From: John Sheehy on
Lionel <usenet(a)> wrote in

> Not that I know of. But all the Canons already offer this amazing
> 'binning' feature anyway - that's what you get when you set your
> camera to JPEG mode, & use a JPEG size smaller than full resolution.
> ;^)

Well, downsampling and binning are a bit different. A downsampling should
use a filter to remove frequencies that complicate and artifact the output.
Binning is just a box effect, with the contents literally added together.


<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS(a)no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
From: Roger N. Clark (change username to rnclark) on
John Sheehy wrote:
> "Roger N. Clark (change username to rnclark)" <username(a)> wrote
> in news:45FC7FEE.7010800(a)
>>Your number for the 20D is incorrect. What you refer to as read noise
>>is not true sensor read noise, but the limitation of a 12-bit A/D
> I didn't say that it was "true sensor read noise". I said that it was
> "read noise" and sometimes I qualify it by saying "blackframe read
> noise".
> Regardless of the source, it is making the shadows of the RAW files much
> less useful than what the sensor wells themselves record.

But you keep comparing this artificial A/D limit to other
cameras (with small pixels) that aren't limited by the
A/D converter.
>>The 20D has a full signal well depth of 51,000 electrons.
> No, it has about 44,000. It has about 26,500 at RAW saturation at ISO
> 200, but ISO 100 has less headroom, and the camera stretches the upper
> highlights to reach 4095 ADUs, like the other ISOs on the camera.

The test data I've seen do not indicate this, whether it is
44,000 or 51,000 makes little difference; that is only 0.2 stop

> Here is the same scene, with twice the exposure time on the left, ISO 100
> on the left, and ISO 200 on the right:
> The ISO 100 comes to clipping faster in the gradient, even though all the
> darker tones are exactly the same.

And is your camera calibrated to better than 0.2 stop to be
sure you are seeing true sensor property. The calibration
includes the f/stop, the shutter speed, the light source
stability, and the ISO gain. Again, it matters little
in the argument.

>>So: 51000/4095 = 12.5 electrons per A/D bit.
> The 20D has 3967 meaningful ADUs; 0 through 127 are only for negative
> noise.
> 51000/3967 = 12.85 electrons per ADU. Read noise is 2.07 ADU at ISO 100.
> Read noise is then 12.85*2.07 = 26.61 electrons.

Your smaller range (3967) only makes your argument worse.

>> Add +/- 1 bit noise on each A/D reading,
> +/- 0.5 ADU, with a global offset of +/- 0.5 ADU, which should be
> accounted for in blackpointing the RAW data, and the former adds like any
> other noise; the square root of the sum of the squares, so you don't add
> it linearly. It sometimes goes positive when the analog noise goes
> negative, and visa-versa.

No, the error in quantization is +/- 1 bit. There are no "half" bits!

>>and noise should be about 1.4 bits, or
>>1.4 * 12.5 = 17.5
> What kind of math is that?

I took the 2-bit error and assumed the RMS would be root 2,
so the factor 1.4. But I have underestimates the A/D noise.
It seems that the minimum noise is more like 1.8 ADU.
This is illustrated in Christian Buil's wen page, see Table 3:
The data for canon sensors clearly shows all the sensors
bottoming out just above 1.8 ADU.

> You're multiplying a term called "bits" which
> should be a logarithm, I think, by a linear ratio. Also, where did the
> 1.4 come from? The noise in an ISO 100 blackframe from a 20D is about
> 2.07 ADU. 2.07 * 12.5 = 25.87. 2.07 * 12.85 = 26.61. Neither is close
> to 17.5.

1 bit = 1 ADU, the smallest unit of change.

>>electrons not including actual sensor read
>>noise. DSLRs are so good at low ISOs, they are limited at the
>>low end by A/D converter electronics, not sensor read noise.
> Whatever it is, it is a problem. 2.07 ADU of noise is not caused by bit
> depth. It is analog noise, regardless of where it happens in the signal
> path.

What is your evidence for the noise source. I pointed you to
Christian Buils web page that clearly shows the read noise
from multiple sensor bottoming out above 1.8 ADU. Then 16-bit
CCDs used in amateur and professional astronomy properly
digitize lower read noise than 26 electrons. So do point
and shoot cameras with smaller well depths. P&S cameras
with their cheaper electronics don't seem limited by
some mythical 26 electron analog noise. To the contrary,
a 1.8 ADU error in A/D conversion best describes all the
available data, from P&S to scientific systems.

>>Small pixel P&S cameras get so few photons, that the entire range
>>is adequately characterized by 12-bit converters.
> No argument there, for high ISOs. The FZ50 has 4800 photons per pixel,
> and 3982 (IIRC) RAW values at ISO 100. 12 bits could not count that
> accurately.

Why not. 4800 electrons/3982 = 1.2 electrons/ADU, pretty good.

> Even at ISO 200, it would cause an uneven histogram, with a
> gap every so often. Electrons need to be oversampled by about 3x or so
> to avoid erratic histograms (the effects, of course, are quite subtle,
> but if you really were just counting photons, it could make a visible
> difference in deep shadows pushed in PP). Of course, read noise make
> exact photon counting impossible with any bit depth.

Sorry, but I see no evidence for the need to oversample
electrons 3x. That would be ISO 4800 on the Canon 5D!
Again, see Christian Buil's web page. He has a section
on optimal gain: see Table 5:

What you need is adequate sampling of the noise, not an electron.
Christian derives, for example the optimal gain for the 5D
at ISO 1100, or 1.5 electrons/ADU.

>>Again, not doing well with 12-bit A/Ds. That is why canon has
>>now announced a camera with 14-bit A/Ds. We'll see more of that
>>in the future,
> Again, probably marketing lies. The read noise is analog. The 1DmkIIIs
> that Canon has released to reviewers has the same read noise, relative to
> max signal at ISO 100, as the mkII does.

Hmmm. Doesn't fit your paradigm, so its marketing lies.
Canon claims significant improvement in shadow detail.
We'll have to wait and see until real tests are performed,
not reviewers using raw converters that mes sup the signal.

> You greatly overestimate the role of bit depth. Its effect is rather
> subtle in the ranges we're talking about.


>>The real world marches to a complete description of physics,
>>not a narrow view to push an agenda.
> The real world has small pixels that bin down to better pixels than the
> big DSLR pixels, and are better also, unbinned, with the same
> magnification of the sensor surface. So to say that larger pixels are
> needed for IQ and SQ is nonsense.

Sorry, not the physics of the real world. If we had perfect
sensors with zero added noise, then small pixels summed
could equal larger pixels, but not better them.

From: acl on
On Mar 18, 8:59 am, "Roger N. Clark (change username to rnclark)"
<usern...(a)> wrote:

> >>and noise should be about 1.4 bits, or
> >>1.4 * 12.5 = 17.5
> > What kind of math is that?
> I took the 2-bit error and assumed the RMS would be root 2,
> so the factor 1.4. But I have underestimates the A/D noise.

He is trying to say that bits=log(counts)/log(2), and you meant
counts, not bits.

> It seems that the minimum noise is more like 1.8 ADU.
> This is illustrated in Christian Buil's wen page, see Table 3:
> The data for canon sensors clearly shows all the sensors
> bottoming out just above 1.8 ADU.

From: Roger N. Clark (change username to rnclark) on
Doug McDonald wrote:

> For sensitivity, one does want, for a given number of pixels,
> the largest sensor size possible. Of course, to use this one
> needs large, fast, expensive lenses ... and the concommittent
> loss of depth of field.

We've been over this one before in this news group.
For the same image quality, there is NO loss in
depth of field by using a larger sensor.
Details are discussed here:

From: Roger N. Clark (change username to rnclark) on
Paul Rubin wrote:
> "David J. Littleboy" <davidjl(a)> writes:
>> The point/claim is that pixel binning (or noise reduction plus downsampling)
>> will result in the same image as the larger pixels would have in lower
>> light. One is collecting the same number of photons, so this should/might
>> work.
> Is there experimental validation for this claim? My experience has
> been not so encouraging but I'm probably not using the best possible
> methods.

Binning methods are used in astronomy all the time.
Sometimes binning is needed to improve the signal-to-noise ratio
on faint subjects.

It is also used in scientific work. Thus it is well understood,
and the claims being made here about equaling or bettering
a larger pixel sensor have never been demonstrated. The one
example being used in this thread to support the claim is
a restricted case where the larger pixel sensor is A/D limited,
not sensor performance limited. If John would do the test
with the same two sensors at a higher ISO where the
large pixel sensor was not A/D limited, he would come to the
opposite conclusion.