Leica M sensor low light performance state of the art?

[JimKasson]

Advertisement (gone after registration)

I finished the real-world image series where the exposure is left constant, the ISO is reduced a stop each step, and Camera Raw Exposure is added to compensate.

 

Conclusions:

 

 

• There’s more noise as the ISO increases
  
• The ISO 3200, and to a lesser extent the ISO 1600 image, look sharper than the low-ISOones. They’re not, but the added noise makes them appear to be sharper just like grainier films sometimes appear sharper.
  
• There evidence in the form of aliasing of the low ISO images that they are sharper than the high-ISO ones.
  
• Camera Raw introduces a slight greenish cast in the shadows with used to push the low-ISO images.
  

Mids and highlights are here.

 

 

Shadows are here.

[JimKasson]

Ok I'm officially lost. the main problem I find with using the higher ISO settings is banding in the shadow areas so is this reduced by keeping the ISO down and pushing processing the image?

 

In researching this issue, I've found the banding at ISOs 3200 and 1600, mostly running in the long direction -- from left to right when the camera is held in the landscape orientation. When I tried to see what happened when I turned down the ISO and pushed in post, I got anomalous results: lower signal level than you'd expect, and a color shift in the green direction.

 

Here are some of the results: Characterizing the Leica M240, part 10 | The Last Word

 

I am now concerned that my camera may be defective in some way, and want to find out if other M240s behave the same way before continuing too far with this work. If you have an M240 running firmware version 1.1.0.2 and are willing to make a series of test images and send me the raw files, please PM me.

 

I would like five images, all of the same scene. The scene should have approximately 6500K color balance (5000 to 10000 is OK), and a normal contrast range. Outdoor sunny scenes are out because of the blue shadows. Outdoor cloudy scenes are fine, but may be too bright if you don't have a ND filter. Outdoor dusk scenes are good, as are indoor scenes light indirectly by sunlight.

 

The first image should be four stops underexposed at ISO 3200. The next four should use the same f-stop and shutter speed at ISOs 1600, 800, 400, and 200,

 

Thanks for your help on this.

 

Jim

[JimKasson]

I've completed a comparison of the shadow noise of the M240 at ISO 3200, ISO 1600 with a 1-stop push in Lr, ISO 800 with a 2-stop push in Lr, ISO 400 with a 2-stop push in Lr, and ISO 200 with a 4-stop push in Lr,For comparison, I've done a similar series with the Sony RX1.

 

The shadow green shift of the M240 images at ISO 800 with a 2-stop push in Lr, ISO 400 with a 2-stop push in Lr, and ISO 200 with a 4-stop push in Lr is evident in these overall images. There is no material shift at ISO 3200 or ISO 1600 with a 1-stop push in Lr. The RX1 doesn't exhibit any shadow color shifts, except maybe a little at ISO 200 with a 4-stop push in Lr.

 

2:1 images of 300x300 pixel crops of the series are here.

 

My conclusions:

 

 

• The M240 at ISO 3200 is noisier than it is at ISO 1600 pushed one stop in post. Therefore, I'd stop increasing the ISO on the M240 at 1600.
  
• The green shadow shift (remember, the brightest part of the images is no brighter than 4 stops below full scale) in the pushed ISO 200, 400, and 800 images is not worth dealing with for any noise reduction that it might offer. Therefore, I wouldn't plan to push M240 images exposed at these ISOs. In all the above, we are pushing to the equivalent of ISO 3200. ISO 100 M240 images may be just fine if only pushed a stop or two.
  
• The RX1 has less noise than the M240 at all settings.
  
• The banding that showed up on the M240 in the flat-field testing is not visible in these images. Although some have said that banding is a problem in low light with the M240, I wouldn't say so from my testing.
  

I still would like to see some test images from other M240s to see if the green shadow shift under post-pushing that I observe in my camera is the result of some manufacturing (as opposed to design) defect.

[jffielde]

Thank you for doing this work, but I am confused at your statement that banding does not appear to be an issue. At least on my monitor, all of the bookcase images look banded beyond use for any purpose - even at thumbnail size. I must be looking at different images or misinterpreting your intent.

[Guest malland]

Jim, thanks for providing this information. Very helpful and interesting conclusions. It would be great if a couple of people could provide test images from other M240s to see if the green shadow shift under post-pushing that you observe is present in their results as well.

 

Are there any observations in terms of comparisons with the M9 that are worth making, such as pushing the M240 from ISO 1600 to 3200 vs pushing the M9 from ISO 640 to ISO 3200? My results in terms of night shots with the M9 pushed 2-4 stops from ISO 640 in LR5 shown in the M9 thread lead me to conclude that I like the M9 for high-ISO photography using this technique — in particular in terms of the color rendition at night. Is there anything you can say about the comparative color rendition of the M240 and the M9, or are there just too many variables involved?

 

—Mitch/Paris

Tristes Tropiques [WIP]

[dwbell]

In researching this issue, I've found the banding at ISOs 3200 and 1600, mostly running in the long direction -- from left to right when the camera is held in the landscape orientation. When I tried to see what happened when I turned down the ISO and pushed in post, I got anomalous results: lower signal level than you'd expect, and a color shift in the green direction.

 

Here are some of the results: Characterizing the Leica M240, part 10 | The Last Word

 

I am now concerned that my camera may be defective in some way, and want to find out if other M240s behave the same way before continuing too far with this work. If you have an M240 running firmware version 1.1.0.2 and are willing to make a series of test images and send me the raw files, please PM me.

 

I would like five images, all of the same scene. The scene should have approximately 6500K color balance (5000 to 10000 is OK), and a normal contrast range. Outdoor sunny scenes are out because of the blue shadows. Outdoor cloudy scenes are fine, but may be too bright if you don't have a ND filter. Outdoor dusk scenes are good, as are indoor scenes light indirectly by sunlight.

 

The first image should be four stops underexposed at ISO 3200. The next four should use the same f-stop and shutter speed at ISOs 1600, 800, 400, and 200,

 

Thanks for your help on this.

 

Jim

 

Can do tomorrow Jim (EU). Will pm you a drop box link.

[JimKasson]

Advertisement (gone after registration)

Thank you for doing this work, but I am confused at your statement that banding does not appear to be an issue. At least on my monitor, all of the bookcase images look banded beyond use for any purpose - even at thumbnail size. I must be looking at different images or misinterpreting your intent.

 

I guess I'm not understanding what you mean by "banding". Maybe you could describe what you see in detail, referring to one of the images. Do you see the banding in the Sony images as well?

 

Jim

[JimKasson]

Are there any observations in terms of comparisons with the M9 that are worth making, such as pushing the M240 from ISO 1600 to 3200 vs pushing the M9 from ISO 640 to ISO 3200? My results in terms of night shots with the M9 pushed 2-4 stops from ISO 640 in LR5 shown in the M9 thread lead me to conclude that I like the M9 for high-ISO photography using this technique — in particular in terms of the color rendition at night.

 

I think the testing on each sensor has to stand alone. The technologies are quite different (CMOS vs CCD), and I don't think you can draw conclusions about one camera by looking at the performance of the other.

 

Is there anything you can say about the comparative color rendition of the M240 and the M9, or are there just too many variables involved?

 

Both cameras are pretty linear, so color rendering differences revolve around differences in the spectral response of the filters in the color filter array. We know that no commercially-available three-color camera sees color like the human eye. The colors of the raw converted image are not only a function of the CFA but of the algorithms used in the raw converter, and the profiles as well. The two cameras see color differently. Which is more pleasing to you will depend on your eyes, your esthetics, your raw converter, your profiles, the lighting of the scene, and the spectral response of the surface reflectivity of the objects in the scene.

 

Yes, it's too complicated for a simple answer, at least for me.

 

Jim

[dwbell]

Can do tomorrow Jim (EU). Will pm you a drop box link.

 

pm sent

[JimKasson]

pm sent

 

Thanks, Duncan. Here's what I see in your images (and they are perfect for this purpose):

 

There is a green shift that gets progressively more extreme as the ISO drops and the push goes up. As I saw in the bookcase images that I did, the lower ISO images are also darker than they should be, which, among other things, leads me to think that there's a negative offset applied before analog to digital conversion.

 

There is a fairly-pronounced shift in the green direction going from the ISO 3200 to the ISO 1600 image, which is something that I didn't see when I did my testing. It looks like my basic exposure is between one-third and two-thirds of a stop more than yours, which might explain this discrepancy.

 

The noise in the ISO 1600 image seems to be greater than the noise in the ISO 3200 one. I saw the opposite. Your ISO 3200 image appears to have slightly more noise than mine, but your ISO 1600 image appears to have quite a bit more noise than mine.

 

The amount of green cast to the ISO 100, 200 and 400 images is very close to what I see in mine. I don't think my M240 is defective. As we engineers used to tell the product managers from time to time when they complained about a feature implementation, it "works as designed." They never liked that, and, frankly, neither did I, but it was sometimes what you had to do to keep your development schedule from being jerked around.

 

I can see the horizontal banding on all of your images. Now that I'm tuned in to it, I can see it in mine, too, although it's less than in yours; it helps if I rotate my stacked image 90 degrees so the main banding runs from left to right. I can't see the banding in the RX1 images. If I take out the Exposure boost layer ( + 4.5 EV in your images, + 4 EV in mine) to get the tonal structure you'd see if you were trying to capture the scene rather than do a test, I can't see the banding in my image, although I can make it out in yours if I look hard.

 

Thanks again for all your help. This has been most instructive to me, and I hope to you as well.

 

Jim

[dwbell]

Most instructive Jim. I now know my camera is much noisier at most ISO's than yours!

 

[JimKasson]

Most instructive Jim. I now know my camera is much noisier at most ISO's than yours!

 

 

I think that's probably the difference in exposure. You don't keep your house really warm, do you?

 

Jim

[dwbell]

Nope, normal centrally heated with an over active five year old leaving all doors open all the time.

[JimKasson]

Nope...

 

I can't find my Macbeth color checker. I've ordered a new one (the Macbeth name is now gone, however). I'll make some images of it at various exposures and see what the color shifts and the noise look like compared to the color checker in your images.

 

I'll probably post them on Wednesday.

 

Jim

[JimKasson]

I now believe that the color shifts I’ve been observing are not the result of my having a defective camera, but rather something inherent to the design. There is another thing to check, though. If there were an error in the Adobe camera profile for the M240, I’d see similar results with any M240, even if the camera itself is fine. I don’t think that’s the case, because I’ve observed anomalous noise floor versus ISO behavior with the M240 using RawDigger only, but it’s something I need to check before I point my finger at the M240′s design.

 

Jim

[kdriceman]

 

I think Leica has taken the better road. Nikon/Canon should at least provide a "off switch".

 

.

 

FWIW... Canon does have an "off" switch... Or at least a menu-accessible soft switch.

[jaapv]

My experiments with the M seem to indicate that ISO 800 is the optimum speed to shoot at.

[JimKasson]

I used Iridient Developer 2.2 to process the M240 DNG images that I’ve been converting in Lr 5. I used two profiles: one called RDv211 Leica M (Typ 240), and one derived from the DNG metadata.

 

The first yielded results different from in general, but substantially the same in the low ISO/push green shift as Lr 5, although I wasn’t able to test the ISO 200 image, since the most that ID 2.2 won’t allow is a 3 stop push. Adding a Ps Exposure layer above the ISO 200 image and setting it to + 1 EV yielded a pitifully washed-out result.

 

 

Using the metadata from the DNG profile produced the same results as above.

 

 

I don't think the green shadows when pushing are a profile issue.

 

 

Jim

[jaapv]

3200 push and 6400 push on the M itself produce a green cast as well.

[Advertisement]

Hallo,

Look here: Leica M

[JimKasson]

3200 push and 6400 push on the M itself produce a green cast as well.

 

I tried, but was unable to verify that.

 

Characterizing the Leica M240, part 16 | The Last Word

 

Jim