wifi signals to see how they are blocked , as a person
walks around an unseen room. From that they can compute where a person
is and where their limbs are, for multiple people and all
through a wall, just mind-blowing. We're safe in here as there is
no wifi. You can see military applications for this , as long as a hotspot
is near them.
So we need people or actors to go and do your stuff . One for a dog ,
synthesised quadraped motion. No real dog involved. Captured loads of
imagery of actual dogs , fed into a computer learning system .
So they now have a virtual dog that wanders around and behaves
according to algorithms. Then that data can be spat out , ready
for a games engineer to use.
So then you can get rid of every person in the pipeline,
all replaced with a computer.
Photogrammetry is another toolset , take loads of photos of
static things , there is free software to do this at home.
Take loads of pics of your teddy bear, feed into a computer
, get 3D model, a data-cloud , which tells you where all
the points are and what colours they are. The result can
look like the actual scene but it has never existed. The down-side
is they can be very expensive models , having 2 or 3 million
points in them , to make them look good. Then algorithms to take the
entire scene and recomopute and optimise it , to be in a game.
The Vanishing of Ethan Carr ? used just photogrametry.
No artists used, just photos of everything and it looks
amazing. The technology is there and the industry is
The other big problem is that coders are expensive,
time-consuming and not 100% accurate. They put bugs in their
code. If we could teach computers to code we could solve a lot
of problems. We could rid the human factor and speed things up
dramatically. Problems are computers can't handle paradoxes ,
they can't handle the halting problem. Alan Turing's experiment
that a computer can't detect a program that does not halt.
P v. NP problem, polynomial time versus non-polynomial time ,
can you solve every problem. Descriptions of Intent,
basically what a coder does, you say to the coder what you
want and he translates that into code. Its your intent and they
wriite code for it. The current theory is that our programming
languages are wrong. Its not that they are impossible,
although mathematical arguments that the halting problem is impossible.
A conjecure has it that its only impossible in the current
paradigm of the way we write languages these days and we should build a
different computer language, that would circumvent
such problems. An interesting branch of doding maths: is it possible
to write a langueage that can just pass the halting problem.
If we could do that , then computers could write code. Currently they
can help, there is IDE Integrated Developement Environment.
It can look a t the first few letters you type and it can make a
suggestion "you are looking for a variable/ function, I'll tell you what it is
so you don't have to spell it. Code prediction is useful, high level
languages are useful , allowing you to make generalised statements and then the comp
will go away and work out exactly you need to be doing.
Then compilers that are optimising code, apparently dominated
by Microsoft currently. Writing a good compiler solves many problems
of writing code. Because you want a good compiler, the langueag
and everything else is just another layer of icing on the cake.
We have a massive skill shortage . The problem with automation,
you can eliminate repitive tasks , low skilled and narrow skilled jobs
and get rid of labour intensive work . We saw this with the industrial revolution
, automating many jobs. The problem we have , as you take away
low skill jobs, the entry level has less and less openings for
graduates and a rising expectation of skill level of what someone
has just to get into the industry. A barrier to entry so the
industry can't grow. Developing and maintsining
these highly developed jobs is in itself a highly skilled job. If you
don't have people who know the basics , you don't have anywhere
to go. But we are actively getting rid of these jobs. Traditionally in
the art pipeline one of the jobs popular for graduates
was modding, taking someone-elses model anf you had to build a simpler
version of it, reducing some of the complexity.
Also retopologising which is similar , you take a very high polygon
model and draw simplicity from that model. But those jobs have
gone, partly through automation , partly off-shoring to Asia where that
Some wild predictions about the future.
Machine learning - a great story about a boy who found his
father's ghost in a vintage vg. The games industry is really keen
to sample your online play and simulate it. Multiplayer games need
an audience , but also need the participants , to create the bots.
Fill the server with bots , machines emulating humans , then , #
providing no one can tell the difference , these bots will be much more
attractive. They don't mind losing. That means lots more people
get playing your game , because they've just kicked everyone's ars.
In reality they wil lbe playing against a bunch of bots that don't
care. A lot of work has been done on the way you play a game,
synthesising it and then replaying it.
Subscription streamed play , no consol required , we had
Stadia coming out as GDC earlier this year. Google into this
now, no console, we just send you everything down your
broadband. Mobiles will win , not my prediction, I don't
like mobiles. However they are prevalent, a capacity to do
more in a simple compact form . The only thing holding them back at tthe
moment is the batteries. I'll say that VR will go mainstream.
There have been at least 3 attempts at VR games industry.
But the pickup over VR has been substantial . Its in architectural
simulation , for training motor mechanics and doctors ,
military. Its big and its growing.
Then when computers build games , we know enough about the method
of building games, procedural generated environments .
Q: Have they already tried?
They have tried but they were awful. There have been a few games where the
entrant was a comp and it looks like a Wolfenstone knockoff.
The building blocks are there . Computers will learnt to play games
, but they will be too good , which is kind of ironic. Trying to generate
interesting opponents for humans is that we actually have to make
them dumb. They have access to all the data, they knw who
you are , know what you're doing, where you're looking. They are the
ultimate ghost in the machine. They know just when you are changing your
clip, and they pop up and shoot you. There is no argument about them
being hard to beat, they don't play like how humans do.
We have things like heat-maps , we watch where people play games,
generate hesatmaps of where people hang out. On top of this we can
build decision loops . Let people play games and there is a hotspot
where people always die, the AI knows to avoid that area.
So a layer on top, a decision map . If there is another place where if you
stand, you always get kills, then another decision map. You can build up
very good systems just based on just these 2 maps.
Then neural networks and deep learning can learn to play
effectively. Then the problem is controlling their developement of
learning because they will reach a point where they flip-flop
from beating them reasonably easily , to not winning at all.
My current research is navigation in non-orthogonal space
and i'm looking for mathematicians to help me out. Its a big
area of problems where we understand how humans navigate
thru space and a good understanding of how effective grids are
in vg, from Astar algorithms , but the problem with grids are
they are too regular and they don't join together easily.
So research in how to go off the grid, so to speak.
Q: With consoles, surely mechanically they are better than
mobiles or PCs, they're made for gaming. No latency
between pushing a button and the code operation.
Or an absolute minimum compared to scanning across a screen
of a mobile say. ?
Yes, I totally agree.
Q: But counter-intuitively things seem to be progressing the wrong way?
Then the people who only look at websites on a mobile, I
will only use a pc. But mobiles are taking over ,
because people have them on all the time.
There is no accounting for people. The culture and society don't
necessarily make the best decisions.
All the effort into hi-def graphics and its all lost on tiny screens.
We're in the wrong age demographic to understand it.
The big money goes where the mobile is. There will still be a hard
core of gamers who want a massive screen in their living room .
We know we need a new controller/console. You can span a lot
of keys on a keyboard quickly, but keyboards are'nt ideal.
There is a divide in the market where there is aplastic thing with a
dozen buttons at the most or a plastic thing with 120 buttons.
We need tactile VR and then we can have as many buttons as we like.
There are hotkeys, but there is a lot to be done in the human/computer
interface. We can make fairly solid predictions on what PS5
would have and yes we'll still have a controller.
A slide you had, based on an old book, where the number of consumers
grow and the number of producers shrinks. If you look at the
music industry , thats not happened. The number of people generating
music has grown exponentially, the tools to do that are there and easier and
in everyone's pocket now. But its harder to make any money, perhaps where
that graph still applies. The number able to make a living shtinks?
THere has to be a threshold of competence. Its like u-tube , massive number
making such videos. You can consider them creators bu t do they have a financial
stability, maybe that was missing from the graph.
For vg some are getting very expensive but some are getting very cheap .
Some will hit on a success and will get catapaulted . The best predictions
on the big games are that they will double their money , wheras predictions
for indy-games , some do 100s of times more revenue coming in than
expenditure. Its a market of competition . Knotch and his Minecraft started i na
bedroom, he didn't really have an audience, he just liked making it.
One of the most successful products in terms of investment you
could ever think of.
Do you happen to know what makes a game addictive?
Lots of possibilities of different gameplay. Someone called Heisenburg
had a theory of hygene for what makes people happy at work,
things like agency and control . There's no amount of pay that makes
people happy. Give people agency, control and challenge , that makes them
happy. So in vg , to change the player, to have agency and control
gets into feedback . We know that if you put those things
into games , its fairly sure they'll be hooked. That cycle of play-reward,
you get into a spiral.
So it could be induced by the marketing deptment?
Allied to that , why so much spent on marketting , when there is so
much "social media" around , to magnify the word-of-mouth?
Some of that marketting budget is social media.
Again comparing to music, in the original vinyl days. Someone gets an
album , and someone else hears it , but had dismissed that genre before ,
he goes out and buys it. These days is it because their friends take it up
, or is it direct marketed to them? What makes them take up a game?
Its direct marketting, we have the data, not via friends.
The problem with vg , if something is not great, you don't get
word of mouth. This is why direct marketting can compensate for
shoddy games. In 2016 I was surprised to see signs for Doom and
Red Redemption, who need to be told that Red Redemption 2 is
coming out. Every fan out there , knows this is coming out. Why
do they need to market it, all very strange, spending as much
on the marketting as on the game. Thats insane if it doesn't
work. I'm convinced that advertising doesnt work.
They don't have to reach the fandom, just the people on hte
fence. Plant the idea in hteir head, repeatedly seeing the mentions,
then when they see it come up on their media feed, perhaps via an
influencer, they'll go "i'd better check it out, see what its like".
I've talked to people in marketting firms and they
don't know what works that well. We know what doesn't work.
Its not possible to do a blind-test. So do a test, don't
advertise RedDem3 , will we still sell lots of copies. No way of filtering
that to see if it works. You spend 500 million on making a
game and then another 500 mill on marketting .
I'm intrigued in the concept of watching others. I like
watching snooke rbut try and play yourself, its hopeless,
you can't make a decent break. You need someone who does it well.
I'm not familiar with Minecraft is it popular because its so difficult to play, that its
more enjoyable to watch others who can play it well?
No its easy to play. It can be fun to watch others play it badly.
So is it like the TV series Gogglebox, getting people to watch other people
looking at TV?
Yes and because of advancements in video-editing , there is comedy that
comes out of the video-editing.
For a particular streamer of video-gaming his performance is much more
entertaining , because he does ??.
So many reasons, just having 5 minutes to watch something, not play
the full game, or pick up tips .
I scoffed when I first saw others watching vgamers but then I
scoffed when I saw the first version of Minecraft, thinking who would
buy that. I don't think we know enough about human nature , to
understand why people watch others doing vgs.
I've seen u-tube commentary about others on u-tube, so how far off
for u-tube commentary on a player of a vg. Dr eus and the bee-watchers.
Is it because we have too much free-time. Instead of mining coal
we watch dancing pixels.
Gaming of use in medicine and education. I wondered , as I',m
a psychiatrist, how things are moving in that direction, the
cross-over. A colleauge is using it for phobia control ,
developing VR , works quite well , as you're living that
experience, but not having any danger attached. ?
There's some great work in mixed reality VR. That is where
we see the real world and the VR at the same time.
A great BMW one , for training mechanics on their vehicles
with VR/AR specs . You can see all the car and it will light
up the nut to undo next, saying which spanner and torque settings.
That makes experts out of novices, a great way of expanding
knowledge transfer and technical abiliies. You could send a VR/AR
kit out to a very remote place and guide a doctor
through a very delicate appendicitis. We has humans
have the ability to do delicate things, but don't necessarily
have the understanding to do it.
A couple of my students have worked in an area of
de-sensitising , curiously, a bunch of people with a phobia
about using telephones . A researcher with me has been
doing work on autism , agsin with AR/VR. I think the
sky's the limit for VR because you can do so many things, but limited
at the moment be lack of tactile feedback. I think we've satisfied
latency and issues over resolution . Motion-blur is another
interesting sticking point, crack that and it will expand to a whole
new group of people, because a lot of people are immediately
sick due to it.
By tactile feedback you mean goggles plus some sotrt of ha[ptic
You could feel the pressure . Microsoft have been working on
haptics and hand-tracking systems and vibration systems.
I can't imagine being able to grab something in VR.
At the moment i'm happy that the controller vibrates , telling me
I've got hold of something, that will do .
With VR can you grab something and feel how heavy
No, everything feels like polystyrene in VR , all weird. Your brain
sees a bowling ball but your arm feels nothing .
You'd probably need a whole body suit, to resist your whole
arm movement. You can get used to many things, I can walk thru walls in VR
. I said recently to my daughter, remember you can teleport,
a strange thing to find yourself saying.
In the early days of BBC beeb pc, the kids learnt to
programme by liking to program games and then
move on. When you're looking for students for your courses ,
what sorts of qualities or skills or minds to lok for,
they don't have to be mathematical , but perhaps be logical?
Numeracy and literacy.
That all? numeracy is just arithmatic.
I teach art predominately. The fundamentals are that I
can go a long way with someone who understands the maths
of art. If you hsve a good appreciation of art but not good
maths , then its a real struggle. Because you are in what we call
"The Glass" , you're separated from your artwork by a sheet of glass
and the only way you can communicate with that art is if you understand how
a computer works. Understand the maths behind making that picture appear.
Don't understand that maths and unfortunately the industry
ddoes not have a place for you.
Would you need an Alevel in maths, ? be able to program before joing a course?
Yes, ideally . We used to have an interview process , a thorough way
of filtering students . Nowadays a complete political/economic
spectrum on uni-level education.
So take someone like me doing physics, and did not realise how much
heavy maths there was. For someone on a gaming course ,
can they end up dropping out because they were not expecting the
As an educator , I'm disappointed in the underaluing of numeracy in the uK.
There is aeven a large body of people that take pride in that they
don't get on with maths. Computing is basically glorified
mathematics. A lot of the future we can see is software
programming and engineering. If you cant do maths, you'll struggle.
We don't teach the interesting parts of maths , the exploration,
the excitement , patterns, the discovery . We teach rote-learning
, do 100 subtractions, now additions , now multiply. But a
computer does that . The first thing I showed my own kids
when they started at school , was how to use a spreadsheet.
Boomph, ther's all your homework done , in one mouseclick.
Then you can do the interesting stuff, look for patterns.
Maths is looking for patterns, finding them .
On computers doing coding, it reminded me a bit of doing Scratch,
scaled up to 3D.? Small kids manipulation of physical blocks
to program robots etc?
Scratch is great and MIT have another version called AcIinventor?
a more grown up version. Someting called Gadoe?, not as efficent .
With computers cant program concept , fundamentally is because
we've written a bad programming language.
By bad, is our logical constructs lend themselves to the problems
of paradoxes and the halting-problem. We need to get thru to
writing structured commands that cannot have a paradox within
them. We're entering philosophy here. Inately we're not
good at it because our language is so mesy, so full of
abstraction and ambiguity.
If we end up with a paradox, tough luck, it won't do us
any harm , lets move on to the next thing, it
doesn't make us fall to the floor?
Occasionally we as humans do crash. There are problems with the
human computer . But if we want AI working
safely , safety is the key aspect, then we need to solve some
of these problems. You cannot afford to have a computing
device having a problem with a paradox .
Say, I make a really good stamp-collecting programme and the computer
decides the best way to do that is to kill all humans
and turn them into stamps.
Program in moral thinking?
Then you get HAL in the 2001 film, would not detonate itself .
It all lies at the human end not specifying clearly enough what they
want. I'm a programmer as a living. The biggest problem
is teasing out of my clients exactly what they want? Most of the time
they don't know , until they see it?
They have a picture in their head , but its not solid. Then there is the
trolley problem. You have a large wagon going along a rail track.
You have a split , on one track there is 5 people and the othe rtrack ,
one person. The trolley is going to hit the fivesome, do
you flick a lever , save 5 but kill 1. It emerges in the world
of computers driving cars. The creator of Resource3 , Mika?,
he conducted an experiment live recreating this. The psychological
effect on the people who flicked the switch . The people who
didn't flick the switch didn''t seem to care. Those people died ,
but I don't take responsibility. 2 people in that experiment,
who flicked the switch , were traumatised. This is the complexity
of moral decision making. Imagine a computer being upset because
it had run ouver a dog. How do you build in empathy in the
With morality. If someone does something bad to you, is your
response , do something bad to them. To teach them a lesson
or do you give them so many chances. Someone called something
like Axylrod . He got a number of programs to see which woulfd
give the best response and it turned out to be tit-for-tat?
An early example of trying to get computers to handle moral solutions?
I believe th ewinning strategy was I'll do something nice
and if you do something nice back, I'll reciprocate.
Do somethin bad and I'll reciprocate doing bad . That program
won out over all the other very complicated ones.
So logivally we stop having guns in vg?
We'd love to get rid of guns in vg , but cant find a solution .
Guns and vg are a kind of hybrid of what is the simplest
interaction . It would be nice to talk to a computer ,
have sentences and have a conversation, but we can't do that.
So apparently we use guns. Its an interesting deconstruction,
that the easiest way of resolving any conflict is to overpower
your opponent. Same if an alien came into this room, we
cant have a conversation.
There was ages ago a faux psychiatrist program called Lisa.
"I'm feeling really depressed today."
"oh, you're feeling really depressed"
just reflection. But there are now sophisticated programsd that
you would not know you were not talking to a trained therapist or
(From the psychiatrist) It is a developing area, developing apps
and programs like that , as there are not enough psychiatrists
or social workers , in the UK.
If its helping the patient , then its useful. Some people react to
computer CBT programs more positively than to a therapist.
There are lots of facets to the games industry and technology
which are external to the industry, which are used in the industry
and influences other industries. We've talked a lot here on
non-vg , but still relevant. What other stuff gaming tech does
, moral questions as well. Does this go back into the vg industry?
Yes. One of the good things about the vg industry is it makes a lot
of money, which ironically , makes money in other areas of
research. The big thing about GPUs leading to machine learning
which is a huge boon generally, compared to if everyone said, I'm
not buying a graphics card. Its pleasant to know there are
positive aspects to vg. One other thing to use vg for, is to
support anaesthesia , because if you can distract someone by playing
Mario , you can do more to them, before they register pain.
Computers playing games and internal AI , you didn't talk about
external AI, Go and Chess games etc , what are your thoughts?
Once you start remote computing stuff , you can spend a lot of
time on it. There isalmost a branch-point between logic AI
which is how to develop winning strategies and the human
perception AI , I'm more interested in the human perception
stuff. Basically AI cheats, because it knows everything and the
only thing it doesn't know , in say chess, is your next move.
We've tried brute-force i nthe past and its pretty much incomputable.
The Deep Learning has had to come up with better algorithms.
For me, getting past the computer cheat of just looking thru
all the data it possibly can , then good chance of knowing your next move
will be because it has so much access to data. Rather than looking at the
screen and make some predictions , based on fair information .
If you flip this round, you were playing a game against a computer
and it let you dump out all the memory and scan thru , so you
know what its thinking , it will be much easier for you.
Perhaps 100s of people combing through the computer, so someone
knows what it will do next.
Thats why we use computers, because their better at that than we are?
Go was interesting, like chess a game of limited inputs but so many
many outputs. The Delta2? exampe was mind-blowing because its
almost infinitely complex? The computer started doing moves , that were counter to the human move , that were way too quick for me to counter
and doing strategies a human just would not think of?
Its almost verging on creativity , but not , as its only analysing stuff?
For me, chess , showed great scope . How many infinites are there, chess
is difficult , but then Go is phenominally difficult , that no one thought
computers could be doing, even just 8 years ago.
That Go program self learnt compared to Deep Blue that was trained
by humans. Now its gone on to plau Doehow?, now its limitless , what is it
capable of doing. Doing things that world experts in Go thought , thats not a good move , but 12 moves down the line, was a good move.
One of the classic signs of intelligence , is questions. I try and teach
my students, the more fundamental the question , the harf=der it is
to answer. Most of these questions have been high . But students often
find they get down to a really fundamental level and think this sa
a stupid question. But it can be profound, almost to the level of
what is it "to be".
I assume its still the same as my college days that , even if the
lecturer asks if there is any questions, no one puts their hand up
as all too scared to be shown-up amongst their peer-group. ?
Yes. If you get to the end of your lecture, ask if any questions,
there ar enone, then you've failed.
Can you pick on someone and ask them to explain something?
You can but they might not come to yoyur classes any more.
I have a giant pink inflatable beach ball and pass it around the room.
Thwere isa bunch of rules , but basically you get to ask questions.
Even if they know the answer, it still encourages them to ask
questions , because they can then ask questions of other people.
It breaks the tension, I can ask a question now.
Silly quaestion, Whats Minecraft?
Look on you-tube and you'll find people pklaying it.
Its like Lego . There isan eductional version of it used in schools.
I could give an hour lecture on what Minecraft is.
A simple explanation , break a tree, use the tree to make a thing to
break a rock , use the rock to make a thing to break an Awk,
use the awk to make another thing to break more Awks.
Its like 2001 A Space Odyssey, you start as a primitive creature with
very primitive tools and hopefully you will be a wizard one day.
Tuesday 21 May 2019 Peatland Ecosystems: Proxies for environmental change and current management. Anna Leveridge, Jordan Paker, George O'Ferrall of Soton Uni
Peatbogs are very wet , generally found in colder places. Not a lot
appears to be going on in terms of biodiversity but later on we'll
find it is important.
Peat is the layer of sediment that builds up , when organic matter
decays. It can be thick and so useful for re-constructing
past environments. There are 2 main types of peat bogs, blanket
bogs which look pretty homogenous when looked at from
afar. They hug the landscape and look like grassy hills etc.
Then reaised bogs that occur mor ein isolation , accumulations
of peat rising above the water table. So they are fed only by
precipitation rather than precipitation and gtound water.
Another one is the acrotelm which is the active peat layer at the
surface , where things get deposited and all the plant matter
decays. The sediment corer, a metal cylinder thrust in the
ground and can pull out a really long circular
core of the sediment.
The distribution of peatlands worldwide. Mainly found in Asia
and N America . Globally 4.2 milion sq km is peatland ,
which is about 3% of all land area. Europe has quite a lot especially
in Scandinavia. A significant difference between the peatlands
that build in the north of Europe and the south, because of patterns
of rainfrall and temp. They prefer colder temps and more
precipitation to build up. In the UK blanket bog forms the largest
semi-natural habitat. Important for bio-diversity and conservation,
because they are good at regulating C. They are eco-systms that
have high acidity and should be waterlogged if in good condition .
They are anoxyc which means no oxygen in there at depth.
This explains why peat accumalates rather than decomposes
unlike othere systems like forests. Without oxygen the plants canot respire .
Important habitat for lots of wildlife, species adapted to acidic
conditions. So endemic plants, ie plants only found in that kind
of ecosystem. A key species is spagnum moss , which produce
phenolic compounds which produce acidity , the main driver of the
conditions there. They are conserved under UK and EU law.
The Habitats and Wildlife Directives of EU law , transfered to
UK law on Brexit. Specialist species live here because of the
niche conditions. Not many species could cope with the low
pH and constant waterlogging. So GreenShank, Dunlin , Red Throated
Diver and Woodlark. And similar variety of invertebrates .
As these peatlands change potentially due to anthropogenic
forcing like extracting peat for peat-turf fuel , the ecosystem
changes and as it changes sometimes it becomes more like
heathland and changes the species variety supported.
So the benefits that humans can derrive from nature , 4 main types.
Culteral services, can't be easily defined in monetary terms.
Like taking a walk within nature and cnsequent well-being.
And the aesthetic value , wild and atural. Provisioning
services, that can more easily have a price tag attached.
eg fuel in Ireland 17C, animal bedding and as soil
conditioning for horticultural purposes. Regulating
services, benefits to water quality alone are worth 15 billion pounds
in the UK. So the way the environment reacts and responds to the
wider environment, such as sequestering C , mitigating
floods as peatland is essentially a sponge, protecting
The historical archive that the sediment core can produce.
Using a combination of plant remains and pollen, its possible
to reconstruct pictures of past landscapes and past climates.
In the case of the UK peatland as far back as 10,000 years.
So why bother reconstructing past enviro. If you can reconstruct the
past vegetation of a site, then you can reconstruct its climate
and how the vegetation changes hsppen according to the climate
change. Of relevance now , because we can look back into the
sediment history and see how bio-diversity has responded
to potential climate change in the future.
Marine sediment cores and ice cores are used also, and for
much longer timescales . But on the local scale lots of benefits as being
more accessible and economic as no boat required and don't need to go
to Antartica. There isa large range f proxie measures. High temporal
resolution , so you can study many changes over a relatively short time
frame. Autochtonous, developed at site, not transferred from elsewhere or
redeposited. Maybe every 20 years rather than ice-core more like every 100,000
years. What can a tube of mud tell us, they contain , pollen , micro-fossils
and tephra. A proxy is something you can measure , and from that
measurement tell what something else is doing.
You can find traces of things like lead and titanium within cores,
and this can show human influences carried into the air. Traces of lead
can show the timing of the building of roads and the shift to unleaded petrol.
Titanium is a measure of erosion , can see when changes of land use
occured , just from measuring titanium.
Pollen under a microscope has different characteristics
for each species. They can be difficult to tell as not always nice
and flat, can be folded or torn. possible to create pollen diagrams, bottom
scale is percentage of what pollen species in the core . So birch, pine ,
alder ,beech, lime. Each species likes specific conditions. Godwin in
1940 saw this and that in some areas these patterns of
change are similar, he created a pollen-zoning system. Can be used to
compare with different areas, as C dating is expensive. Micro-fossils
are basically small pieces of plant matter , like seeds ,fruits ,stems .
These enhance the pollen record, as you can only see species that
produce pollen . There is good taxonomic precision , compared to
pollen which is generally limited to only family level , you could
say its Pine, not Scots Pine or another pine.
Different cores from different sites will have different depths.
Say in the New Forest , more organic deposits available, so
may get deeper faster than areas of ?? may be more shallower.
So how can we take cores from different regions around the world
and look at them in the same contexts. For that we use tephrochronology.
Tephra is volcanic ejecta , any ejecta, could be large boulders
, volcanic bombs, lakoli? smaller rocks and fragments. But here its
ash, tiny shards of minerals and rocks. Such ash gets pushed up
in the atmosphere , sometimes the stratosphere in large events.
Then its deposited around the world. This occurs on a rapid
timescale , in geological terms. This tephra deposit occurs in
about 1 year, so that one tephra deposition event in one
location will be the same time as any other location.
A thick layer would be alocal eruption If a very thin
layer of ash ,often hard to identify , appears as these shards
in otherwise organic. Each volcano has its own individual
fingerprint of spectrum of minerals. Find one mineral makeup in a
tephra location and the same or very similar mineral makeup in another location ,
they probably came from the same eruption.
So you can draw a line between them, although apearing at
different depths in the cores. Take the different depths and you can
create a model for age v depth per location. We can date the tephra
layers by likes of argon-argon dating but usually using mineral dating.
Tephra dating is the main technique for synchronising cores
from different locations, its not always plain sailing.
Tephra deposition is effectively synchronise in geoplogical
timing sense. Near volcanoes give deeper layers of tephra , but farther
away may get a thin layer or no layer. In extreme cases with cores
at the same site which don't have tephra layers in them
but move 10m away and have a different layer profile to it, less
shards . So we have to take multiple cores from
the same sites and average out what you see. You have to check that
cores that show no layers, from the same same site, are representing the
same time series. Also animal activity and plant growth within the
deposition area, so the often fragile ,thin tephra layers ,
especially in the northern hemisphere , where there is less volcanic
activity , can be disturbed . So faults can occur where material
slides down , so can appear younger or older simultaneously .
Also vertical mixing from tephra being deposited on top of a living
plant producing a more diffuse tephra layer. You want to identify the
paleosurface , the surface where you can say at this time , where we
dated this to , is the surface as far as internal composition.
The plant matter and the geological matter at that layer.
You cant say , this time period looked exactly like this,
just a geeral idea. With tephrachronology we can date volcanic
eruptions very specifically to each volcano. We also get apicture
of what disasters wer elike . I the last 10,000 years there has been a
lot of eruptions and disasters can define eras in geology, extinction
events define new eras of life on Earth. We come up with a very
real picture , that we can compare to today, to understand how
disasters happen. A core with thick deep tephra layers in a short
space of time , so many eruptions in one area could be hostile to life.
In recent history its possible to track very recent eruptions.
A study looking at tephra layers in Ireland , could track eruptions of a
singl;e volcano in Iceland over 1,000 years. For recent Icelandic
eruptions we'll be able to see those layers of ash in only 100 or 200 years .
So future geologists and geographers will be able to re-establish
our current environments.
Its possible to use volcanic eruptions to date eruptions on other planets,
not using peat cores of course. Coring on Mars would show ash
An important efect of peatland is carbon sequestration. Possible to us e
natural systems to take our C overload back out of the atmosphere.
Peatlands are possible the largest C store on the land of Earth.
Northern Hemispheric peat alone has 25% of the world soil carbon.
These peatlands can exchange 20 to 30 gram of C per m square, per year.
England alone has about 14,185 sq km of peatland, so many tons of
C has been sequestered in a year. A large ampount of C stored in these
peatlnads. Globally estimated to be between 400 and 600 GigaTons.
So compare with what we have of estimates left in our C budget
of about 375GTons. So they contain as much C
, than we have left to put into the atmosphere , before
we do irreversible climate change. Around 1000 Gtons released since
the start of the industrial revolution. There is difficulty in measuring this
potentail, due t permafrost. Occurs primarily in the N hemisphere
, as most land mass is in the north. Permafrost is a layer of
frosted ground that pretty much never melts season to season.
Methane is trapped in the frost as bubbles and in the micro-organisms in the
soil that is frozen. When unfrozen , will decay and release
methane and othe rC compounds , back into the atmosphere. Similarly with the
bubbles. Video of igniting and exploding methane previously
trapped in permafrost. We're already seeing permafrost melting
rapidly in most of the world. This is just the surface, way more buried
deep under the surface , trapped there while frozen. There is twice as
much C in permafrost buried , as there is in the entire
atmosphere right now. If we melt all the permafrost and put all that
C back into the armosphere, then a serious negative impact on
our climate. A positive feedback loop , that melting, C release, increasing
heat in the atmosphere causing more permafrost to melt.
We're not so sure how much C will be released when it melts.
Its so recent a phenomenon, we've not got good measures of it yet.
Organisms decay at different rates , some may not decsay, some
might re-activate and start sequestering C again.
Wet peat acts better at sequestering old c than has been emitted,
so could end up overall more sequestering than releasing.
The worst case scenario , the Clathrate Gun or Clathrate Canon.
A hypothesis that methane from permafrost will be released
and once released can never be sequestered back because the
offsets from revised peatlands will not be enough to resequester.
So we may already have set off something, we won't know
until its too late. There may be clathrate methane released from clathrate
mounds at the bottom of the ocean. They are under continetal
shelves, that s something we cannot deal with. But regards
peatlands we should now be ensuring we can sequester as mmuch C as possible,
we can help there. Mitigating against further peatland permafrost melting.
Peat is a good source of fuel, its C rich. People living in upland
regions, harsh environments, historically extracted it for fuel.
Then unintended consequences that so often come along with environmental
issues. To extract peat you need to drain it completely , the C stock
changes. Change 1 thing , there will be a series of knock-on effects.
Lowering the water table, damages the sphagnum moss , it cannot
grow, no peat accumulation, and total loss of the peatland habitat
and loss of its species. Instead of individual crofters, the issue now is
commercial extraction, less frequent but larger effect. Sometimes for fuel
or more often for horticulture.
So drains started on the sides , containing a raised hummock of
peat. Eventually the sphagnum stops growing as the water table is lowered,
C starts to be released. Forestry can take over peatlands as its very
flat , easy to work upon. Trees are very good at tsakig up water,
lowering the water table and peat growth stops totally.
Many thought planting trees in such places was very "green" but
ended up losing the unique habitat of peatland.
For agriculture again ideal landscape for likes of rice growing,
which needs flat waterlogged land. The worst comes from livestock
farmig , they graze nd will easily over-graze and trample
the key spagnum layer on the top. The natural slight undulations
will flatten out and loose structure, water table will
lower and trees and shrubs will be encouragesed to grow and again peatland loss.
Groouse shooting has had a large cultural and economic effects.
A large part of upland economy comes from grouse shooting.
To rais grouse requires a very specific management regime , involving
drainage, fires and sometimes kill off predatory species of the grouse.
Again you tend to loose peatland and get a heather dominated landscape.
Wildfires - benefits and problems.
Fire is part of the natural ecosystem. Fire causes a successional
change in vegetation. This can be benefit but also encourages the growth
of small shrubs and heather that are much more flammable than spagnum.
Try to set fire to a wet sponge. There are threats particularly , Malaysia, Indonesia
anf Peru, where peatlands are often near large urban areas.
So when they burn , very harmful air pollution to the humans.
Nearer home, Saddleworth Moor near Manchester burnt last
year. This highlighted the deposition of metals in heathland. When they burn
the metals and toxins are released into the air. Linked to respiratory
issues for the young and old particularly.
Wildfires are predicted to increase under climate change. We will
see more Saddleworth Moor events. The effects on peatlands from
climate change. Increase the temp of the atmosphere , the risk
of wildfires and peatland loss increases. Therefore more c in the atmsophere
and then more warming , positive feedback again.
Peatbogs are more vulnerable to drying than to warming.
However with climate change in the UK precipitation is likely
to increase in most areas. So in the UK would be beneficial
but in other areas like Indonesia likely to be detrimental.
Most upcoming effects wil bwe negative with climate change.
One threat is invasive species. If temp does change , it alows the
climatic range of species to encroach. Spagnum moss is what we want to retain
but is easily out-competed by othe rspecies. Reproduce quickly, spread quickly
and more likely to use the new resources that may come with climate change.
While peatland vegetation will struggle to adapt. Things like rhodedron
will quickly invade. Humans laying down pylon lines across
heathlands , is a good nursery for invasive species. They grow to a point
where they can then spread out. People inadvertently carrying spores
and seeds on themselves, and even domestic pets is an issuue.
A lot has been done and a lot of research, because they are so important
but so vulnerable. THey are so sensitive to change,
once humans start meddling , then unintended consequences.
Moist layer transfer, you take some acrohelm and sphagnum
and move onto areas to start the peat productiuon again. Things like
peat mosaics will start it of. If you can convince a farmer
to create such stepping stones for species to move acrsoo.
Also long strips of linear habitat within agriculture
improves connectance and reduce isolation effects.
Counter though is improve the connectance and you increase the
chance of wildfire and invasive species spread . So management
plans can have negative consequences.
Removing trees where they start to invade. Improving the
water table is important, will help restore dried out peatland.
Fires do break a dominance that can grow and give other species
a chance. We don't know what level of fires is natural
in origin. A simple remedy is just blocking drains.
Walking by humans and vehicle traffic can be managed
Legislation varies between countries, but protection of heathlands
is not number 1 priority. In Ireland there's no specific law to protect
peatland. Indonesia 2017 , protection legislation there was shot doewn be the
supreme court because of effects on local likelihoods.
As often 2 sides to these stories.
Some f the things you can find in peat bogs, Tolland man found in
1950 and he lived in 4C Denmark. His preservation is superb , due to phenolic
compounds . Bog-butter, people in Ireland used to bury theor
butter in peatland for preserving. It is sometimes found and dug out
and tastes like butter still. Plastic is preserved in peatland ,
including microplastic because its ever present now. Mammoths,
best chance of DNA cloning a live mammal from the extinct past, as
so well preserved a full genome from the body.
With Cathrate analysis, with the melting of permafrost, methane would
be released. Some tens of millions of years ago, the Earth was a lot
warmer than today. You did mention it was irreversible but with the Earth
hotter at one point in time, if there is methane in permafrost now
doesn't it mean it is reversible and if the case, how severe if it does contine
to melt. ?
Its reversible in a geological timescale, but climate change only really
started in the 20C. If permafrost melts now then more than
doubling the amount of C released into the atmosphere.
As for if reversible, there isan intereseting argument that if we can stop it
from hap[pening at all. The world was very different 10 million
years ago, it was very different 3 million years ago. The current ppm of CO2 has just reached the
the level it was 3 mya, but the Earth was very different then.
Continue with C release and Earth will become something we will not
recognise. If permafrost melts it will go beyond the current situation.
For IPCC RCP scenarios , no consideration is made of future
permafrost melt and extra C release, it only considers human released C.
So worst case scenario would be entire equatorial belt countries being uninhabitable
and mass extinctions. By the time we could possible reverse it all , all too late.
We must keep an eye on it but the fingerprint to it will only turn up
after its started. Scientists go out there and test the bubbles for
methane. Peatland fires in Russia released such methan and the fires burnt for
many year, because so much of the gas beneath them. We should
be looking at peatlands for indicators , if we don't change our ways.
There was a period of 5 to 7 degree C warming during the mid-pleistocene,
about 800,000ya . That has been hypothetised to have been caused by this
clathrate release. So we've seen it before but it doesn't really
matteer if its reversible or not because we've seen it before and in that earlier
period something like 70% of terrestrial life got wiped out , due to that
temp increase. S owe want to prevent it rather than reverse it in the
For the marine deep deposits of clathrates , is it normally kept there
by both temp and pressure?
Its the calcification depth . The threat is sea levels decreasing which will
not be happening.
Are methane hydrates a different process?
Methane hydrates are more like trapped gas but clathrates are solid.
It then dissolves and releases gas. In the mid-pleistocene it was
changes in sea level that caused release. Sediments on the continental
shelves were disturbed and as they fell , they shifted the deep ocean
With the tephra layers , do they spread far and evenly enough for
you to sequence them like dendrochronology?
Thats what we use them for. Some eruptions don't go far
enough. You have to account for the fact that the eruptions that
do turn up tend to be very large ones . Sometimes the layers are so thin, they
are missed, in cores where you should be getting tephra.
That particular area could have been protected by wind shadow.
For pollen analysis, how do you tell the difference between say one pine tree 10m
away depositing pollen and 10 acres of pine trees 10 miles away? and pollen generally can
travel hundreds of miles?
Some species have longer dispersal distance than others. By using pollen which
they know where its come from and work out what percentage of
pollen can be found in a core. Some pollen from grasses and
gramaloid? can travel very far. They know that if there is more than
5% of that pollen , then it indicates more long distance dispersal
than immediate dispersal. They can apply arbitrary units on whether
something has been dispersed far or just local, a reference.
What is generally the state of peatlands in the New Forset with all
the cattle , horses and people trampling around?
Historically I think the most effect would have been the rearing of
agricultural livestock. Nowadays they are stringent about what
is allowed on the land.
About 5 or 6 years ago the commission had a change of policy.
They started actively blocking some of the drains. Certainly in the
south of the forset, so the ground could be rewetted, not for any
agricultural reason. It has a greater conservation value and
significance as bog or semi-natural bog than it does as very
poor quality grazing land.
Would you know what is happening in the very North, Caithness
after the war, a program to introduce fast growing
conifers, big plantations of Sitka Spruce. Lots of drainage went on ?
Not specifically. I cant believe it would be good.
Basically juniper tundra , and when the wind got up , they just all fell over?
Unintended consequence. If people want timber they'll plant
You're not aware of any programmes to re-purpose.?
There are always vested interests involved .
Are you aware of our local mystery of the bog bodies found at Fawley?
When Fawley Power station was built 50s/60s. Men in white coats
took away the bodies. The official record has it they were bits of old wood
not bog bodies. Experienced groundwork crew , fullly familiar with
bits of old wood, recognised the leathery appearance. Someone contacted
the "Ministry" and instead of the men in white coats dumping the
so-called wood on the site bonfire , took whatever it was away and nothing heard
(Ironically) In my experience you typically wear whitr coats to remove
bits of wood. It may have been considered a bio-hazard.
It would be posible that some experts would come along and take
away just pieces of wood for some further analysis, dendrochronolgy or
Yes, entirely possible.
There's quite an extensive layer of peat under a lot of the
Soton area, deposited meso-lithic sort of times, at the Boldner Cliff
sort of ground level of human occupancy, now well under water in the Solent?
Tuesday 18 June 2019 Florentin Bulot, Soton Uni: Low-cost sensors for
monitoring air-borne particulate matter (PM), using Raspberry Pi .
The 6 St Denys monitors , including live data,
Every year there is about 9 million deaths associated with
air pollution. Thats similar to cigarettes. In the UK its associated
with about 40,000 deaths a year. What is special about air pollution is that
it varies from one street to another and time of day , like at rush hour.
That makes it hard to assess what people are inhaling.
Map of PM around the world, in the UK its not too bad compared to
some other parts of the world.
PM pollution comes from many types of sourcres.
Building and construction pollution is mainly fine PM.
Agriculture pollution is mainly from ammonia.
Fine particles are basically dust that is smaller than 10 micron.
Anything less than 10 micron is PM10. Anything smaller than
2.5 is PM2.5. They come from traffic, shipping , industry , natural
sources like dust/fine sand from the Sahara. So not necessarily
harmful dust, it depends on the source. The smaller it is, the
further into your body it can go and potential health effects.
In the UK the predominent source of the finest pm is the burning of
wood and coal, 38% whereas transport is only 12%.
Woodburning stoves have been promoted recently , but its becoming
the predominent source of fine PM in cities of the UK.
Fine PM when lodged in th elungs cause a range of respiratory
diseases, like asthma. In the blood can cause blockages of blood
vessels leading to cardio-vascular diseases. It is being linked to
diabetes and dimentia.
In the UK we have about 300 monitoring sites called the Automuted Urban
and Rural Network, AURUN. They are reliable monitors and its what is
used to show whether we are complying with legal limits. The gold
standard. It provides 2 kinds of reading , some are roadside
and some are background. Roadside gives us an idea of what people are
inhaling walking along that road. Bckground are located away
from any local sources of pollution , and give a more general
idea of the concentration in a city.
One of these stations cost 100,000 GBP to install and 3,000 GBP per year
to run. Only 6 such stations around Soton, you cant say this street is
worse than that street .
The sensors used measure the mass of the particles directly. The air enters and
is heated , because there is a humidity problem, hence different weight.
Particles collected o na filter , linked to a vibrating plate, the frequency
of vibration is linked to the mass of the particle. And so hourly data
of the mass of PM. Low cost sensors for PM work differently.
For fine particles , inside one is basically a laser and a photo-receptor.
So an optical sensor. Every time a particle pases in front of the
laser the sensor infers the number of particles. Then transforming that
number of particles into a mass. So not direct reading of mass , which
is how the legal limit is defined. Because of that , there are more
limitations. Its sensitive to aerosol composition in the air, its
sensitive to the concentration of PM. So if little particles , it is less
precise. Here we are measuring near the limit of detection so a serious
issue. Also sensitive to humidity and temperature.
The costs of these sensors are lower, the technical requirements
are lower , so can be operated by a larger range of people.
Because of that its possible to create dense networks of sensors
around a city. To start to understand what people are
actually inhaling. The reference stations are good to
have but to understand what people inhale , it only shows what a
person would inhale if he stood all the time near a reference station.
It becomes possible to identify sources of pollution . With a network of
sensors deployed around a city , you can start seeing pollution move
around the city and start to understand where its coming from.
It is important to involve citizens. The st Denys project is creating dat a
that people can relate to. There are mitigation measures we want to impliment
, but knowledge of hotspots/times is helpful in that.
It compliments the existing city network . But the big question is
what is the quality of this data produced, its reliability.
When can we trust the data the sensors produce , just knowing that would
be a great progress, but we don't yet.
Some readings taken at a site in St Denys bewtteen 24 April
at 4am to 2pm. Between 10 and 11 all the sensors follow the same peak
and then another peak later, they all show the same trend but looking at
each value individually recorded , there are big differences.
So these sensors are good for reporting a trend in pollution
but absolute values recorded should be handled wiht more caution.
These sensors cannot be deployed on their own, they need some electronics
for the comms, protected in a box .
So the solution I've developed to put the sensors out around the city.
The Pi takes power and has an internet connection via one lead,
its modular so can just swap out
a sensor. Today we're looking at PM but diferent sensors could go inr
there. If fine particles are no longer a problem in the city ,the
sensors could be replaced with something else and contine their use
for monitoring. There are 6 of these boxes arounfd St Denys to spot.
The public access plots are for readings every 15 minutres.
Difficult to interpret as different peaks occuring at different times.
If you see a truck passing , you can see if anything is happening on the plot
nearby. Shows the morning rush-hour peak, a peak at 4pm so perhaps school-run
then the evening rush hour till 7 or 8 pm.
Comparing my boxes to the reference station , the average of the 6 boxes
locally , they agree quite well in values. The box is waterproof
and removing lid ,shows the 10 sensors iin one box. 10 because of the
variability of output , the spread of data for different sensors.
We take the average of the 10 sensors and 6 boxes .
My earlier deployment was in 2 Soton schools. The reason behingd that study
was to compare different kinds of low cost sensors in a year long
comparison. The outputs are dependent on the composition of the
air being monitored and the time of year having more or less sources
of pollution. The cruise ship season is in the summer mostly
and winter there is wood-burning. We also wanted a wide range of
temp and humidity . One question we don't have the answer to
yet is are the sensors degrading over time, or do they need replacing.
So we compared sensors against each other of the same model and
different makes. We compared them against the reference station.
The 2 schools were St Johns primary school , French St near the
docks and Sholing school . Only one ref station in the city gives
PM2.5 readings which is the one we used.
At each school we deployed 3 boxes and 4 sensors in each box.
One box near the school entrance, one in playground and a third one .
Comparing the sensors against each other , the variation between
different sensors . If 2 sensors have a coefficient of variation of 0
then they perfectly agree. The threshold , from the US EPA is 0.1.
If 0.1 then good for fine particles. We had over 0.1 , close but over.
So 1 of these sensors cannot be used to monitor air pollution.
Thats why we moved to 10 sensors in one box. Was 4 ,now 10 per box,
it is also easier for us to do comparisons with 10. We still don't know
how many sensors are really needed in one monitor ,somewhere between 2 and 10.
We studied 4 types of sensors , the Plantower5 against Plantower7 and the
scatter plot is almpost a perfect line. Both from the same maker, different versions, PMS5003 type used in the St Denys boxes and LoRaWAN quarter-hour data feed for the
public internet .
But looking at other sensors Alphasens compared to Plantower5
there are 3 modes within the scatter plot. That shows if we apply a
correction method to one type of sensor we would need a different
correction applied to another type. These 3 modes are
partly explained by different responses to humidity.
They don't report the same values , in high humidity. Humidity doesn't
explain everything though.
The biggest difference between ref station and the sensors is that the
ref station is giving a reading every hour. While my ones are giving a reading
every minute. The absolute shortest time interval is every second , but that has a
lot of noise , so upped to every minute. All the city sites follow generally
the ref station . For that we've studied the correlation and obtained
? coefficient between 0.61 and 0.88 which is quite good for the sensors.
When we look at the correlation for different levels of background pollution
for the ref station fine particles below 8 . For the lowest levels the correlation
is less and above about 17 the correlation is better. So the sensors are
better at monitoring high levels of pollution .
25 is the legal limit here , so we want to be below that. So it is harder to
use the sensors in the UK , not that we can't use them, just harder to apply here.
So do the sensors degrade over time. So we looked at the correlation between
ref station , in different months of the study.
A pronounced dip on all sensors in August , but after that it returns.
For 2 sensors a straight line showing the performance does not degrade over a
year. For the AlphaSense one you see a lowering in performance. Maybe due to
dust accumulating in the sensor or component degradation .
The dip in Auguist has 2 potential explanations , the wind was mostly
blowing from the coast and we also see a general dip in correlation
for days when the wind was coming from the coast. Maybe more sea-spray
with the particles . The other reason for the dip was barbecues , around 6 or 7
pm and mainly fridays , I'm convinced registered barbies but were not
registered by the ref station.
These sensors can track very short lived pollution events , like barbies.
Also for fires in the city , like the Waterstones bookshop fire in Soton city centre
we could see it in the readings.
A reading at the Sholing school 25 August and the wind was blowing
between sensors , the peak registered on P2 can be seen a few minutres
after the peaks at P1 and P3, so giving a direction of track of pollution, in this case a local barbie.
If we take proper care of the data produced then these sensors can give us
more localised and persom=nalised data on air pollution. To understand
what people are breathing in , while travelling in the city,
because currently we don't know. Another concern, currently unknown is what
peole ar ebreathing while indoors at home.
It will be possible to monitor this high frequency data , wind direction and
data about ship movements and general traffic , then we can get a better
understanding of sources of pollution. We can interact with the populace
to act on air pollution , with the help of these sensors.
The current correction method we use. A paper only published last week.
2 ways. 1/ to know the characteristics of the p[articles , know the
composition, know the density and a few other parameters about them.
Someone else in Soton is working on this .
The second is mor eempirical , based on a regression model , based on
humidity and temp. At the moment no consensus on how exactly to
correct the sensor data.
The next considerations in my PhD , so the schools sensors, the st Denys
network , I'm intending to expand the network to get a better understanding
of pollution in the city. The intention is to produce a map of pollution
in the city. At the moment , the only maps are based on models
and only a few measurement datapoints. We hope for better
precision to the map , with the city-wide network of sensors.
One place of placement of sensors is within the port, I've an agreement with
ABP who are willing. Also sensors at Marchwood so we've an idea
of what is upwind of the port as well as city side of the port.
I'll be characterising the sensors in the lab , controlling the air composition
, temp and humidity . Up to now all my sensors have been
some distance from the ref station and hoping I was measuring the same.
I'll be placing some of these sensors in Vietnam and Morocco. where there is
higher levels of pollution and the sensors would be more reliable there.
In Morocco there is a lot of dust from the Sahara , dirtier cars with a differnt
sort of fuel to here. A totally different environment, and so better to
understand how to correct the sensors.
When you described the sensor box you had data and power along a cable to the
box, but also a wireless antenna. Why 2 data routes?
Every time we are studying a sensor , we need to find a local internet
connection and drilling thru walls, all rather messy. This is a back-up solution to
collect data. With the LoRaWan network you can only send little
data, the maximum rate of output we ar eallowed is every 15 minutes,
sending just the average. Ultimately want to get rid of the hard-wired internet connection.
You don't have an SD card in the Pi?
At the moment 15 locations of these boxes and 150 sensors so I don't want
to go around opening boxes and collect SD cards.
But wouldn't the SD cards store data for each of the 10 internal sensors every minute?
Actually every second . With the wired connection and SD cards we get all the raw data
, but in the end , when we've sorted out sensor correction it will be redundent.
Say you find sensor6 is definitely erroneous you can ignore its data rather than
have it contaminating the average output?
In the end we intend to reduce the number of sensors , currently these are boxes for
testing proposals. Another project just started is how to get this box smaller
. The sensors themselves are low cost between 10 and 100GBP but the whole
box is 1000 GBP. The project just started 2 weeks ago would only have 2 sensors per
Do the sensors have a fan, or natural ventilation?
The sensors have a fan , very small , but otherwise no fans. We could have a
fan , but cost and power consumption goes up. At the moment consumption is
What about contamination of the sensors , they must eventually
Yes, thats why we wanted to study degradation. When you open up thse
sensors after a year you see see dust in there. The sensors are facing
downwards , so dust tends to fall off .
When i first came across this project i thought you must use reference smoke of
some sort. Plumbers and gas-fitters use test-smike for checking flues and
chimneys. They set fire to a little preformed tablet of something that
generates blue or orange smoke. I'm assume the smoke output is fairly
standard in its gradation of particle size. But you don't seem to use a standard reference
smoke at all, for initially setting up these sensors? Perhaps set off one in a metre cube box with a fan and introduce to a sensor?
You are purely comparing to the ref station in the city?
There is no standard smoke in a can . Some studies use incense , some use cigarettes,
polystyrene vaporised and Arizona Dust which is becoming a standard
in the USA, basically dust from Arizona . But no consensus on any
standard dust or smoke. NPL produces gases in cans for standardising and are
just working on cans of standardised smoke. What you said is what I'll
be doing in the lab . Placing a number of boxes in an hermetically sealed environmental
chamber and introduce some PM. I'm not settled on the source of PM
that I'll be using . Also varying the humidity and temp inside the chamber.
Do you have any problem with condensation inside the boxes in UK winters, causing problems with the electronics etc.?
So far no problem. So far I had to change 1 of the 6 . The boxes are warm
from the electronics inside and no condensation issue.
We are running the LoRaWan network, the base-station owned by the uni.
One of those we opened and water flowed out of it, but not these sensor boxes.
Assuming you get the number of sensors reduced and overall cost less,
where do you see the uptake of them? Informing public policy perha[s?
There is a website where you can find the cleanest walking and
cycle routes , is that the sort of application?
Thats one but also public bodies generally , checking that measures
they take actually work.
The council is considering a clean-air surcharge on vehicles entering the
city. So they will only be tackling 12% of city pollution at the best?
There are groups involved with air pollution but so far have
had little data to underpin any actions. If say a group gets 200
residents to go to the council , over some measure to take .
There is "clean Air Southampton" I work with. All the readings are
publically available on the breathingspaces website for this St Denys project.
Large ships in the docks and the container port straddle carriers
ar eworking 24 hours a day. ?
We don't have actual measurements of it yet. NO2 pollution is mor elinked
to road traffic. For shipping its NOx, CO2 and SO3. For NO2 ships are
less significant than cars, just by the models, as we don't have real data on
it. Ships burn much dirtier fuel, so PM output can be much more
toxic . So say the pollution from ships is 5% it might be mor e
harmfull than tghe other 95% around the city.
another uni group is collecting data via a different monitor, like a big
vacuum cleaner, er inside and do a chemical analysis on
the filtered . Thet found there were more harmful particles in the summer
than in the winter, work in progress .
Do you measure the temp thru the sensors as well?
Yes and the humidity.
Why did you go for PM rather than gas sensor?
Gas sensors have been more studied, spo for PhD purposes it made more sense
to go for PM . Gas or PM need completely different techniques , so I
had to concentrate on one only. And fine PM is the more harmful
part of pollution , compared to gases. Now we are cleaning up the pollution process ,
ridding larger PM,
fine PM will come to the fore.
There is no regulation at all at the moment for ultra fine particles, quite concerning in
Have you picked up pollution from other places? I was looking at your local
data and one night about midnight there was a peak at about midnight but nothing on
the other local boxes. Then i was thinking, the same with the one on this hall,
its quite near a gas-burning boiler flue. Do gas burners output PM?
Yes , any combustion process will produce PM. Its also possible for a sensor
to be simply wrong.
Do you know how far such particles do travel, say from a major fire say in Winchester
and the wind coming from that direction?
Some people in London can phe signatiure of ships ick up , all the way from
Soton. It may be just traces , but it can be found.
I was surprised to see that on one random sampling weekday day recently that
161 diesel trains passed thru St Denys.
What about the big petroleum industrial complex at Fawley?
Its upwind of the city so yes. There is a big mix ofg sources of pollution in
Soton. Traffic, the port, airport, Marchwood incinerator, Fawley
refinery. Its not good to live here but its useful to study pollution
here. It makes some sort of sense.
I was doing this sort of thing about 30 years ago , our low
cost sensors were over 5000GBP 30 years ago. It emerged that
November 5th is the time that industry burns anythng it possibly
can . We had 3 sensors around Fawley and nov 05 stood out. ?
A study in Amsterdam found similar peak at New Year.
I'd not thought about people deliberately burning nasties at those times.
With your high frequency data have you found anything that could be precisely
tied down to say planes passing over or mainline train service at a regular time,
My work is to characterie the sensors and less to make sense of the data.
I don't have the time to interpret the data for that sort of purpose,
much that I'd like. I'm sure it would be worth somone trying to
correlate repeated minor peaks with say a regular diesel train service.
It would need a longer tmeperiod of 2 years or so , to rule out
other possible repeating factors,. The high frequency data is available on request.
Are we supposed to discourage friends and families from having barbecuses?
Yes ,and wood-burning stoves. Even the best such stoves , one such is equivalent
to 6 trucks running. All very dirty but its good for the environment
in the sense of being C neutral. Unforseen consequnce again.
Ships burning gas rather than diesel is better for pollution
but not necessarily for the climate. The average distance of a car travelled
in a city is less than 1km. Do we need a car to travel 1km or could we
Tuesday 16 July 2019 Dr Christine Currie, Soton Uni Mathematical Sciences: The Mathematics of Pricing
The work we did on ferries was with Red Funnel in Soton. No commercial price sensitive info here, they liked our vehicle packing info but they have other ways of doing their
pricing. We initially worked with P&O ferries bu tthen we moved to
Red Funnel. Most of this maths work relates to Red Funnel but it will
be generally applicable. S opricing according to ease of packing and increasing the
efficiency of the packing. Awhole research area described as cutting and
packing originally relating to minimising material wastage for cutting out dress
patterns. We will be using the similar techniw=ques for optimal
packing of vehicles on to a cardeck of a ferry.
Big vehicles take up lots of space and awkward to pack so you might think
we should charge them more. Often the big vehicles are freight that come
day in, day out all thru the year. So charging those users a lot of money is
not necessarily a good idea. Wheras tourist traffic is just for Easter break say,.
Red funnel has a ferry terminal in Soton and run to Cowes , Isle of Wight.
31,000 sailings between mainland and IoW and back.
3.4 million passengers transported and about 873,000 vehicles
since 2016 and still increasing , so they have introduced a freight ship as well.
In some ways that has solved some of the problems as that means they can rid
some of the awkward trafic onto the freight only ship.
A related problem is how to arrange the cuing of vehicles on th eland
prior to embarking. There are also WightLink ferries and hoverspeed hovercraft
services to the IoW, so there is competition. There is pros and cons for
choosing relating to road access on the mainland. Soton is easier to get to
but a longer crossing time. So how canwe improve efficiency for Red
RM= Revenue Management. How to maximise revenue by dynamic pricing.
RM appeared in the 1970 for airplane costing after deregulation of the
airline market in the USA. Prior to that prices were highly regulated and
you could not vary them much. Low cost carriers came in and American
Airlines had to compete with these low cost carriers.
So they started to put conditions on their cheap prices. They could sell
tickets for the back of the [lane cheaply, undercutting the competition
because they knew the high payers would then carry over for the whole operation.
Its easier to imagine this in terms of planes than ferries, so the diversion to planes.
Planes have fixed seats and csan only have 1 person sat on 1 seat. They take up the
same seat space whether in economy or business class.
We know we have X seats to sell on this plane. THere is whats called
perishable inventory , meaning after the flight has gone you cant sell it again.
There is fixed capacity, very unusual for a carrier to find another plane
of squeeeze in some more seats. Then the allocation problem we are trying to
solve. If we charge nothing, the demand for those seats will exceede
the supply. So more people than seats, if not, then it becomes a different
problem. So how do we set the prices so we get the right number of people
on the plane. Set the price too high and I wont sell all theseats.
Set the price too low and there would be people who would be prepared to
pay more are paying a lower price. What price do I offer and when.
For dynamic pricing we are changing the pricing going thru time. Its
not one price for the whole season , typically it increases. The close rto
a time of departure, the more a person is prepared to pay. There is
randomness in there, how many people actually turn up for a flight.
More people will be thinking of purchasding nearer to departure time,
giving a peak. That peak will move dependendent on whether a leisure
or business flight. We know that but we cannot quantify how many
will turn up for that one flight. Forecasting comes in there and also
statistics. To get a feel for what this variation is. Also getting an idea
of what people are prepared to pay and how that varies with time.
Then 2 questions almost the flip side of each. What price do I offer
and when , then how many seats do I offer at a price. So we can have 2
different models, one the price varies over time, and the other a set
of price points and a set of seats I will sell at those prices. Closing off one
price point ater selling so many. As a consumer all you see is that the
price keeps changing. We tend to look at dynamic pricing, the price changes.
Customers arrive into the system , so they want to buy a ticket. We assume this a
is a ??? term,
random process with a time varying rate. They arrive randomly but price
changes over time, generally increases closer to departure. Thats while people
think of buying , check prices of competitors , phoning up for offers etc.
Then another term , the probability that they purchase a tickeet. An asumption we
make that this is dependent on the price being offered and the time at which they
look. The closer the time to departure, the more likely they will buy
at a given price. The highr tthe price, the less likely they will buy.
There is a term called Reserve Prices , the price you have in your head
,the most that you are prepared to pay, say 150. Someone else would have a different maximum in their head of 50 and someone else a maximum of 500.
So a distribution across the population , Reserve Price is related to Probability of
Purchase term. We cannot see these prices in their heads, having to rely
on historical data for that, a tricky problem. So you will have some
distribution of prices and rates lambda of t, then we need a method to
optimise the revenue ,for given prices. With the important capacity
constraint , you wont get an infinite number of people buying.
The problem and the assumptions going on uinderneath.
Now look at packing relating to ferries, a bit like playing Tetris,
then how to combine the two.
For a Red Funnel ferry , the drivers drive on and drive off, RoRo ferry.
No turning round of vehicles. There is a mezzanine deck which can be
raised or lowered , dependent on what turns up and can be altered between
sailings, quiote a quick process. It increases capacity for smaller cars.
You effectively get another lane of cars. The space below the mezanine is quite
low , limiting what can be placed there. There are 2 of those mezanine as one on
each side. To choose to use 1 or both of those moveable decks is another
optimisation problem. There is lift acess and some vehicles need to be
close to the lifts for people who cant walk far. There are lane markings on the
decks which makes it easier to solve packing but reduces the efficiency of
packing . A free-attack to the problem of different width vehicles i easier without
lanes. We created 2 solution versions , 1 with lanes and 1 without.
This is the main deck which has mixed traffic, then the top deck
which is cars only. Less interesting as the cars are much the same size. Even so
its possible to squeeze 1 more on by parking it diagonally.
Most of the traffic to the IoW comes on the ferry, so supermarket
delivery lorries, building supplies with width overhanging loads ,
and even large yachts on trailers. There are prison vans with special
requirements as the prisoners are locked in during the passage,
so they have to be placed under the sprinkler system . Sometimes hazardous
loads cross. all sorts of complications. The loads on the main deck are
kind of rectangular . Manoeverability has to be taken into account
, so some spaces cannot be used because of turning circles.
The experienced loaders sort this all out in their head, but have a new loader
and you can see the packing difference, less vehicles loaded on.
So some of our product is for training of loaders. Gives them some practise
before they get started for real. There are 2 ways of looking at this and relates
to the painted lanes. Called the Exact Way and a Simulation Heuristic.
When selling tickets and looking at packing , how do I define the space of a
vehicle. The way of solving , by the most accurate process is to say
I've 3 of type 1 , 2 of type 2 and 1 of type 3 and 4 of type 4.
Each number is effectively a dimension . As you get to 20 different vehiclr types
, which Red Funnel easily have, then you have a 20 dimensional
problem which gets huge to handle as a huge number of vehicle
combinations to fit. So this works , but grinds to a halt over a certain
number . For a 1D problem of fitting to lanes , each lane is a bin
and we're packing htose bins as efficiently as possible.
The other way , allows for all the complications of a massive number of
vehicle types. Instead of this number of each type of vehicles , we look at how
much empty space we have left. Thats how we can define the system.
This speeds things up because we've 3 dimensions . So you have some
area left, what if I pack a big thing next , will I still be able to put
it on. This is why we have simulation heuristics. This packs across the
whole space and not constrained by lanes , makes it more efficient.
The exact method where we are looking at the numbers of each vehicle
type. To maximise useable space we call U, U is space that hasn't been
trapped between 2 lorries. If vehicles were at an angle you would
end up with unusable deadspace between. We are looking for useable space we
can park something in. Plus a symmetry breaker, which increases efficiency
, otherwise if you have lots of lanes all looking much the same , the optimisation
gets confused as to which lane to place into. Its an integre program as we're looking
at integres. We have integre number of variables . All the terms are linear as they
are linear in the variables. Such as a given weight is one variable.
So we are trying to optimise amongst a lot of constraints.
1 constraint , I;ll have to activate an extra lane, a "big M" constraint.
Basically its saying try not to add he extra lane if you don't need to.
2 constraint, limited to lengths, cannot hang over the edge at the end.
3 use the lane with most space available, use it sparingly.
On the other side is the simulator, it doesn't have a set of equations in there,
less mathematical in a sense. Looking at how the ferry is being loaded, looking more heuristically, for optimisig. Its assuming ??? loading on to the ferry, it has the lanes on
the upper deck and then the 2D packing problem on the main deck.
Where our main interest is with the bigger mix of types.
For the simulator, like human loaders , its assumed to take account of the
real world constraints mentioned, lift access , mezzanine deck position,
dropped trailers (lorry beds without traction units which are left on-shore and use the ferry
companies highly manoeverable tractor units) , reverse gaps where something has to slightly
reverse before puilling out. The result is a quick and effective method to optimise
packing efficiency for any particular set of loading types . Also gets used to
prevent over-selling , what we and RF would like is iif they sell a load
then it will all get on the ferry. Having such a simulation package , it
can also function as a training tool for loaders, loaded on my laptop to play with.
Its calculating the remaining space that can be used. It moves the shape along and
calculates the remaining space , the sum of the rectangles left behind.
The heuristic element loks at all the vehicle types that re there on the
dockside . For each of those types it looks at where I could load it, for the
best placing. Then which vehicle type to load next, its not caring where they
lay on the quayside. It assumes it can get to any of them at any
place in the queing on the dockside, that is the next piece of work for us,
where this becomes a constraint on accessibility prior to loading.
It makes its decision by looking at various arttributes and run lots of times
with lots of different scenarios and see what weight we should give each of the
attributes. Tightness is a factor, packing a vehicle in where it just fits
rather than having lots of space around it. Space loss due to stagger parking,
needs minimising. Overlap adjascency ratio ???. The weighting factors we do as
a result of data done in advance. If we are running it for real we'd have the
weightings in there , sum the result and pick the one that works best.
Part o the process is a simulation annealing algorithm.
Q: There is no issue around mass balancing the ferry, say all the lorries on one side?
There ar eissues like that , the most recent version of this has a centre of gravity
that moves around and is output as a visual display , but its not really part of the
optimisation as its not a massive problem as the ships are quite heavy in itself
and they don't know how much these lorries weigh. They ar enot weighed o nthe
dockside, a lot is what the driver says . When we first showed this program to the loaders a
few said we wouldn't do that configuration because all the lorries are down one
side. So we try and balance a bit better. Also if you load something heavy at the
bow , then the stern will rise and can mess up the loading ramp positioning.
The CoG doesn't move around a great deal.
So we've maximised space and now the pricing objective. A decision
variable would be the price offered to each customer and type of vehicle in
each space in each time period. This is how I define space comes in.
Having a nice clear description of the space is important as this will
affect what we know and what we think we know about the system.
Then the capacity constraints , we also need to know whether we can fit these
vehicles on the ferry. Unlike the airline example of 100 seats and so
100 tickets can be sold, there is an area constraint . They can overbook but thats another
talk. With a ferry you can overbook if the estimate is not right. We need a
reliable packing algorithm to allow this.
We split into lots of discrete time periods and go with one booking into each
of those discrete periods, that assumption makes the maths easier.
Q: A minute time period?
The whole selling period is about 1 year. Each discrete period tends to be a
few hours long but it depends on demand. High demand leads to shorter period.
So at least 1 cutomer will buy one ticket in that period. We ignored complications
like group bookings , which tends to be rare for a ferry.
So the model may have a couple of empty time periods where nobody
turns up , then a car with a ticket turns up. Then you have to decide at this
point , an empty state how much time is left, you know what price you
are offering for that vehicle type. Wiht ticket purchase that changes the
the status of that discrete period and that may change the price to the
next offering. You keep going like that and at a later period you
have a probability of a certain type of vehicle ticket purchase.
2 things can then happen buy or not buy. So a tree-like structure ,
does something turn up, if something does turn up it may buy or
may not. You are loking ahead and how you set your price is then
dependendent on how much you coul;d get if you didn't sell to
that person. The technique we use there is Dynamic Programming
, its looking ahead to what could happen . My expected prifit
at atime t and a particular space, I'm trying to find a price that maximises
that. They arrive at a probility lambda(v) for v vehicle type
and this one will purchase with a probability a. If they purchase I
get an immediate revenue P but it won't be giving me any revenue in the
future. The next time period has the same equation
but it will be in a different state as one vehicle space is sold.
This is sort of the opportunity cost, the cost of having sold that ticket
in that time period.
The next situation is arrive but not purchase (1-a ) probability in which
I'll progress to the next period but will be in the same state.
I'll be one step ahead but the state remains the same.
Also a possibility of no one turns up and no purchase. Dynamic
programming can look at this sort of situation very efficiently
almost working your way backwards to get to the answer.
The diagram of the probability structure. If we get to
high price and we're a long way out , very few people will purchase.
If at a low price , even zero , and right at the time of departure
people are more likely to pay.
The exact packing with hte slidefull of compound summation
equations, can be high dimensional for all vehicle types and ccan
be long computer time. We use that set of equations to
define a Pareto? front, where if we put 1 more of any vehicle
on that ferry , then it would be overfull. This is the efficient vehiclre
combination and there's no point in looking outside that front as we
know it will not fit. This limits our safe-space being able to work out what that
is. With the lane structure formulation we use the heuristic structure
when we are a long way departure, we don't need to solve this big
complex program. Only nearer capacity or departure time do we need to
worry about solving the exact equations.
If there are 2 vehicle types there are 2,271 different states . 5 different
vehicle types its close to a million and that is reflected in the time
to crunch the problem. We can solve quickly for 2 but increase to 5
then solving the combined packing and pricing problem the processing
time is getting huge. Go to 20 and its becoming impossible.
The expected revenue , you get better returns if you split the
vehicles into a higher number of classes. For 2 classes , we say anything
less than that size iclass 1 and bigger than is class 2. So you end up
putting 5m vans in a class that goes up to 11m and a lot of wasted space.
Thats why we get different revenue results. For 3/4/5 types we can get a
reasonable revenue from the eact solutuion. For the heuristic one and packing
over the whole space rather than just lanes we can get much higher revenue emerging.
It givesa more efficient atack and a more efficient understanding of what is
Another project was with a big E-grocer 5 or so years ago before
delivery to your door had taken off. We wanted to have cutomer
based pricing optimisation as to time slots of delivery. We wanted to
encourage customers to pick a slot where the a delivery van was already
delivering just down the road. Imagine being i na far off village
then you would be given a cheap slot if you chose similar time to
a delivery to a neighbouring village.
A www page showing the green vans showing the most efficient
slots. A customer would get a whole suite of options for times
and also include green-van slots , with hte prices .
How can w euse the price and maybe the green-van to
encourage people to go for slots that are beneficial ot the
deliverer. Cheeper , more efficient and generally better all
round. So we had to build a model around how a
customer veiws the slots. We would design a tentative route plan
based on the forecast and then inserting the cost of an additional
drop. Used an almpst identical Dynamic Program as the ferries one, just there
was a cost function in there , how much cost to serve that 1 customer.
So modelling customer behaviour, how would they choose between
slots and would they purchase if the delivery cost was too high.
So optimising, then update your forecast routing plan as a result.
So almost the same method as the ferries but on a completely different
problem. These programs improve outcomes, but only based on
simulations, so who knows what happens in the real world.
Sometims a high-value customer would get a negative delivery cost ,
because you could haul them in and keep them shopping with you as the
overall profit is worth it, and also ensure they are in the slot you
want them to be in.
Future work ;- ferries again but multiple drop-off like Channel
Isles ferries, or Scottissh Island ferries where there is a sequence of
drop off at different ports and also other traffic comes on,
all potentially out of sequence. Put something on for a short hop
and you've perhaps lost space for something to doa long journey
as well as the packing problem.
Air-cargo is another one , you're never sure what will turn up
with air cargo, very uncertain so a lot about balancing uncertainties.
Delivery of large awkward shaped packages , so can be plant-hire
vehicles on the back of a lorry , like diggers and ecavators, how do you
do the deliveries. A nice one will be how to deliver packages to
space-ships. For a future with more commercialised activity up there
how do you price the delivery of such material, you've the weight , the
volume , watch this space.
As a non-mathematician , seems a lot of the parameters you're wrestling
with overlap with the container business. I know they have a fied box
, but they do vary in size , and fridge-box containers , multiple
destinations . ?
The regulariy of containers makes a big difference, although there are
differences in size , its not too different. There is stacking as well
though. It comes under a big category called rectangular packing.
Drop trailers problem?
Therre is an issue about how many such trailers you put on a
crossing because RF has had problems with not enough time to board and clear
lorries in the turn around. Drop trailers are kept separate and only a couple can
be moved at a time.
The booking system should be taking in info about your car or vehicle including
dimensions and place straight into the program?
At the moment they give everythinhg a car-equivalent unit and they will look
at keeping that nimber below a certain level , somewhere aroud the 200 mark.
They know that if they had loads that was all cars ,they cannot fit them all
in with that number of car equivalent units. If they have a mixed load they
can fit them in a bit more easily because they are a bit over-generous on some of the
car eqivlaent units. Thats the model they use at the moment but its not
Can they refuse a booking and wait for something better to come along?
We are trying to use price as a lever, so if something big comes along and its looking
quite full , then they will see a high price. So a maed out pricing.
Its rude to say no but set them a high price , thats acceptable.?
Presumably there's a sort of mathematical relationship between supplier
and regular customer on that as well. those customers will have some
insight as to when they will get a better price or some sort of feedback perhaps.
Wheras an onspec car customer would not have.?
Definitely. The regular customers are looked after very well. Other than
freight contract, most of them are island carriers so they will see
slightly different prices going that way. Game theory comes
into this .
Also an element of c-operation as well as competition?
Yes. There are ways of modelling those interactions. We've not done it
here but generally yes. It varies product to product , how important
those interactions are. There is a perhaps mythical story of a
USA professor who would try and book flights that he thought would be
over-booked, so he could get free airline vouchers. The worst that could happen
is he'd have to fly somewhere .
Have you looked at digital ? pricing ?
We've looked at pricing for web pages , clicking-on and also
radio adverts. Radi9o is quite highly regulated and restricted on the
number of assets you can fit in.
Is there rule-of-thumbs with the time-served loaders knowing say a particular
vehicle type turns up , they know a load of motorbikes can go in a specific
space next to it, that sort of knowledge base?
Yes they know a lot of that sort of thing. We keep saying we're
not trying to do them out of a job . They will even assess you as a driver
as you drive up as to how tight a spot they will put you into.
They must have collected data on regular users. Can they look that up
later, yes we know this vehicle , we know this driver?
No . I think its more assesment on the day.
So not a customer database?
There is but not used for loading purposes.
We're trying to develop something so the result works like how the
loaders would fit . They know what they couldn't do , so when it comes to
setting the prices , then yes or no. And also for trainee loaders.
I'm always surprised the short turn-around they have , 55 minute crossing
and most of the time 1.5 hour with turnaround at both ends, no
spare ship to call on. I've never seen breakdowns causing problems or
It can happen . At Cowes in particular they are trying to improve
the situation there . The loading yard used to have a road in the
middle of it, and their vehicles had to cross it.
Also vehicles coming off the ferry went in front of vehicles boarding ,
so no loading and unloading at the same time. so they'd load the car deck
while unloading the main deck for eample. They're currently redesigning the
I've a motorhome ,ring up to go to IoW in July
then something like 300 GBP return but book a weeks camping and
bring my motorhome then about 90GBP, etraordinary gymnastics in
Yes they have arrangements with different companies on the island ,
the loyal customers bit. The campsite becomes the customer and not you.
We did some related work on customer lifetime value . Then the switch point
between RF and WightLink say.
Any eamples of loaders doing favours for their friends?
No , none seen.
For the airline problem it used ot be , start with a high price , then
effectively a Dutch Auction and a standby price a tthe end. Your work
seems to be much the same as all online retailers and travel companies have
these days. That then obviates the need for standby pricing or still
a role for that , if unfilled capacity at the end, sell at a cheap rate. ?
It often depends on the market , its knowing when is the last
minute for airlines . You could have someone who really needs to get to
a meeting and they may be prepared to spend 5000 for that, right at the
last minute, but if you sell it for 200, you've lost , so its often orth keeping it
for that high price , perhaps 1 time in 100 one will appear but
onaverage you're no worse off. Its balancing off the probabilities
and the costs.
For your dynamic programming of pricing, how dynamic is it in
allowing someone to input new variables if something changes like a change of
fleet freight operators .?
I had a PhD student who had a project estimating these models as you go along.
If you have pricing models , say entering a new market or the market is
changing , what prices do you put out inthe market. So you are learning
but making money at the same time, so you have to be strategic as you'd
like to know their reaction to any price. Charge something crazy you'll not
make any money. Then the parameters get updated.
How low a level of staff can change your programming structure?
They would tend to be updated automatically via management systems
constantly feeding dat ain .
You don't have to have a coder come in?
generally not , only if the logic of the process changes.
The pricing algorithm is looking for strange behaviours. Like
every year there is a moped festival on the IoW , can RF all of a sudden
get all those on one ferry.
For the ioW music festival there is a lot more car traffic but because of the
demographics for that , they tend to be smaller cars so can get more on.
Not big family estate cars. Things like that, the model cannot capture
but they will know about certain special ecvents.
Tuesday 06 August 2019 Dennis Doerfel of Reap Systems , Southampton:
Developing hybrid marine-diesel/electric propulsion systems .
first is the concept of hybrids, second is results of the ferry deployment we made
in Falmouth, Cornwall.
In 2015 I met somone at a conference and we discussed what we could do to
improve marine transportation, and he said about the hybrid concept .
We started in Venice , thousands of boats in Venice busily polluting the
water and the air, smoke and noise. We started with the Venetian
taxi boat , that can carry 10 to 12 passengers , usually from airport to
the centre of Venice. This Venetian taxi boat is now in Soton. We converted it
to hybrid , so we could learn about the technology . Its 9m long, not
very tall so it can go under Venice bridges. About 550 such boats in
Venice. The second Falmouth project we had a number of
partners including government funding.
Reap systems is based about 10 minutes walk from here on the river
Itchen. We've been passionate about electric cars for a long time
, buying my first one in 1998 and decided on this area as a career ,
doing a PhD in soton on Li-ion battery tech for cars. We started the
company in 2003. Other projects with another company in Farnham which
makes motors that sit inside the wheel, and we supplied the battery
tech and integration. That car was 300hp and are mainly in China now
because Chinese bought the company. Since 2004 we've supplied
systems to a German company that makes electric boats , many lakes in
Germany are not allowed to use combustion engines on them, they must be electric.
For the hybrid boats we made the whole system , excluding the
manufacture of the engines. We started with electric cars but as a
company we're too small an operation , so we decided to move into the
marine market. We made a small electric RIB and put it on the
boat show in 2013 to gauge the market. Everyone loved the boat
but no one wanted to pay for it, too expensive.
One problem with e-boats is the price of batteries, although coming down now.
Benefits are , no local emissions , even in global emission terms it is lower C
than combustion engine. It is near enough silent , regenerative breaking
is possible , on slowing down some of the energy can be back converted
to the battery, it is very efficient. Another advantage is if you want to rely
entirely on renewable energy , most people would go for electric
not liquid fuel . Take solar or wind power generation or other renewable
structures go for producing electricity first. I*ts moree efficient to
store in a battery than to create a liquid fuel as the energy store.
It is possible to produce methane from CO2 and energy but its not
been widely adopted yet, likely to come soon.
Diagram showing different energy storage devices , for storing in car or
boat for the propulsion. on one axis the energy density how much
energy for the weight versus the power density . Different battey
types on the market , some are high energy batteries for e-cars like
Tesla and a very high energy density but the power density is low.
For a hybrid car you would aim for high power density but you
don't need so much energy. In 5 years ahead this performance might double.
At the cell level 690 W-hours per Lt but at the battery level of combining lots of
cells about 400. Comparing that to diesel fuel , instead of 400 it is 10,000
,so you can store a lot more fuel . That may not be a fair comparison I
thought as a diesel engine is not very efficient compared to electric motor.
Allowing for the lack of efficiency it comes down to effecdtively something
like 4,000 for storage, but its still a factor of 10 better than for battery.
Hence the impetus to consider a hybrid , battery for short distances around
town with much breaking and regenerative recharging but use liquid
fuel for longer journeys.
The speed profile of Thames Clippers in London , on some parts of the Thames there
is a speed limit of 10 knots but other parts up to 30 knots. The boat needs a
lot more power to do 30 kn that 10 knots, not linear , not 3x more like 10x.
For 30kn the battery would have to be enormous but at 10 kn diesel is
very inefficient as its not operating at enough power output. In between there are
stops , a long stop at Greenwich and often the diesels tick over continually there,
to keep the generator and other auxillaries running..
So for a hybrid , run diesel mode when fast running and charge the batteries
and then for slow mode run on electric only, no diesels running and the battery
will supply all the auxillary power needs while in that mode.
But the question is will there be a fuel saving. So could there be a fuel
saving. The efficency plot of a diesel engine , pretty hard to find.
Everyone uses diesel and a lot of talk about efficency but try and get an
efficency plot from a manufaturer, virtually impossible.
Shows the rpm of the diesel, between idle and max rpm.
Shows the torque and the extra curve shows what the diesel is capable of ,
how much torque it can produce at different rpm. Shows up lines of
equal efficiencies, so the highest efficencies at specific fuel consumptions in essence.
So how much fuel does it consume to produce 1KW-hr for example.
The most efficent area is at a relatively low RPM and relatively high torque.
From that area fuel consumption goes up. A factor of more than 2 across
the plot. For a boat , has a propellor, and will run at any rpm
and the boat will accelerate. There will be more torque and more thrust on the
prop. So run the prop at idle speed, you have a torque and it follows
the curve up to full speed of the engine. The aim is to end up at the
maximum power point , because people tend to want maximum speed.
There are parts of the curve where a boat hardly ever operates .
Most of the time a diesel would be on the curve that is far off the
optimum. A simpler diagram just looking at the prop curve
, what is my fuel consumption at different rpm, high fuel consumption at
low rpm , medium rpm is good and then at high rpm it goes up again.
Most boats are operated around 80% maximum, the better
parts are not used routinely.
So for the Venice taxi boat its got to go really fast ,about 30
knots near the airport but in the small canals 2,3 or 4 knots.
So the boat will be operating in the non-efficient areas for very
long times. So the idea is to operate one section on electric
and diesel for the other section. When you operate on diesel ,
you will recharge the battery. So what is the effect on the diesel operation.
Running the generator forces the diesel up into a more efficent operational
zone and that works.
There are different types of hybrids in the marine industry.
Simplest is the diesel boat with engine and gearbox , prop
shaft and prop. Then we are working with mechanical power all the time.
Series-hybrid is the next easier. Electric motor and mechanical
power , gearbox, prop shaft and prop. To run a motor from a
battery you need an inverter because the battery is DC and you need
3-phase to drive the motor. The battery will be empty at some
point . The engine is diesel usually in boats , electric motor working
as a generator , with inverter and charger . Series hybrid because you
have to produce electric power first and then use electric power to
make mechanical power. On big ships there is already big generators
, for them to add a battery would not be too difficult.
The disadvantage of series-hybrid , from fuel you convert that into
mechanocal power, convert to electric power, convert to dc power,
store that as electrochemical power , take it back out and convert to
ac electric power , to mechanical power, convert to lower rpm ,
convert to thrust. A lot of conversions with losses each time ,
so potential fuel savings are probably eaten up in all those
The parallel setup tries to overcome some of these issues and does look
a bit simpler, although more difficult to impliment.
Normal diesel enging, gear box , prop shaft , in between is a
clutch and you can take mech power off and connect an electric
motor to this . Via the clutch the engine can be disengaged and the
boat can be operated from the electic motor or close the clutch
and diesel driving the prop, the motor acting as a generator and
recharge the battery. So less components , so smaller and
cheaper system and most of the components can be smaller.
The electic motor needs to povide full power.
Q: Can you run both motor and engine at the same time?
Yes. you can run as electric only, or diesel only , diesel
with the motor acting as a generator. What we've done, during
acceleration the electric motor is helping the diesel ,
reducing emissions, it sdoesn't really reduce fuel
consumption. The emissions are reduced because the diesel is working
harder. There are also othe reffects. So can do both and add a
generator and you can do more. Most bigger boats already
have a generator, often 2. The generators ar eusually inefficient ,
put on one lamp and the generator has to fully run. Thee it would be
much more efficient to use a battery for that situatin.
So a hybrid generator. You can do mor eon boats than on cars.
The Venice taxi isa planing craft , when goinf fast
a lot of it is coming out of the water, with lower friction.
You can see that in the fuel consumption . Generally a boat is
low efficiency at low speed but this boat , when it starts
planing its quite efficient for a while, till wind resistance and
efficiency drops again. So optimum is probably when just planing
and so reasonable speed but most efficient. Its a bit doubtful
about fuel savngs but there are other benefits of a hybrid.
Reductionof local polution for example , which is pretty
bad in Venice. There is less vibration and low noise from an
electic motor. We tried to create a drop-in hybrid , so can
be retrofitd to old boats. Boats can be quite old, the ferry is from 1934
and the Venice taxi 1982. Just the engine tends to be replaced and on the
next refit , could include a hybrid, thats the idea. Also as the
engine is not running so much , less wear and less maintainence
so longer life a nd a saving there. The engine would be operating at a better
operating point on the curve. Electric propulsion allows you to go
very slowly , so manoeverability goes up. eg going in and out of a lock
they have to continally put in and out of gear , a lot of
clonking going on. Don't need to do that with electric because a motor
can turn very slowly. So much smoother operation .
We had to tart up the boats as they are both old, as part of using them
as test-beds, what are any potential issues and benefits. People underestimate
what is involved. They see all these components are easily available
, so surely it should be easy to put that lot together , but its not easy.
But control systems are not really available off the shelf, a lot of the
money goes into developing software. Imagine you're an operator
of 30 ferries and you want to make them more fuel efficient.
You would want to know how much you will save, what othe rbenefits
and what is the best system . This complete aspect does not exist at the moment.
Systems integration is the name. Before someone commits themselves to buying
an expensive system . The hybrid system is about 3 tmes more expensive than
#a normal diesel only system . Wil la certain system be reliable , which
gives the best results, cost-benefits, what other options are there, can I
use this for PR for instance. Will it consume more fuel . Take a hybrid
car used on a motorway most of the time it will consume more fuel than a diesel
only car. Unfortunately there is a lot of green-wash around this.
Someone suggested to me that I should be selling just the product sticker,
rather sad. Our system we try to make it as simple to install
as possible. Its very short, for the taxi stern drive and no space in the engine
bay, so we have to make our bit small to fit in what space there is.
Clutch and electric motor can do 40KW and the battery is under a
bench. So it takes up no space that would be used by passengers or their
luggage. I think we're the only ones can do this at the moment.
So system consits of the control, high voltage Li-ion battery and its control
unit and clutch+electric motor and lots of cables. Its important that the
interface for the driver has no change, so no confusion introduced.
The only change is a little display , tucked away, that can be used to change
modes. We've used the taxi around Soton for testing and demoing
and we now have to upgrade it to the system on the ferry.
Then take it to Venice, find an investor and supply ?? because the hybrid
system is new tech and we cannot provide world-wide support.
About 2,000 boats in Venice wiith the same drive system and about
20,000 more that are slightly different.
The next presentation relates the results on the ferry, the fuel consumption etc?
Q: What voltage cells are you using?
380V which is more or less what electric cars are using, so we can tap
into the same sort of components.
Q: The taxi , is that the original engine or did you change the engine ?
It would work with almost any othe rengine , but we bought the taxi boat without
an engine and got a new engine for it. The ferry had a very old engine ,
maybe 50 years old. With an old engine we cannot electronically control.
Q: Does stop and start cause a problem.? The taxi starts from the
airport and then its off for 10 minutes , not like the stop/start of a car?
Modern engines can cope better with stop/start, not so much heat mass?
so designed for start/stop. Also allows electronic control of the engine becaus ewe
cannot be utilising mechanical levers. Old engines tend to be dirtier, but we
prefer to go with relatively clean engines.
Q: So the offer would be new engine and battery for a taxi driver, because they
would be in the market because of a worn out engine? The whole thing in one go as part of
normal investment point.?
Yes, In venice these engines tend to last only 3 or 4 years. The cost of retrofitting
a hybrid system to an existing system and running for 2 more years is not
viable, maybe possible in the future.
Q: How long does the new system last com[ared to the existing types?
For the ferry project in Falmouth ,about twice as long, because half the
time it will be on electric. For Venice taxi boats , it would be more as they
are running on diesel a lot, very slowly . The ferry is going rapidly back and forth.
I even had doubts whether hybrid would make any sense , for such ferries, but it does seem to
Q: Scaling up?. Is this just the beginning, can you scale up to quite
I think large vessels are possible. We've been here before. There's not much
market yet but a lot of competition already. Because on a world-wide scale
there are lots of companies and investors who think this has potential.
So we had to look at what is potentially our little niche. There are big
companies doing big ships.
Q: Did you study the Isle of Wight ferries at all. THe 2 big companies
are both saying they are green and hybrid , but don't seem to be in
port long enough to have any payback, I'm confused by this.
Hybrid in the sense that that is what is painted on the side,?
I think it makes not much sense. You can enter the port on both
sides in electric mode , for less emissions , but otherwise
pretty constant operation .
Q: In Corsica this year, tourist boats with 20-30 people on them
, left the harbour with no sound . THe PR value of this was
fantastic . It went into various caves then up comes the
diesel for the main crossing. A delightful experience.?
That is what we were hoping for on the Falmouth ferry.
But the engine is quite loud and an old boat so a lot of
Q: Compaired with a car , the dead weight of a non-planing
boat would not matter, but what about the planing taxi?
Put it at the rear and it would plane all the time. Having a
battery , and its potential variation of placement , helps t balance the
Q: A benefit for leisure craft , if you want to slip in or out
of port at night and not disturb people , that would be delightful,
at the moment we are shouting crew commands over a noisy diesel?
On the Falmouth ferry they do alot of commentaries and previously they
had to slow down at those times because the diesel was so noisy.
Q: There's lots of marine engineers familiar with diesel engines
but what about the electic side of hybrids , a problem of finding engineers
capable in that area for maintainence and repair?
Cumin-Raye? engines are not common at all, we have an issue with that
engine and no one can fix it.
For the fast section Venice taxis are there any hydrofoil hulls?
Not yet, some ideas around hydrofoils and different hull shapes
but at the moment they are pretty conservative. People like the looks of
their existing boats and it will be difficult to change .
Q: It strikes me it can be a minor alteration but a large power advantabe.?
The water buses in Venice are crazy old technology , the marine industry
does not seem ot be fast on the uptake. We have worked in racing , but that is very different.
The Falmouth ferry project.
The official launch was may this year and operational in July.
This ferry has operated between St Maurs and Trellisik in Flmouth.
Its a new route because they did not want to put this piece of innovative new tech
on a very important route . They thought this route, no one is going to use it
anyway. On the first trip we had 35 on board and 160 people transported
on the first day, which was a surprise how well the route was received.
Trellisik is National Trust gardens , a very nice area, and the reason for
making it hybrid. The river is noce and green and people want to
watch the birds and fish , so switching to electic in those areas is much less
noise and no fumes. Its an open boat , during a nice sunny day its
very busy in the open part and no one under the canopy part.
When a diesel boat comes alongside or leave the pontoon ,
they have to reverse, or the wind may blow from the stern and the
passengers are sitting in diesel fumes, not nice. With electric ,
no problem of that.
THe boat itself is called Kingsly 2 ? , its on the historic ship
register and built in 1934. It can carry 50 and up to 60.
We managed to get MCA approval for a hybrid boat , Lloyds register has
endorsed the battery system . Its 50 foot long, longer than the taxi boat.
So what is the fuel saving. The fery leaves St Moors ,goes to Trellisik and
returns, the speed variation is pretty small. It has to come alongside and has to
wait . It may have to slow down in the river going past moorings ,
and other manoevering. But what about any fuel savings.
The engine is directly connected to the gearbox and the prop. In between is
the clutch and a motor that can work as a mtor or as a generator to
charge the battery. ??? . The boat has much mor espace than the taxi ,
the battery is aft of the engine , best for weight distributuon.
There is a dc-dc converter for in the electric mode to charge 12V batteries for the
radios and lights. Otherwise much the same as the taxi boat. The motor is
40KW but only weighs 11Kg. We use the same batteries as Mitsubishi
plug-in car hybrid, of Japanese manufacture Lithium Manganese cell
type which is safer than laptop or mobile types. If it burnt it cannot
release its own oxygen, so you can extinguish it. Compared to cells based
on cobalt which can release oxygen when burning and impossible to extinguish
such fires. Its a little heavier but no matter on a boat.
The voltage is between 290 and 430 volts, 18.3KWh , thats always a lie,
from all manufacturers as it depends on how you use it. You should never
fully use it. You could fully use it but it would probably get too warm
during an 8 hour day of operation of charging/discharging repeatedly.
Q: Would Lithium-ion phosphate be a contender?
Yes but it has other disadvantages , its difficult to determine th state of
charge in a hybrid , because the voltage does not depend reliably on the
state of charge. Most phosphate ones are coming from China and manganese or
cobalt from Japan. So there are commercial disadvantages of dealing
with Chinese and Korean suppliers compared to Japan.
They are also difficult to balance, becasue of the voltage/state of charge problem
as you think they are the same state of charge but they're not.
Q: I bought a leisure one for my motor home , it works very well
but it does seem to loose track of what its doing and needs resetting,
probably what your saying?
Yes, they have some issues. I'm nit sure whether cooling is easy
with them either. Our decision was more based on having known them for years
and hoping to make a business case and working with the Chinese was never going
These are automotive cells , so have received a lot of safety testing, which is
important for insurance and Lloyds Register approval. The Chinese say they cannot
send any certification. Same with Korean batteries, we cannot transport Korean
batteries any more because of dodgey certifications found out by a UK
transportation body. 3 independent test results looked suspiciously the same .
All this sort f activity puts lithium batteries as a whole into a bad reputation.
Its a safe battery , liquid cooled . For leisure boats , used twice a
year but a ferry is used a lot. Put this sort of tech on a leisure boat
it can be made cheaply as it doesn't have to last that long.
Q: I have a Nissan Leaf and its surprising how much the capacity
changes with temperature. Basically the hotter the better ,
then you run up against being a danger above a certain temp. Wht sort
of policy do you have. Also can you charge it up at low temperature
, then the temp goes up , without charging and you get more out of it or
charge it while its hot, I've no idea, a lot of info is just not provided
to the owner of such batteries. ?
Yes, all interesting topics and as usual it all depends. One great advantage of
this type of battery is you can charge it at sub-zero. Whereas all
the others normally theu say don't.
Q: If you charge sub-zero and warm it up , do you get more out of the
battery without charging it any more?
No, its just that chemical reactions work better at higher temps. Yo're getting
mor eenergy but you don't necessarily get more capacity, because the voltage
drop is lower . The over-voltage you need to drive the chemical reaction
can be smaller at higher temps but the aging accelerates, that the problem.
You want to keep it between 20 and 40 degree C , above 40 and aging
accelerates pretty fast. At 60 they will perform really well , in racing they
operate at 85 or 90 which is where you get most power out but they
won't last very long.
Q: Is the cooling internal or external to the battery case? and is that something
your control system takes care of?
Liquid cooling with inlet and outlet, 2 batteries split into 2, so for smaller
boats we can nicely position them. Its not salt-water cooling , there is a
heat exchanger for running the salt-water through there. We are actually
changing that, on the next Venice water taxi we will use sea-water
cooling for the batteries.
Q: For the batteries in theory you could use all sorts of electrolytes , I was wondering about
using salt water and the boat being one of the electrodes, and the other one being
something dangling in the water. Are you having to use Li-ion because
everybody does and therefore the most developed , because your not that
weight sensitive, so there must be other technologies possible. Just thinking
you're floating around in an electrolyte , why not use it? a daft question?
Generally the marine industry is not apt to change . The MCA is familiar
with lead-acid batteries on boats but put a Li-ion battery on board
and they only developed some rules a year ago for that. We had to look
at the system and writ ea long letter saying about safety etc.
I suspect you stating the use of lithium because everyone uses it is probably correct.
For a number of reasons , firstly people know how it works , know whats
available , no patent issues , more than 1 supplier is very important.
Another battery technology had one suplier who sat on it
and its patent but people want 2 or more sources. So there is some competition
and people can't fool around with pricing , you want a replacement, that will
cost you 5 times as much.
Sony? was invented by EA Technology ? near Oxford mainly for space applications
and Sony could see opportunity in this and were the first to put it in mobile phones
Q: It seems extremely narrow that society should just take a Walkman
battery and use it for powering cars , it seems amazing that no one has broadened it out?
We used lead-acid batteries for a very long time , NiCad came in,
that was replaced by NiMH . Lithium is not replacing Lead-acid so
If its a cold rough-sea day and a battery dies , you would be in trouble, wheras a car in that situation can just coast to a stop on the side of the road?
For the automotive industry, they can try 4 or 5 types, spend a fortune on
engineering and validation programmes. The marine industry they don't.
They rely on replication of where its been done before and somewhere else.
Q:Sea-water sloshing over the batteries and extra electronics?
Just the air is pretty corrossive.
Q: But you do have 2 mechansms for powering as long as they work when
one packs up, is better than just having one power source. ?
We've used that redundency a few times , both ways. In Soton we had the diesel
of the taxi pack up just in front o the fast ferry. So I was glad we had an electric
motor that could take us out of the way.
On the ferry we implemented a backup, for the crew. There is a lot
of software , so no longer a hammer and spanner will work . You have to
have 5 laptops and 3 ? at least. So they were all afraid of this complexity.
So its a matter of changing a few relays and its back to their old diesel
only, the whole hybrid structure is disconnected, is a feature that
people like. For the ferry we've over-sized the engine, sounds a bit
counter-intuitive. At one point on the operation curve we need
60hp but we went for 200 hp engine instead and we changed the prop
and gear ratio so the prop thrust curve has changed , overloading
its sometimes called. When you put the gear in , the engine is
ticking over , its not a smooth clutch , so clunk. But as we've overloaded
it , its too much for the engine during an emergency stop and it
stalls. So we've told the electric motor to help it so
it doesn't stop. If it was not a hybrid you could not do this,
another way to make it more efficient.
So you have a fuel powered engine in your boat so
you are already paying for friction in the engine as overheads.
Can you now add the generator and take energy out for battery
charging - how efficient is that because your already paying for the
overheads. So this curve shows this horizontally speed in knots
and vertical axis the engine efficiency. At low speed the engine
is inefficient and becones more efficient as speed goes up.
The top efficency is about 37% for the diesel engine ,
but now add in the generator then we are pushing the engine into a
more efficient region. That means the elctric energy is generated with
an efficency of 35 to 45% from the already running diesel engine ,
which is pretty good.
Q: Do any of your rojects rely on external ?ing?
Its not a plug-in hybrid, its self-contained. Normally thats a
disadvantage because there are just no recharging points on the Falmouth
Q: With that efficency increase are you getting something for nothing?
Not for nothing but you get it pretty cheap , much more cheaply than
running a generator. If your boat needs a generator anyway, diesel
propulsion and a generator it would be better to use the diesel
just for propulsion , than the battery-hybrid system with no generator.
That sort of system exista for sailing boats now. Thats a pretty clever
system especially for larger sailing boats. People have a diesel then
a generator and they don't then need a generator any more.
So the marine market is more diverse and interesting when it
comes to electic propulsion , than automotive as so many different
aspects to consider. The ferry performance, x-axis speed in knots
v the power required to push that boat at that speed.
So if you do 2 knots you need next to nothing like 1KW but goto
4 or 5 knots, the speed limit on the Test and Itchen is 6 knots, you
need 10 kW for a boat carrying 60 people. So a boat is pretty
efficient at low speed. But a diesel cannot operate a boat slower than
this , improves at higher speed where the boat gets worse.
So what overall efficiency. We had a number of laptops logging the
results on a rainy day in Falmouth. The fuel consumption in terms of
equivalent litres per nautical mile.
x-axis vessel speed i nknots , one curve shows the fuel consumption
the boat could have theretically if the propulsion system was 100%
efficient. A propellor good be improved but take it as a boat from
1934. One curve is the fuel consumption if it was purely electric,
pretty close to being optimum. The other curve is diesel only and quite
bad in comparison, because diesel is not very efficient.
The boat is good at one point on the curves but the diesel is bad,
and elsewhere the boat is bad but the diesel is good , so
no obvious sweet spot. But if you have a strong tide , things change
though. The hybrid curve, difference between diesel and the electric
, I cannot go electric all the time and have to recharge and so using fuel.
I'm using more fuel than if I was not charging , so I don't get that
for free , just cheaper . When we go to diesel we will be pretty fast.
The curves show it makes sense up to 7.5 kn it makes sense to go electric
because you're consuming less fuel. One question when demonstrating the
boats is , whats the range. It depends a lot on the speed as the boat is very
efficient at low speed. Go at 3.5kn you can go 12 nm with the ferry
in electric mode. But if you want to go fast 7.5kn then only 3 nm.
How fast does it recharge. Running on diesel at 6kn it takes about
8nm torecharge and go at 10 kn its much the same. Little difference on
what speed you do in diesel mode, dependent on the setting of the boat though.
Its about 30 minutes about 15 min on electric and 15min on diesel.
Going slow is always electric as is manoevering alongside , very slow ,
so disproportionate amount of time , where the diesel is pretty bad for this.
So the fuel saving of combined compared to the same system running
just on diesel.
Go slowly in electric mode about 20% saving , go quite fast and the
saving is only 5%. If you go too fast in electric mode you start to require
more fuel. This does not take into account ,in normal operation ,they
often leave the engine running . It also does not take into account
the ability to overload the engine , there are othe rsavings.
Skippers are now a lot more interested in all this , asking about
fuel saving possibilities as well as how to save the planet, which they
didn't do before. So that is the worst case in terms of fuel saving because that
ferry is not the most amenable to such fuel saving as its just going back and forth.
Venice taxi boats would be saving more than that .
Q:If you could do the whole route under solar panels, would it
make a negligible contribution or some contribution?
I think it would make quite a good contribution because , 2 to 3KW with
panels on the cabin top. Something you'd have to worry about
is spray on the panels. If not marine quality there would be galloping
corrossion and salt deposits , so a bad environment for PV panels.
Q: At either end of the route , is there any shore power?
With appropriate PV panels and daily surface cleaning and assume 3KW
that boat can do about 4.5kn so noticeable. Also its only operating
during the summer.
People don't seem to use the term Lt/nm , they use the term Lt/hr
a lot in marine matters because they use that for calculating
range and fuel requirements. So its pretty hard work to save just that
20%. If people reduced their spped from 9 to 8 knots they would
also save 20% . Much easier to do that . Perhaps they could improve the
hull , use a different prop or just clean the prop and the hull , much
easier perhaps in terms of fuel saving. But a hybrid system does have othe r
Q: less stress on the diesel engine and maybe extent maintainence intervals?
Q: You could be bringing the world of smart meters into the marine world.
Providing something for skippers to aim for.?
Non technical aspects of this should not be under-estimated. With hybrids
and so electrification it enables a lot of other things. It enables solar,
shore-power , cappoccino machine on board and other life benefits.
So the question asked, why should we change propulsion system to hybrid
and just save 20%, does it really matter. Its not just the fuel saving , but all
the other that comes with it, looking for different fuels which could be methanol
or hydrogen for example . Electricity all very nice but hybrid for the range.
Some applications , purely electric would make sense. Other applications
series hybrid or fuel-cell hybrid makes sense. Some applications a hybrid
makes no sense and I'd say don't by hybrid. We had this where someone in Canada was operating 20 boats in a nature reserve and he was keen to buy our dream hybrid
system. But the way he operated the boats he'd use more fuel .
Q: Is the ferry a "Little Ship" one of the fleet of WW2 Dunkirk evacuee ships, they used a lot of boats like that ferry, right age?
This one came from the Scilly Isles and always on Scilly .
Do you see any inerest in having mainly electric but sort of limp-home
diesel that won't give your normal cruising speed but would get you home
eventually, with failed battery capacity? Often big boats with a big engine have a small one for if the big one breaks down?
Peace-of-mind , yes .
I went to a marine talk and someone in Holland had installed hybrid in tugs,
any thoughts on that?
Daemon? has made a few. I know they had asn issue as they gave skippers the
option of using the hybrid or the diesel and given the choice they were familiar
with diesel and switched off the hybrid.
Our ferry skippers are more the other way, keen to go electric, but they
go a bit too fast on electric for my liking .
A featur efor electric cars is supposed to be superb acceleration, is there a
Yes, it really related to manoevering where you have instant response.
You could have bow thrusters electric as well.
A lot of motor-boats have 2 engines anyway, one port and one
starboard. With your system could you have one electric and one diesel ?
Possibly, they'd have to be matched in all performance modes .
Whats the function of variable pitch props ?
I'm not a prop expert , I'm told you can easily save fuel by having
variable pitch props , similar to all this hybrid stuff but you'd not
be able to improve your emissions. On emissions there is argument
as to whether to install a big and expensive scrubber on a boat,
or install a hybrid , but no one seems to have done comparisons.
If you had shore-power then presumably it could be scrubbed further
up the line?
Yes, that comes up a lot. It depends what sort of electricity you buy or
use. Most people who drive electric cars would buy green electricity
from wind+solar or their own solar PV. There is similar issues
with recycling schemes. If no one puts glass into recycling then #
recycling makes no sense. Thinks don't happen nicely synchronised.
We have to clean up our grid and it should all be CO2-free eventually.
At the same time using more electricity. Electricity is the easiest
form of energy we can create from renewable resources.
Liquid fuels have the great advantage of high energy density
but its difficult to produce sustainably, like
methanol from CO2. But if you produce it from CO2,
you're just delaying the release of CO2. At least you
use it twice, so not so bad in that sense.
On speed controls, I recall boats being banged into and out of gear
for minimum speed. But hybrid allows nice precise speed control
on electric. Would skippers accept that? Do you have a sort
of exponential throttle , to take advantage ofthat preciese control at lower speeds?
Yes we have a control that is not linear, so they get more control at
low RPM and also at high RPM.
Q: Do you get unsolicited positive feedback on that?
We went through control system changes in developement with a skipper
settling on what was quite nice and disliking other settings.
On this ferry the gearbox cannot operate at low RPM , so it still
has to be banged into gear. But they are developing hybrid -
ready gearboxes which are supposed to be finessed. Its hydraulic clutch control
and the hydraulic pump is driven off the shaft which has to
rotate at a certain rpm or the hydraulics don't work so then the clutch
You've only got a small sample of uses, 2 boats and a few skippers .
Comparing with electric cars , so many other benefits , no usage stuck in traffic
and other things coming out, a completely different mind-set. You've
not yet been able to cultivate those present unknowns in your audience. ?
We've used questionaires amonst the ferry passengers and thatgiven us some
answers on thinking. The ferry company has a number of skippers , not just 2.
The Venice boat will go to Venice and be shown to as many people as possible.
Irs got to get to the satage where a skipper will buy a hybrid
system because of all the other features he was not aware of previously?
A trouble with the commercial market is the 2 layers, who is your
cutomer , is the passenger or the purchaser . Especially the Venice boat we have
to show to the local council what is possible. Because only then
canthey produce the restrictions for diesel use, at the moment there is no
demand, the demand has to be created. This is a general problem with
sustainability. In general there is no demand , the tech vto fulfil a demand
is there . Killing the planet , new tech developement , but who is
asking for this tech.
I was involved with the developemnet of the first colour IT display
and we deliberated over how many will we sell , as a prestige item
costing 4 times as much as the existing green-screens. So a room of
200 users, but give one to the supervisor as a hierarchical thing.
We thought the market would stand 85,000 , but once they started
getting out there, no one wanted a green-screen. Our customers
buying these , could not recruit staff unless they changed ovver to
colour ones. You only use monochrme these days to signify stuff thats
out of date?
I bought the first battery car in 1998 and loved it, and thought why
is no one else buying these. We were the first in Europe to try
Li-ion batteries in electric cars and hybrids and there was some interest
from the press but no interest from investors then, because they would ask
how big is the market, how many can we sell . Then Elon Musk
believed in it , used his own money, and the rest is history. That is the sad
thing people don't believe in something . The Chinese are believing in this
and they are investing .
For a replacement technology you need a disruptor like Musk?
At the moment people are more reluctant tobuy diesell and hoping electric car
prices will come down ?
A chicken and egg problem with e-cars, they get their ecar ready
Mitsubishi and GM not selling well and the CEO is in trouble.
Something that has worked is Formule-e racing?
In the states with Tesla they go mountain racing at weekends, get to top and then
on the way back regenerative recharging for the braking .
Tuesday 17 September 2019 Malc Pearson Action on Hearing Loss volunteer, talk on hearing loss.
We have support groups in Soton
We have lots of small groups with us, we have training and I'm part of that.
We give training, mainly to corporates , for diversity and deaf-awareness.
We have campaigns, we lobby . We were very active lobbying for subtitling
on TV. We do research , including at Soton uni
on likes of cochlea implants and tinnitus . We also sell eqipment, there is so much stuff
to support you when you are deaf.
Firstly can you hear my voice, because I can't hear my own voice.
Sometimes I shout, sometimes I go quiet, sometimes it goes off pitch,
give me a nudge if I do .
For people who are totally deaf, right thru to the other end of the
spectrum with mild hearing loss , 1 in 6 people.
11 million people in 2018. It goes up in over 35s to 15.6% because its in the
older generation . We're living longer , therefore more people with
hearing loss in the future.
Profoundly deaf, probably sign-language users or lip-readers.
Sign-readers 24,000 but its not taken up by youngsters. Wefind sign language is a
very efficient language. But you need 2 people to tango, if you sign you need someone
on the oteher side to receive signing. So when you interface with an
employer or the general public, there is a problem.
6.7million could benefit from a hearing aid HA , the NHS say they've issued about
2 million, so a lot of people out there who could benefit from a
HA or two.
How to you individually recognise deafness, discuss with your neighbours
for 3 minutes. Deafness is invisible, look at me and how did you know I
I might tell you, I did tonight but normally I would not, if I
met you on the street. If I do then you start to behave very peculiarly.
You start to give me very strange mouth patterns. I need normal
speech to be able to read it, not emphasised. People tend to
react and over-react. So I shuffle along a bit, perhaps flash my
hearing aids at them and slowly go into it.
I have 2 HAs, 2 NHS ones and with short hair they are
very visible, I quite like that as it helps me tell people that I'm deaf.
ts fashionable now to have things stuffed in your ears. It used to be
that people did not want to be seen with things in their ears, not any more.
I don't respond if people talk behind me. I do hear bass notes ,
so something dropped with a loud bass note, I'll hear that.
So people walking along a footpath who want to overtake and say something to
me , i don't get it. My vision is what I hear.
When you're deaf you observe things in a different way. You get glances from
people walking past me , because they've said something, and I've not
picked it up.
The different voice. I've been hearing for most of my life , therefore my
voice is reasonable. Now I don't hear it properly I usually shout .
Usually people who ar edeaf have a different kind of voice.
A colleague of mine who gives talks has never heard her own voice,
has to modulate it by touching her throat to gauge the volume.
Once set for the day , she's usually alright.
Go somewhere and the TV is very loud , they probably have hearing loss.
Similar with loud volume in a car, could be the fshion or compensating
for lack of hearing.
Lip-reading. I only get about half of lip reading content, I have to make the
rest of it up as I go along. SoI get it wrong very often ,misunderstanding
things. If I go wrong like that I can tell from people's eyes that I've
gone wrong, we have to back-up a bit and restart.
As a lipreader, I don't have eye contact, I look a little lower.
I'm actually looking at your teeth and you tongue , so that sort of focus
is slightly off and that does worry people. They get a bit uncomfortable becaus eI'm
not giving them pure eye contact. It can get a bit embarassing at
times, when I'm tired I beging to stare. If I don't know th eperson
and they don't realise what I'm doing , its a bit unnerving for them. But I'm aware of this and have to compensate. So if someone seems to be staring at your lips they're
probably deaf and are lip reading.
As you get older, bits wear out .
I can't go to dinner parties becasue when people are eating and drinking
and also talking , it all goes completely out the window. Equally any
social interaction where you have to interact with people , lots of
people trying to talk to me, it gets very difficult.
I only go out to dinner with people who know me and they can compensate
and help me thru it. If someone is withdrawing from you in a social
situation deafness could be the reason why.
If you want to learn something different , sign-language. It makes you lok
at languages in a very different way. But you need 2 to tango , so you don't usually
see it in action unless 2 people are deaf together or going to a club where
they know deaf people will be present. Some deaf clubs, sign language is the
only langueage that is used inside those clubs. So a place of silence, and if
you tried anything else in there you'd get some very strange looks.
The red and white stick, that means visual impairment and hearing
impairment. Not a barbers pole but red/white/red/white, they won't see
you very well and wont hear you very well. So you have to think about
how you will communicte with them. A light touch on the shoulder and then
move in front of them and then let them drive it, as they know what to
do. They will guide you from then on. If its an emergency , say someone is
stepping into a road, grab them by the arm, if an adult.
Q: I've never seen a red and white stick, I'd immediately associeate
it with a surveyor's staff. Would those poles have one of the
sounding lumps on the end,for echo gauging purposes , looking like a golfing stick?
I've seen the plain white but not red and white around?
They are around Soton.
Q: I've a problem understanding women rather than men, because of
upper frequency sound loss. So if you find male voices easier to understand than
female , that can be a signature of deafness?
It is , when I show you the innards of the ear , you'll see why that happens.
There are 2 big reasons why you can go deaf. I'm not a medic .
Conductive deafness , where you hear everything but at a lower level,
just as turning the volume down of a tv. You hear everything but have to
strain to hear it. The second is perceptive loss, where you've lost a
band of frequencies . Usually that frequency is in the higher band,
but it can be anywhere. My loss was due to an illness and it caused me to
loose several bands going along the route of the cochlea.
This gives you distortion of the heard sound. As you get older you get a mixture
of both. I started off with the second but with age then the first as well.
Ear flaps are designed to catch the moving air. It sticks out and funnels
the moving air down the ear canal to the ear drum. This is a living
membrane that goes across the ear. Inside of that there are 3 small
bones , the smallest bones in the human. The first touches the back of the
ear drum , picks up the vibrations that are now physical
movement and passes thru the next 2 to the end of the cochlea.
Its like that because it needs to move , because your jaw
moves and your whole head moves so it has to be a flexible joint.
As you get older your bones change , these bones change as well
and they're not so conductive , therefore its like turning the volume down.
The transmission of vibrations across that bridge are not as efficient as they
used to be.
The cochlea, thumbnail size and snail shape. Its a hollow tube, 2 tubes
in fact. The centre of the tube is fluid filled . If I straighten out the snail
shape , a tapered tube. Inside the tube are thousands of hairs ,
pointing inwards that pick up the vibrations. At the hair ends are
nerves , which all join together and become the auditory nerve tha t
send electrical impulses toyour brain . Responds best to high frequencies
at the narrow end and bass frequencies at the wide end. All the vibrations
start with most amplitude at one end , takes a lot of work , because all the
vibrations have to pass through those hairs, you loose those, they wear out
and they won't replace. Thats why you start to hear mens voices
better than women's. The illness I had, got into the fluid of the ear
and it destroyed the hairs in some patches, so I've lost those
frequencies. They can't be replaced, nothing you can do about it.
This is all vibrstions that the hairs pick up , converted by the
nerves into electricity and the brain does all the work of converting into the
sensation of hearing, same with eyes.
Why is one tube larger than the other?
Just a schematic , the main tube is what I,m describing here. There is a
second tube that picks up secondary waves , and the brain compares the two, so
more complex than what I was saying. For the purpose of understanding ,
just considering one.
What s the normal hearing frequency range of ordinary people?
About 20 hertz to 20 Kilohertz, only for very young people, it drops off
quite quickly as you get older.
So the communication needs of different people. If you're a hearing aid
user , you need signs and subtitles as we are using our eyes as well as
the small bit of hearing we have got.
If we're lip-readers , we need note-takers and speech-to-text to
help us. Note-takers are professional peole who will come with me to
a uni lecture , because I can't look at a lecture and write notes at the
same time. Speech-to-text where someone with a special keyboard
would type in what is said and it would come up[ as large font print
on a screen. You see this in courts where there are people with
hearing loss involved. There are signer-readers and interpreters,
people who will sit between and a hearing and a deaf person who
is using sign langauge. You can engage such people by booking thru
the charity Action on Hearing loss. With an interpreter its very eady to
end up talking to the interpreter and not the real interviewer.
Aubtitling , I use it all the time. It started with BBC doing some programs
and now its across most of the companies. Usually subtitling is pre-recorded
bu tsomething live like the 10 o'clock news , they don't know what they will say until
being on air. You can tell the difference, because it comes up in blocks
and often a wrong word in there or mixed up . The software they use protects from
strange words coming up .
Q: I noticed on BBC reletext this week that they're stopping the red button business.
I've always liked that facility , compact concise precissed text and if anything
piques my interest I look on the internet for it, armed now with a few keywords. Snippets of news in that
structured /teletext/ceefax way I'd have thought was very useful for a deaf person?
Its wonderful for us as an information source and as a communication source.
Q: Is that text-biased news/weather etc very useful to the deaf in the same manner as
It is important to us. I've not heard any comments , no lobbying that I've
seen about the red button , its only just come up in the news, early days.
We've had a long campaign to support subtitling ampng the various
companies . When we have a problem reported to us we go and try to
sort it out. A problem with subtitling is its often imported films
, which is inadrqute, they need to put it on for an English voice.
Everybody lipreads to some degree. Have a conversation with a
friend and put a hand in front of your mouth and I guarrentee that your
friend a few minurtes later will ask you to move your hand.
Its also reassuring to see the lips as people get the emotions
along with the speech. If the conversation
is going well , it flows well, a nice smile on your face, a warm look
that people acknowledge. Lip-reading isn't just about the lips,
its the whole image of the face.
Lip-reading needs a lot of concentration, I can only lip-read
for about 30 minutes , after that I've lost it, I have to go outside.
All a bit strange as I do most of the talking , so lip-reading
can be a challenge to do at the same time.
I get to a point where things become jumbled and I go out and have a coffee,
unfortunately the peole come with me. I just want to go away and have
some time on my own for a few minutes. About 50% accurate.
Its a bit like predictive text, people say things and you respond to things
in set patterns . When your close to someone you can almost tell what they will
say before they say it. A lot is inteliigent guess work, but that also
creates mistakes and you've got to acknowledge that.
I come wiht a kitbag and when you go to lip-reading clases , they tell
you to carry a kit bag with you. I'm happy to share some of my tricks with
you but not all of them. The first one is I start conversations, you
don't stand a chance of starting a conversation with me, because I will
start. When I meet you ,I'm in there . Then I know what we're
talking about. If i let you start , I'd have not idea what the
topic is, it could be M£ traffic, a sunny day or whatever. If I start , it
lets me relax and it lets me know what we're talking about.
Some people think I'm pushy because of that, but hey, thats the way it works.
We have thinking words and one of them is really!. If you say really to someone
it answers most things. So if I'm stuck, I say really and we can move on,
without shocking?. As soon as I shock?, then the funny mouths and all
the rest of it. I like people to speak a little slower than normal.
I can slow you down, by making my speech slower , you'll slow down.
Try this with someone in conversation. You need good lighting
, I can just about manage with the low level lights here for the projector.
Often the new downlighter LED lights in houses now are terrible because
your nose shades your lips. Often people are standing under one of those
strong lights and I can't see a thing. I move to the
side and the other person moves with me. I don't have ot say , lets move,
they come with me without explanation. We do come out of lipreading
training with a load of tricks.
As a demo of confusing groups say amongst yourselves as pairs while looking at
Say ma (as in murder), be ( as in Bert) , pe (as in pearl).
You should see they all look the same shape, known as confusion groups.
Most of you were smiling then, as soon as you smile , it distorts the lips so
you have to do this with a flat face.
Some more Che (as in church), Sha (Sherbert) and ga (as in gerbil)
you can feel your tongue going into different positions from the previous group.
The and Fe (as in fir) is tonge position.
Finally Ta (as in turban) and De (as in dirt) your tongue is behind
your top teeth and again your jaw shape is diffferent.
Its not just your lips , its your whole face shape, that changes.
If you say, tea on Saturday and dinner on sunday, look at how those
appear as lipreading, very close lips similarity, so easy confusion.
When you're communicating with someone with hearing loss, the image
your portraying becomes part of the language.
So drink a cup of tea with me look at my face appearance . Or driving
along a road in the country without a care in the world compared to
facial appearance when driving a bus in the rush hour.
How can you help with hearing loss. Find a quiet place, this room is
wonderful, lots of rooms have fans and other noises going on and
traffic noise coming into the room. So how would you get my
attention there. If I'm focused over my laptop say, bring your hand
into my vision from the side. A light touch on the shoulder , a heavy
touch if there isa fire. Face the person, a lot of people try to face the
person but end up half way, come all the weay round.
If you're going to talk for any length of time , try to include the
subject in your first sentence. Slightly slower than normal but try
to keep the normal rythm going. Going slower tends to exagerate the
features on your face around the lips. Take turns in talking,
if you talk over , it will all fall apart. If you shout at me
and I understand it , I'll shout back, other than that your face looks
angry. So if someone shouts I immediately see anger and therefore
I assume there is a danger.
Its quite panicking when you see someone you don't know shouting at you.
Don't waffle. When you study sign-langauge its amazing how many
words are unnecessary. Keep to plain language. Re-phrase something
if its not going well . If it gets to the time where I can't take any more,
pack it in and restart in half an hour. But please don't give up and walk
away. Worse comes to worse write down on a bit of paper.
If you walk away , I think he doesn't care and this errodes friendships.
I have a sheet here with communication points from both points of
view, hearing and hearing-loss.
You can check your own hearing by our health check ,
phone the number 0845 600 5555
and press the keypad for different tones for about 10 minutes
and use a landline phone not a mobile . It will say your ok
or need to have it checked . Also repeat again for the other ear.
NHS service are via your GP or you can go the private route.
Learn to lip-read, quite a few free tutorialsd on the internet.
There are some on our websirte as well but if you ant to do it
seriously , you join a group, often attached to libraries,
so you go to a community, you just practise. There is a structure
to it but most of the time its quite social .
There's never been a better time to be deaf with all the tech
around. You could spend a fortune on it if you want.
Communicating text , there is Whatsapp, all these apps gives a deaf
person access to communication that was not possible
before. THe leaflets I brought along , do get out of
date very quickly , so go to the website where the pdf files
are updated . There are hundreds on there. Any sort of issue,
you'll find something about hearing loss there. We have an
email newsletter as well , once a month.
Q: A language spoken with a lot of sibilance is very difficult to
When you talk vto people who are not deaf , they do pick up
a lot from what we do not say. Particularly youngsters, I'll
be talking at the Solent uni on friday , they ar eso interested
on hearing loss because its such a common thying.
Q: I went to have my ears tested and he said a lot of people have
accelerated hearing loss because they have these things in their ears.
I'm told that when you're on a train you can hear the hedphones of
someone two people forward, buzzing around. That is so
dangerous - also the noise in a car.
Earlier when you all were discussing how you would recognise
deafness in someone, you registered 80 dB, via an app on my
phone. 95dB is the starting threshold for hearing damage.
That sort of damage is recoverable if you go somewhere quiet
afterwards, but do it continuously or for a long time, damage
will be done.
Could you saya bit about signlanguage. My daughter is teaching
my granddaughter a simplified sign langauge, macaton?
Often used in schools especialy children with spacial needs as well.
Is it a stripped-down Brittish one?
The language structur eis different. If you are asking what is my
name, it is structured name/me/malcolm. Word order is changed from
English and trivial words are just lopped off.
The important stuff comes first, incredibly basic like that.
Why is there no international Esperanto sign-language?
No one could agree. Not enough people because to develop
you must have someone to sign to. Lip-reading is very
common now. If you can look at signed TV program
you see the word order is completely different to what is being spoken.
For TV signing they use exagerated facial expression?
Yes, a lot of the separation is the face, different expressions have different
I've heard they use male speakers for older viewers and then there was complaints
about gender equality .
Very true . There is AD audio description , like teletext but
describing the scene. Where this is done by a woman they choose someone
with a lower pitch of voice . But they also needed a voice that did not
clash with what was being spoken . Have a go with the AD button
on your TV controller, not available on every program.
The aspect of everyone can lipread to some extent. I used to know
someone with cerebral palsey and because of the distortion of his
speech , I found I dropped into a mode of looking away
from him and then I could understand him. If I looked at him, somehow
the distorted mouth-form distracted or confused my underestanding of his speech.?
It just shows that we all must have some background ability at
lip-reading , by this converse.
Its easy to have a hearing test, these days its almost automated.
It comes up with a graph that shows your hearing loss. Then they can
set up any HA to your needs.
Please make emails plain text only , no more than 5KByte or 500 words.
Anyone sending larger texts or attachments such as digital signatures, pictures etc will have
them automatically deleted on the server. I will be totally unaware of this, all your email will be deleted - sorry, again
blame the spammers. If you suspect problems emailing me then please try using
keyword for searchengines , scicafshadow, scicafsoton, Southampton Science Café, Café Scientifique, scicaf, scicaf1, scicaf2
, free talks, open talks, free lectures, open lectures ,