M9 - Best ISO For SNR

[willie_901]

Disclosure: I do not own a M9, so many of you will want to stop reading now.

I thought this information might be of interest to M9 owners who face situations where shadow detail is important in their work. It is important to acknowledge that while SNR is a significant factor in image quality and dynamic range, other factors are also relevant. For instance M9 does not bit shift to simulate increased brightness. This is a good thing. There are other factors as well.

A very recent thread on Luminous Landscape discusses ETTR and the effects of ISO on signal to noise. Post number 46 in this thread shows results for the M9 (along with ther cameras which are irrelevant to the goal of my post) based on analysis of the linear raw data.


http://www.luminous-landscape.com/forum/index.php?topic=78677.msg632965#msg632965

The data (last graph in post no. 46) indicate the M9 SNR drops significantly above ISO 640. All this means is if you need more brightness at ISO 640 it is best to just let the camera underexpose (according to the meter). During post processing use the exposure slider to achieve acceptabe brightness instead of increasing ISO above 640. The data also implies when shutter speed or aperture requires ISO above 640 to achieve adequate brightness, reording jpegs is a significant handicap.

 

[Exdsc]

With digital cameras, when you underexpose you increase ISO sensitivity.

In a digital camera the ISO range is an indication of how many stops you can underexpose before clipping the shadows, in other words, when in doubt underexpose and then remedy with RAW software.

ETTR is an impractical concept for anyone who works with natural light.

 

[MCTuomey]

Thank you Willie. I am an M9 user, often in dim venues. This is very interesting. Practically, not being tech knowledgeable, I'm finding that underexposed M9 files corrected in pp do give superior results, at times, compared to my prior practice of trying for "correct" exposure by cranking up iso and then relying on noise reduction in pp. My problem, of course, is shutter speed when the subjects are active in very poor light.

 

[willie_901]

Thank you Willie. I am an M9 user, often in dim venues. This is very interesting. Practically, not being tech knowledgeable, I'm finding that underexposed M9 files corrected in pp do give superior results, at times, compared to my prior practice of trying for "correct" exposure by cranking up iso and then relying on noise reduction in pp. My problem, of course, is shutter speed when the subjects are active in very poor light.

Mike,

I'm happy you found this information to be useful.

 

[flyalf]

Thank you Willie. I am an M9 user, often in dim venues. This is very interesting. Practically, not being tech knowledgeable, I'm finding that underexposed M9 files corrected in pp do give superior results, at times, compared to my prior practice of trying for "correct" exposure by cranking up iso and then relying on noise reduction in pp. My problem, of course, is shutter speed when the subjects are active in very poor light.

Agree, but I'm a bit puzzled by "My problem, of course, is shutter speed when the subjects are active in very poor light.". Imo the underexposure is solving this problem. Lets assume that you need 1/30 @F/1,4 and ISO 640. You need to boost the shutter speed. You can a) increase to ISO 1280 to have 1/60 @ f/1,4, or you can b) and c) underexpose -1 or -2 by 1/60 or 1/125 @F/1,4 and ISO 640. For all you want to apply (different) noise correction in pp, but b) will give you best SNR.

 

[Richard G]

High ISO's work OK when you don't also underexpose. A recent post on the M9 suggested using ISO 500 and underexposing, and then fixing in post. Maybe that was Mike. I was intrigued by this but haven't yet tested the strategy.

 

[icebear]

I am not a wizard in post processing noise reduction ... but logic tells me :
The signal is captured with the exposure and it is at it's best when you expose properly.
Where do you see noise ? In properly exposed areas of the shot or in underexposed shadow areas ?
To underexpose on purpose seems to be utter nonsense to me. I have several shots with the M9 from the NYC "Fashion's Night Out" last year, shot at ISO1250 and to me these are perfectly acceptable in terms of noise, as always: YMMV.

 

[Richard G]

Yes, Icebear's night ISO 1250 and 1600 stuff has been impressive. In daylight I was amazed by ISO 1600 correctly or slightly over-exposed. Fraser on here, I think it was, recommended deliberately overexposing slightly with the M9 at high ISOs. I have seen a wedding photographer describe the M9's ISO 2500 as usable, unlike most opinions. In good light, I suspect 2500 is useable.

This was taken in the dimmest light in a basement bar. I guessed a generous exposure, 1/4s at f2.8. It's M9 at ISO 2500:

 

[willie_901]

Icebear,

In brighter regions you will see shot noise. The shot noise level is camera/sensor independent as it is intrinsic to the measurement of light. Shot noise rarely affects prints, especially when they are viewed from reasonable distances.

In darker regions the noise will be dominated by read noise. This does vary significantly from camera to camera. The read noise level is due to the sensor and analog to digital converter electronics. When the sensor is undeerexposed for practical reasons (preserve highlights, DOF and motion) the analog signal level drops while the read noise remains constant. Increasing ISO amplifies both the signal and the noise. The lower S/N is most obvious in the shadows. In some cameras (mostly Canons) the shadow read noise can impove as ISO increases. In most cameras it does not.

In your photograph you made a wise choice. You let the shadows remain dark. In this case, especially for prints, the reduced S/N can not be seen. While the detail in the shadow regions is compromised, that detail was not important to the photograph.

The reason I started this thread was to share how analysis of the M9 raw data files indicates the best performance for shadow regions, when base ISO is impractical, is at ISO 640 and below.

At higher ISOs the read noise increases much more than one might assume. This does not mean the results are not "usable" as Richard G mentions. It only means at ISO 640 or less, the shadow regions have more information compared to higher ISOs. This is not an esoteric point for raw. With raw you will always do better by never exceeding ISO 640 and intentionally underexpose the sesnor in order to use the required shutter speed or DOF. Then increase the brightness during post processing.

 

[icebear]

Yes, Icebear's night ISO 1250 and 1600 stuff has been impressive. In daylight I was amazed by ISO 1600 correctly or slightly over-exposed. Fraser on here, I think it was, recommended deliberately overexposing slightly with the M9 at high ISOs. I have seen a wedding photographer describe the M9's ISO 2500 as usable, unlike most opinions. In good light, I suspect 2500 is useable.

This was taken in the dimmest light in a basement bar. I guessed a generous exposure, 1/4s at f2.8. It's M9 at ISO 2500:

A generous helping of light i.e. signal will always yield a better result - was true with HP5 or TMY when I still shot film and is still true for digital, as long as you watch the highlights with the MM 😉. And btw for 1/4 s in a bar, I guess you didn't have too many beers😀.

 

[MCTuomey]

Agree, but I'm a bit puzzled by "My problem, of course, is shutter speed when the subjects are active in very poor light.". Imo the underexposure is solving this problem. Lets assume that you need 1/30 @F/1,4 and ISO 640. You need to boost the shutter speed. You can a) increase to ISO 1280 to have 1/60 @ f/1,4, or you can b) and c) underexpose -1 or -2 by 1/60 or 1/125 @F/1,4 and ISO 640. For all you want to apply (different) noise correction in pp, but b) will give you best SNR.

What I said is puzzling, re-reading it. Sorry. What I meant to say was, in general, I'm almost always needing more speed, so a stop or more gained through underexposure can be helpful. I'm often at iso 1250, f/1.4, and 1/30, metered, in a couple bars where several bands play that I follow. In this situation, 1/30 will give me lot of blurred subjects. I don't really like iso 2500 shadows on my M9. So I will underexpose a stop or more: iso 1250, f/1.4, 1/60, sometimes 1/90 or 1/125 if there's much musician movement. What I'm going to try more carefully is shooting these venues with the same shutter speeds and f-stops at a max iso of 640 and also at iso 1250 and see how they look in post after bringing them up. I've done this loosely and, despite the additional underexposure at iso 640 and need to bring the files up that much more, I like what I see.

 

[flyalf]

@icebear: Its great that you are satisfied with you way of doing things. Stick to what suits you. But that doesn't mean your way is the best. And to disregard other proposals as "utter nonsense" without having tried them seems to me like, well not utter nonsense, but at best disrespectful.

Here is a graph showing SN of M and M9. As one can see the noise increases form ISO 500 and upwards.

Here is my understanding based on facts and works of others (thanks):
There is also some indications that one could (should) treat M9 as ISO-less from 640 (or better 500) and do post-prosessing push in LR instead of increasing ISO in M9. Please see: ETTR — Just crank up the ISO? Part 13 | The Last Word
http://blog.kasson.com/?p=2824

Some more information:
http://www.l-camera-forum.com/leica-forum/leica-m9-forum/272817-does-m9-have-good-bad-isos-3.html