M9 Colors at Night — Best Way to Shoot High ISO?

[Guest malland]

Advertisement (gone after registration)

Guys,

 

On the German forum, where there is a parallel thread, someone asked me how to process in Lightroom 4/5 in order to push Exposure from ISO 640. I thought it might be useful to post here the steps that I posted in the German forum, although I gave less detailed description of this earlier in this thread:

 

Exposure

At ISO 640 start (with the Elmarit 21mm ASPH) by exposing at f/2.8 and a shutter speed of 1/60 or 1/90, and (with a Summicorn-28) by exposing at f/2.0 and a shutter speed of 1/120 or 1/180. The aim is not to blow out the highlights. On a dark night, I find that I don't need much depth of field because of the rapid drop-off in light intensity.

 

Lightroom 4/5 Post-Processing

1. Click Auto in the Exposure Panel and use this as the starting point.

2. Adjust the Exposure Slider to the point at which you like the look.

3. Press "J" and see whether any of the highlights are blown out.

4. If necessary (from Step 3), pull back on the Highlights Slider.

5. I like to pull back the Black Slider to negative numbers so that picture has a look that I like.

6. Click White Balance Eye Dropper in an area you want to be neutral grey, but I don't like to neutralize the colors completely because I want a "real" look from the lights in the picture.

7. Try moving the Clarity Slider between +10 – +30. (I find that increasing Clarity creates a good feeling of light in a high contrast scene, but sometimes this may not be necessary or even may not look good.

8. If you increased Clarity, you can probably pull back a bit on the Exposure.

9. In the Noise Reduction panel, after setting the View to 100%, move the Color Slider to the right until the color noise disappears.

10. Still at 100% View, move the Luminance Slider to the right if necessary to remove more noise, but be careful no to go too far. Some pictures will not need any Luminance Slide increase.

 

—Mitch/Pak Nam Pran

Bangkok Obvious [WIP]

Eggleston said that he was "at war with the obvious"...

[jaapv]

If it's the same sensor and the only change is the removal of the Color Filter Array, then the ISO where the noise starts to fall off relative to pushing in post should be twice as high, since the removal of the CFA makes the sensor twice as efficient at converting photons to electrons. Therefore, I'd expect that the point where you should stop increasing the ISO is 1250.

 

If it's a different sensor, I don't know the answer, since I've never tested -- or even seen -- an M Monochrom.

 

Jim

Yes, but you have no luminance noise and by leaving out the colour interpolation steps the noise is not blotchy and more grain-like leading to higher acceptable noise levels. So the value of 1250 is subject to experimental confirmation. Actually it - or 2500 if possible- is a value I like to shoot at to get some structure in my image.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[Guest NEIL-D-WILLIAMS]

Looking at the settings you had on most of the pictures I think you could have got better results slowing the shutter speed to 1 stop over focal length. That way reducing the ISO number…………………correct me if I am wrong but I thought you could set the auto ISO to the lens that you have screwed onto the body i.e if you are shooting with a 50mm then the shutter speed will also be 50mm minimum and ISO set accordingly

My 2 cents

Shot at 10,000 ISO by mistake http://www.neilsphotography.co/mabul-sunrise/ Dfine kills noise in Nik Software

Edited July 27, 2013 by NDWgolf

[Guest malland]

Neil, I think you can ignore what shutter speed I used because, of course, I could have used other speeds. Reading what you wrote, it seems that you haven't read post #1 explaining the technique and the technical article by Jim Kasson linked in the post. The point is that tests by Jim Kasson (and others?) show that the M9 gets better image quality (less noise) if you increase ISO in-camera up to ISO 640, and from that point on if you keep on shooting at ISO 640 and push the Exposure in Lightroom.

 

Jaap, similarly, for the M-Monochrom, we can assume for the time being, until someone correctly tests this, the the breakpoint is ISO 1250, after which it's better to increase Exposure in post rather than in-camera.

 

—Mitch/Bangkok

Bangkok Obvious [WIP]

Eggleston said that he was "at war with the obvious"...

[Guest NEIL-D-WILLIAMS]

Neil, I think you can ignore what shutter speed I used because, of course, I could have used other speeds. Reading what you wrote, it seems that you haven't read post #1 explaining the technique and the technical article by Jim Kasson linked in the post. The point is that tests by Jim Kasson (and others?) show that the M9 gets better image quality (less noise) if you increase ISO in-camera up to ISO 640, and from that point on if you keep on shooting at ISO 640 and push the Exposure in Lightroom.

 

Jaap, similarly, for the M-Monochrom, we can assume for the time being, until someone correctly tests this, the the breakpoint is ISO 1250, after which it's better to increase Exposure in post rather than in-camera.

 

—Mitch/Bangkok

Bangkok Obvious [WIP]

Eggleston said that he was "at war with the obvious"...

Mitch,

I did read your post, my point is, is that had you decreased the shutter speed then you wouldn't have needed as much ISO in turn less noise

[Guest malland]

Sorry, Neil, I misunderstood. Yes, of course, I could have used lower shutter speeds and then pushed less in post. The times I used 1/125 sec is in the earlier examples, before I had settled down to a methodology on how to expose using this technique.

 

—Mitch/Pak Nam Pran

Bangkok Obvious [WIP]

Eggleston said that he was "at war with the obvious"...

[Guest NEIL-D-WILLIAMS]

Advertisement (gone after registration)

Sorry, Neil, I misunderstood. ...

No loblem.......as the say in Thailand

[Guest malland]

Neil, I'm very slow today: I should have added that often I would not want to limit the shutter speed to the reciprocal of the focal length of the lens because of possible, or actual subject motion. That may have been a reason for using 1/125 and 1/180.

 

—Mitch/Bangkok

Bangkok Obvious [WIP]

Eggleston said that he was "at war with the obvious"...

[Guest NEIL-D-WILLIAMS]

Neil, I'm very slow today: I should have added that often I would not want to limit the shutter speed to the reciprocal of the focal length of the lens because of possible, or actual subject motion. That may have been a reason for using 1/125 and 1/180.

 

—Mitch/Bangkok

Bangkok Obvious [WIP]

Eggleston said that he was "at war with the obvious"...

I thought you could set the minimum shutter speed in the auto ISO settings............I dont have a Leica right now or I would check. Meaning if you screw on a 50mm you can set the shutter to say 60/sec or a 28mm say 40/sec and so on

[Guest malland]

Neil, now your losing me again: I certainly cannot use AutoISO for this technique, which is based on shooting at ISO 640 and pushing in post.

 

—Mitch/Bangkok

Paris au rhythme de Basquiat and Other Poems [download link for book project]

[Guest NEIL-D-WILLIAMS]

Neil, now your losing me again: I certainly cannot use AutoISO for this technique, which is based on shooting at ISO 640 and pushing in post.

 

—Mitch/Bangkok

Paris au rhythme de Basquiat and Other Poems [download link for book project]

Download Nik software and use Dfine 2 to get rid of any noise.............it’s the dogs bollocks mate trust me:D:D

[Guest malland]

Neil, I don't agree and much prefer LR4/5 Noise Reduction, used in the manner stated in post #101. However, I think comparisons of noise reduction software are OT for this thread: if that is what you want to discuss, please start a new thread on that topic.

 

Mitch/Pak Nam Pran

Bangkok Hysteria (download link for book project)

[baci]

OK, so I did some more “flat” lighting tests and the results were frankly atrocious.

 

Let me tell you what I think is happening. The whole point of ETTR is to utilise the maximum dynamic range of the sensor – the full “bit depth” (which I believe to be 22.5 bit).

 

In uniform low light conditions, when I choose to disregard ISO and set the camera for depth of field (smaller aperture, less light) and shutter speed (faster shutter, less light) I end up letting so little light through to the sensor that I am only using a small fraction of the available dynamic range. This translates to the image being maybe four or five bit, instead of the full 22.5.

 

In this scenario, sensor noise is the least of my problems. There’s just little definition in the image and the NR treatment tends to reduce this still further. Hence a nasty looking image.

 

Shooting in higher contrast situations with artificial light piercing the darkness enables the sensor to operate well in the highlights, and the dark areas require little definition anyway, so voila – an attractive image.

 

Just a thought...

[JimKasson]

The whole point of ETTR is to utilise the maximum dynamic range of the sensor – the full “bit depth” (which I believe to be 22.5 bit).

 

You are absolutely right about the whole point of ETTR. However, cranking up the ISO doesn't make more light fall on the sensor. Once you've set the ISO dial to anything but vase ISO -- 160 in the M9's case -- you've set up a situation in which you can't achieve all the dynamic range of which the sensor is capable.

 

My testing confirms that the M9 has a 14-bit ADC, and substantial noise in the least-significant bits, for a dynamic range of just under 13 bits. So I can't see why you say is has a 22.5 bit dynamic range. A 22.5 bit DR is about six million to one.

 

Look at the noise floor versus ISO graph (third graph from the top) here:

 

ETTR — Just crank up the ISO? Part 13 | The Last Word

 

and you'll see the dynamic range starting at under 13 bits at ISO 160 and dropping to eight bits at ISO 2500, and that's assuming ETTR exposures.

 

In uniform low light conditions, when I choose to disregard ISO and set the camera for depth of field (smaller aperture, less light) and shutter speed (faster shutter, less light) I end up letting so little light through to the sensor that I am only using a small fraction of the available dynamic range...

 

In this scenario, sensor noise is the least of my problems. There’s just little definition in the image and the NR treatment tends to reduce this still further. Hence a nasty looking image.

 

I think the point that the OP was making, and the point that I was making in the blog posts the OP linked to, was not that pushing in post was better than letting more light on the sensor, but that, after ISO 640, that, for the same exposure, pushing in post gave measurably better results than dialing in the same amount of in-camera gain with the ISO dial.

 

You are entirely correct in that there is no substitute for more light on the sensor, right up to the point where those little electron wells get near saturation.

 

Jim

Edited July 29, 2013 by JimKasson

[JimKasson]

Yes, but you have no luminance noise and by leaving out the colour interpolation steps the noise is not blotchy and more grain-like leading to higher acceptable noise levels.

 

I think you mean that you have no chroma noise in the Monochrom; since the image is monochromatic, all the noise is in the luminance dimension. You are correct in pointing out that demosaicing, which is not necessary in the Monochrom, tends to spatially spread sensor noise, resulting in relatively less energy at high spatial frequencies than in the Monochrom images.

 

I have performed some experiments with demosaicing versus luminance equalizing deep-IR images.

 

Processing raw B&W infrared images | The Last Word

 

I didn't look at noise, but I'll bet if I did I'd get the results you are talking about.

 

So the value of 1250 is subject to experimental confirmation.

 

I think all this is subject to experimental confirmation, so I agree with you there. If you are saying that the interpolation of demosaicing can cause what used to be the minimum-noise ISO to no longer be the minimum-noise ISO, I don't see how that could happen.

 

Actually it - or 2500 if possible- is a value I like to shoot at to get some structure in my image.

 

I believe the thrust of this thread was how to minimize noise and maximize dynamic range. It seems that you're saying here that, under some circumstances, you actually like more noise. If that's the case, strategies to minimize the noise are solving a problem that you don't have, and may move you further from your goals.

[Guest malland]

...I believe the thrust of this thread was how to minimize noise and maximize dynamic range. It seems that you're saying here that, under some circumstances, you actually like more noise. If that's the case, strategies to minimize the noise are solving a problem that you don't have, and may move you further from your goals.

Agreed. My own view is that at speeds like ISO 5,000 and 10,000 the dynamic range is so severely reduced that I would prefer to shoot the M-Monovhrom at the "breakpoint" and then push in post. It would be good if someone would run the tests to confirm whether this breakpoint between increasing the ISO in camera or in pushing in post is indeed ISO 1250.

 

Also, as stated in earlier posts, for me, an important advantage of the technique being discussed is that exposure is essentially determined in post by moving the Exposure Slider to the minimum necessary to achieve the look one wants, This is also an important in that shooting at night, usually in high contrast and mixed light situations, often makes the "correct" (i.e., desired) exposure difficult to measure a priori.

 

If I had an M240, I would also want to know where this breakpoint is.

 

—Mitch/Pak Nam Pran

Paris au rhythme de Basquiat and Other Poems [download link for book project]

[sblitz]

in a nutshell the way i understand is this -- you can achieve gain either in camera by raising iso or increasing exposure pp. each method has its practical break points so they are not the same at all times. we have, for the purposes of this thread, decided that iso 640 is as high as you want to go with the m9 and from there use pp to increase exposure. of course we are trying to replicate what was photographed and not make it look like it was shot at noon. it would be seem to me that carrying a small reflective meter set at what iso we want should be the bench mark for how to set f/stop and exposure in the camera. i set meter iso to, say, 2500, meter the area, set the m9 as if it were set at 2500 to the meter's metrics and then "push" to 2500 in pp. i am not sure how you can do this without a meter unless you simply set a exp & f/stop to a constant and then, in effect, fix it in pp which is essentially allowing iso to be the variable in the equation. do i have this right? is there a way to know where 640 is indeed the breakpoint? i appreciate the work done and the comments made, just wondering there is some more scientific arbitrary way to come up with the answer. kind of like those web sites that give you the true iso of a film. the concept should be applicable to all digital cameras.

[JimKasson]

it would be seem to me that carrying a small reflective meter set at what iso we want should be the bench mark for how to set f/stop and exposure in the camera. i set meter iso to, say, 2500, meter the area, set the m9 as if it were set at 2500 to the meter's metrics and then "push" to 2500 in pp. i am not sure how you can do this without a meter unless you simply set a exp & f/stop to a constant and then, in effect, fix it in pp which is essentially allowing iso to be the variable in the equation. do i have this right?

 

My suggested way to set the exposure with the M9 when conditions are such that you can't do an Exposure To The Right (ETTR) exposure at ISOs below 640 is: just leave the f/stop and shutter speed at what's acceptable to you and the ISO dial at 640 and take pictures. You are correct in pointing out that, with no metering, you won't know haw far you have to push in post. But then again, what use is that information to you under the circumstances I'm describing? You've already set the camera to the slowest shutter speed and the widest f-stop you think you can use for the shot. The only use I can think of is that the light is so dim that the Image Quality will be unacceptable, and you shouldn't even bother tripping the shutter, and I'd rather try to get the shot and fins out later that I can't make it work than not even try. You miss 100% of the pictures you don't take.

 

Here is a more complete description of the ways to use this technique:

 

ETTR and ISO settings | The Last Word

 

is there a way to know where 640 is indeed the breakpoint?

 

The method I used was to measure the signal-to-noise ratio (SNR) in the raw files with a flat target at various ISO settings, keeping the mean raw values approximately constant, and comparing the measured SNRs to those of an ideal, photon-noise-limited camera. When the SNR at the higher ISO is worse than the photon-noise-corrected SNR at a lower ISO, you're just adding noise by turning the ISO up.

 

The M9 is unusual in that the point where increasing the ISO adds noise is easy to measure because the quality loss above that point (if you don't like noise) is much larger than in cameras with modern Sony CMOS sensors like the Sony RX-1, the Sony NEX-7, the Nikon D800, and the Nikon D4.

 

i appreciate the work done and the comments made, just wondering there is some more scientific arbitrary way to come up with the answer. kind of like those web sites that give you the true iso of a film. the concept should be applicable to all digital cameras.

 

I'm not sure what you mean by "more scientific". The SNR is an accepted measure of noise. The physics of photon noise are well established. All the software I used to make the measurements and calculations is readily available (and, in the case of RawDigger, free). You can read what I did and replicate the results for yourself if you'd like to. I did control for the thermal noise effects of repetitive exposures, which, in the case of the M9, is minimal, partially due to the small buffer and low write speeds.

 

Here is a step-by-step walkthrough of the measurements for the M9:

 

ETTR — Just crank up the ISO? Part 13 | The Last Word

 

Other work that could be done is repeating the testing with 1/3 stop increments. If you'd like to try that, I'd appreciate seeing the results. The testing is somewhat time-consuming because of the necessity of making 16 shots per test case so that random variations can be averaged out.

 

Jim

Edited July 31, 2013 by JimKasson

[JimKasson]

i appreciate the work done and the comments made, just wondering there is some more scientific arbitrary way to come up with the answer. kind of like those web sites that give you the true iso of a film. the concept should be applicable to all digital cameras.

 

Here is an approach that would bring the user's preferred raw developer into the mix. Maybe this is what you mean.

 

Make a series of exposures of a featureless target at the same shutter speed and f-stop, but with varying ISOs a full stop apart. Do this with the highest-ISO image exposed for 2 stops under full scale and 5 or six stops under full scale. Make 16 exposures per test case.

 

Bring the raw files into the raw developer of choice and develop the highest-ISO image at Exposure = o stop adjustment, the next-highest at Exposure = +1 stop adjustment, he next-highest at Exposure = +2 stop adjustment, and so on. Turn noise reduction off, and/or leave it at your preferred setting.

 

Export the developed files as 16-bit TIFFs.

 

Bring the files into your image analysis program of choice (mine is MatLab). Measure the standard deviation of the noise in a 200x200 or 400x400 pixel central area. For each color plane, average the standard deviations over the 16-shot sequences, and compute the standard deviations of the standard deviations to make sure that 16 was enough exposures.

 

Plot the two curves, the highlight and the shadows, with the standard deviations of the noise on the y-axis and the ISO setting on the x-axis..

 

Find the ISOs that produce the minimum noise.

 

I would have done this originally, but I didn't want to have the results only apply to one raw developer. If someone wants to do this, I'd be interested in the results.

 

Jim

[Advertisement]

Hallo,

Look here: Leica M

[Guest malland]

Jim Kasson has, in the last two posts has covered comprehensively and articulately the technical aspects of the "shoot at ISO 640 and push in post" technique that is the subject of this thread. I'll only deal with some of the practical aspects of shooting and processing, which are listed in post #101 above.

 

Like a lot of people, I had trouble making sense in practical terms of how to expose, because for a photographer it seems counter-intuitive not to consider ISO or worry about metering the light. But as soon as I tried it all made sense — so I encourage anyone with an M9 interested in high-ISO photography to try, and hope that we'll get some people to post their results here. Just start with ISO 640 f/2.8 and 1/60 or 1/90 sec with a wide angle lens of 21, 28 or 35mm. You can try a 50mm lens as well, but initially things will be easier with wide angle lenses.

 

One issue in Lightroom 4/5 processing is how increase Exposure. In the 10 steps listed post #101, step 5 is pull back the Black Slider to negative numbers so that picture has a look that you like. In picture Nos. 20–23 below, when I pressed Auto in the Basic Panel, the Black Slider went to +26 — and the pictures looked a bit washed out and flat; I pulled the Black Slider back to –23 to –26 for these pictures. I find that pulling the Black Slider to the left to get negative number provides a look consistent to the rapid fall-off of light in partially lit, dark scene.

 

Incidentally, you'll see below that I shot No. 21 at 1/750 sec: a simple mistake, as this picture would not have required any pushing if I had exposed correctly at a lower shutter speed; but I had my camera set to 1/750 sec from another shot and, turning around and seeing the subjects grouped in the way they stood, I wanted to get a shot off before they changed position.

 

No. 22 was a shot did not want to take because the light was so dull and flat and low: the two women, who were standing just in front of a bar that has the whole front open to a narrow, dark unlit lane asked me to take a picture of them and insisted that I do so when I had refused. After, I took the picture they wanted to see it on the LCD, but I explained that with this camera the picture would be too dark to see. The woman on the left did not believe me, but the one on the right said to her, "Don't you see? he takes it like this and then lightens it in the computer." Bingo: she understood the "shoot at ISO 640 and push in post" technique.

 

What I did in No. 22 is to push and then use the new Radial Filter of LR5 to burn in (with feathering) the area around the women by 2 stops, after having pushed the whole picture 3.65 stops, the latter being equivalent to an effective ISO of 8,448. Compare No. 22 to No. 23 (posted earlier in this thread as No. 6). No. 23 is pushed 3.2 stops (equivalent ISO of 6,144). Because the light on the subjects is so much better in No. 23 than No. 22, the image is much easier to process and the results more interesting. No. 22 is here as an example of how to deal with some shots in dull, flat and low light — photography is all about light and very poor light is difficult to deal with.

 

 

 

No. 20 | Elmarit-21 ASPH | ISO 640 pushed 0.9 stops | 1/60 sec

Hua Hin

 

 

 

 

No. 21 | Elmarit-21 ASPH | ISO 640 pushed 1.65 stops | 1/750 sec

Hua Hin

 

 

 

 

No. 22 | Elmarit-21 ASPH | ISO 640 pushed 3.65 stops | 1/60 sec

Hua Hin

 

 

 

 

No. 23 | Summicron-28 | ISO 640 pushed 3.2 stops | f/2.8 | 1/90 sec

Bangkok

 

 

—Mitch/Hua Hin

Lanka Footsteps [M-Monochrom/Sri Lanka]