S1II has been added to photonstophoto / looks promising

[Quentinquirelino]

Today, measurements for the new S1II has been added to the photonstophoto database. The measurements are looking pretty good. Not only it doesn't show the loss of DR in lower ISO due to the partially stacked sensor, like the Z6III does. It also is in line or slightly better compared to it's competition and predecessor.

Photographic Dynamic Range versus ISO Setting

Looks like the partially stacked sensor isn't the drawback in terms of image quality as many of us has been afraid of.

 

[xaviergut]

Thank you for the information. The people from cined.com published 4 days ago aswell the results if their Lab Test and are very positive:

Panasonic LUMIX S1II Lab Test, Part 2 - Dynamic Range and Latitude Using ProRes RAW 5.8K with DR Boost "ON" | CineD

Part 2 of our Panasonic LUMIX S1II lab test, using 5.8K ProRes RAW DR Boost "ON". Dynamic Range and exposure latitude are tested.

www.cined.com

"For the first time, we have a consumer full-frame camera that is on par with the ARRI Alexa Mini LF in our standardized latitude test. The LUMIX S1II exhibits a very robust image pipeline with a fine, pleasing noise pattern that survives after noise reduction when heavy shadow lifting is required. All of that without large color shifts. Hence, outperforming all other consumer full frame cameras, but also professional ones in completely different price brackets like the URSA Cine LF 12K, or the RED V-Raptor [X], or the Sony BURANO".

 

[HarisO]

Thank you for the information. The people from cined.com published 4 days ago aswell the results if their Lab Test and are very positive:

Panasonic LUMIX S1II Lab Test, Part 2 - Dynamic Range and Latitude Using ProRes RAW 5.8K with DR Boost "ON" | CineD

Part 2 of our Panasonic LUMIX S1II lab test, using 5.8K ProRes RAW DR Boost "ON". Dynamic Range and exposure latitude are tested.

www.cined.com

"For the first time, we have a consumer full-frame camera that is on par with the ARRI Alexa Mini LF in our standardized latitude test. The LUMIX S1II exhibits a very robust image pipeline with a fine, pleasing noise pattern that survives after noise reduction when heavy shadow lifting is required. All of that without large color shifts. Hence, outperforming all other consumer full frame cameras, but also professional ones in completely different price brackets like the URSA Cine LF 12K, or the RED V-Raptor [X], or the Sony BURANO".

Truly a Pocket ARRI. Lumix really been hitting it out of the park this year.

 

[pdk42]

Wow!

 

[CharlesH]

The people from cined.com published 4 days ago aswell the results if their Lab Test and are very positive:

I hadn't seen the cined report. This is serious stuff.

 

[Quentinquirelino]

Someone in the dpreview forum is currently taking a closer look on what's going on and compares it to the behavior of the Z6III. According to some of his first results, it seems Panasonic is using DR Boost in photo mode when using mechanical shutter. He compared the noise levels when using mechanical shutter and electronic shutter and it seems like the S1II produces a lot less noise in mechanical shutter but in e shutter it's finding are a very comparable to the Z6III.

 

[pdk42]

Someone in the dpreview forum is currently taking a closer look on what's going on and compares it to the behavior of the Z6III. According to some of his first results, it seems Panasonic is using DR Boost in photo mode when using mechanical shutter. He compared the noise levels when using mechanical shutter and electronic shutter and it seems like the S1II produces a lot less noise in mechanical shutter but in e shutter it's finding are a very comparable to the Z6III.

Sounds similar to the G9ii behaviour.

 

[Quentinquirelino]

Panasonic has confirmed, the S1ii is using DR Boost in some photo modes. Petrapixel Poster an article about the matter and directly asked some Panasonic representatives.

Quote:
“DRBoost is active during stills shooting in some cases. While we cannot specify the detailed conditions under which it works at this time, we can confirm this functionality provides the increased Image Quality performance observed while capturing stills with the LUMIX S1II.”

The Panasonic S1 II Is Secretly a Groundbreaking Full-Frame Camera

The magic of dual gain output.

petapixel.com

 

[GeorgeHudetz]

This is quite interesting (Horshack's video is worth a watch). So the S1II should give noticeably better raised shadows than the S1RII, at least on very high DR scenes, and at base ISO. But, it's not clear yet how significant the advantage would be for scenes where you push the shadows more like 3-4 stops, rather than the 8 stops that Horshack used. I suppose downloading some of the DPR test shots could prove insightful there.

It's also interesting the Panasonic has chosen not to disclose right now when DGO is active, although Horshack does a good job describing it. Seems like Panasonic would want to market this feature. Perhaps they are waiting until they deliver a bit more refinement in the menus, such as a DGO on/off switch, or making it available with ES - particularly with HR mode. Imagine that kind of resolution coupled with bullet-proof shadows!

I also wonder if DGO could be enabled on the S1RII, given that it (presumably) uses the same ISP as the S1II, and would - again presumably - have the additional logic needed to support the feature, as Horshack alluded to. Of course, the higher resolution files would take more processing power to accomplish the blend, so perhaps it's not practical. But I would certainly embrace a "High DR" mode - which would sit adjacent to the "High Resolution" mode - that would require user activation and an acknowledgement that shot-to-shot times might suffer a bit.

Anyway, I have validated that on the S1RII, a +5 stop push at ISO 640 gives better shadows than a +8 stop push at ISO 80, which is as-expected. It would be interesting to try to blend an ISO 80 shot with an ISO 640 shot in PS using luminosity masks or some such. Not sure how hard that would be. Too bad ISO bracketing is not supported.

 

[CharlesH]

It would be interesting to try to blend an ISO 80 shot with an ISO 640 shot in PS using luminosity masks or some such. Not sure how hard that would be.

I think Horshack did that with a program he wrote in Matlab. Perhaps if you sent him RAW files he could so that simulation.

Seems like Panasonic would want to market this feature.

Sometimes if you have better performance than competitors you don't want to disclose how you do it. But with Horshack letting the cat our of the bag, Panasonic may change their marketing approach.

 

[Andreas]

Sounds similar to the G9ii behaviour.

No, the G9ii results are essentially identical for the electronic and mechanical shutters. Bill Claff's Photons to Photos site initially reported a substantial difference, but this was due to a testing error. He was eventually convinced to redo the results with new test images, and his revised results show no difference. This correction came after several months and the damage was already done. Most people read of the initial test results but missed the correction. Here is a link to a mea culpa by PetaPixel, which explains the change in Bill Claff's analysis and provides the updated results:

G9ii Dynamic Range Performance

The new results seem to be a well-kept secret!

It's amazing to me that none of the reviewers who reported these results ran any tests of their own to confirm that a difference exists. Given the rather substantial difference in the initial report they should have been able to see photographic evidence with a few tests. When they failed to find a difference--because there isn't one--they would then be left to question either the accuracy of the results or the relevance of the test. Either way it might have led to an interesting discussion, and perhaps a more timely correction.

 

[GeorgeHudetz]

I think Horshack did that with a program he wrote in Matlab. Perhaps if you sent him RAW files he could so that simulation.

Yeah, I think he did use Matlab. Seems like you could do it in PS as well - basically, raise shadows on both in a consistent manner (which means less raising on the high ISO shot) and then create a mask for the high ISO shot that covers a certain luminosity range on the low end. Then copy the image data in that mask and paste it into the base ISO shot. Blending could get tricky. It's unclear to me if one could determine a repeatable set of steps that work for most images (i.e., avoiding tweaking it for each image).

But then, Panasonic figured out a way to do it using a far less powerful processor, so it seems like it could be done. But maybe it's easier to just do a traditional exposure bracket.

Sometimes if you have better performance than competitors you don't want to disclose how you do it. But with Horshack letting the cat our of the bag, Panasonic may change their marketing approach.

Yeah, I wondered about that too. And, I suspect lots of people are now going to try to trip it up - look for obvious signs of blending, etc. Can't have Panasonic showing the best IQ!

 

[CharlesH]

Yeah, I think he did use Matlab.

Sadly I've never used Matlab; I do wish I had experience with it. The basic price is something over $2,000, although there is a home (personal) version for only $149.

It would be great to have an AI program that would script Photoshop to do what you are asking. In my case whenever I set out to do something different in Photoshop it's like I have to learn Photoshop all over again. But the trend in programming is to have AI write the program. I expect there is work to have AI write scripts for Photoshop, but I don't know if there is anything up to what you are asking for.

 

[GeorgeHudetz]

So, here is a comparison via a screenshot from Capture One. Both shots are from an S1RII, 1/30 of a second, F8. All NR & sharpening turned off. View is at 200% on a MBP.

Top shot is ISO 80 with the shadows pushed 5.3 stops. Bottom shot is ISO 400 (which is the second native ISO for the S1RII) but with the shadows pushed 3 stops, which gives an equal exposure to the ISO 80 shot.

It's a pretty significant difference, IMO. Somebody on DPR said they'd pay for a firmware update/extension to the S1RII to get this, and I tend to agree. With the default NR switched on, the ISO 400 shot looks nearly noise-free, even after pushing 3 stops.

 

[GeorgeHudetz]

OK, that was a crappy crop to get it to fit into the upload limitations. Here is a better screen grab, but with default NR on in C1.

 

[pdk42]

Thanks for doing this George. Very interesting result.

 

[Babylonia]

@GeorgeHudetz

Really interesting testing of the files by your new S1RII camera body, by two different "basic" ISO settings, ISO 80 versus ISO 400.
And pushing up the exposure levels of the shadows by software of Capture One.

However Capture One is a very nice RAW converter in general, (I also do use it).
At the level of noise reduction, it is outdated by more new RAW converters, having specialised algorithms for noise reduction.

To get a better output for high ISO images, (or pushing up the shadow areas),
not long ago I purchased DXO RAW / Photolab "8" (+ extra edit functionalities).
But seems, there is already a new update Photolab "9": https://www.dxo.com/dxo-photolab/

I got an update of my Photolab "8.8" just these days.



Looking to the Panasonic S1RII
I could not find an update "yet" for the so called "including those from 5 generation sensors." for the latest Photolab version.

Unterstützte kameras/objektive - DxO

DxO PhotoLab und DxO PureRAW verarbeiten Bilder im RAW-Format von über 500 Digitalkameras, die in unseren Labors kalibriert wurden.

www.dxo.com

However I get wonderful high ISO quality by my own used Lumix S1R camera, going far beyond what is possible by Capture One.
Maybe you can get a free demo working version for 30 days of Photolab 9
To experiment what it is bringing to your Lumix S1RII camera.

(If necessary, you can contact me via a private message to provide your RAW files,
so I can edit them here in DXO Photolab 8.8 for comparison with what you have done in Capture One).

For best result, you should be making a "mix" between what can be performed by the "DeepPrime 3 (or version 2") noise reduction,
that can be looking a bit "over smooth" as a separate file, and blending in an untouched layer file.
Or add a minimal extra "film" noise after the noise reduction is done, so less "clinical".

And keep in mind some extra attention to use a more small "enlarging" box to see the result of this "DeepPrime 3" noise reduction
in a more small area of the image. As the full image takes to far CPU time to process the full image immediately.
(All is processed at the end when saving a file, as e.g. done by Capture One to).



-