LUMIX ambassador: S1R-II Dynamic Range measures are wrong

[Interceptor121]

....

While with my Sony I would trust the camera to recover colors underexposed I would not do it with the Panasonic S1RII

That is directly at odds with Richard Wong's tests, published in his S1Rii review on his YouTube channel. He compared the S1Rii to the 60mp Sony sensor in his A7CR, as well to the S5ii and original S1R. His test images were underexposed (by 7 stops and 8 stops), and processed in Capture One. He then recovered the shadows, and compared them at a common resolution. The images from his Sony camera had more color shift and more banding than the images from the S1Rii, so the opposite of what you are claiming.

While the S1Rii did better than the Sony, the S5ii did the best overall, although it was only a bit better than the S1Rii. In the past I have always found Richard's tests to be reliable, in the sense that whenever I try to repeat them I get very similar results.

Richard is a nice guy but studio scenes only measure snr

the photonstophotos tests is looking at snr ar latitude with quantitative measurements not you choosing where to look

landscapes are different as they look at scenes with lots or light and darks

to evaluate that at test level you need to look at p2p input referred noise

There you can see that the gap between the s1r2 and the a7cm2 is 0.8 ev of which 0.7 ev is due to higher read noise and 0.1 to saturation however the s1r2 meter is 0.3 up so this shows equally 0.4 ev each side weaker

To see this easily you need to take backlit shots as those I have taken at the time

OK, let's see them! Show us a case where the shadow recovery of the S1RII is problematic relative to the a7RV.

Am not sure what is shadow recovery did someone get lost in the shadows? If it is you bring a torch. Perhaps you are referring to shadow improvement with ISO? This is not what I am talking about here. In a scene that clips there is nothing you can do with exposure you can only lift shadows in certain part of the image so playing with i take a shot 3 stops down on the dpreview scene and then i lift it up is pointless. That is a low dynamic range scene is not a proxy for a real landscape where you will be near clipping with nowhere to go

Now a few considerations on shadow improvement and why playing with dpreview scene is not a proxy for DR or else

1. The studio scene is affected by the camera metering and cameras have different offset. In short ISO 100 on camera A may be different from camera B

2. Camera base ISO also differ in the specific example the S1RII has base ISO 80

3. Pixel peeping on an image at 100 or 200 is impacted by pixel size you should scale the images to the same resolution using tiff however this will be impacted by adobe processing of the raw; the pdr test uses non demosaiced data and therefore is robust compared to adobe

4. Shadow improvement is impacted by dual gain and by general strategy of the camera maker. In the specific Sony make pretty much ISO invariant cameras i.e. other than stepping out dual gain nothing changes with ISO. Panasonic instead is not so ISO invariant. Sony camera do not change behaviour with ISO by design

5. Ultimately this does not matter because exposure is a relative concept what matter is the range where SNR is acceptable and the camera does not clip you the human being make a choice in a landscape shot not the camera meter

6. You are taking a visual examination of a sample of your choice you are not evaluating the entire frame. So your evaluation is not objective.

So that should be sufficient to send to the bin your attempt and others to play with dpreview scene. That is just meant to give an idea of image quality however is impact by many factor sand the lens is certainly one of those. Some site test a camera with the same lens using adapters to eliminate the bias but lets not go there

There is one website that is trustworth on camera review from a technical point of view and that is optyczne let's hope they review the S1R2 . They also test matrix resolution which is something nobody does anymore as well as tonal depth and long exposures.

Now what you want to do to see how a camera really behaves without going to great lenghts and with one or two simply shots. Place a person in front of a window on a complete overcast day where the sky looks white and take a shot with the highlights not clipping. You will need to bracket on the S1R2 as the zebra are not realiable. in addition you need to consider that lightroom is pushing up exposure already on its own so to understand if it really clip you need something like rawdigger.

then in post adjust lifting shadow and highlights you can play with the exposure but is not a situation of 1 stop or similar. Once you have finished see what you get. Thats is a dynamic range test. While you playing with the dpreview scene is just an intrascene latitude test at best and is not objective as you are not averaging the scene you are looking where you prefer to do

You can appreciate that your theory is quite flawed and you want to argue with the P2P testing method with those arguments? You are basically going in battle naked with a stick while the others are using drones and lasers

Again, show us your S1RII test image that convinced you that you cannot trust it "to recover colors underexposed."

Let's see it, side by side with the Sony image.

Sorry you brought in the comparison scene not me this not my requirement I was just explaining to you that do not have the means how to do it. But this is still a craft method it does not compare with the accurate testing already done it was just to show the point your method is pointless am not suggest this is a good method

The data is there and is accurate until proven differently. Have you sent the files to Bill? What did he say? I guess he said they are identical or worse and therefore you already know the answer

You cannot fight a detail test scene with silly argument the trick did not work you are just failing on the whole field time to retire

On that note I go and watch some tv farewell

Ah, I see, no photographic evidence of your extraordinary claim. As I expected.

p.s. I have sent files to Bill. And: I am in no way saying that the current data is incorrect, or that Bill is incorrect. Only that your interpretation of the practical meaning of that data is incorrect, or at least greatly exaggerated.

Sorry where is the exaggeration? There are 2/3 stops difference at base ISO that is a fact

I also said the camera does very well at high gain (ISO 400-5000) before noise reduction ruins it

I never said is crap I said considering how it goes at high gain it is disappointing what happens at base ISO

That the camera is 0.5 Ev better at ISO 5000 it does not matter much to me to be fair as 70% of my photos are ISO 100 and 500

This camera could be interesting to someone shooting relatively high shutter speeds in low light for sports the autofocus kills it but for nightime street it should be very good and of course studio work product photography still life

While look elsewhere for sport and action nature and wildlife and landscape more options elsehwere there

anyway my hope was for a top notch landscape camera and that did not happen the high speed action was always going to be problematic I was not really expecting anything there

I hope panasonic sorts out the hum at base ISO it seems very very strange what goes on

 

[Melidonia]

For what it's worth and after some weeks with the S1RII, I think Interceptor is mostly right. The dynamic range is clearly not the best, the highlights clip very soon, to the point I thought I had a defective camera at first. Sure, the shadows are better but still, I think the dynamic range is worse compared to the A7CII or S5II.

There is not a big difference to the S5II actually, 0.13 Ev is difficult to see

But there is 2/3 ev to the A7C II

Panasonic clipping is due to their choice of eliminating highlight protection however this can be fixed by the user

https://eww.pavc.panasonic.co.jp/dscoi/DC-S1RM2/html/DC-S1RM2_DVQP3245_eng/0129.html#00013

Set -0.3ev in all modes and it will behave like Sony however this will impact jpegs and you need then to work out what to do in your raw processor because that will also expect a bright image to start

The issue however is that zebra maximum level is 95% if you set them for highlight warning you may get quite a lot of false positive for clipping

Indeed the max value on the S1RII is 95% compared to the S5II, but we can select the option "base/range" in the zebras settings to select the value we want (like 100 or 105%), this is the option just below the "50%" value in the bottom of the zebra settings, I chose 98% but I'm not sure it is accuracte.

100% is what you need if the video luminance is 16-235

Thank you, I use 0-255 with Cine A2 or Standard mostly. Which zebra value is the best for photo to prevent highlights clipping ?

 

[Melidonia]

For what it's worth and after some weeks with the S1RII, I think Interceptor is mostly right. The dynamic range is clearly not the best, the highlights clip very soon, to the point I thought I had a defective camera at first.

How has the highlight recovery been for you? I have found it to be quite flexible:

Capture One, all defaults:

With highlights reduced:

Overall, I find the flexibility of highlights to be similar to my S1R, which I always considered quite good.

Shadows - and in particular deep blacks - on the S1RII are much more pliable than on the S1R, however.

Sure, the shadows are better but still, I think the dynamic range is worse compared to the A7CII or S5II.
The AF in both photo and video is really not on part with Sony, sometimes I even think the S5II is better, I hope it will be improved with a firmware update.

The IBIS is stellar, the colors are good, I always prefer Panasonic colors over Sony. The overall colors and WB are better controlled than the S5II most of the time, however, the tint is maybe a bit too green and I was surprised to prefer the colors of the S5II a lot of time when I compared both. The new A2 profile for video is really great though.

Indeed I tried Capture One and it makes a big difference compared to ACR even with the last update. However I still struggle with the highlights clipping in video when using most profiles. Only V-log with DRE-ON is good.

 

[Interceptor121]

For what it's worth and after some weeks with the S1RII, I think Interceptor is mostly right. The dynamic range is clearly not the best, the highlights clip very soon, to the point I thought I had a defective camera at first. Sure, the shadows are better but still, I think the dynamic range is worse compared to the A7CII or S5II.

There is not a big difference to the S5II actually, 0.13 Ev is difficult to see

But there is 2/3 ev to the A7C II

Panasonic clipping is due to their choice of eliminating highlight protection however this can be fixed by the user

https://eww.pavc.panasonic.co.jp/dscoi/DC-S1RM2/html/DC-S1RM2_DVQP3245_eng/0129.html#00013

Set -0.3ev in all modes and it will behave like Sony however this will impact jpegs and you need then to work out what to do in your raw processor because that will also expect a bright image to start

The issue however is that zebra maximum level is 95% if you set them for highlight warning you may get quite a lot of false positive for clipping

Indeed the max value on the S1RII is 95% compared to the S5II, but we can select the option "base/range" in the zebras settings to select the value we want (like 100 or 105%), this is the option just below the "50%" value in the bottom of the zebra settings, I chose 98% but I'm not sure it is accuracte.

100% is what you need if the video luminance is 16-235

Thank you, I use 0-255 with Cine A2 or Standard mostly. Which zebra value is the best for photo to prevent highlights clipping ?

It clips at 109% normally if you use all bitcodes

 

[PhotoHawk]

I'm coming late to this thread. However, there are a few things to mention.

First Bill Claff's site is an excellent source of information. Thanks Bill! But we should understand that Bill is using manufacturers ISO. He does not attempt to normalize them. So when comparing cameras, particularly from different makers, while the shape of the DR curve won't change the shift to the left or right relative to another camera, can change if the curves are normalized. How much of a shift depends on how the manufacturer measures ISO.

Second with regard to the SR1MKII we do not have a readily available normalized ISO measurement. At the DXO Mark site we can see (for example to take one point for discussion )that the in the DR curve comparison of the A7RV and the S1R that the values at a manfacturers value of 800 are 681 measured for the S1R and 589 for the A7RV. DXOMark uses measured ISO - their measured ISO. That difference can shift Bill's curves a bit.

Third note that we can't assume that the S1R MkII is similar to the S1R Mk1. If you note the S1 ISO curve its measured ISO at a manufactuers ISO of 800 is 568. This is quite a bit different than the S1R. Thus there is reason to believe that the normalized ISO of the S1RMkII may not be the same as the S1R. Thus the difference shown in Bill's curves between the S1RMkii and the A7R5 could be less if normalized ISO's are used.

Lastly the difference doesn't likley matter to the vast majority of users. Current advanced HDR computer monitors can display about 10 stops and a print is less. So with proper exposure and good NR in the shadows as the ISO gets higher any current model camera (Z8, S1RMk2 or A7R5) should be good for the vast majority of users.

 

[Interceptor121]

....

While with my Sony I would trust the camera to recover colors underexposed I would not do it with the Panasonic S1RII

That is directly at odds with Richard Wong's tests, published in his S1Rii review on his YouTube channel. He compared the S1Rii to the 60mp Sony sensor in his A7CR, as well to the S5ii and original S1R. His test images were underexposed (by 7 stops and 8 stops), and processed in Capture One. He then recovered the shadows, and compared them at a common resolution. The images from his Sony camera had more color shift and more banding than the images from the S1Rii, so the opposite of what you are claiming.

While the S1Rii did better than the Sony, the S5ii did the best overall, although it was only a bit better than the S1Rii. In the past I have always found Richard's tests to be reliable, in the sense that whenever I try to repeat them I get very similar results.

Richard is a nice guy but studio scenes only measure snr

the photonstophotos tests is looking at snr ar latitude with quantitative measurements not you choosing where to look

landscapes are different as they look at scenes with lots or light and darks

to evaluate that at test level you need to look at p2p input referred noise

There you can see that the gap between the s1r2 and the a7cm2 is 0.8 ev of which 0.7 ev is due to higher read noise and 0.1 to saturation however the s1r2 meter is 0.3 up so this shows equally 0.4 ev each side weaker

To see this easily you need to take backlit shots as those I have taken at the time

You are back to arguing total DR. The statement to which I was responding had to with recovering colors from underexposed portions of an image, which is relevant in many different contexts, not just cases where you have a high DR landscape that is pushing the limits of the sensor. For example, you are shooting wide open with a modern wide angle lens that has 5-6 stops of falloff in the corners, and your exposure leaves the corners darker than is optimal because of a bit of an exposure error (the highlights are well within range). Mistakes like these happen. How well can you recover the underexposed corners? The S1Rii will do a bit better than the Sony, because it is SNR that matters. This is a far more common scenario than the case you are imagining, where you have a high-DR landscape where the DR just happens to fall right within the delta between the DR of the two cameras. That's the only case where the Sony would be at an advantage, and then only in the sense that you would only need a single exposure. Mountains out of molehills.

Mistakes like that do not happen you review your photos and correct

Instead if the camera runs out of DR there is nothing to correct

An you got your theory wrong SNR and DR are the same thing PDR is the number of stops at SNR=20

The DxoMark index DR is SNR=1

I does help to know what you are actually talking about before coming out with stuff

Latitude is the tolerance to exposure errors this depends from other factors than SNR