How to find your camera's unity gain setting. [guide]

[pipnina]

I imagine one or two people here are tarting to realise I think about things a lot. This is another one of those things.

I wanted to know what the unity gain for my camera was (Nikon D3200) so I wouldn't end up wasting precious photons out in the field. It's a useful setting to know since going higher is useless and just narrows your dynamic range and going lower is just plain wasteful!

So, what did I do to find mine? I set my camera to a very low level of exposure- so low that the highst ISO setting scarcely produced an image at all. (for me, that was f22, 1/4000 shutter) Following this I took the same shot at each ISO setting until I reached the lowest my camera wold allow- 100 ISO

Following this, I converted these RAW files into 16-bit TIFFs (my image editor doesn't open Nikon RAWs) and opened each of them, in order or ISO, in Paint.NET. In Paint.NET I strectched the histograms (dragging the white level down in the levels tool, under adjustments) so that all of the images were roughly the same brightness. Now I had all of these different ISO images of equivalent exposure open on my computer, it was possible to analyse the results

First, here are the images (sadly only in .jpg format, the TIFFs were rather large):

http://i.imgur.com/GpNFpUg.jpg(12800 ISO (High-1)

http://i.imgur.com/cnAeMaY.jpg(6400 ISO)

http://i.imgur.com/rHT2byK.jpg(3200 ISO)

http://i.imgur.com/6zK2EKG.jpg(1600 ISO)

http://i.imgur.com/siiThcV.jpg(800 ISO)

http://i.imgur.com/vttRuyN.jpg(400 ISO)

http://i.imgur.com/TYiwjEI.jpg(200 ISO)

http://i.imgur.com/HNIOMND.jpg(100 ISO)

Phew. Now, to begin with, I'm gonna say that 12800 ISO and 6400 ISO seems to have 0 difference between them. But out of all of these images, 3200 ISO seems to produce the "Best" image out of the lot. Lower than 3200 ISO and the image starts to fade into the read noise, above and it looks like demons have begun posessing the camera to the extent that the image starts to suffer and data appears to be lost as a result.

If you haven't already, I encourage you to find your own unity gain ISO using this method as it could make your images just that little bit better. I have to admit, 3200 ISO is not only convenient (model of my camera) but also higher than I had expected.

Hope this helps someone! It was definately interesting for me.

(p.s. if you have any suggestions regarding making the process more accurate or if you realise my method is completely wrong, do tell!)

        ~pip

[kalasinman]

Sorry, but I can't agree with your results. I believe unity gain is not the brightest or best image, but the point where one additional photon produces one count. In your example, that appears to be ISO 800. This is what would be expected for your camera. Past that the image appears "brighter", as amplification is applied, doubling the count again and again.

[kalasinman]

Although the Nikon D5100 and D7000 are said to be "Isoless", with most DSLRs the read noise decreases with ISO.

I'm just thinking that to get a good test, you would have to take a standard light source and take a series of increasing exposures at each ISO, and then plot them to determine when the increase in SNR became linear. That would be your unity gain point, it seems to me.

[pipnina]

Although the Nikon D5100 and D7000 are said to be "Isoless", with most DSLRs the read noise decreases with ISO.

I'm just thinking that to get a good test, you would have to take a standard light source and take a series of increasing exposures at each ISO, and then plot them to determine when the increase in SNR became linear. That would be your unity gain point, it seems to me.

I have 71 images of this at every ISO setting on my camera, going from f22, 1/4000th upping the exposure by 4 shutter settings each image until the histogram hit 1/2 a stop before 2 stops over.

Not sure what you're asking me to do right now, though. How am I going to graph the SNR?

For a bit now, though, I'm going to take some bias frames for eah ISO setting.

[sharkmelley]

Yes ISO 800 gives more or less unity gain on the Nikon D3200.

There is an easy way to determine unity ISO for any camera, assuming it appears in the list at http://www.sensorgen.info/ and assuming the accuracy of those figures.

For a 12 bit camera the unity gain occurs at the ISO whose saturation is nearest 4096

For a 14 bit camera the unity gain occurs at the ISO whose saturation is nearest 16384

For a 16 bit camera (do they exist?)  the unity gain occurs at the ISO whose saturation is nearest 65536

So, for the Nikon D3200, being a 12 bit camera, ISO 800 gives a saturation at 3717 electrons (http://www.sensorgen.info/NikonD3200.html).  So this 3717 electrons converts into a pixel value of approx. 4096 i.e. ISO 800 is giving, very roughly, 1 digital unit for each electron - which is the definition of unity gain.

Mark

[pipnina]

Yes ISO 800 gives more or less unity gain on the Nikon D3200.

There is an easy way to determine unity ISO for any camera, assuming it appears in the list at http://www.sensorgen.info/ and assuming the accuracy of those figures.

For a 12 bit camera the unity gain occurs at the ISO whose saturation is nearest 4096

For a 14 bit camera the unity gain occurs at the ISO whose saturation is nearest 16384

For a 16 bit camera (do they exist?)  the unity gain occurs at the ISO whose saturation is nearest 65536

So, for the Nikon D3200, being a 12 bit camera, ISO 800 gives a saturation at 3717 electrons (http://www.sensorgen.info/NikonD3200.html).  So this 3717 electrons converts into a pixel value of approx. 4096 i.e. ISO 800 is giving, very roughly, 1 digital unit for each electron - which is the definition of unity gain.

Mark

Hmm, there must have been something wrong with my method. I was well out. 800 ISO it is then!

[kalasinman]

Yes ISO 800 gives more or less unity gain on the Nikon D3200.

There is an easy way to determine unity ISO for any camera, assuming it appears in the list at http://www.sensorgen.info/ and assuming the accuracy of those figures.

For a 12 bit camera the unity gain occurs at the ISO whose saturation is nearest 4096

For a 14 bit camera the unity gain occurs at the ISO whose saturation is nearest 16384

For a 16 bit camera (do they exist?)  the unity gain occurs at the ISO whose saturation is nearest 65536

So, for the Nikon D3200, being a 12 bit camera, ISO 800 gives a saturation at 3717 electrons (http://www.sensorgen.info/NikonD3200.html).  So this 3717 electrons converts into a pixel value of approx. 4096 i.e. ISO 800 is giving, very roughly, 1 digital unit for each electron - which is the definition of unity gain.

Mark

Thanks Mark. Very interesting. Looking at the sensorgen data for the D5100, the highest DR is at ISO 100. With a camera that's had the hack, doesn't that imply that the best images, in the case of dim objects are going to be long exposures near ISO 100? I've read elsewhere this is the case. Hard to accept, but if that strategy, or any, helps DR and keeps the thermal noise as low as possible, I'll do it.

[sharkmelley]

Thanks Mark. Very interesting. Looking at the sensorgen data for the D5100, the highest DR is at ISO 100. With a camera that's had the hack, doesn't that imply that the best images, in the case of dim objects are going to be long exposures near ISO 100? I've read elsewhere this is the case. Hard to accept, but if that strategy, or any, helps DR and keeps the thermal noise as low as possible, I'll do it.

Unity gain for the D5100 would be around ISO 200.  Hence, using it at ISO 100 introduces some extra quantisation error.  But since this quantisation error is much smaller than the read noise (which is around 4e) then there is nothing to be lost by using it at ISO 100.  However there is plenty to be gained at ISO 100 in terms of dynamic range.  So my choice would definitely be ISO 100.  The joys of ISO-less cameras!

Mark

[kalasinman]

Unity gain for the D5100 would be around ISO 200.  Hence, using it at ISO 100 introduces some extra quantisation error.  But since this quantisation error is much smaller than the read noise (which is around 4e) then there is nothing to be lost by using it at ISO 100.  However there is plenty to be gained at ISO 100 in terms of dynamic range.  So my choice would definitely be ISO 100.  The joys of ISO-less cameras!

Mark

As I recall, the hacked D5100 goes to 2 -e

[chrisvdberge]

Not sure how valuable the whole concept of Unity gain still is. 

This is what Clark himself says about it:

NOTE: Unity Gain is a flawed concept in my opinion. It is included here for historical reference. Contrary to some posts on the net, I did not start this concept. It seems like a good idea: that the fundamental counting unit is one quanta: an electron. It seemed that one should not need to digitize a signal finer than 1 electron. In the early days of digital cameras (before about circa 2008), the electronics in digital cameras were too noisy to counter the theory. But since then digital cameras have substantially lower noise. Read noise at high ISOs are commonly less than 2 electrons. But this is only obtained when ISOs are much higher than Unity Gain. Clearly there are advantages to ISOs beyond Unity Gain. The fundamental reason Unity Gain is not relevant is because the sensor in a digital camera is an analog system, not digital. The signals from the sensor are analog and only after amplification is the signal digitized.

http://www.clarkvision.com/articles/digital.sensor.performance.summary/#unity_gain

[kalasinman]

Not sure how valuable the whole concept of Unity gain still is. 

This is what Clark himself says about it:

NOTE: Unity Gain is a flawed concept in my opinion. It is included here for historical reference. Contrary to some posts on the net, I did not start this concept. It seems like a good idea: that the fundamental counting unit is one quanta: an electron. It seemed that one should not need to digitize a signal finer than 1 electron. In the early days of digital cameras (before about circa 2008), the electronics in digital cameras were too noisy to counter the theory. But since then digital cameras have substantially lower noise. Read noise at high ISOs are commonly less than 2 electrons. But this is only obtained when ISOs are much higher than Unity Gain. Clearly there are advantages to ISOs beyond Unity Gain. The fundamental reason Unity Gain is not relevant is because the sensor in a digital camera is an analog system, not digital. The signals from the sensor are analog and only after amplification is the signal digitized.

http://www.clarkvision.com/articles/digital.sensor.performance.summary/#unity_gain

As mentioned prior, Nikons with EXMOR sensors have the same read noise at all ISO, "isoless" and with hack 2 -e, much lower than a CCD . Test data indicates no quantization error with the hacked firmware even at ISO100. My understanding of dynamic range indicates that since DR is full well capacity divided by read noise, as the well depth is constant, then operating nearer the tead noise at ISO 100 or 200 would make the DR more useful, below the thermal noise. Maybe I'm wrong, but these are pretty good sensors.

[chrisvdberge]

As mentioned prior, Nikons with EXMOR sensors have the same read noise at all ISO, "isoless" and with hack 2 -e, much lower than a CCD . Test data indicates no quantization error with the hacked firmware even at ISO100. My understanding of dynamic range indicates that since DR is full well capacity divided by read noise, as the well depth is constant, then operating nearer the tead noise at ISO 100 or 200 would make the DR more useful, below the thermal noise. Maybe I'm wrong, but these are pretty good sensors.

Thx. Seems I really need to test out my D600 at ISO100 vs ISO800