LIttle ISO test EM1.3 vs S1R vs S5

[pdk42]

I was curious to see how well these three cameras performed now that we have modern NR techniques to help with noise. So, I took shots (in raw) at increasingly high ISOs with each camera and then used LR's "Enhance" NR feature to clean each image up as best I could. The scene was a dimly-lit interior since I think this best tests high ISO.

Applying the NR, I went for a balance of reducing noise vs preserving detail. It's always a compromise I think, but I prefer to have a little noise left than to smear detail too much. I then resized them all at 20Mp (5184 px long edge) to match the EM1 and went pixel peeking.

Whilst the order is totally expected (EM1.3 worst, S5 best), I was surprised at how much better the S5 is:

Camera	High quality	Just about acceptable	Never exceed !
EM1.3	800	3200	6400
S1R	1600	6400	12800
S5	6400	12800	51200

In general the S1R beats the EM1.3 by 1 stop and the S5 beats it by 2-3 stops. And the S5 beats the S1R by 1-2 stops.

Of course, at base ISO they all do a great job, but pushing shadows to 100% on them all revealed a noticeable level of noise in the EM1.3 (even with NR) whilst both the FF cameras remained essentially noise free.

 

[dirk]

Great idea, thank you for testing this.

How would the results change, if you would not resize the higher MP images to 20MP?

As far as I read it, the newer Olympus OM1 and OM1 Mkii have better high ISO quality than the EM1.3

 

[Markuswelder]

Yeah, my S5 was a real eye opener when I first got it, & shot a few frames with it. I owned a Canon 6D in the past, never really gelled with it, and didn't miss it at all when I tried out a GX8. So much so that I sold the 6D, and dived right into m4/3. First and only camera I've ever sold lol.
The S5 is a completely different story. For me, not so much the high ISO performance -which I think is outstanding by the way, but more the ability to lift shadows & dark areas at lower ISO's, without any real penalty in the form of noise, colour shifts or other weirdness. Very very much like my S5. It's a keeper for me. 24MP is just the right balance of file size, image quality and file malleability for my tastes.
Interesting results by the way. Thanks for posting

 

[pdk42]

An interesting corollary to this test is to ask the question "is the S1R really any better than the S5/S1"? It's a question that's been buzzing around in the back of my head for a while. Based on the down-sized images here, the obvious answer is "no". And to echo what @Markuswelder said, even if I stick to base ISO, I can abuse S5 images much more than S1R images in terms of exposure correction (shadows etc).

So, unless I'm using those extra Mp, the S1R is actually a worse camera in terms of output than the S5. I can perhaps see some slightly better colour/tones with the S1R on base ISO shots - but if it's there it's subtle, and could well be confirmation bias on my part rather than anything real.

So where do the extra Mp help? Obviously it's about cropping and printing big. I'm not convinced by the printing big argument since (a) I think you need to be getting REALLY big before it matters (certainly in excess of a meter) and even then there are tricks like upscaling that can be done; and (b) no one except a pixel peeker goes looking at a 1m+ wide print from 10cm away!

So we're left with cropping. This can be a great tool to give more reach - as is used in the Q3 and SL3. I think this is where I'm starting to re-align my thinking on high Mp. Ideally the new S1ii will be a 60Mp camera but with the ability to crop to APSC at 24Mp or to down-sample the full frame image to something similar.

 

[pdk42]

Great idea, thank you for testing this.

How would the results change, if you would not resize the higher MP images to 20MP?

Viewing 100% on my monitor, the S1R gets to "just about acceptable" at ISO 1600.

As far as I read it, the newer Olympus OM1 and OM1 Mkii have better high ISO quality than the EM1.3

Yes, that's what I've been told too. The big problem with the EM1.3 is that over ISO 6400 there is a strong magenta cast. The OM1 doesn't do this.

 

[xaviergut]

This can be a great tool to give more reach - as is used in the Q3 and SL3. I think this is where I'm starting to re-align my thinking on high Mp.

Paul, I wonder what happens with this triple resolution feature of the Leica SL3. According to Leica: "The Leica SL3's 60-megapixel sensor offers a dynamic range of up to 15 stops, 14-bit colour depth, and exceptionally good noise reduction, guaranteeing breath-taking images. Available in three different resolutions: 60, 36, and 18 megapixels, in both DNG and JPG formats. Its ISO sensitivity ranges from ISO 50 to ISO 100.000, ensuring outstanding results in all lighting conditions".

My question is: is the ISO and noise performance much better in the 18 mpx mode than the other two?

Do we want that triple resolution feature in the next Lumix S1 camera?

 

[xaviergut]

Paul, I wonder what happens with this triple resolution feature of the Leica SL3. According to Leica: "The Leica SL3's 60-megapixel sensor offers a dynamic range of up to 15 stops, 14-bit colour depth, and exceptionally good noise reduction, guaranteeing breath-taking images. Available in three different resolutions: 60, 36, and 18 megapixels, in both DNG and JPG formats. Its ISO sensitivity ranges from ISO 50 to ISO 100.000, ensuring outstanding results in all lighting conditions".

My question is: is the ISO and noise performance much better in the 18 mpx mode than the other two?

Do we want that triple resolution feature in the next Lumix S1 camera?

By the way, I just found a review of the SL3 and the reviewer says that the ISO performance is like the SL2-S (24mpx, like the S5 OG), but with more resolution:

"In fact, the noise performance of the SL3 is very close to that of the SL2-S, yet it offers much more resolution. Imagine the low-light performance of the SL2-S with 60MP—it’s that good!"

REVIEW: Leica SL3 - My third Leica SL in a decade (part 2) — Kristian Dowling

Resolution - Historical Relevance Back in 1913/1914, when the Ur-Leica was conceived, Oskar Barnack and Ernst Leitz II relied heavily on the company’s optical expertise to realize the revolutionary concept of "small negative-large print." Max Berek, a visionary optical engineer, was taske

kristiandowling.com

 

[pdk42]

Available in three different resolutions: 60, 36, and 18 megapixels, in both DNG and JPG formats. Its ISO sensitivity ranges from ISO 50 to ISO 100.000, ensuring outstanding results in all lighting conditions".

My question is: is the ISO and noise performance much better in the 18 mpx mode than the other two?

I'm absolutely sure that the lower res raw files will have lower noise since they will be over-sampled from the full 60Mp - and the over-sampling will result in an increase in SNR. I think if I had an SL3, I'd normally be shooting in 33Mp mode most of the time, or 24Mp APSC when I wanted my 24-105 to be 36-160!

Do we want that triple resolution feature in the next Lumix S1 camera?

Yes, absolutely.

 

[pdk42]

By the way, I just found a review of the SL3 and the reviewer says that the ISO performance is like the SL2-S (24mpx, like the S5 OG), but with more resolution:

"In fact, the noise performance of the SL3 is very close to that of the SL2-S, yet it offers much more resolution. Imagine the low-light performance of the SL2-S with 60MP—it’s that good!"

REVIEW: Leica SL3 - My third Leica SL in a decade (part 2) — Kristian Dowling

Resolution - Historical Relevance Back in 1913/1914, when the Ur-Leica was conceived, Oskar Barnack and Ernst Leitz II relied heavily on the company’s optical expertise to realize the revolutionary concept of "small negative-large print." Max Berek, a visionary optical engineer, was taske

kristiandowling.com

I've played with SL3 raw files and they are impressive, but I'd not say they were as good at high ISO as the 24Mp sensor when viewed at 100%. Reduced to the same size, they are probably just as good, or even better.

 

[dirk]

Viewing 100% on my monitor, the S1R gets to "just about acceptable" at ISO 1600.

That is something to think about. If an old MFT 20MP sensor achieves at ISO 3200 "just about acceptable" results and the S1R without resizing only up to ISO1600, MFT still has a bright future.

My question is: is the ISO and noise performance much better in the 18 mpx mode than the other two?

I did not do tests for this with my Q3 43, but the noise level with 60MP is not exciting, if you are used to the Lumix S5.

Do we want that triple resolution feature in the next Lumix S1 camera?

Yes. Absolutely. I love this feature on the Q3 43. This makes a camera so much more versatile. I do not need 60MP for my kind of photos, but I have it with me in case I need it for a specific situation.

I use on my Q3 43 always the 36MP setting. The file size with 36MP in DNG and Jpeg are the same as with 60MP by the way.

Read here:

Leica Q3 43 - resolution depending on DNG size and crop factor

Hi, I did some snapshots to get a better understanding how the image quality might suffer from cropping from 43mm to 75mm with the Leica Q3 43. I shoot always DNG & Jpeg at the same time. First I did 2 shots with the DNG setting to "L" (=60MP). I shot the scene first with 43mm and then with...

leicaqforum.com

 

[GeorgeHudetz]

So we're left with cropping. This can be a great tool to give more reach - as is used in the Q3 and SL3. I think this is where I'm starting to re-align my thinking on high Mp. Ideally the new S1ii will be a 60Mp camera but with the ability to crop to APSC at 24Mp or to down-sample the full frame image to something similar.

Yes, that is how I commonly use the 47 MP S1R for birding, but of course there is that tension with ISO since you often have to push the ISO to keep the SS high. But in decent light it's normally not an issue.

And yes, the 60 MP sensor should give both more cropping AND better high ISO, so it's a win-win.

BTW, I do think that Leica is doing something beyond just downsampling for the lower resolution images. They are doing pixel-binning, which (as I understand it) does an averaging of multiple adjacent pixels to increase SNR above and beyond what simple downsampling does. I think there is a review out there somewhere that looks into it's effectiveness.

And yes, I hope the S1RII (or whatever it is) does this. I know that manual pixel-binning is a technique used by some of the more advanced astrophotographers; it will be interesting to see how well the 36 or 18 MP "binned" images will work for astro.

 

[RuleOfThirds]

How does the hand held resolution mode of the S5ii stack up to the S1r? I printed my Malbork castle picture taken in this mode at 100cm and that print is absolutely gorgeous. I don’t know if the s5 has that feature. For action photos or birding this method is a no-go. But for your landscape type pictures it could be.

If I would want a higher mp body I would look at Fujifilm GFX.

 

[pdk42]

How does the hand held resolution mode of the S5ii stack up to the S1r? I printed my Malbork castle picture taken in this mode at 100cm and that print is absolutely gorgeous. I don’t know if the s5 has that feature. For action photos or birding this method is a no-go. But for your landscape type pictures it could be.

If I would want a higher mp body I would look at Fujifilm GFX.

The S5 had tripod hi res, as does the S1R. In fact, I took shots on a recent outing using the tripod hi res feature with all my three cameras (EM1.3, S5, S1R). They are all very impressive, especially the 190Mp file from the S1R, but they are all overkill for what I need.

As to HHHR, I’ve used it on Olympus cameras and on the S5ii and I’m sold on its ability to crush noise, but not on its ability to add much in the way of extra detail.

 

[pdk42]

Yes, that is how I commonly use the 47 MP S1R for birding, but of course there is that tension with ISO since you often have to push the ISO to keep the SS high. But in decent light it's normally not an issue.

And yes, the 60 MP sensor should give both more cropping AND better high ISO, so it's a win-win.

BTW, I do think that Leica is doing something beyond just downsampling for the lower resolution images. They are doing pixel-binning, which (as I understand it) does an averaging of multiple adjacent pixels to increase SNR above and beyond what simple downsampling does. I think there is a review out there somewhere that looks into it's effectiveness.

And yes, I hope the S1RII (or whatever it is) does this. I know that manual pixel-binning is a technique used by some of the more advanced astrophotographers; it will be interesting to see how well the 36 or 18 MP "binned" images will work for astro.

Pixel binding sounds like the way to go then. Having the ability to deliver 60Mp images will be great for some things, but 24 or 36 Mp files with lower noise will be great for when you need to raise the ISO.

 

[dirk]

We are really spoiled nowadays. 60MP fullframe sensors, high resolution mode, upscaling, pixel binning, antishake, autofokus, matrix metering etc.

It is almost impossible to take a photo which is not technically perfect and can not be printed 1m x 0,7m. Even with 6 year old cameras. Even with small sensor sizes.

I am sorry for Ansel Adams that he lived in the wrong century....

 

[GeorgeHudetz]

I did a little base-ISO comparison - using the DPR comparison tool low-light images - where I pushed base ISO images from the S1R, S5II, G9II, and Sony A7R5 (assuming it has the same sensor as the next ri-res body from Panasonic). Looked at all at 100% - no image reduction used. Using Capture One I pushed shadows 100%, and blacks 50%. It's pretty predictable:

S5II > Sony A7RV > S1R > G9II.

The (downward) jump from the S1R to the G9II was substantial - the biggest "delta" of the test. That G9II image did not hold together very well relative to the FF images.

The S1R & A7RV were fairly close, with the edge to the Sony, but the S5II was the only image that didn't show an obvious green tinge - grey stayed grey.

So yeah, similar results to your test.