From: John Sheehy on
"acl" <achilleaslazarides(a)yahoo.co.uk> wrote in
news:1174048371.412781.12090(a)n76g2000hsh.googlegroups.com:

> On Mar 16, 7:04 am, "Roger N. Clark (change username to rnclark)"
> <usern...(a)qwest.net> wrote:
>
>> 2) There is a difference. The signal you record has added
>> read noise. A larger pixel collects more photons
>> so the signal is larger compared to the read noise.
>> Thus you can detect fainter things, or have better high
>> ISO performance. If you sum the signal from a smaller
>> pixels to equal the area of a larger pixel size,
>> you are also adding read noise, so you don't gain
>> as much as having the larger pixel with one read noise.
>>
>
> Is read noise fixed per pixel, per unit area, or something else?

It's something that happens when reading the pixels, and has more to do
with the electronics involved than anything else. It includes noise from
all parts of the readout chain (at the pixel, in the amplifier, in the
readout wiring, in the ADC, etc. Canon is very good at keeping it low,
relative to signal, at high ISOs compared to most other companies, which
seem to just amplify the same noise to a higher amplitude.

There are basically three categories in the read noise vs ISO category;
Cameras like Canon DSLRs that have less read noise at ISO 1600 than ISO
100, in electrons, cameras like the Pentax K10D which have the same read
noise in electrons at all ISOs, and cameras like my FZ50, which does a
very good job of readout at ISO 100, but gets more read noise in
electrons at ISO 100.

Of course, the most relevant part of read noise at the pixel level, as
measured in units of electrons, is the ratio of the maximum number of
electrons digitized at the ISO, and the read noise in electrons. That
determines the lowest signal level, relative to saturation, where a 1:1
S/N could be obtained, if blackframe read noise was the only noise.

--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS(a)no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
From: John Sheehy on
John Sheehy <JPS(a)no.komm> wrote in
news:Xns98F58CDA865F2jpsnokomm(a)130.81.64.196:

> Blackframe read noise on my FZ50 is about 3.34 electrons at ISO 100
> (4800 electrons at saturation), and about 2.71 electrons at ISO 1600
> (about 300 electrons at saturation). Binned down 3x3 (to DSLR size),
> that's about 0.9 and 1.11 electrons (43,200 and 2700 max),
> respectively.

Sorry, I had just woken up after an 18-hour emergency shift at work.

I divided by three when I should have multiplied by three. I was thinking
in terms of the noise-to-signal ratio, and applied it to absolute photon
counts.

That should have read:

Blackframe read noise on my FZ50 is about 3.34 electrons at ISO 100
(4800 electrons at saturation), and about 2.71 electrons at ISO 1600
(about 300 electrons at saturation). Binned down 3x3 (to DSLR size),
that's about 8.1 and 10 electrons (43,200 and 2700 max),
respectively.


--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS(a)no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
From: acl on
On Mar 16, 8:53 pm, John Sheehy <J...(a)no.komm> wrote:
> "acl" <achilleaslazari...(a)yahoo.co.uk> wrote innews:1174048371.412781.12090(a)n76g2000hsh.googlegroups.com:
>
> > On Mar 16, 7:04 am, "Roger N. Clark (change username to rnclark)"
> > <usern...(a)qwest.net> wrote:
>
> >> 2) There is a difference. The signal you record has added
> >> read noise. A larger pixel collects more photons
> >> so the signal is larger compared to the read noise.
> >> Thus you can detect fainter things, or have better high
> >> ISO performance. If you sum the signal from a smaller
> >> pixels to equal the area of a larger pixel size,
> >> you are also adding read noise, so you don't gain
> >> as much as having the larger pixel with one read noise.
>
> > Is read noise fixed per pixel, per unit area, or something else?
>
> It's something that happens when reading the pixels, and has more to do
> with the electronics involved than anything else. It includes noise from
> all parts of the readout chain (at the pixel, in the amplifier, in the
> readout wiring, in the ADC, etc. Canon is very good at keeping it low,
> relative to signal, at high ISOs compared to most other companies, which
> seem to just amplify the same noise to a higher amplitude.

Maybe I didn't phrase my question carefully: Is it dependent on the
area of the pixel? For example, something like thermal noise should be
proportional to the active volume of the photodetector (to a first
approximation at least), while noise which is produced by the
amplifier should be independent of the actual area. The point of the
question is to find out whether it will scale down with the pixels or
not, and if it does, do we gain or lose? eg if it was proportional to
the volume, then scaling down to half the linear dimensions would
increase the s/n ratio, as the number of photons detected would be 1/4
but the noise would be 1/8 (this is supposed to be a sketch to explain
what I mean, not a serious argument).

Of course it is obvious that eventually there's a limit, for small
enough structures everything changes, but this should be at the scale
of a few nm at most, not microns, so irrelevant for us.

>
> There are basically three categories in the read noise vs ISO category;
> Cameras like Canon DSLRs that have less read noise at ISO 1600 than ISO
> 100, in electrons, cameras like the Pentax K10D which have the same read
> noise in electrons at all ISOs, and cameras like my FZ50, which does a
> very good job of readout at ISO 100, but gets more read noise in
> electrons at ISO 100.
>
> Of course, the most relevant part of read noise at the pixel level, as
> measured in units of electrons, is the ratio of the maximum number of
> electrons digitized at the ISO, and the read noise in electrons. That
> determines the lowest signal level, relative to saturation, where a 1:1
> S/N could be obtained, if blackframe read noise was the only noise.
>

Yes and the question relevant to our subject is how the two scale. The
signal scales with the square of the linear size, but the noise? That
is the point.

From: John Sheehy on
"Roger N. Clark (change username to rnclark)" <username(a)qwest.net> wrote
in news:45FA1739.7040208(a)qwest.net:

> If you sum the signal from a smaller
> pixels to equal the area of a larger pixel size,
> you are also adding read noise, so you don't gain
> as much as having the larger pixel with one read noise.

No. Read noise doesn't add linearly. 9 pixels binned into one triples the
pixel read noise in electrons, while multiplying the signal electrons by 9,
resulting in 3x the signal-to-readnoise ratio. Have you ever binned
blackframes?

With CCDs, on-chip binning can actually increase the ratio further, as they
can add the same read noise to 4 pixels as they can to one pixel. There's
a Dalsa paper about this technique. Such hardware binning is the only
reason besides storage/speed issues to do the binning in-camera, IMO;
otherwise, more and smaller pixels are better (especially if RAW data is
not clipped at the blackpoint).

--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS(a)no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
From: John Sheehy on
"acl" <achilleaslazarides(a)yahoo.co.uk> wrote in
news:1174058129.036168.91300(a)e65g2000hsc.googlegroups.com:

> But personally I hope this kind of noise could go down enough so we
> can have high pixel density sensors which will give more flexibility
> in trading off noise for resolution. Hopefully with built-in binning
> for the raw files too, as always having 60MB raw files seems a bit
> wasteful (but the again 300KB for an executable seemed huge to me in
> 1988).

The smaller the pixels become, the less bit depth you need to record them
at the same level of accuracy.

--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS(a)no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><