Naked Science Forum
General Science => General Science => Topic started by: Andrew on 09/11/2008 22:50:08
-
Andrew asked the Naked Scientists:
My digital camera can act as a video recorder. It can record a
bandwidth of about 6.5 megabytes a second of visual data. The camera
is "writing" this amount of information through to the memory card of the camera.
What is the comparable bandwidth of human vision? That is what kind of
digital camera or digital display would reproduce human vision? For that matter what is the bandwidth of each of the senses, that is how much information is being "written" by each sense to the brain?
Thanks for such a quality show. I learn much from you.
Andrew Griffin
Peterborough, Ontario, Canada
What do you think?
-
You can check this out.
http://www.wdv.com/Eye/EyeBandwidth/
But the eye works differently I think.
Its software, the brain, is so much more advanced than any software we developed yet.
And it is adapted to that special creature it serves.
For example, the speed of the praying Mantis relies on its eyes signals.
"The mantid [mantis] uses two primary hunting methods," Feldman says. It usually waits "motionless" until prey "comes within strike range" or it will occasionally pursue, in fluid slow motion, its hapless target until it gets within range. The strike takes "30 to 50 thousandths of a second."
There is no animal I know of that can counter react in that time.
There is another difference too.
Our brain is not 'digital'.
it's not 'bit by bit' serial..
Even when you read about 'quadra core' CPU:s
It's still serial signal processing inside it and out on the buss(es).
The brain is more like analogue.
Containing a high ratio of simultaneous 'noise' aka information.
And the brain sorts that out.
At all times.
But adapting to your 'needs' like sleeping hunting etc.
Reading Mister Warren you might find him confusing at times:)
I did too. So I tried to translate his equations into pixels here.
1 Petabytes = 1000 terabytes where every terabyte represents 1000 gigabytes
So 4000 Terabytes times 1000 = 4 000000 gigabytes, where 1 gigabyte is 1024 Megabytes
So counted that way we have four millions times 1024 = 4096000000 mega bytes
4096000000 megabytes times 1024 is 4194304000000 kilo bytes
And (as one Kilobyte = 1024 bytes) then 4194304000000 times 1024 will give us 4294967296000000 bytes
And it takes approximately 3 bytes to characterize each pixel if I got it right.
That as every pixel is made up of R,G,and B channels and requires one byte for each channel.
("Therefore, one pixel is 3 bytes, 1 megapixel is 3 megabytes etc."
But that is not really correct as one megabyte is 1024 times 1024 bytes = 1048576 bytes (times eight bits) digitaly.
And one megapixel is 1000000 pixels so one million pixels should then be three million bytes.
That translates to 145728 bytes less than three megabytes.
If one had said megabit instead of megabyte it would be correct though.
That as one megabit is 1 000 000 bits)
So when we split 4294967296000000 bytes in three it will give us 1431655 765 333333,3333333333333333 pixels for a two hour movie.
And that I won't even try to write out in letters.
If I got it right this time?
and two hours is 7200 seconds split with 1431655 765 333333,3333333333333333 give 198841078518,51851851851851851852 pixels per second.
Please check my numbers though, it's getting late here:)
-
The cable that connects my old PC to the monitor has a specified bandwidth of 15MHz Similar bandwidths are used for ordinary TV. The idea that you need 198,841,078,518 Hz (btw, do you know what spurious accuracy is) is clearly suspect.
It might be better to look at the number of light sensing cells at the back of the eye (about 100,000,000) and the refresh rate (about 20Hz). That gives 2GHz (still rather high) rather than 200GHz.
-
But the sensors are not bistate. They are analogue and so you can't suddenly start talking in terms of Bytes.
Also, there is a lot of signal processing (parallel) going on on the retina itself so there is vastly more potential information available than a simple TV scanning system can carry. You can recognise patterns, movement etc. at a rate much higher than simple calculations might suggest.
It's really comparing chalk and cheese and the numbers don't mean much.
I suppose that, if you reckon that you could get by if the only visual input you had was aTV monitor then the 5.5MHz bandwidth of a PAL coded signal gives you a rough value to work with. But, if a tiger were creeping up on you from the left, you wouldn't be aware of it with your TV picture alone (you get less than 20 degrees of visual arc for 'five times picture height', which is the viewing distance which gives you about the same resolution as the eye) . The sorts of data rates achieved with MPEG coding would be a hopeless comparison figure - it falls down all the time (can't even cope with the grass on a football field when it's panning).
-
But the sensors are not bistate.
The sensors are not 2 state, but the signals that come out down the nerves are, it's roughly a pulse frequency modulated signal.
-
BC I agree to the 'strange' exactitude of those numbers:)
But it's still the best I've seen in trying to cover all possible variables.
Maybe there are more?
Correct me if I'm wrong but don't we use a electrochemical communication?
Isn't it at both time a electrical field/signal and a chemical substance transmitted?
-
But the sensors are not bistate.
The sensors are not 2 state, but the signals that come out down the nerves are, it's roughly a pulse frequency modulated signal.
But the pulse width is still an ANALOGUE quantity - there is no quantisation. PM is not a digital technique.
-
True, but it's bistate.
I also forgot that most people have 2 eyes which doubles the bandwidth that the brain has to process.
I'm fairly sure that you can talk about bandwidth without having a digital system. Analogue TV systems use about 6MHz, FM radio varies but it's often 34KHz (IIRC) and old fashioned telephones use about 3kHz.
-
True, but it's bistate.
That is neither here nor there. The frequency, amplitude, phase or pulse width are only analogue versions of the original signal. That doesn't make anything digital. To be digital, it has to be quantised into discrete values and I am not aware that the nerves do that. They just use an analogue modulation system so that the channel will pass DC values.
Yes, of course you can talk of bandwidth. I was objecting to the use of Bytes per second - which is a data rate - a very different concept which involves signal to noise ratio and all the factors involved with DSP.
Analogue (PAL) TV uses 5.5Mhz, which uses a colour subcarrier to carry colour information, each channel of colour being about 1.5Mhz.
In any case, what we see and what we SEE are two different things. How much of our field of view is being used at one time? What information are you going to include in your calculation? The raw info is processed extensively between the retina cells.
-
Read this and enjoy :)
http://hubel.med.harvard.edu/b2.htm
Or for a zipped bok (It's 'Huge':)
http://hubel.med.harvard.edu/bcontex.htm
And impress your fellow physician with your newfound knowledge
-
Just thought I'd point out that although a reaction time of between 30-50 thousandths of a second sounds really fast it's only between 1/33 and 1/20 of a second. Humans have a reaction time better than 1/5 of a second, and sometimes approaching 1/10 of a second, so it's only between 4-7 times faster than us.
-
I'm fairly sure that you can talk about bandwidth without having a digital system. Analogue TV systems use about 6MHz, FM radio varies but it's often 34KHz (IIRC) and old fashioned telephones use about 3kHz.
I just thought I'd clear up between a radio-engineer's use of the word "bandwidth" and the digital info-systems use...
In radio, we talk about bandwidth as the "spread" of frequencies occupied by a transmission, measured in Hz, kHz, MHz etc. The amount of information (bits/bytes, or kilobits/sec or megabits/sec) you can carry over that channel depends on the sophistication of your modulation scheme and fundamentally on the noise-level in the channel. FOr example, you could sample audio at the CD standard 44.1kHz (44100 samples per second) at 8-bit (44100*8 = 352kbits/sec), which would still have 22kHz audio bandwidth but would sound rough, or at 16-bit CD standard which contains twice as much information (but still audio 22kHz bandwidth) and sound perfect (44100*16 = 704kbits/sec).
The picture-part of an analog TV signal occupies about 6MHz of radio spectrum and is coded within the television company networks as a signal digitised at 13.5MHz. The limiting noise with a good signal is probably around 40dB meaning you've got about almost the equivalent of 8-bit sampling (maybe "7.5" bits) of data. That's a raw data rate of roughly 100mega-bits-per-second. Forget about colour - in a PAL signal that's carried within the same bandwidth as the luminance part of the signal (and if not filtered out, simply appears as fine cross-hatch patterning on a black and white TV).
Now although that's the raw data rate, there's a lot of redundancy there as most of the time one frame is very much like the next.
MPEG digital compression (bit-rate-reduction) aims to chuck out all the details you don't see, and eliminate most of the frame-to-frame redundancy, to give something more representative of the "useful" or "unique" information-content.
Unfortunately, when you scrunch 100Mbps down to something like 2-3Mbps as we do for digital TV broadcasts in the UK, discarding 97-98% of the original information, the result looks pretty rough.
Realistically, if it were compressed to only say 8-10Mbps (mega-bits-per-sec) the images would be visually virtually indistinguishable from the original.
The eye is non-uniform in resolution and in response-time, being higher resolution and full colour (but slower response) in the fovea (central couple of degrees of vision), and lower resolution, monochrome and faster (much more sensitive to flicker) in the peripheral vision.
Which makes estimation of the "effective" information content of the whole of human vision somewhat tricky to judge.