How does the D3 achieve such high ISO?

Started Feb 17, 2008 | Discussions
PerL Forum Pro • Posts: 14,464
Re: I want high ISO

SDRebel wrote:

It isn't the end of the world; however, I want Canon to provide me
the high ISO option the D3 achieves.

If that requires a tradeoff, then I may find the price of high ISO
isn't worth paying, e.g., in lower quality files at other ISO
settings.

But, if higher ISO can easily be provided by Canon with no
significant tradeoff, then I would like to see a firmware upgrade, if
that is possible.

I follow your point, yet I see it as engaging in a discussion similar
to one I remember of whether or not carburators would inevitably be
replaced by fuel injection.

The first thing I needed to know in that discussion was what fuel
injection did that carburators didn't, if anything. Or, was fuel
injection just marketing hype? Well, the answer has been in for a
long time now; but in the beginning it generated quite a bit of
discussion and required a learning curve.

So, again, how has Nikon accomplished the high ISO provided by the
D3? Can Canon provide it in the 1Ds3 with a firmware upgrade or will
it have to await a new sensor design? Is there a tradeoff that makes
high ISO ultimately undesirable for most shooting conditions, given
even the Nikon technology?

I still am not certain that I've read any definitive answers to my
initial questions.

Many thanks for your thoughts -- I enjoy learning as well as I enjoy
shooting with my 1Ds3 with the technology is presently offers even if
that is limited for the moment to ISO 3200.

If you want to shot at the highest ISOs a Canon 1D3 or a Nikon D3 will both give you less noise even if you downsample the 1DS3 to the same resolution. You will not get the same low noise at ISO 6400 or above with a firmware upgrade of a 1Ds3, even with an "artificial" ISO 6400 setting.

PerL Forum Pro • Posts: 14,464
The bottom line....

ohyva wrote:

Marianne Oelund wrote:

ejmartin wrote:

Marianne, where are you getting your figure of 22K electrons at
saturation for the 1D3 at ISO 200? Mine has 71K electrons at
saturation for ISO 100 and about 38K at ISO 200. 22K seems awfully
low, more like the value for the 1D2.

Can you reconcile your values to Roger Clark's 12-bit unity gain ISO
value of 1000 for the Mk III? That would imply 20,500 electrons
full-scale at ISO 200.

First it's good to remember the full-well capacity in Canon cams are
reached at ISO100 while in D3 it's at ISO200. So comparison at ISO200
is apples against oranges. What's then more important, low noise in
general or low noise at certain expo-time&aperture settings, depends
then on the aimed use of the gear. I persponally prefer speed for
sport photos, so the new Nikon would be tempting if I'd not be so
committed with my current Canon gear.

Second I'm still sceptic about these well capacity "measurements". No
direct measurement method AFIAK (unless you probe into the chip which
I think has not been done (and would probably to some extend
interfere with the masurement). If done indirectly based on noise
figs, then I'd really love to know how all the different noise
sources are separated. And how to ensure there is no in-sensor
digital domain NR applied - which is reported by Sony to be one of
the methods how they have reduced the noise levels in their new CMOS
sensor technology.

As for the change in Clark's data, he had measured the gain at ISO
50, not realizing that ISO 50 is really ISO 100 overexposed by a
stop. So values for gain above ISO 50 are double what he intially
reported, and read noise in electrons consequently also doubled for
all ISO's above 50. (There was clearly a problem, since if you took
the original figures seriously, the 1D3 had the worst photon
collection efficiency of any Canon DSLR since before the 10D).

Thanks for the explanation - I was wondering. Has this also caused
an error in his unity-gain ISO calculation?
If the Mk III unity-gain ISO is actually closer to 2000, then the D3
and Mk III would be back to rough parity regarding read noise, with
the D3 retaining an advantage in shot noise.

...is the images. There have been many high ISO comparisions made between 1D3 an D3, but very few low ISO comparisions between these two.

In a swedish test (the magazine Kamera&Bild) the D3 at low ISOs could take more punishment in PP (lifting extreme shadows) than the 5D and the 1D3s. Here I suppose the larger photsites gave the D3 an advantage over the 1Ds3, and vs the 5D it had better shielded electronics.

It would be interesting to see a similar low ISO/postprocessing-ability comparision between the D3 and a 1D3 where the size of the photosites should be reasonably close and the shielding and the quality of the electronics probably are on the same level.

D3 pusher Regular Member • Posts: 250
Re: I want high ISO

well I have no doubt that Canon will produce sensor that will equal or better D3, its given that they will. If you are getting the results with your current equipment you have no problems. If you want something now then D3 is it, again as long as you are looking for good all around pro body that you can take anywhere. If your needs are more studio oriented then 1DS should do wonders.

I use both D3 and 5D and other cams I have friends who makes living doing this for me its a hobby. What is that one picture worth to you can you afford to wait or do you get the shot others didnt. Again its question you can only answer, you could get the camera and be making money at it and when something better comes out get that doesnt matter really whether its Canon or Nikon. Good luck with your quest!

Tannin Senior Member • Posts: 1,448
How many times do you have to be told?

SDRebel wrote (about 17 times so far in this thread):

How did Nikon do it, whether or not it works to
everyone's satisfaction? And, can Canon provide us 1Ds3 owners with
higher ISO via a firmware upgrade?

(1) Nikon used bigger pixels.
(2) No.

Everything else is minor, small-change stuff. Bigger pixels = lower noise. Period. End of story. Nothing more to discuss. Do you get it yet?

You want lower noise, get a camera with bigger pixels. (Like a D3.) You want higher resolution, get a camera with smaller pixels, and accept that you will get more noise. Laws of physics.

Those two questions remain unanswered in any definitive fashion;

They have been answered clearly and accurately and in very simple language so many times in this thread that I lost count.

If you want low noise, get something with bigger pixels - a 5D or a D3 or a 1D III.

5D had big pixels: it's got good low noise. 1D III has fairly big pixels and latest sensor tech, it too has very good low noise. D3 has big pixels and latest-generation sensor tech, so you would naturally expect it to be the low noise champion -which according to most reports it is. IDs III is latest generation tech but it has (relatively) small pixels to give it high resolution at the cost of higher noise. It's a design trade-off. In the real world (hey, remember the real world?) engineers make design trade-offs all the time. If you don't like hte design trade-off in your your 1Ds III, get a D3 or a 5D or a 1D III.

Bigger pixels = lower noise. Period. End of story.

Do you get it yet or should I write it in bigger letters? Oh, and please quit whining about the laws of physics. You can whine about your boss or your wife or your football team, but when you whine about the laws of physics god stops listening.

 Tannin's gear list:Tannin's gear list
Canon G9 X II Canon EOS-1D Mark IV Canon EOS 7D Mark II Canon EOS 5DS R Canon EF 35mm F1.4L USM +9 more
TheVoIP Contributing Member • Posts: 883
Nikon is the best in noise reduction. That's all

SDRebel wrote:

The jury seems to be in on the Nikon D3's successful implementation
of significantly higher and usable ISO than I can achieve with my
1Ds3.

Has Nikon made a breakthrough in sensor design? Or, is the Nikon ISO
range a result of a breakthrough in post processing?

Thanks to Nikon bad past experinces with CCDs that are terrible with noise. Thier engineers have no choice but to researce heavily on noise reduction post processing in order to get 'accpetable' image.

When they have similar CMOS sensor as Canon, thier advantages of having previous noise reductions algorithm are rewarded, with the so call, low noise at ISO6400.

The fact is that.... However, to me excessive noise reduction destroy image quality. I wonder why in the first place people buying great lenses that insist on sharpness, where they are willing to scarify that for low noise, which can be done on computer software as well?

Anyway, the success of D3 works for most people because they were not from photography background and do not have the capability to judge image quality using thier own eyes. Thus the cheat from Nikon works for them.

I am not saying that D3 suck... I am saying it is not better than 2 years old EOS 5D nor 1D mark III in actual image quality. I am still with Canon because it clearly provide me with some of the best lenses for the job; EF85 f1.2 L II, EF135 f2 L and EF100-400 IS L that Nikon do not have.

Shaun_Nyc
Shaun_Nyc Senior Member • Posts: 2,327
Marianne

Marianne Oelund wrote:

Marianne, I would have thought I would have seen your congrats to this fella with the the IDIII ice skating photos. ??? Maybe you could post a few with the mighty D3 so we can see how bad it out performes it..Btw, when are you sending yours in for recall repair so some fair judgement can be passed on your copy ?

http://forums.dpreview.com/forums/read.asp?forum=1032&message=26836992

 Shaun_Nyc's gear list:Shaun_Nyc's gear list
Nikon D50 Nikon D810 Canon EOS R6 Nikon AF-S Nikkor 70-200mm f/2.8G ED VR Nikon AF-S Nikkor 14-24mm f/2.8G ED +15 more
PerL Forum Pro • Posts: 14,464
Ha, ha

Anyway, the success of D3 works for most people because they were not
from photography background and do not have the capability to judge
image quality using thier own eyes. Thus the cheat from Nikon works
for them.

So most people who buys a D3 for 5000 dollars doesnt have any photographic background? Sounds probable.

ejmartin Veteran Member • Posts: 6,274
Who doesn't get it

Tannin wrote:
SDRebel wrote (about 17 times so far in this thread):

How did Nikon do it, whether or not it works to
everyone's satisfaction? And, can Canon provide us 1Ds3 owners with
higher ISO via a firmware upgrade?

(1) Nikon used bigger pixels.
(2) No.

Everything else is minor, small-change stuff. Bigger pixels = lower
noise. Period. End of story. Nothing more to discuss. Do you get it
yet?

You want lower noise, get a camera with bigger pixels. (Like a D3.)
You want higher resolution, get a camera with smaller pixels, and
accept that you will get more noise. Laws of physics.

Well, here you would be wrong. The 1Ds2 has smaller pixels than the 5D, but collects more photons per unit area than the 5D. The 40D has smaller pixels than the 1Ds3, whose pixels are smaller than the 1D3, yet they all collect the same amount of light per unit area of sensor. And photon noise is related to the number of photons collected. Read noise per unit area is similar.

So if one is not fixated on the pixel (which corresponds to a different spatial scale depending on the pixel size) but rather looks at a fixed scale, pixel size is essentially irrelevant within a wide range, its major effect is on resolution, not noise. Unless of course you print images based on pixel size and so different sizes for different cameras, rather than some standard size where smaller pixels are a smaller percentage of the image.

-- hide signature --
ejmartin Veteran Member • Posts: 6,274
Re: 1D3 full well?

ohyva wrote:

Second I'm still sceptic about these well capacity "measurements". No
direct measurement method AFIAK (unless you probe into the chip which
I think has not been done (and would probably to some extend
interfere with the masurement). If done indirectly based on noise
figs, then I'd really love to know how all the different noise
sources are separated. And how to ensure there is no in-sensor
digital domain NR applied - which is reported by Sony to be one of
the methods how they have reduced the noise levels in their new CMOS
sensor technology.

There are three common noise sources -- the electronic noise in the readout and amplification circuits, thermal noise in those same circuits, and photon shot noise. They are all independent and can all be measured by varying the exposure.

The thermal noise is proportional to the time of exposure, and can be measured in a series of longer and longer exposures with the lens cap on; it is typically irrelevant until exposure times exceed a second or more.

Electronic noise is there independent of exposure, it just comes from reading out the sensor, and so can be measured at very short exposures with the lens cap on (negligible thermal noise, no photon signal).

Photon noise varies proportional to the square root of the number of photons collected, and can be measured by a set of exposures at higher illumination levels where the other noise sources are negligible (of course, one can subtract them out for a slight improvement in the accuracy).

So they can all be separated from one another by varying the exposure, since they all vary differently with exposure.

As for whether NR is applied, all Canons and probably other CMOS sensors use a method to cancel noise in the readout of a pixel. This is not NR as most people think of it, ie smearing out neighboring pixels to smooth out the noise grain; rather it is done for each pixel independently and involves no comparison with neighboring pixels. Noise reduction that involves averaging over nearby pixels is easily detected by looking at the Fourier transform of the image; if this sort of NR were being performed one would see the noise at high spatial frequencies take a nose dive relative to lower frequencies. I have examined raw images of the D3 in this way and see no evidence of such NR -- the noise spectrum is flat all the way out to the limits of the sensor resolution (Nyquist frequency).

-- hide signature --
sandy b
sandy b Veteran Member • Posts: 9,723
Nothing personal, but ignorance is best hidden

not paraded out for all to view. The nikon has great images because of sensor technology. They had a long time to think about it. And of course, canon will catch up and surpass them. Ad infinitum. Thats good, makes for better cheaper technologies.

 sandy b's gear list:sandy b's gear list
Nikon D750 Nikon D7500 Nikon Z6 Nikon Z50 Nikon AF-S Teleconverter TC-14E II +12 more
guycollier Contributing Member • Posts: 573
Re: Nikon is the best in noise reduction. That's all

Anyway, the success of D3 works for most people because they were not
from photography background and do not have the capability to judge
image quality using thier own eyes. Thus the cheat from Nikon works
for them.

You know, I read some absolute drivel here from time to time but this is perhaps the single most moronic statement yet.

-- hide signature --

Can't sleep. Clowns will eat me.

OP SDRebel Senior Member • Posts: 1,489
Re: How many times do you have to be told?

I hear you.

I had a 5D and sold it upon taking delivery of my 1Ds3.

My 5D did not allow me to set ISO at 12800. Nor did it allow, so far as I knew when I owned it, to get the results I've seen posted for the D3.

Please show me 5D results that match the D3 at these high ISO conditions. I may not have known how to the use the camera and may have made a great mistake selling it. If you can demonstrate that to me, I can probably sell my 1Ds3, given current availability, for more than I paid for it and buy back a 5D.

As for definitive answers, you were obviously satisfied by the first comment that the 5D and 3D have larger sensors -- a fact of which I was well aware.

However, contrary to your assertion, some people have suggested that with my 1Ds3 I can obtain high ISO results. I've been experimenting and am so far amazed at what I can obtain. I just don't know how those results compare to the D3.

Finally, if I understand you correctly, in your opinion Nikon has produced nothing in the way of new technology with the D3. It's just a 5D at a higher price as far as ISO goes. Am I correct?

Anyway, no doubt I am much slower than you in understanding these issues. Please have patience. Surely in school you had to put up with people like me. Sorry. Thankfully others are more compassionate.

ohyva Veteran Member • Posts: 6,342
Re: 1D3 full well?

ejmartin wrote:

ohyva wrote:

Second I'm still sceptic about these well capacity "measurements". No
direct measurement method AFIAK (unless you probe into the chip which
I think has not been done (and would probably to some extend
interfere with the masurement). If done indirectly based on noise
figs, then I'd really love to know how all the different noise
sources are separated. And how to ensure there is no in-sensor
digital domain NR applied - which is reported by Sony to be one of
the methods how they have reduced the noise levels in their new CMOS
sensor technology.

There are three common noise sources -- the electronic noise in the
readout and amplification circuits, thermal noise in those same
circuits, and photon shot noise. They are all independent and can
all be measured by varying the exposure.

Are they? I did study for 3 years semiconductor technologies as my minor in university, and I'd not dare to make this sort of statement - but that was now close 3 decades ago when those technologies were quite young. Perhaps you can point me some more recent technical studies clarifying that.

And you should not forget the dark current noise which is the dominant noise source in long exposures.

The thermal noise is proportional to the time of exposure, and can be
measured in a series of longer and longer exposures with the lens cap
on; it is typically irrelevant until exposure times exceed a second
or more.

Have you counted the increase of thermal noise flow with exposure time due to sensor heating? What anout all sort of paracitic leagages which may be severy (and non-linearily) dependent on the collected charge, the trasmitted signal level and of course sensor temperature?

Electronic noise is there independent of exposure, it just comes from
reading out the sensor, and so can be measured at very short
exposures with the lens cap on (negligible thermal noise, no photon
signal).

I'd not be too sure of this either, as some of the paracitic components (being the deliver mechanisn on the intra-component transition noise) in the chip may vary on the signal level - and thus cause unknown quantities.

Photon noise varies proportional to the square root of the number of
photons collected, and can be measured by a set of exposures at
higher illumination levels where the other noise sources are
negligible (of course, one can subtract them out for a slight
improvement in the accuracy).

So they can all be separated from one another by varying the
exposure, since they all vary differently with exposure.

Probably true to some accuracy level. But I cannot say if that within unit of percentages or tens of percentages.

As for whether NR is applied, all Canons and probably other CMOS
sensors use a method to cancel noise in the readout of a pixel. This
is not NR as most people think of it, ie smearing out neighboring
pixels to smooth out the noise grain; rather it is done for each
pixel independently and involves no comparison with neighboring
pixels.

Yep, the coherent doubly sampling can reduce, typically not eliminate the transition based noise. Or when poorly designed it actually double the noise level (2nd outcome from the same equations), but I guess Canon has not made a bad design.

Noise reduction that involves averaging over nearby pixels
is easily detected by looking at the Fourier transform of the image;
if this sort of NR were being performed one would see the noise at
high spatial frequencies take a nose dive relative to lower
frequencies. I have examined raw images of the D3 in this way and
see no evidence of such NR -- the noise spectrum is flat all the way
out to the limits of the sensor resolution (Nyquist frequency).

As NR is typically decision controlled median type of local filtering I very much doubt you can see that clearly in any fourier based analysis. But I have to admit it's quite long time since my last math lessons.

I don't have either any evidence of the D3 sensor what it has or what it does - only the material published by Sony of their new sensor technology, and even that is pretty high level illustration.

megapixeldave Senior Member • Posts: 1,746
how ironic...

Honor wrote:

Marianne Oelund wrote:

Honor wrote:

The Canon 1Dm3 sensor has better high ISO than the Nikon D3. period.

I own both cameras, and have tested both extensively. I do not rely
on other's test results, which are too easy to misinterpret.

You credibility is about the same as a cow in a pasture who tells me
it saw an ufo last night. You have published zero to substantiate
your fallacies yet you accuse other people of tainting their results
when they at least have made the effort to publish their findings.
This foremosts tells me you could well be a Nikon troll. While I am
ready to believe you own a D3, I highly doubt you own a 1dmk3.

Here is what you hit:
credibility 0%
arrogance 80%

Honor...for someone that has only been on dpreview for a short few months... no website/pbase to validate that you even know how to shoot a camera...in contrast to Marianne's 4 years of posting history... real name ....legit website...

your the one lacking in credibility as i see it. as for arrogance...you seem to have no shortage given that you supply nothing to illustrate your own credibility but are quick to point the finger at Marianne.

david
http://www.davidprobst.com

Marianne Oelund Veteran Member • Posts: 7,788
Something else to try

SDRebel wrote:

Please show me 5D results that match the D3 at these high ISO
conditions. I may not have known how to the use the camera and may
have made a great mistake selling it. If you can demonstrate that to
me, I can probably sell my 1Ds3, given current availability, for more
than I paid for it and buy back a 5D.

Don't do anything hastily. Another technique you can try is down-sampling. This allows you to effectively trade your higher pixel count for lower noise. [Strictly speaking, the result won't be exactly the same as using a camera with fewer and larger pixels, but you can get quite close.]

Your 21Mpix camera can do a much better job at emulating a lower-noise 12Mpix camera, than the 12Mpix camera can do emulating the resolution of a 21Mpix camera by up-res techniques.

Now you have quite a bit of experimentation to do, before you will have a thorough understanding of the ultimate capabilities of your camera. Give your 1Ds Mk III some time, especially if you've been very happy with its images so far.

Hans Kruse Veteran Member • Posts: 3,432
Re: Something else to try

Marianne Oelund wrote:

Don't do anything hastily. Another technique you can try is
down-sampling. This allows you to effectively trade your higher
pixel count for lower noise. [Strictly speaking, the result won't be
exactly the same as using a camera with fewer and larger pixels, but
you can get quite close.]

I did a comparison between my 5D and my new 1Ds3, see http://forums.dpreview.com/forums/read.asp?forum=1032&message=26828157

After I shot the pictures I noticed that I had to adjust to make the brightnes exactly the same. You can see on each how much is adjusted in Lightroom. So for ISO 3200 the 1Ds3 had to be adjusted 1/2 stop up to match the 5D picture in brightness. The metering is different on the two camera. I could have repeated the whole test, but decided that the test still proves the point of the 1Ds3 being very close to the 5D when down sampled to the same resolution and this was slightly better than I had expected.

Your 21Mpix camera can do a much better job at emulating a
lower-noise 12Mpix camera, than the 12Mpix camera can do emulating
the resolution of a 21Mpix camera by up-res techniques.
Now you have quite a bit of experimentation to do, before you will
have a thorough understanding of the ultimate capabilities of your
camera. Give your 1Ds Mk III some time, especially if you've been
very happy with its images so far.

That is true and I would say based on my experience with the 1Ds3 so far that even iso 3200 in many cases can produce very nice and detailed pictures, that the 5D would be challenged to do. So was this important enough for me that I expected this before I bought it? No, I was pleasantly surprised. I bought the camera since I wanted a really all round camera with high resolution, absolute top detail and resolution, good iso performance, very good AF performance, robust, fast, enough frame rate for wild life, plenty of pixels to crop from and still have good resolution for A3+ sized prints. The 1Ds3 have so far lived up to all that and more. I only had it for a bit over a week now and have shot about 2500 shots. I have figured out how auto iso works (not documented well) and it is great. Just wonder why it took Canon that long to get it done.

 Hans Kruse's gear list:Hans Kruse's gear list
Canon EOS 5DS
Hans Kruse Veteran Member • Posts: 3,432
Re: Reduce your 1DsIII image to 12MP (about 50%)

Eric Sorensen wrote:

...then compare with the D3 and tell us how the noise looks at the
new and equal 100% level. This is the difference you will see when
the same framed image is printed the same size. I bet it's close, if
not in Canon's favor.

As mentioned in a previous post I see the noise when down sampling to the same resolution of the 5D, that the 1Ds3 and 5D is pretty close. So not quite up the level of the D3.

A German Photo Magazin made a test with results that can be shown in the following charts

and on the pixel level

This basically shows that there is no free lunch. The D3 achieves the performance at the expence of details and resolution on the basis of this test. RAW files from the D3 look less detailed and sharp than pictures from the 5D is what I have seen. But this is an area where I have seen no camera reviews really look at (other than the German Foto Magazin). This is a bit surprising as one would think that a camera reviewer would be critically looking at this area.

 Hans Kruse's gear list:Hans Kruse's gear list
Canon EOS 5DS
OP SDRebel Senior Member • Posts: 1,489
Thanks - plus I enjoy experimentation

I hadn't in the past experimented with high ISO. The D3 is what has gotten me thinking what high ISO might do for me.

So, thanks for the suggestions. In the one experiment so far I've gotten a taste for what might be possible.

OP SDRebel Senior Member • Posts: 1,489
Auto ISO

Could you take the time to elaborate on your findings regarding how auto ISO can be implemented.

I had it enabled briefly but didn't see it working and so moved on to other aspects of the steep learning curve for the 1Ds3.

Thanks.

Hans Kruse Veteran Member • Posts: 3,432
Re: Auto ISO

SDRebel wrote:

Could you take the time to elaborate on your findings regarding how
auto ISO can be implemented.

Yes, I did it here http://forums.dpreview.com/forums/read.asp?forum=1032&message=26819975

I had it enabled briefly but didn't see it working and so moved on to
other aspects of the steep learning curve for the 1Ds3.

It works great and by having the needed parts in MyMenu it is very quick to enable and disable it.

 Hans Kruse's gear list:Hans Kruse's gear list
Canon EOS 5DS
Keyboard shortcuts:
FForum MMy threads