From: ejmartin on
On Jul 20, 9:48 am, Bob Newman <bob.csx...(a)gmail.com> wrote:

>
> We agree that in real, present day cameras ISO doesn't occur in the
> photosite. In fact, it doesn't really occur at all, what's happening
> is that the capture system is being finagled to best read the bit of
> the sensor DR you're actually interested in. As you point out, with a
> better capture system, no need for ISO. What I was correcting was your
> statement that no amplification occurs in the photocell of an active
> pixel sensor. It does (although current rather than voltage
> amplification) and its noise is probably the irreducible minimum read
> noise. Luckily, if electron referred, it scales down with cell
> capacitance, and therefore area (unless you or someone shows
> otherwise)

OK, perhaps one should separate two potentially distinct issues:

1. How does pixel read noise scale with pixel pitch?

2. Are John Sheehy's claims that FZ50 pixels are, per area, better
performers than 1D3 pixels, supported by the evidence?

On question 1, I'll have to think about your little treatise. One
thing confuses me, however, and that is your lumping the noise of the
programmable gain amplifier into the "middle read noises" Nm. This
means that they get multiplied by the amplifier gain Gi in your
formulae; but then you say that "This assumes that the variable gain
amplifier is a well designed feedback controlled amplifier, and its
noise is somewhat independent of gain." But in effect, according to
your expression, the noise output by the amplifier attributable to the
amplifier itself IS proportional to the gain, since it's Gi Nm in your
expression. So I'm perplexed as to your meaning there.

As to question 2, we agree that, for current DSLR's, the downstream
circuitry is limiting the DR. A dual amplification scheme recovers
that DR (BTW, Eric Fossum said there might be issues with CDS; do you
know what he was referring to?), and so the comparison between the
FZ50 to the DSLR should be made using that fully recovered DR, not
using the limited DR of fixed ISO. Because, while the FZ50 does do
better than the 1D3 numbers at low ISO, it is about a stop and a half
worse in DR per area than the 1D3 with its sensor DR fully recovered.
So, on that basis, I think John's claims are incorrect, though it is
surprising how favorably the FZ50 pixels compare on a per area basis
with *currently realized* DSLR's such as the 1D3.

There may be in some hoped-for future a means of lowering the small
pixel read noise to about 1 electron (input referred), which is not
simultaneously available for bigger pixels; perhaps the reason will be
the sort of capacitance arguments you have put forth. At that point,
small pixel DR on a per area basis will equal that of the 1D3's fully
realized sensor DR, and small pixels will be competitive on SNR and
DR. But there is no such pixel like that among current examples.
From: ejmartin on
On Jul 20, 10:39 am, John Sheehy <J...(a)no.komm> wrote:
> ejmartin <ejm_60...(a)yahoo.com> wrote innews:d35ccaac-b701-44df-bd71-f767b7169d79(a)c58g2000hsc.googlegroups.com:

>
> A problem I predict with your system is noise-contouring.  I'm not quite
> sure I'm ready to look at pulled-up shadows where the read noise at one
> level 4.5 stops below saturation is 4 to 7 times as high as a stop below
> that level.  

The fact that the red ISO 1600 and blue ISO 100 SNR curves (see the
below-linked figure) are quite close to one another at the upper end
of the ISO 1600 curve's range says that read noise is quite small
relative to photon shot noise in that range. So while the ISO 100
read noise is about 6 times larger than its ISO 1600 read noise in
absolute terms, this is still a negligible contribution to overall
noise, otherwise the SNR curves wouldn't be close to one another. And
if that small difference is really bothersome, we can use ISO 800
instead (shown in black on the linked figure below) to get a truly
seamless match of SNR at the crossover, with only a slight penalty in
SNR in the very lowest stops of EV:

http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/dpr/blend1d3-3.png

Since read noise is so subdominant a component of total noise, there
won't be any "noise contouring" at least from the component of read
noise that is white and gaussian. I think the only issue along these
lines, so to speak, will be pattern noises. I should probably do some
sample blends to see if this is going to be an issue; I doubt it,
since pattern noise is so well controlled on the 1D3, and I certainly
haven't found any situation where it affected IQ in the tonal range
4-5 stops down from saturation at ISO 100.

> If the photosite can be read multiple times
> non-destructively, why not just read it multiple times and add the
> results together, in an extended-DR mode?  You could tell the camera how
> many times to read, depending on the amount of time you can allot between
> shots, and it would add them in a 16-bit or deeper buffer.  That would
> give a natural SNR curve with no contouring bands in the noise.  Or, if
> you can't afford the extra read time and you want different gains, do a
> spread of 3 or 4 and blend them into a better curve.
>
> 4 stops below saturation at ISO 100 (which would be on the upper side of
> your blend, since the top level of the blend must be comfortably below
> ISO 1600 saturation) still results in things like chromatic noise in
> solid colors in the best DSLRs, unless you have something like a 1Ds3
> where it dissolves in the resolution, or print small.
>

From: ejmartin on
On Jul 20, 10:39 am, John Sheehy <J...(a)no.komm> wrote:

Oops, I hit the send button before I was ready...

> If the photosite can be read multiple times
> non-destructively, why not just read it multiple times and add the
> results together, in an extended-DR mode?  You could tell the camera how
> many times to read, depending on the amount of time you can allot between
> shots, and it would add them in a 16-bit or deeper buffer.  That would
> give a natural SNR curve with no contouring bands in the noise.  Or, if
> you can't afford the extra read time and you want different gains, do a
> spread of 3 or 4 and blend them into a better curve.

Because of the only sqrt increase of the SNR as a function of the
number of reads, that is a very inefficient way to improve SNR.

>
> 4 stops below saturation at ISO 100 (which would be on the upper side of
> your blend, since the top level of the blend must be comfortably below
> ISO 1600 saturation) still results in things like chromatic noise in
> solid colors in the best DSLRs, unless you have something like a 1Ds3
> where it dissolves in the resolution, or print small.
>

Yes, the appearance of noise is somewhat different with big pixels vs
small pixels, due to the spatial frequency where it has its support.
But now we're starting to talk about the tradeoffs of big pixels
(better SNR) vs small pixels (better resolution), aren't we?
From: Bob Newman on
On 20 Jul, 16:46, ejmartin <ejm_60...(a)yahoo.com> wrote:
> On Jul 20, 9:48 am, Bob Newman <bob.csx...(a)gmail.com> wrote:
>
>
>
> > We agree that in real, present day cameras ISO doesn't occur in the
> > photosite. In fact, it doesn't really occur at all, what's happening
> > is that the capture system is being finagled to best read the bit of
> > the sensor DR you're actually interested in. As you point out, with a
> > better capture system, no need for ISO. What I was correcting was your
> > statement that no amplification occurs in the photocell of an active
> > pixel sensor. It does (although current rather than voltage
> > amplification) and its noise is probably the irreducible minimum read
> > noise. Luckily, if electron referred, it scales down with cell
> > capacitance, and therefore area (unless you or someone shows
> > otherwise)
>
> OK, perhaps one should separate two potentially distinct issues:
>
> 1. How does pixel read noise scale with pixel pitch?
That's the nub of the question on which hangs the whole 'are small
pixels better/worse (for general photography)'.
> 2. Are John Sheehy's claims that FZ50 pixels are, per area, better
> performers than 1D3 pixels, supported by the evidence?
I'm not sure that's the essence of John's claim. He's claiming that so
far as noise at low ISO's are concerned, on a per area basis imaging
chains based on small pixels can perform better. The choice of actual
hardware is constrained by what he had available. I think he has
demonstrated the 'can' pretty conclusively. Its the 'in all
circumstances' and the 'best possible' that are still up for grabs.
> On question 1, I'll have to think about your little treatise. One
> thing confuses me, however, and that is your lumping the noise of the
> programmable gain amplifier into the "middle read noises" Nm. This
> means that they get multiplied by the amplifier gain Gi in your
> formulae; but then you say that "This assumes that the variable gain
> amplifier is a well designed feedback controlled amplifier, and its
> noise is somewhat independent of gain." But in effect, according to
> your expression, the noise output by the amplifier attributable to the
> amplifier itself IS proportional to the gain, since it's Gi Nm in your
> expression. So I'm perplexed as to your meaning there.
Bad wording on my part. The noise in a feedback amplifier is mainly
due to the first stage, since noise thereafter gets corrected by the
feedback (the first stage adds in the feedback noisily). That means
that first stage noise gets multiplied by the gain, but not that you
get more and more amplifier noise the more gain you have. If I were to
design a sensor (which I'm beginning to convince myself might be fun)
one design avenue that might be worth pursuing is actually feeding
back to the per pixel front end amp, then all the subsequent gain
would essentially come for free, noisewise.
>
> As to question 2, we agree that, for current DSLR's, the downstream
> circuitry is limiting the DR. A dual amplification scheme recovers
> that DR (BTW, Eric Fossum said there might be issues with CDS; do you
> know what he was referring to?),
I think that the problem is you'd need to subtract the reset noise
signal from both channels. Whether that was difficult depends on where
the read channel splits into two. The most common scheme seems to be
to store the reset sample from the source follower on a capacitor,
which is the negative input of a differential amplifier, then the
image signal is sampled and fed into the positive input, to give a
sample less the reset. Your scheme would split the channel there, and
would lose the opportunity to do all the gain in one feedback
amplifier stage, with the noise advantages above. Still, Canon seem to
use at least two separate gain control stages in their high end models
and still get good performance. The Sony IMX021 does CDS in a
different way. It has counter based per-column ADC's. These sample the
reset signal counting downwards, then count upwards again to sample
the image signal. This wouldn't be compatible with the dual channel
idea. Either of those could have been what Eric was referring to. Then
again, he might have something else completely in mind, based on his
extensive experience.
> and so the comparison between the
> FZ50 to the DSLR should be made using that fully recovered DR, not
> using the limited DR of fixed ISO. Because, while the FZ50 does do
> better than the 1D3 numbers at low ISO, it is about a stop and a half
> worse in DR per area than the 1D3 with its sensor DR fully recovered.
> So, on that basis, I think John's claims are incorrect, though it is
> surprising how favorably the FZ50 pixels compare on a per area basis
> with *currently realized* DSLR's such as the 1D3.
I think currently realised is the issue at the moment, both for sensor
technology and signal processing chain. In the best of all possible
worlds many things might be possible on both fronts. EF certainly
seems to have something up his sleeve, and from his slide show, it
seems to include really tiny pixels and really big DR. In the limit,
when you get to a true digital sensor, when each pixel has a FWC of
1e, read noise ceases to be an issue. Near that limit, if each pixel
has a FWC of 2e, it's not much of an issue. This is another thing that
makes me think that you and Roger are not right, fundamentally, on
this. Somewhere between here and there, there would need to be a
turning point when the read noise issue stopped getting worse and
started getting better. In fact, I'm beginning to think I could mount
an inductive proof that you are wrong.
> There may be in some hoped-for future a means of lowering the small
> pixel read noise to about 1 electron (input referred), which is not
> simultaneously available for bigger pixels; perhaps the reason will be
> the sort of capacitance arguments you have put forth. At that point,
> small pixel DR on a per area basis will equal that of the 1D3's fully
> realized sensor DR, and small pixels will be competitive on SNR and
> DR. But there is no such pixel like that among current examples.
John and I would say, because no-one has bothered to develop it,
because it lies so far off the accepted orthodoxy of camera design. I
think that's the way Eric's going, though.
From: ejmartin on
On Jul 20, 11:30 am, Bob Newman <bob.csx...(a)gmail.com> wrote:

>
> I think currently realised is the issue at the moment, both for sensor
> technology and signal processing chain. In the best of all possible
> worlds many things might be possible on both fronts. EF certainly
> seems to have something up his sleeve, and from his slide show, it
> seems to include really tiny pixels and really big DR. In the limit,
> when you get to a true digital sensor, when each pixel has a FWC of
> 1e, read noise ceases to be an issue. Near that limit, if each pixel
> has a FWC of 2e, it's not much of an issue. This is another thing that
> makes me think that you and Roger are not right, fundamentally, on
> this. Somewhere between here and there, there would need to be a
> turning point when the read noise issue stopped getting worse and
> started getting better. In fact, I'm beginning to think I could mount
> an inductive proof that you are wrong.> There may be in some hoped-for future a means of lowering the small
> > pixel read noise to about 1 electron (input referred), which is not
> > simultaneously available for bigger pixels; perhaps the reason will be
> > the sort of capacitance arguments you have put forth.  At that point,
> > small pixel DR on a per area basis will equal that of the 1D3's fully
> > realized sensor DR, and small pixels will be competitive on SNR and
> > DR.  But there is no such pixel like that among current examples.
>
> John and I would say, because no-one has bothered to develop it,
> because it lies so far off the accepted orthodoxy of camera design. I
> think that's the way Eric's going, though.

I just remembered, while we are waiting for production small-pixel
CMOS sensors from Canon, there is one further data point: that 52MP,
APS-H sized sensor that they made a prototype of:

http://www.imagesensors.org/Past%20Workshops/2007%20Workshop/2007%20Papers/076%20Iwane%20et%20al.pdf

3.3µ pixels, with 5.5 electrons of read noise. Still not getting
smaller in proportion to pixel pitch (in fact, a bit worse; granted,
it's preproduction, but if it were easy to beat down the read noise by
making the pixels smaller, shouldn't they have been able to at least
match the performance of pixels with 4 or more times the area?).