From: acl on 17 Mar 2007 00:14
On Mar 17, 7:09 am, Lionel <use...(a)imagenoir.com> wrote:
> On 16 Mar 2007 20:54:04 -0700, "acl" <achilleaslazari...(a)yahoo.co.uk>
> >On Mar 17, 6:48 am, Lionel <use...(a)imagenoir.com> wrote:
> >> On 16 Mar 2007 20:33:30 -0700, "acl" <achilleaslazari...(a)yahoo.co.uk>
> >> wrote:
> >> >On Mar 17, 5:55 am, Paul Rubin <http://phr...(a)NOSPAM.invalid> wrote:
> >> >> Precisely. Thus the question about experimental validation. I think
> >> >> we've pretty much established that with large pixels, shot noise
> >> >> dominates. I don't think this has been established for small pixels
> >> >Well for high signal levels I certainly hope it does. But what
> >> >dominates per pixel isn't the point, the point is that if we read the
> >> >pixels, each having read noise r, and bin them nxn, the read noise per
> >> >binned pixel will be effectively n*r. So we'd need lower read noise.
> >> Your read noise is going to be multplied by the number of photodiodes
> >> you're binning. Worse, because the signal level is proportional to the
> >> size of the photodiode, your read noise for each pixel increases (ie;
> >> the signal to read noise ratio decreases) with decreases in photodiode
> >> size.
> >No it's multiplied by the sqrt of the number of pixels you're binning.
> >This is exactly what I say in the part you quoted. I also do not see
> >why the absolute value of the read noise will decrease with decreasing
> Huh? I said the read noise would /increase/ with smaller photodiodes.
Oops typo, the question was "I don't see why it would increase".
> (This, BTW, is for the simple reason that it's harder to read smaller
> signals than larger one, & because the amplifiers also suffer from
> thermal & shot noise that must be included in the calculations.)
But it will not increase in absolute value, only in relation to the
signal. Look, write it down and try it (add the squares of the noises
and take the sqrt), you'll see what I mean. And when you bin, you
increase the signal (yes, the noises too, but, erm, read my post :) )
> >True, the s/n decreases because the signal decreases, but blah
> >blah. Please, read the post to which you replied a bit more carefully,
> >I think that would be the answer to what you say.
> Hm, I think I could say the same thing to you. ;^)
Ah, but you'd be wrong :). Seriously, I think you misread my post.
From: John Sheehy on 17 Mar 2007 00:24
Paul Rubin <http://phr.cx(a)NOSPAM.invalid> wrote in
> John Sheehy <JPS(a)no.komm> writes:
>> Your assistant is me, and I decide to replace the 16 containers with
>> 64 square containers, 4 of which fit in the same space as 1 of yours.
>> I come back to you with a list of results from 64 smaller containers,
>> instead of the 16 you asked for. The list is longer, and the total
>> count is the same as it would be if I used your original containers,
>> but have I created any *NOISE*?
> Well, how to you measure how many raindrops fell in each container?
This is just a thought experiment, not a real-world problem. Change it
to marbles, and actually count them. Happy? The thought experiment is
about shot noise. Read noise is another issue, but reality works in the
favor of small pixels there, really.
> Let's say exactly 100 raindrops fell in each of the original 16
> containers. Your counting method is accurate to within +/- 5
> raindrops, so you get counts from 95 to 105 inclusive, not too bad for
> this type of measurement; each count is within a 5% error band.
You do understand that this is analogous to read noise, not shot noise,
right? There really is no counting error in shot noise. Shot noise is
not really noise; it is an accurate count of the *REAL* nature of light.
We just don't like it in our images, because unlike the shot "noise" in
our retinas, our perceptual system doesn't buffer it from us in our
experience of media. Our brains only do noise reduction on what it
thinks is direct experience of light.
> Now you use the 64 smaller containers, and your counting method is
> still accurate to +/- 5 raindrops.
That's not a very likely scenario. Error is likely to be proportional to
amount, at least in part, not a fixed number. But keep going...
> But now you only have 25 raindrops
> per container on average, so now your +/- 5 raindrop error is a 20%
> band. And the total count for each 4 raindrop cluster can be anywhere
> from 80 to 120 instead of 95 to 105.
> In this case I would say you have added more noise by using smaller
> containers, since you haven't decreased the counting error per
Then I think you'd be wrong.
Real World: My FZ50 has a read noise of 2.7 electrons at ISO 100, with a
full-well capacity of 4800 electrons. 9 of its pixels, binned together,
will have a read noise of 8.1 electrons, and a full count of 43,200
photons, about the same as full-well capacity in my 20D, but the 20D has
a read noise of about 26 electrons at ISO 100.
43200/8.1 = 5233.33. The 20D has a full-well capacity of about 44000
electrons. 44000/26 = 1692.31. The binned FZ50 pixels have over 3x the
dynamic range of a 20D pixel at ISO 100 (where DR is the ratio of max
signal to the level where 1:1 S/N lies); they cover about the same area,
and collect about the same number of photons.
> You've taken a +/- 5% accurate measurement and replaced it
> with a +/- 20% measurement.
No. Noise or error doesn't add like that. Cases of 4 +5s and 4 -5s
will be rare , even with a linear distribution, but in the majority of
cases, there will be near-cancellation of + and - errors. The standard
deviation would actually be lower, even with your unrealistically high
error rate of 20% in the smaller containers, when binned down to the
bigger virtual containers.
<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS(a)no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
From: acl on 17 Mar 2007 00:37
On Mar 17, 7:28 am, Lionel <use...(a)imagenoir.com> wrote:
> On 16 Mar 2007 21:03:19 -0700, "acl" <achilleaslazari...(a)yahoo.co.uk>
> >On Mar 17, 6:13 am, Lionel <use...(a)imagenoir.com> wrote:
> >> On Sat, 17 Mar 2007 02:21:02 GMT, John Sheehy <J...(a)no.komm> wrote:
> >> >Even if you lose some photons due to miniaturization, remember, the shot
> >> >noise is only proportional to the square root of the signal.
> >> IIRC, it's proportional to the temperature of the sensor, & the bias
> >> across the photodiode, which is not the same thing. When we start
> >Shot noise in this thread refers to fluctuations in the number of
> >detected photons due to the random intervals between their arrival
> >(hence poissonian); it has absolutely nothing to do with temperature
> >etc. There are other noise sources that do.
> True, I've been conflating shot noise & thermal noise, which is
> incorrect of me, & a result of me pushing the boundries of my
> electroics knowledge. However, both noise sources are present, & must
> be taken into consideration when we're discussing real sensors.
They are taken into consideration!
> >> discussing things in this sort of detail, hand-waving & analogies stop
> >> being useful, & we need to refer to the underlying physics & the
> >> characteristics of real photodiodes.
> >These aren't hand-waving analogies, it's the real thing. The only
> >quantum aspect of the photons relevant to shot noise here is that
> >they're discrete particles.
> Yes, exactly. What do you think 'quantum' means?
It means a lot more, but this is not the place to go on about it, and
anyway it's irrelevant for what we're discussing.
> >> There's a very good reference
> >> available from one manufacturer of photodiodes at:
> >> <http://sales.hamamatsu.com/assets/applications/SSD/photodiode_technic...>
> >> You'll find the noise characteristics in section 2-4. Do bear in mind
> >This is about read noise and similar noise sources (the ones that will
> >be a problem in this scenario, indeed).
> ...and thermal noise, & shot noise. Read the whole section.
I did. Look, what I am saying is that the shot noise will be exactly
the same if you bin as when you used a bigger pixel (except for the
effects of the fill factor, which you have already mentioned), while
the read noise will increase in comparison to the small pixels. So it
will be necessary to have lower constant noise sources
(constant=independent of the signal level). In fact, if we must bin
nxn pixels to get the same pixel size, we need n times less constant
noise sources to get the same s/n ratio. I don't think anybody said
> Well, the point is that when you're talking about real image sensors
> where, (when all other characteristics are the same), the S:N ratio
> for an image from a 4MP sensor will be different to that of an 8MP
> sensor, even if you bin them both down to the same output resolution.
Look I think we're talking at cross purposes, we're saying the same
things but reach opposite conclusions. Better to let it rest, we're
just going to both end up with carpal tunnel syndrome and still
From: acl on 17 Mar 2007 00:41
On Mar 17, 7:37 am, "acl" <achilleaslazari...(a)yahoo.co.uk> wrote:
> (constant=independent of the signal level). In fact, if we must bin
> nxn pixels to get the same pixel size, we need n times less constant
> noise sources to get the same s/n ratio.
Argh, what did I write? I meant that the total constant noise in the
binned pixel will be n times that of the smaller ones.
See? Attrition is already setting in :)
From: acl on 17 Mar 2007 00:51
On Mar 17, 7:05 am, "Roger N. Clark (change username to rnclark)"
> Paul Rubin wrote:
> > John Sheehy <J...(a)no.komm> writes:
> >>Your assistant is me, and I decide to replace the 16 containers with 64
> >>square containers, 4 of which fit in the same space as 1 of yours. I come
> >>back to you with a list of results from 64 smaller containers, instead of
> >>the 16 you asked for. The list is longer, and the total count is the
> >>same as it would be if I used your original containers, but have I
> >>created any *NOISE*?
> > Well, how to you measure how many raindrops fell in each container?
> > Let's say exactly 100 raindrops fell in each of the original 16
> > containers. Your counting method is accurate to within +/- 5
> > raindrops, so you get counts from 95 to 105 inclusive, not too bad for
> > this type of measurement; each count is within a 5% error band.
> > Now you use the 64 smaller containers, and your counting method is
> > still accurate to +/- 5 raindrops. But now you only have 25 raindrops
> > per container on average, so now your +/- 5 raindrop error is a 20%
> > band. And the total count for each 4 raindrop cluster can be anywhere
> > from 80 to 120 instead of 95 to 105.
> > In this case I would say you have added more noise by using smaller
> > containers, since you haven't decreased the counting error per container.
> > You've taken a +/- 5% accurate measurement and replaced it with a +/- 20%
> > measurement.
> Excellent analogy. I was going to write something similar.
The analogy might be good, but the reasoning and conclusion is not. He
takes errors of +/- 5 per container and concludes that the total error
in the binned case is 4*that, ie he adds the noises instead of adding
the squares and taking the square root. In reality, the "read" noise
will be doubled, not quadrupled, so long as the noise sources are
uncorrelated. Still, of course, the s/n ratio decreases.