Which camera has the most dynamic range in your opinion?

PM01 wrote in post #9024040
Why the D3x?

For any given CoC (i.e. peak spatial frequency of the viewer), it has the lowest scaled read noise and highest full well capacity.

Collin85 wrote in post #9014355
Uh oh, here comes the graphs.

The pixel peepers are here! (Fires up "Ride of the Valkyries".)

http://en.wikipedia.or​g/wiki/Dynamic_range

Daniel Browning wrote in post #9024503
For any given CoC (i.e. peak spatial frequency of the viewer), it has the lowest scaled read noise and highest full well capacity.

Are there any websites that will have data on the D3x? Full well pixel depth, noise, QE, etc.?

PM01 wrote in post #9024761
Are there any websites that will have data on the D3x? Full well pixel depth, noise, QE, etc.?

Yes, but you have to know how to interpret them. For one, most sites don't take pattern noise into consideration at all, even though it may affect dynamic range by 1-3 stops. For that it's best to inspect the raw files visually.

Roger Clark's web site (clarkvision) offers some great info on a lot of cameras, but he doesn't have data on the D3X. (Just don't use his "sensor" dynamic range numbers, because they are based on using the read noise of ISO 1600 and the FWC of ISO 100, which is nonsensical.)

Gabor has some good measurements, too, but not the D3X.

DxOMark gives good general information, but not the underlying data. Fortunately, DxOmark uses a simple (and correct) model for their data, like the one described in the EMVA 1288 Standard, which assumes that the sensor is linear. Because of that, it's possible to work backwards to get something like the original data. Someone did just that for a few cameras, including the D3X:

Read noise and full well capacity derived from DxOMark

There you can see that the D3X has a full well capacity of 48k e- and read noise of 6.0 e-, a 13-stop dynamic range per pixel. In an 8x10 150ppi print, that becomes 15 stops, which is quite an achievement.

I'd agree with the opinions on DxO Mark info - very thorough.

Daniel Browning wrote in post #9024948
Gabor has some good measurements, too, but not the D3X.
There you can see that the D3X has a full well capacity of 48k e- and read noise of 6.0 e-, a 13-stop dynamic range per pixel. In an 8x10 150ppi print, that becomes 15 stops, which is quite an achievement.

So where does the 2 extra stops come from? 15 stops on print, but 13 stops on sensor?

PM01 wrote in post #9025506
So where does the 2 extra stops come from? 15 stops on print, but 13 stops on sensor?

Reproduction magnification (i.e. the viewer's peak spatial frequency or CoC). It's the same reason that small prints have deeper DOF, less aberrations, and less noise than large prints, viewed from the same close distance.

Noise always rises with spatial frequency (i.e. level of detail). If you take a 24 MP picture at ISO 100,000 and resize it to a teensy tiny 60x40 thumbnail web image, it will not look that noisy. That's because you're only looking at the very coarsest spatial frequencies (less than 1 lp/mm). When you look at 100% crop, you're examining very, very high spatial frequencies, which correspond to 6-foot-wide prints. That's a high spatial frequency, say 80+ lp/mm. Since noise rises with spatial frequency, the visible noise will be higher in a 30x20 print than a 3x2 wallet-size print. Since the noise is less in smaller images, the dynamic range is higher.

This applies to all sensors, not just the D3X. For every four-fold reduction in megapixels (e.g. 24 MP to 6 MP), there is a one-stop increase in dynamic range.

D3x - sensor ONLY. About 13 stops?

Wonder what the 1dm4 will churn out.

For DxO, do they measure the actual RAW file or are they geared more towards the jpeg (processed in camera or otherwise) file?

Also, I've noticed that pattern noise only exists on cameras with an ABG array. Scientific cameras used for astronomy don't need to worry about the ABG array since they don't need to bleed off excess signal to prevent blooming or electron spillover. Pretty much all DSLRs have the ABG array.

Example below.

http://images.google.c​om …e%3Doff%26sa%3D​N%26um%3D1

Now are there any references with true dynamic range utilizing the RAW file measurements?

PM01 wrote in post #9031457
For DxO, do they measure the actual RAW file or are they geared more towards the jpeg (processed in camera or otherwise) file?

Raw, of course. You insult my intelligence . If I linking to anything that was based on JPEG, that would be almost as misleading as linking to review by DPR!

Not that DxO is perfect. There are lots of flaws (some of which I already mentioned), but at least they are consistent and measure a wide variety of cameras. They also post a bit about their methodology on their web site.

PM01 wrote in post #9031457
Also, I've noticed that pattern noise only exists on cameras with an [RGB] array.

Not at all. In fact, the color filters have no effect on it at all, except they may tend to coincide with sensors that have 4+ channel readouts, which may correspond with 4-line periodicity in the pattern noise.

PM01 wrote in post #9031457
Scientific cameras used for astronomy don't need to worry about the ABG array since they don't need to bleed off excess signal to prevent blooming or electron spillover.

First of all, correlation is not causation. Just because some non-RGB cameras do not have pattern noise does not mean RGB is the cause. It's just coincidence. Scientific cameras are different from DSLRs in many, many ways, and you are picking the wrong characteristic as the reason why most of them do not have pattern noise. The primary cause in that case is the readout speed, but there are many others as well.

Second, many non-RGB cameras do indeed have pattern noise, particularly among the ones that are in the same price range as DSLRs or are intended for fast frame capture (e.g. to overcome seeing limitations through light frame stacking).

PM01 wrote in post #9031457
Pretty much all DSLRs have the ABG array.

Yes, that's correct.

PM01 wrote in post #9031457
Now are there any references with true dynamic range utilizing the RAW file measurements?

I already gave you three.

Daniel,

You misquoted me on the ABG array. Anti Blooming Gates, not RGB. [RGB] in your quote.

I'm not talking about the RGB filter, I'm well aware of their characteristics. What I'm talking about is their ABG, or anti blooming gate array. It all depends on the sensor, ABG or NABG (NON anti blooming gate) on the quick observation of pattern noise.

The ABG gate usually covers a fair amount of the pixel, thus it's reduced QE. (Quantum efficiency) It's main purpose is to bleed off any excess electrons. This is commonly used in DSLR type cameras, whether it be monochrome or color with a RGB array.

The NABG variety comes in during scientific measurements that are required to be linear throughout the collection of electrons, that is, there is no cutoff or ramp off point. ABG type sensors have the ramp off/bleed off point to prevent spillover of electrons into adjacent pixels.

If you take a KAF1602LE sensor (ABG) vs a KAF1603 (NABG) sensor, you will often get pattern noise with the ABG sensor. The 1603 however, exhibits random noise, which every imaging device has, but no pattern noise. Both sensors are manufactured by Kodak and have the same physical dimensions, with the exception that one is ABG, and the other is NABG. The NABG also has a deeper pixel well depth due to the space saved by eliminating the ABG array.

Getting back to DxO, do they have D3x tests utilizing mode 3 operation? Apparently there is processing of the RAW file within the D3/D3x even with the noise reduction set to off. Perhaps this would change the dynamic range number?

Links here..
http://www.astrosurf.c​om/buil/nikon_test/tes​t.htm

Even with the older D70..
http://astrosurf.com/b​uil/d70v10d/eval

PM01 wrote in post #9031792
Daniel,

You misquoted me on the ABG array. Anti Blooming Gates, not RGB. [RGB] in your quote.

I'm sorry. You are right: I misunderstood you.

PM01 wrote in post #9031792
It all depends on the sensor, ABG or NABG (NON anti blooming gate) on the quick observation of pattern noise.

That may be one cause of pattern noise, particularly in CCD sensors, but the best DSLR sensors have been using CMOS for the last three years or longer. And as I'm sure you know, CMOS does not require any extra gates to avoid blooming like CCD does, so there is no ABG.

Furthermore, the worst pattern noise occurs after (or during) the ISO gain amplifier. For an example, compare the pattern noise of the 7D at -10 EV in ISO 100 compared to -6 EV in ISO 1600. When raw with high analog gain has no pattern noise at all compared to very terrible pattern noise in the unamplified raw. That means something post-amplification, such as the ADC, is adding the pattern noise. Furthermore, the periodicity of the pattern noise is 8, corresponding to the 8 readout channels used by the ADC. So in that case, the pattern noise is not caused by the ABG.

PM01 wrote in post #9031792
The ABG gate usually covers a fair amount of the pixel, thus it's reduced QE. (Quantum efficiency) It's main purpose is to bleed off any excess electrons. This is commonly used in DSLR type cameras, whether it be monochrome or color with a RGB array.

I'm sure it's common in DSLRs built with CCD, but the best DSLR sensors (e.g. D3X) are CMOS.

PM01 wrote in post #9031792
Getting back to DxO, do they have D3x tests utilizing mode 3 operation?

If I had a nickle for every time I had to debunk this particular myth...

No, they do not use the so-called "mode 3" (ugh, terrible name), nor do they need to.

PM01 wrote in post #9031792
Apparently there is processing of the RAW file within the D3/D3x even with the noise reduction set to off.

Yes, but it only occurs in long exposures! (Between 2 and 10 seconds.) And even then it only occurs when dark frame subtraction is disabled.

It affects only *one* type of photographer: astrophotographers who use their own dark frame library. Astrophotographers that don't build their own dark frame library are not affected because they can use the in-camera dark frame subtraction.

Yes, that includes me, and yes, it's annoying. In fact, it's one of the reasons why I have a Canon 5D2 instead of a Nikon D700. But I understand why Nikon doesn't bother to change it, because 99.999% of photographers don't care about it, and neither does it have any effect on the measurements that DxO does.

That's the kind of stuff we astrophotographers have to put up with.

PM01 wrote in post #9031792
Perhaps this would change the dynamic range number?

No. You would have to ask "Which camera has the most dynamic range for astrophotography when using 10-second exposures with my own custom dark frame subtraction?"

Most people will not even know what that means, let alone care that Nikon applies noise reduction that affects that case.

Daniel Browning wrote in post #9032315
Furthermore, the worst pattern noise occurs after (or during) the ISO gain amplifier. For an example, compare the pattern noise of the 7D at -10 EV in ISO 100 compared to -6 EV in ISO 1600. When raw with high analog gain has no pattern noise at all compared to very terrible pattern noise in the unamplified raw. That means something post-amplification, such as the ADC, is adding the pattern noise. Furthermore, the periodicity of the pattern noise is 8, corresponding to the 8 readout channels used by the ADC. So in that case, the pattern noise is not caused by the ABG.

***Good point. With the astro CCD the noise itself is more than likely caused by the ABG, a completely different pattern than the CMOS.***

No, they do not use the so-called "mode 3" (ugh, terrible name), nor do they need to.

***In terrestrial applications, I agree. Though the article shows that the dimmer stars are obliterated, even without the dark frame sub.***

Yes, but it only occurs in long exposures! (Between 2 and 10 seconds.) And even then it only occurs when dark frame subtraction is disabled.

***Article claims that anything more than 1 second will cause a non "voluntary" noise reduction algorithm to be implemented. With regular bright day photography this may be a no issue.***

It affects only *one* type of photographer: astrophotographers who use their own dark frame library. Astrophotographers that don't build their own dark frame library are not affected because they can use the in-camera dark frame subtraction.

***Dark Frame Libraries...they don't work with me very well either! Yes, it's supposed to able to scale according to temperature, but even with MaximDL, I could never get it to play nicely. I usually do the dark immediately after grabbing the light frame.***

Yes, that includes me, and yes, it's annoying. In fact, it's one of the reasons why I have a Canon 5D2 instead of a Nikon D700. But I understand why Nikon doesn't bother to change it, because 99.999% of photographers don't care about it, and neither does it have any effect on the measurements that DxO does.

***Agreed. I'm using 1dsm3, 1dm3, 5dm2, 20d, 40d for astro work also. Much easier. Now if they got rid of the AA filter or made it removable the resolution would increase. The AA filters are kinda strong!***

That's the kind of stuff we astrophotographers have to put up with.

***I know! Been doing the astro for about 2 decades now. Very challenging and you have to get everything right, and have some extremely good optics.***

No. You would have to ask "Which camera has the most dynamic range for astrophotography when using 10-second exposures with my own custom dark frame subtraction?"

***Let's add to this..."good for between 20 degrees F to 30 degrees F with a humidity of 90%...etc etc"***

Most people will not even know what that means, let alone care that Nikon applies noise reduction that affects that case.

But then most aren't crazy enough to use the cameras for astrowork. Sitting around in -40F/-40C temps to gather the photons of faint nebula isn't exactly most peoples idea of fun.

Canon people (like myself...though I shoot with any system) are generally crazy. And quite proud of it!

PM01 wrote in post #9032643
But then most aren't crazy enough to use the cameras for astrowork. Sitting around in -40F/-40C temps to gather the photons of faint nebula isn't exactly most peoples idea of fun.

Not to mention driving 4 hours into the remotest deserts every new moon . Our central Oregon skies are pretty awesome.