Special ISO?

[PhotoGramly]

A while back one of my friends who is a photographer asserted that using ISO intervals that are slightly higher than the "standard" ISO numbers most people would use (100, 200, 400, 800, 1600, 3200 etc.) would actually result in less noise on the higher end of the scale. So for example, using ISO 2500 instead of ISO 1600 (and adjusting other settings to maintain the same overall exposure) would lead to less noise.

It doesn't sound right to me, but it's proving difficult to look into cause how do you search for *that* with a succinct phrase? So I thought I would check here and see if anyone knows anything about it.

 

[weepete]

I think they may be getting a bit confused. The ISOs you mention are in full stops (100, 200, 400, 800, 1600, 3200, 6400 etc). Native ISO is a thing that's talked about a bit, is dependant on the camera but I think that's only to do with one setting. There is a technique of "Expose To The Right" or ETTR where you can minimise noise by overexposing in camera (as long as your dynamic range is compressed) and bringing the exposure down in post. ISO invariance is the newer thing, basically meaning theres no noise penalty from raising shadows in post compared to shooting at a higher ISO but not all sensors are ISO invariant and if yours isn't there does tend to be a greater noise penalty from underexposing and raising shadows in post than shooting at a higher ISO in the first place.

 

[wfooshee]

I don't think it leads to less noise SOOC, but weepete's explanation about post-processing sounds spot-on to me; you could end up with less noise after your processing work if you're bringing things down rather that bringing them up.

My D7200 isn't as bad with this, but my D7000 was horrible for trying to raise shadows to bring out detail.

 

[Ysarex]

A while back one of my friends who is a photographer asserted that using ISO intervals that are slightly higher than the "standard" ISO numbers most people would use (100, 200, 400, 800, 1600, 3200 etc.) would actually result in less noise on the higher end of the scale. So for example, using ISO 2500 instead of ISO 1600 (and adjusting other settings to maintain the same overall exposure) would lead to less noise.

It doesn't sound right to me, but it's proving difficult to look into cause how do you search for *that* with a succinct phrase? So I thought I would check here and see if anyone knows anything about it.

I suspect your friend is a little confused. But that's par for the course if the topic is ISO. So the simple answer is no. My best suspicion is that your friend may have encountered or was told this because of the way Canon used to engineer their cameras. Canon's signal processing (we're talking more than a decade ago) produced an uneven read noise output on the low end of the ISO scale that some folks noticed. It was always a minor difference but once turned loose on the internet it got a life of it's own and mutated into the rule (for all cameras) that the noise output was better if you used an intermediate 1/3 stop ISO value. (It was actually both better and worse). Look at this chart that plots read noise for an old Canon 5dmkII and a newer 5DS: Read Noise in DNs versus ISO Setting Note the jagged plot for the lower ISO values on the old mkII. Note also that the plot straightens out once you head into the high ISO values which is what you're talking about. That jagged low end of the plot isn't apparent in the newer camera.

The most important point about this is: It has nothing to do with what ISO is and has only to do with the way one camera manufacturer used to engineer the sensor signal processing in some of their cameras which they don't do any more.

ISO and noise: ISO is either noise neutral or it suppresses noise. Another Internet myth is to assign a cause and effect relationship to ISO relative to noise in a digital photo. There are numerous causes of noise in a digital photo, but for the most part we are concerned with two: 1. shot noise and 2. read noise. The noise plotted in that chart and that spawned the intermediate ISO rule is read noise. Read noise is the lesser problem. With modern cameras it's almost become a negligible problem. Read noise comes from the camera's processing electronics. When ISO is implemented in electronic hardware on the camera the effect is to suppress read noise.

When we photographers see noise in our photos and find it objectionable we are almost always looking at shot noise. Shot noise is in the signal itself and results from inherent photon randomness. Shot noise is always present but gets lost in a strong signal. In a weaker signal (that has to be boosted) we start to see it. Visible shot noise in our photos then is a function of exposure. Expose for a strong signal and no visible noise. Expose less and the signal weakens until we see the noise. ISO's relationship in here is relative not causal -- less exposure that produces a weakened signal is the cause. ISO is related in that we raise ISO as a way of instructing our camera's metering/exposure system to reduce exposure. ISO as implemented in the camera then boosts the weak signal so we see a "normal" lightness image and there we see the noise caused by the reduced exposure.

Read: What's that noise? Part one: Shedding some light on the sources of noise

Joe

 

[PhotoGramly]

Thanks for the responses. Makes me glad I never took that advice onboard wholesale.

I suspect your friend is a little confused. But that's par for the course if the topic is ISO. So the simple answer is no. My best suspicion is that your friend may have encountered or was told this because of the way Canon used to engineer their cameras. Canon's signal processing (we're talking more than a decade ago) produced an uneven read noise output on the low end of the ISO scale that some folks noticed. It was always a minor difference but once turned loose on the internet it got a life of it's own and mutated into the rule (for all cameras) that the noise output was better if you used an intermediate 1/3 stop ISO value. (It was actually both better and worse). Look at this chart that plots read noise for an old Canon 5dmkII and a newer 5DS: Read Noise in DNs versus ISO Setting Note the jagged plot for the lower ISO values on the old mkII. Note also that the plot straightens out once you head into the high ISO values which is what you're talking about. That jagged low end of the plot isn't apparent in the newer camera.

That would make sense, I know she used to use a Canon 5DmkII before she upgraded so maybe that's where it came from.

When we photographers see noise in our photos and find it objectionable we are almost always looking at shot noise. Shot noise is in the signal itself and results from inherent photon randomness. Shot noise is always present but gets lost in a strong signal. In a weaker signal (that has to be boosted) we start to see it. Visible shot noise in our photos then is a function of exposure. Expose for a strong signal and no visible noise. Expose less and the signal weakens until we see the noise. ISO's relationship in here is relative not causal -- less exposure that produces a weakened signal is the cause. ISO is related in that we raise ISO as a way of instructing our camera's metering/exposure system to reduce exposure. ISO as implemented in the camera then boosts the weak signal so we see a "normal" lightness image and there we see the noise caused by the reduced exposure.

I have to admit that's the first time I've heard of that concept, so I guess I'll be doing a lot of reading to try to really understand it myself.

 

[Ysarex]

Thanks for the responses. Makes me glad I never took that advice onboard wholesale.

I suspect your friend is a little confused. But that's par for the course if the topic is ISO. So the simple answer is no. My best suspicion is that your friend may have encountered or was told this because of the way Canon used to engineer their cameras. Canon's signal processing (we're talking more than a decade ago) produced an uneven read noise output on the low end of the ISO scale that some folks noticed. It was always a minor difference but once turned loose on the internet it got a life of it's own and mutated into the rule (for all cameras) that the noise output was better if you used an intermediate 1/3 stop ISO value. (It was actually both better and worse). Look at this chart that plots read noise for an old Canon 5dmkII and a newer 5DS: Read Noise in DNs versus ISO Setting Note the jagged plot for the lower ISO values on the old mkII. Note also that the plot straightens out once you head into the high ISO values which is what you're talking about. That jagged low end of the plot isn't apparent in the newer camera.

That would make sense, I know she used to use a Canon 5DmkII before she upgraded so maybe that's where it came from.

When we photographers see noise in our photos and find it objectionable we are almost always looking at shot noise. Shot noise is in the signal itself and results from inherent photon randomness. Shot noise is always present but gets lost in a strong signal. In a weaker signal (that has to be boosted) we start to see it. Visible shot noise in our photos then is a function of exposure. Expose for a strong signal and no visible noise. Expose less and the signal weakens until we see the noise. ISO's relationship in here is relative not causal -- less exposure that produces a weakened signal is the cause. ISO is related in that we raise ISO as a way of instructing our camera's metering/exposure system to reduce exposure. ISO as implemented in the camera then boosts the weak signal so we see a "normal" lightness image and there we see the noise caused by the reduced exposure.

I have to admit that's the first time I've heard of that concept, so I guess I'll be doing a lot of reading to try to really understand it myself.

You got to be careful what you read. There's more misinformation on the Internet than there is good info. That article I linked by Richard Butler is good. Look at this photo. I did this for my class to demonstrate that it's not ISO that causes noise it's exposure. This photo is taken with an APS-C (Fuji crop sensor) camera at ISO 12,800. You can download it and look at the exif data and verify the ISO setting. It doesn't exhibit the noise you'd expect from that high ISO value. That's because I exposed the sensor as much as possible given the camera settings. (I got an over-exposed looking camera JPEG but processed the raw file). Normally when we're forced to a very high ISO value it's because we can't expose more. We're trying to hand-hold the camera and not drop the shutter speed. The noise comes from the weak signal -- not something that ISO does. If your camera is in any kind of auto or semi-auto mode ISO will spin the meter and so the camera will reduce exposure accordingly. That reduced exposure will cause the noise and then ISO in the camera electronics will lighten up the photo which of course shows you the noise.

Joe

 

[photoflyer]

Am I interpreting this correctly: overexpose a stop or two at high ISO and then pull it back in post?

 

[Ysarex]

Am I interpreting this correctly: overexpose a stop or two at high ISO and then pull it back in post?

Not exactly. What I did was a parlor trick of sorts. In a real-world situation where you're forced to raise the ISO sky high you can't over-expose. That's why you're raising the ISO in the first place. If you can expose more then you might as well lower the ISO and expose more. The point is: the noise is a function of the exposure.

What you don't want to do is what I catch my students doing all the time. Their faulty thinking goes like this:

1. I want to avoid noise in my photo.
2. Raising ISO causes noise.
3. I'm hand-holding the camera and need at least 1/50th sec. shutter speed.
4. Cr*p! I need ISO 3200 to get 1/50th sec. shutter speed.
5. I'm going to set the ISO to 1600 and use 1/50th sec. to try and keep the noise down.

That doesn't work and may in fact make the noise worse since ISO suppresses noise. Better off to raise the ISO as high as needed -- that won't add more noise over using less ISO and under-exposing. The noise is from the exposure. So always expose as much as possible and raise the ISO as needed.

Joe

 

[photoflyer]

OK. Got it. Something to experiment with.

 

[adamhiram]

Something you might find useful is this site that shows the read noise at different ISO settings for many popular cameras. For mine (D500), the line is fairly linear, as expected, but it also shows that if I need another stop above ISO 100, I might as well skip to ISO 400 because of how it amplifies the signal and where it switches over with a Dual Gain architecture.

Photonstophotos.net Charts:
Read Noise in DNs versus ISO Setting

Tony Northrup explanation:

(From an interesting but a bit cringe-worthy series of videos where people argued over whether ISO was "real")

 

[Derrel]

A good Tony Northrup video. ( YES,I just typed that. Lol.)

 

[fmw]

Also you should know what the native ISO of your camera is. My cameras have a native ISO of 200. So strangely, the cameras produce slightly (very slightly) more noise at ISO 100. You can probably guess I never use ISO 100