How does read noise actually affect image quality?

[Marsu42]

Since somehow every sensor thread seems to degrade quickly into gibberish and brand fanboyism, I'm a bit embarrassed to admit that I still don't get how the real world impact of a camera's read noise is.

My current understanding: traditional non-Exmor sensor designs have higher read noise at low iso which also affects max. dynamic range (but this thread isn't about dr). Also read noise doesn't equal banding artifacts, see the lower 7d figures - but this model has higher banding due to the dual readout channels.

sensorgen.info says: @iso 100 / 200 / 400 / 800:
7d: 8.4 / 4.7 / 3.3 / 2.8
60d: 13.2 / 8.4 / 4.4 / 3.2
5d3: 33.1 / 18.2 / 10.6 / 6.1
6d: 26.8 / 14.6 / 7.9 / 5.1

Question: Except for long time astronomy exposures, when (if at all) does this mean that shooting at higher iso might/will give better iq than at the lowest iso setting? Thanks for explaining!

 

[privatebydesign]

This is a good link to start at.

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

As you can see, the effects of one type of noise can easily be overwhelmed by a different type of noise! It is very counter productive to concentrate too hard on any one aspect of a single components performance when the output of the complete device is the sum of how lots of components work together.

You could have a sensor with considerably better read noise, but have other issues that impacts image quality even more.

 

[Marsu42]

You could have a sensor with considerably better read noise, but have other issues that impacts image quality even more.

Thanks for the link explaining the concepts - but as far as I see it doesn't sum it up: What's the combined snr for say the 6d that I just boldly ordered for different iso levels, i.e. where is the "sweet spot" and break-even point of different noise curves?

For my 60d/crop I know by no matter what any numbers say iso100 is the setting with the highest iq if shutter speed doesn't matter like on tripod. But is this the same for ff, or is iso200 or even iso400 "better" and iso100 only a good idea if you have reached max. shutter speed (x-sync) or are doing long time exposures?

 

[ajfotofilmagem]

You could have a sensor with considerably better read noise, but have other issues that impacts image quality even more.

Thanks for the link explaining the concepts - but as far as I see it doesn't sum it up: What's the combined snr for say the 6d that I just boldly ordered for different iso levels, i.e. where is the "sweet spot" and break-even point of different noise curves?

For my 60d/crop I know by no matter what any numbers say iso100 is the setting with the highest iq if shutter speed doesn't matter like on tripod. But is this the same for ff, or is iso200 or even iso400 "better" and iso100 only a good idea if you have reached max. shutter speed (x-sync) or are doing long time exposures?

I also could not come to any conclusion, reading the various graphics and text link clarkvision.com. In fact, all cameras I have ever used, always had better image quality with the lowest ISO available, with the exception of extended low ISO. The unexpected differences become visible in fractions of a point. Canon 60D, for example ISO1000 has more noise than in ISO1600. There is no guarantee that Canon follow the same logic in all models, but if kept this behavior, we should avoid the ISO that are digitally pushed as ISO1000 in 60D.

 

[ajfotofilmagem]

The readout noise, specifically seems to be a problem only in low ISO. I believe that should be avoided Canon 60D at ISO125 by the logic of pushing digitally. But honestly my eye could not see the difference between ISO100 and ISO125. Maybe if I make a picture purposely dark, and and raise the shadows four points of light in Photoshop. : I leave it to the users of Nikon.

 

[Marsu42]

Canon 60D, for example ISO1000 has more noise than in ISO1600. There is no guarantee that Canon follow the same logic in all models

All Canon models except the 1dx which received additional tweaks are the same: for optimal iq use only full (analog) iso stops, everything else is digitally modified like iso1000 is iso1600 underexposed. Intermediary iso steps are only for jpeg, video and auto iso with fixed shutter/aperture.

I also could not come to any conclusion, reading the various graphics and text link clarkvision.com. In fact, all cameras I have ever used, always had better image quality with the lowest ISO available

Hmmm, I hope someone did extended test of this, I really wonder looking at the ff figures if iso200 isn't a viable alternative...

 

[privatebydesign]

http://www.sensorgen.info/CanonEOS_6D.html

These readings indicate that there is no disadvantage to any performance metric when used at indicated 100iso (that tested as actual 80).

 

[ajfotofilmagem]

What the eye does not see, the heart does not feel. : The tests, charts, theories, and marketing are leaving photographers with inferiority complex. Be strong, brother. Have faith in your art and move on. If you can not see the weak point in your camera, consider the strengths. ;D Amen!

 

[Marsu42]

What the eye does not see, the heart does not feel.

That's why this thread is about the "actual", not theoretical iq - and even if not imho it doesn't hurt to know the insides out, even if it doesn't always translate into practice.

These readings indicate that there is no disadvantage to any performance metric when used at indicated 100iso

My question is the other way around: What (if any) is the disadvantage when used at iso200 or iso400, as in real life shooting ?

 

[ajfotofilmagem]

The other day I saw hundreds of photos that a photographer has done with his 5D Mark I ???. I do not know what she did wrong, besides the fact that she wore a Nikon TTL flash. Sometimes the flash failed, and it does not always repeat the photo. :-X When I saw the pictures I was disappointed because of the noise in the shadows that was very, very ugly. Then I remembered every time someone asked for advice on Canonrumors, and told him that the best cheap camera that has ever existed is Canon 5D Mark I. : At that point I was so happy with my "terrible" APS-C 18 megapixel Canon. ;D

 

[privatebydesign]

My question is the other way around: What (if any) is the disadvantage when used at iso200 or iso400, as in real life shooting ?

In real life shooting at 200-400-800 you will have progressively more issues pulling detail from shadows and underexposed areas. Now how you deal with that is up to you, and there is a train of thought that suggests a +1 exposure at 200 and then lowering the exposure but holding the shadows in post will reveal more shadow detail than a correctly exposed image at 100 and trying to lift the shadows one stop. This, essentially, is an ETTR question using iso instead of aperture or shutter speed to gain your stop of over exposure.

The results will be sensor specific so testing in your personal common shooting scenarios when the 6D is in your hands is very much the way to go.

As for your heart not feeling what your eye can't see, tell that to somebody with angina!

 

[ajfotofilmagem]

My question is the other way around: What (if any) is the disadvantage when used at iso200 or iso400, as in real life shooting ?

In real life shooting at 200-400-800 you will have progressively more issues pulling detail from shadows and underexposed areas. Now how you deal with that is up to you, and there is a train of thought that suggests a +1 exposure at 200 and then lowering the exposure but holding the shadows in post will reveal more shadow detail than a correctly exposed image at 100 and trying to lift the shadows one stop. This, essentially, is an ETTR question using iso instead of aperture or shutter speed to gain your stop of over exposure.

The results will be sensor specific so testing in your personal common shooting scenarios when the 6D is in your hands is very much the way to go.

As for your heart not feeling what your eye can't see, tell that to somebody with angina!

I know some photographers who experienced angina chest after he read the DXO tests.

 

[luciolepri]

All Canon models except the 1dx which received additional tweaks are the same: for optimal iq use only full (analog) iso stops, everything else is digitally modified like iso1000 is iso1600 underexposed. Intermediary iso steps are only for jpeg, video and auto iso with fixed shutter/aperture.

Actually, ISO 1000 is 800 pushed, so is much better to use 1250, wich is 1600 pulled.
Anyway, in my experience too, Lower ISO = better IQ, I just avoid pushed ISO's (125-250-500-1000...).
In theory, even ISO 640 has less noise than ISO 100, being 800 ISO pulled, but as far as I'm concerned, images at 100 ISO are much better. Colors and nuances are cleaner and DR is wider.
If you use Magic Lantern, I suggest you to use ML ISO's, highlights rolloff is smoother and even shadow noise is reduced. The advantage is not always noticeable, but sometimes it is. What's more, you can use ISO 80, which can be useful in strong light situations.

 

[Marsu42]

Actually, ISO 1000 is 800 pushed, so is much better to use 1250, wich is 1600 pulled.

Wupps, of course you're correct, I just mixed that up.

In theory, even ISO 640 has less noise than ISO 100, being 800 ISO pulled, but as far as I'm concerned, images at 100 ISO are much better. Colors and nuances are cleaner and DR is wider.

I know about the ml theory and often post the link http://magiclantern.wikia.com/wiki/ISO - but I have to admit I switched to using only full iso stops, and since ff has better iq I was wondering about the snr @full low iso stops. But as the replies so far indicate, also for ff lower iso is "better".

 

[luciolepri]

as the replies so far indicate, also for ff lower iso is "better".

I'd definitely say so, expecially if you're talking about RAW.

Just for nerdiness purposes:
http://home.comcast.net/~NikonD70/Charts/PDR.htm
http://home.comcast.net/~NikonD70/Charts/RN_ADU.htm

 

[Sporgon]

I'd have preferred the question to have missed off the 'How'.

Fact is read noise has no effect on the image most of the time; if it did everyone who demands top drawer IQ would use an Xmor type sensor - and they don't - by far.

It's a case of 'what lies underneath'. Some like to go fishing for it. It will surface with others when they try to over lift shadows, lift under exposure etc, using poor or flawed technique.

There is some evidence to suggest that shooting at an ISO which offers best balance between read noise and saturation has some benefit if you are needing to lift shadows, so there is some relevance to knowing your cameras read noise @ each one thirds ISO.

PBD- thanks for the link.

 

[Pi]

Since somehow every sensor thread seems to degrade quickly into gibberish and brand fanboyism, I'm a bit embarrassed to admit that I still don't get how the real world impact of a camera's read noise is.

My current understanding: traditional non-Exmor sensor designs have higher read noise at low iso which also affects max. dynamic range (but this thread isn't about dr). Also read noise doesn't equal banding artifacts, see the lower 7d figures - but this model has higher banding due to the dual readout channels.

sensorgen.info says: @iso 100 / 200 / 400 / 800:
7d: 8.4 / 4.7 / 3.3 / 2.8
60d: 13.2 / 8.4 / 4.4 / 3.2
5d3: 33.1 / 18.2 / 10.6 / 6.1
6d: 26.8 / 14.6 / 7.9 / 5.1

Question: Except for long time astronomy exposures, when (if at all) does this mean that shooting at higher iso might/will give better iq than at the lowest iso setting? Thanks for explaining!

No. You have lower read noise at higher ISO in absolute units, electrons, not in relative ones like SNR. If you shoot at ISO 200 with the 7D, for example, the saturation level halves, but the read noise almost does but not quite. As a result, you get a slightly higher read noise relative to the signal, and still more shot noise. If the numbers were 16/8/4/..., then the read noise would be the same relative to the signal. With Canon, they are "close" to that, on a log scale, at least, so the read noise stays more or less flat at lower ISOs. With Nikon, it goes up. This does not make the Nikon worse, of course, they just start at a much lower level.

BTW, banding is due not only to multiple readout channels (and I have seen really bad one with defective units) but also to random fluctuations with no apparent pattern (yes, I am talking about pattern noise )

 

[ajfotofilmagem]

Since somehow every sensor thread seems to degrade quickly into gibberish and brand fanboyism, I'm a bit embarrassed to admit that I still don't get how the real world impact of a camera's read noise is.

My current understanding: traditional non-Exmor sensor designs have higher read noise at low iso which also affects max. dynamic range (but this thread isn't about dr). Also read noise doesn't equal banding artifacts, see the lower 7d figures - but this model has higher banding due to the dual readout channels.

sensorgen.info says: @iso 100 / 200 / 400 / 800:
7d: 8.4 / 4.7 / 3.3 / 2.8
60d: 13.2 / 8.4 / 4.4 / 3.2
5d3: 33.1 / 18.2 / 10.6 / 6.1
6d: 26.8 / 14.6 / 7.9 / 5.1

Question: Except for long time astronomy exposures, when (if at all) does this mean that shooting at higher iso might/will give better iq than at the lowest iso setting? Thanks for explaining!

No. You have lower read noise at higher ISO in absolute units, electrons, not in relative ones like SNR. If you shoot at ISO 200 with the 7D, for example, the saturation level halves, but the read noise almost does but not quite. As a result, you get a slightly higher read noise relative to the signal, and still more shot noise. If the numbers were 16/8/4/..., then the read noise would be the same relative to the signal. With Canon, there are "close" to that, on a log scale, at least, so the read noise stays more or less flat at lower ISOs. With Nikon, it goes up. This does not make the Nikon worse, of course, they just start at a much lower level.

BTW, banding is due not only to multiple readout channels (and I have seen really bad one with defective units) but also to random fluctuations with no apparent pattern (yes, I am talking about pattern noise )

I remember being frightened by the noise of Nikon D70, and at that time I decided for Canon Rebel XT. Since then, Nikon has evolved more in low ISO, and Canon has evolved at high ISO. I consider that improvements in high ISO really allows me to do some photos today, would not be possible eight years ago. The evolution in low ISO also happened with Canon, though it is not the leader in this area. Frankly, all current cameras seem capable of making great images at ISO 100 or 200. Glad to be able to use ISO 3200 today, if I need to. However, I would like to have available also ISO 50, and even ISO 25 for long exposures.