OM-1 Mark II - ISO testing

[NotThePainter]

I'm loving my 10 day old OM Systems OM-1 Mark II. My previous experience was mostly nightime landscapes and I was always frustrated by noise. (This was over a decade ago, I was shooting a Canon 5D2.) My new camera has renewed my wife's interest in photography, so now she's shopping. Her current gear is a Canon 7D, shooting wildlife with the 70-200mm f2.8.

I'm seeing amazing things out of my camera, at ISOs I wouldn't have considered using 10 years ago. Post processing with DX0 just blows away Lightroom from a decade ago. (None of this is a surprise.)

But I got to thinking, how much better has it gotten? I know there are a million review sites out there, especially our own dpreview's lab Image Comparison Tool. But that doesn't take into account my post processing and I'v gotta tinker, so I set up a test in the basement "studio," aka my woodshop.







The hardest part was finding incandescent light bulbs, my first run was with the LEDs and when I got into the shorter shutter speeds I was getting brightness changes as the shutter and the LEDs didn't match.

Camera was set to P mode, mounted on a great manfrotto tripod, using a 2 second shutter delay. I ran the ISOs from 200 to 102,400, approximately doubling each time.

Photos were all processed identically in DxO PhotoLab8, using all defaults except for their DeepPrime XD/XD2s denoise and demosaicing. I export to JPEGs using quality 90.


200


400


800


1,600


3,200


6,400


12,800


25,600


51,200


102,400

Now, what conclusions can I make. First, ISO 102,400 is absolutely fine for bird identification. And that's an important use case for me. But going as high as ISO 12,800 results is pretty good images. Not great, but certainly good. Above that and things really start to look artificial and plasticy. For example, pixel peep the 3,200 image, look at the combination square in the lower right. That's about the last "usable" image, higher ISO distort the ruler markings, presumably the AI noise gizmo failed there.

But wow, things have changed in the last decade. Hubba hubba!

 

[johnCam]

Glad to hear you like your OM-1ii. I bought mine about two months ago. I too have truly enjoyed it and my picture taking frequency has increased dramatically. I use auto ISO with the upper limit set at 2000. So, according to your tests, I am well within the limits of getting excellent image quality. Thank you for doing your tests. I am not much of a post processor, so I imagine it is good that I keep my top limit where it is. Here in sunny Florida I am most often shooting at lower ISO numbers too.

 

[Quailane]

Thank you for sharing!

The thing that stands out to me is that the AI noise reduction is making up a lot of fake details in the higher ISO images. To me that makes it look worse when I look at the image at 100% magnification.

I noticed that the F-stop was variable, and that might play a small role in image sharpness. Also did you use the electronic shutter or the mechanical shutter? I've read that high ISO performance is a little better with the mechanical shutter.

 

[NotThePainter]

The thing that stands out to me is that the AI noise reduction is making up a lot of fake details in the higher ISO images. To me that makes it look worse when I look at the image at 100% magnification.

Yeah, AI is amazing until it isn't, and worse, you sometimes don't notice until it is too late.

I noticed that the F-stop was variable, and that might play a small role in image sharpness. Also did you use the electronic shutter or the mechanical shutter? I've read that high ISO performance is a little better with the mechanical shutter.

I was in P mode, so I could just vary the ISO while keep image brightness the same.

But, I could re-do in Aperture mode.

Mechanical shutter, but that was only accidental. :- )

 

[cr0100]

As the photos I'm taking are not meant to be forensically detailed, I've got my max ISO up to 12,800. Seems extreme, but even without tossing in AI noise reduction in Lightroom, I find the results quite acceptable. "Getting the shot with some noise/grain" is better to me than something blurry because of too long of a shutter speed.

For live concert photography, I have the shutter at 1/160 or so to keep motion relatively frozen, aperture wide open (f/2.8 or f/4.0 depending on lens) and let the ISO float automatically up to 12800. Very few of the thousands of images I shoot at a show are bad because of motion blur, so that's a win. Unlike bird photography, I don't have to do a lot of cropping on the image, so keeping the full frame definitely helps the noise level remain acceptable.

Here's a non-concert shot taken during a book club meeting at my house a while ago. I'll include before-and-after AI noise reduction. Really, either of them are perfectly fine as long as you don't go pixel-peeping. First is no noise reduction, second with light AI noise reduction applied in Lightroom Classic.




ISO 12,800 with no noise reduction




ISO 12,800 with light AI noise reduction applied in Lightroom Classic.

 

[NotThePainter]

Here's a non-concert shot taken during a book club meeting at my house a while ago. I'll include before-and-after AI noise reduction. Really, either of them are perfectly fine as long as you don't go pixel-peeping. First is no noise reduction, second with light AI noise reduction applied in Lightroom Classic.

Pixel peep to the black and white thing to the subject's left, in front of the bike helmet. I'd say Lightroom did a great job!

 

[NotThePainter]

I noticed that the F-stop was variable, and that might play a small role in image sharpness. Also did you use the electronic shutter or the mechanical shutter? I've read that high ISO performance is a little better with the mechanical shutter.

I re-ran the tests, constant F6.3, but I went with the electronic shutter, no the mechanical one. I did this because I noticed the very low ISO images above were slightly less sharp than the low medium images. My unsubstantiated theory is that the mechanical shutter is causing a slight vibration. (I almost typed mirror bounce, LOL.)

I also moved the $100 bill to the center so it is easier to pixel peep, at least on my laptop.

This makes the AI distortions far more apparent. If you want to see more of that, scroll down to the combination square in the lower left and see what happens to the ruler markings.

It looks like 6,400 or 12,800 are the tops for "scientific" quality images. By scientific I mean I wouldn't mind uploading a bird photo to ebird's db. Anything higher that that is inviting AI distortions, which I'd be hesitant to to go higher. But, if you are only looking for Bird Identification, this is one of the primary reasons I bought this gear, you can go as high as you want.

But still, wow, a lot has happened in the last decade.


200




400




800




1,600




3,200




6,400




12,800




25,600




51,200




102,400

 

[Quailane]

It's interesting to see how the higher ISO levels wipe out the detail, especially in the blacks and shadows. Thanks for moving the 100 dollar bill, that does stand out as one of the main areas where the AI artifacts crop up. The images are very impressive and I agree that ISO 12800 does look really good after your processing. I think ISO 25600 is usable as well as a whole image. It is above that where things start to fall apart.

I think the moral of the story is to not be afraid to shoot as at high of an ISO as you need to.

Thanks for your hard work!

 

[2ndact scene1]

As the photos I'm taking are not meant to be forensically detailed, I've got my max ISO up to 12,800. Seems extreme, but even without tossing in AI noise reduction in Lightroom, I find the results quite acceptable. "Getting the shot with some noise/grain" is better to me than something blurry because of too long of a shutter speed.

For live concert photography, I have the shutter at 1/160 or so to keep motion relatively frozen, aperture wide open (f/2.8 or f/4.0 depending on lens) and let the ISO float automatically up to 12800. Very few of the thousands of images I shoot at a show are bad because of motion blur, so that's a win. Unlike bird photography, I don't have to do a lot of cropping on the image, so keeping the full frame definitely helps the noise level remain acceptable.

Here's a non-concert shot taken during a book club meeting at my house a while ago. I'll include before-and-after AI noise reduction. Really, either of them are perfectly fine as long as you don't go pixel-peeping. First is no noise reduction, second with light AI noise reduction applied in Lightroom Classic.


ISO 12,800 with no noise reduction


ISO 12,800 with light AI noise reduction applied in Lightroom Classic.

I use a pair of GH5S for theater and events plus a concert photography workshop I did in Nashville last month. I can set maximum ISO to 12,800 on these units becuase they are designed to handle low light, though at the "cost" of only having 10MP. Prior to that, I used various Olympus cameras including the EM1X and EM1.iii. For those cameras, I set maximum ISO at 6400. In all cases, I would process in DXO.

That extra stop of light on the GH5S makes a huge difference. At ISO 6400, f/2.8 zooms are outside their performance envelope at shutter speeds of 1/250 or 1/500. I thought for a while I would have to switch to only using fast primes, such as the Sigma 56mm f/1.4 (which I did buy and works great). Or maybe go back to full frame. But the GH5S solved that problem.

I have tested the G9ii, GH6 and OM-1, and they are all about equal to the GH5S's ISO performance, but obviously with more MP.

I learned some time ago that for the type of shooting I do, you strive to make people look good and to capture important moments. But it is not fine art. That doesn't mean that you are not being creative in the way you shoot your subjects. But you have to keep in mind that the end product is for social media or a printed program or flyer. I do sometimes submit images in our club's print competitions and they sometimes score well. But intense pixel peeping is a bit irrelevant.

 

[cr0100]

I learned some time ago that for the type of shooting I do, you strive to make people look good and to capture important moments. But it is not fine art. That doesn't mean that you are not being creative in the way you shoot your subjects. But you have to keep in mind that the end product is for social media or a printed program or flyer. I do sometimes submit images in our club's print competitions and they sometimes score well. But intense pixel peeping is a bit irrelevant.

100% this. These 12,800 shots have less grain/noise in them than I used to get with my Tri-X 400 B&W concert shots back in the 1980s... and nobody looked at those 1980s shots and said "nice photo, but sure has a lot of grain"! For example...

 

[Asherax]

I learned some time ago that for the type of shooting I do, you strive to make people look good and to capture important moments. But it is not fine art. That doesn't mean that you are not being creative in the way you shoot your subjects. But you have to keep in mind that the end product is for social media or a printed program or flyer. I do sometimes submit images in our club's print competitions and they sometimes score well. But intense pixel peeping is a bit irrelevant.

100% this. These 12,800 shots have less grain/noise in them than I used to get with my Tri-X 400 B&W concert shots back in the 1980s... and nobody looked at those 1980s shots and said "nice photo, but sure has a lot of grain"! For example...

THIS! <3

--
Photography is poetry made visible; it is the art of painting with light!

 

[James Stirling]

I'm loving my 10 day old OM Systems OM-1 Mark II. My previous experience was mostly nightime landscapes and I was always frustrated by noise. (This was over a decade ago, I was shooting a Canon 5D2.) My new camera has renewed my wife's interest in photography, so now she's shopping. Her current gear is a Canon 7D, shooting wildlife with the 70-200mm f2.8.

I'm seeing amazing things out of my camera, at ISOs I wouldn't have considered using 10 years ago. Post processing with DX0 just blows away Lightroom from a decade ago. (None of this is a surprise.)

But I got to thinking, how much better has it gotten? I know there are a million review sites out there, especially our own dpreview's lab Image Comparison Tool. But that doesn't take into account my post processing and I'v gotta tinker, so I set up a test in the basement "studio," aka my woodshop.



The hardest part was finding incandescent light bulbs, my first run was with the LEDs and when I got into the shorter shutter speeds I was getting brightness changes as the shutter and the LEDs didn't match.

Camera was set to P mode, mounted on a great manfrotto tripod, using a 2 second shutter delay. I ran the ISOs from 200 to 102,400, approximately doubling each time.

Photos were all processed identically in DxO PhotoLab8, using all defaults except for their DeepPrime XD/XD2s denoise and demosaicing. I export to JPEGs using quality 90.


200


400


800


1,600


3,200


6,400


12,800


25,600


51,200


102,400

Now, what conclusions can I make. First, ISO 102,400 is absolutely fine for bird identification. And that's an important use case for me. But going as high as ISO 12,800 results is pretty good images. Not great, but certainly good. Above that and things really start to look artificial and plasticy. For example, pixel peep the 3,200 image, look at the combination square in the lower right. That's about the last "usable" image, higher ISO distort the ruler markings, presumably the AI noise gizmo failed there.

But wow, things have changed in the last decade. Hubba hubba!

I think in moderation NR of late has improved greatly, however it is not a panacea , removing noise is easy , but the fine detail is destroyed if you fly too high . I think your observations are fair

--
Jim Stirling:
"To argue with a person who has renounced the use of reason, is like administering medicine to the dead." - Thomas Paine
Feel free to tinker with any photos I post

 

[NotThePainter]

I think in moderation NR of late has improved greatly, however it is not a panacea , removing noise is easy , but the fine detail is destroyed if you fly too high . I think your observations are fair

Exactly. Zoom in on the ruler in the lower left, the details just get worse and worse.