Unity Gain, what is it good for

[Torbjörn Tapani]

I wanted to answer a simple question, what ISO should I use*.

I stumbled upon this old thread about computing unity gain: http://www.luminous-landscape.com/forum/index.php?topic=76446.0

Is unity gain useful as a technique to improve images in low light, or is it maybe just part of an explanation to why raising the ISO up to a point is a good thing?

* I have a very specific interrest in the best ISO for low light work. I have a fixed exposure limited by the widest aperture and a fixed shutter speed when shooting aurora timelapse.

There seems to be evidence that there is something to be gained by increasing ISO at least until digital scaling kicks in, something like ISO 800 on a D7000 and ISO 1600 on the D800. So is unity gain even relevant in the real world?

[Bart_van_der_Wolf]

There seems to be evidence that there is something to be gained by increasing ISO at least until digital scaling kicks in, something like ISO 800 on a D7000 and ISO 1600 on the D800. So is unity gain even relevant in the real world?

Hi,

Well, that's about the gist of it for most modern camera designs. After raising the ISO setting to 400-1600 (depending on camera model), there is usually no image quality to be gained from raising it any further instead of boosting it in post-processing with the 'exposure slider' in the Rawconverter. It even allows to get better specular highlight headroom/rendition by not raising the ISO beyond usefulness.

The only real drawback is that the JPEG thumbnail in the Raw file will remain underexposed. There may also be some Raw converter noise reduction that's triggered by the ISO setting found in the EXIF data. The artificial higher ISOs could also result in default over-reduction of noise, so this may be a good thing, I usually prefer separate noise reduction programs anyway.

Cheers,
Bart

[Jim Kasson]

I wanted to answer a simple question, what ISO should I use*.

I stumbled upon this old thread about computing unity gain: http://www.luminous-landscape.com/forum/index.php?topic=76446.0

Is unity gain useful as a technique to improve images in low light, or is it maybe just part of an explanation to why raising the ISO up to a point is a good thing?

* I have a very specific interrest in the best ISO for low light work. I have a fixed exposure limited by the widest aperture and a fixed shutter speed when shooting aurora timelapse.

There seems to be evidence that there is something to be gained by increasing ISO at least until digital scaling kicks in, something like ISO 800 on a D7000 and ISO 1600 on the D800. So is unity gain even relevant in the real world?

In a camera system where the read noise is much less than 1 LSB, the "Unity gain" concept makes sense. However, those systems are thin on the ground. If there are noise sources ahead of the ADC whose contribution adds up to half an LSB or more, they will supply dithering. This is illustrated here:

http://blog.kasson.com/?p=6513

So UGISO is not particularly useful in modern cameras, IMHO. I believe you want to look at two things to decide what's the maximum ISO you'll use, and when to just push in post.

The first, and to me the most useful, is total noise normalized for exposure at some dark tone. The sources of noise will include read and shot noise. Here's an example: http://blog.kasson.com/?p=6565

The second is read noise referred to the input to the pre-ADC amplification chain, and often measured in electrons (abbreviated e-). If measured every third stop of ISO adjustment, this will give you a good idea of the noise floor. Warning: some cameras clip the read noise, so you can't just use RawDigger to measure dark-field images. If that's the case, there are curve-fitting techniques that can get you read noise estimates, but they're a little tricky. If you're going to do this, I encourage you to not do what most everybody does and use one-stop deltas to ISO settings. You can miss things. Here's an example, for the D810:

http://blog.kasson.com/?p=6552

And another for the a7S: http://blog.kasson.com/?p=6491

Jim

[Torbjörn Tapani]

Hi,

Well, that's about the gist of it for most modern camera designs.

Thanks, I think I'll just forget about unity gain for single exposures.

I do usually underexpose and push in post, not by choice but out of necessity to keep shutter speed low (2-5 seconds, low hah) to capture a moving aurora. I can see a clear benefit going from ISO 100 to 800 on a D7000 but less or no difference going from 800 to 1600 or 3200.

[Torbjörn Tapani]

In a camera system where the read noise is much less than 1 LSB, the "Unity gain" concept makes sense. However, those systems are thin on the ground. If there are noise sources ahead of the ADC whose contribution adds up to half an LSB or more, they will supply dithering. This is illustrated here:

http://blog.kasson.com/?p=6513

So could you say that as long as we get smooth gradations in shadow areas (with the help of some dithering) we are better off with a little more ISO? Could a look at the dark field histogram be used to tell us how far we want to go?

I'm still struggling with the rest of your reply but I'll get there, maybe. But thank you for your effort

[Jim Kasson]

So could you say that as long as we get smooth gradations in shadow areas (with the help of some dithering) we are better off with a little more ISO?

More the opposite. Once you get smooth gradations in the shadows, stop turning up the ISO. That is, if you're using modern Sonys and Nikons. With Canons, there is often a benefit to increased ISO, although, as you point out, stop when the camera starts to apply digital gain; you can do that better in post.

I see above that you're photographing aurorae (sp? auroras? aurori?). That puts a different spin on it. You'll probably want to minimize read noise, 'cause, unless there's a lot of light pollution, that's what you're going to see in the background. Median filtering may be your friend here, especially since auroras aren't sharp-edged.

If you tell me what camera you're using, I may have tested it, and I can give more specific advice.

Jim,

[Torbjörn Tapani]

I have a D800E that I use most of the time and added a D7000 to shoot timelapse.

Here is the D7000 taking a picure of the D800E from an hour ago.

[Jim Kasson]

I have a D800E that I use most of the time and added a D7000 to shoot time lapse

Then I can help you. A friend of mine, who is also an engineer by training, took my data and used it as the basis for a series of tests he made aimed at optimizing the ISO setting for deep sky photography. His situation wasn't exactly the same as yours, but except for the level of detail required, I don't think the Milky Way is that different from an aurora.

After many nights of testing, Rex's conclusion was that iso-1250 on the D800E was the optimum setting for deep sky photography. Why not give that a try?

Jim

[allegretto]

If you tell me what camera you're using, I may have tested it, and I can give more specific advice.

Jim,

this is exactly what I'm looking for! I have a Canon 6D and a Leica M240. It seems the Leica likes about 800 and the Canon 1600-3200. but that's just rough by looking at pics. What can you tell me about your formal tests if you have tested both or either of these cameras.

[Torbjörn Tapani]

Thanks, I will try 1250 on the D800E, it's more or less what I've seen mentioned before.

From the limited testing specific to the D7000 I have done, pushing underexposed ISO 100 shots just won't work, but ISO 800 looks more or less the same as 1600 or 3200 pushed.

This foray into Unity Gain has been somewhat disappointing but I have learned from it at least.

[Jim Kasson]

this is exactly what I'm looking for! I have a Canon 6D and a Leica M240. It seems the Leica likes about 800 and the Canon 1600-3200. but that's just rough by looking at pics. What can you tell me about your formal tests if you have tested both or either of these cameras.

I have no experience with the Canon. Because of the green shadows problem on the M240, pushing hard in post at ISOs below, say, 1600, is problematical.

There is a fix for the green shadows. If you use that you can push ISOs 200-800 in post with impunity. Then there are reasons to stay under ISO 800.

Jim

[allegretto]

I have no experience with the Canon. Because of the green shadows problem on the M240, pushing hard in post at ISOs below, say, 1600, is problematical.

There is a fix for the green shadows. If you use that you can push ISOs 200-800 in post with impunity. Then there are reasons to stay under ISO 800.

Jim

Wonderful Info... Thanks

h