Lumix S1II Impressive Peak Dynamic Range Performance

[Interceptor121]

There is no DGO in photo

There's no DGO option in photo mode. But that doesn't mean it's not using it. GH7 and G9II enable DGO in photo mode by default without any DR Boost option either.

The PDR chart looks quite like how GH7/G9II behaves.

The ISO-Invariance test results of ISO100+6EV and ISO200+5EV are not in the same league compared to basically all other FF cameras. I don't think it's only due to some fancy NR method.

I'm really curious what's really happening. And it's weird why they don't market it.

I dont believe is using it as that comes with 16 bits raw files

 

[iso rivolta]

It's very likely that S1 II is using Sony's DOL option (Digital Overlap) with frame rates of 30 fps and lower, so it includes here all images taken with mechanical shutter. DOL is basically reading each row twice and is decreasing the maximum framerate to half. I think you should test and compare the mechanical shutter results with those taken at 20-30 fps and at 60-70 fps with electronic shutter, even if Panasonic is mentioning DR Boost only for video.

 

[Horshack]

It's very likely that S1 II is using Sony's DOL option (Digital Overlap) with frame rates of 30 fps and lower, so it includes here all images taken with mechanical shutter. DOL is basically reading each row twice and is decreasing the maximum framerate to half. I think you should test and compare the mechanical shutter results with those taken at 20-30 fps and at 60-70 fps with electronic shutter, even if Panasonic is mentioning DR Boost only for video.

I've been examining the GH7's DR boost implementation. It works with both the electronic and mechanical shutter, which I believe excludes the possibility of DOL since DOL requires two overlapping rolling shutter readouts, which isn't possible with a mechanical shutter. Instead I think it's using a single-exposure dual conversion gain output, where a single integration time is readout twice at the two supported conversion gains and then blended together. This scheme would work with both electronic and mechanical shutters, provided the readout at both gains is concurrent, or each can be done sequentially in a non-destructive way.

I'm guessing the S1 II's DR boost operates the same way as the GH7, albeit only used for video. All the evidence I've seen so far indicates the S1 doesn't support DR boost for stills and is achieving its still DR advantage with NR. I have an S1 II on the way and have some experiments planned which should tell me for sure.

 

[Interceptor121]

It's very likely that S1 II is using Sony's DOL option (Digital Overlap) with frame rates of 30 fps and lower, so it includes here all images taken with mechanical shutter. DOL is basically reading each row twice and is decreasing the maximum framerate to half. I think you should test and compare the mechanical shutter results with those taken at 20-30 fps and at 60-70 fps with electronic shutter, even if Panasonic is mentioning DR Boost only for video.

Totally unlikely as the sustained frame rate keeps the bit depth and the raw files are 14 bits while the gh7 are 16

 

[iso rivolta]

It's very likely that S1 II is using Sony's DOL option (Digital Overlap) with frame rates of 30 fps and lower, so it includes here all images taken with mechanical shutter. DOL is basically reading each row twice and is decreasing the maximum framerate to half. I think you should test and compare the mechanical shutter results with those taken at 20-30 fps and at 60-70 fps with electronic shutter, even if Panasonic is mentioning DR Boost only for video.

I've been examining the GH7's DR boost implementation. It works with both the electronic and mechanical shutter, which I believe excludes the possibility of DOL since DOL requires two overlapping rolling shutter readouts, which isn't possible with a mechanical shutter.

Here I just got carried away. I'm not sure if there could be a way to combine a mechanical shutter with two electronic exposures but the S1 II rolling shutter is very slow compared to the 1/250 s of the mechanical shutter travel time, so you're right.

Instead I think it's using a single-exposure dual conversion gain output, where a single integration time is readout twice at the two supported conversion gains and then blended together. This scheme would work with both electronic and mechanical shutters, provided the readout at both gains is concurrent, or each can be done sequentially in a non-destructive way.

I'm guessing the S1 II's DR boost operates the same way as the GH7, albeit only used for video. All the evidence I've seen so far indicates the S1 doesn't support DR boost for stills and is achieving its still DR advantage with NR. I have an S1 II on the way and have some experiments planned which should tell me for sure.

Bil Claff is testing for noise reduction using a 2D Fast Fourier transform to check for a spatial relationship between pixels. He showed no NR until ISO 6400 using this method. Another type of noise reduction?

 

[Horshack]

It's very likely that S1 II is using Sony's DOL option (Digital Overlap) with frame rates of 30 fps and lower, so it includes here all images taken with mechanical shutter. DOL is basically reading each row twice and is decreasing the maximum framerate to half. I think you should test and compare the mechanical shutter results with those taken at 20-30 fps and at 60-70 fps with electronic shutter, even if Panasonic is mentioning DR Boost only for video.

I've been examining the GH7's DR boost implementation. It works with both the electronic and mechanical shutter, which I believe excludes the possibility of DOL since DOL requires two overlapping rolling shutter readouts, which isn't possible with a mechanical shutter.

Here I just got carried away. I'm not sure if there could be a way to combine a mechanical shutter with two fast electronic readouts but the S1 II rolling shutter is very slow compared to the 1/250 s of the mechanical shutter travel time, so you're right.

Instead I think it's using a single-exposure dual conversion gain output, where a single integration time is readout twice at the two supported conversion gains and then blended together. This scheme would work with both electronic and mechanical shutters, provided the readout at both gains is concurrent, or each can be done sequentially in a non-destructive way.

I'm guessing the S1 II's DR boost operates the same way as the GH7, albeit only used for video. All the evidence I've seen so far indicates the S1 doesn't support DR boost for stills and is achieving its still DR advantage with NR. I have an S1 II on the way and have some experiments planned which should tell me for sure.

Bil Claff is testing for noise reduction using a 2D Fast Fourier transform to check for a spatial relationship between pixels. He showed no NR until ISO 6400 using this method. Another type of noise reduction?

His testing found the same frequency-based filtering as my measurements but he didn't initially interpret it as NR. I think he might be revisiting the data to consider marking the results as NR.

 

[FOTONOTO]

It might be DCG-HDR in the ISO100~400 range, which is a single-frame HDR method, already been used in smartphones for some time. One exposure, two readouts, one from high well capacity mode, another from low capacity mode. Since there's still only one circuit, one ADC, the readout time doubles.

GH7 and G9II on the other hand, are using DAG-HDR. One exposure, two parallel analog amp circuits. Readout time stays the same, but the bandwidth doubles.

I don't have a technical background. This is just a guess according to a whole day of searching. But no matter what, I found it hard to believe it's only due to NR.

 

[Interceptor121]

It might be DCG-HDR in the ISO100~400 range, which is a single-frame HDR method, already been used in smartphones for some time. One exposure, two readouts, one from high well capacity mode, another from low capacity mode. Since there's still only one circuit, one ADC, the readout time doubles.

GH7 and G9II on the other hand, are using DAG-HDR. One exposure, two parallel analog amp circuits. Readout time stays the same, but the bandwidth doubles.

I don't have a technical background. This is just a guess according to a whole day of searching. But no matter what, I found it hard to believe it's only due to NR.

Google or AI are not the truth

Phones have quad bayer cells and can do what you suggest but this camera doesn’t have quad bayer sensor so it cant

 

[FOTONOTO]

It might be DCG-HDR in the ISO100~400 range, which is a single-frame HDR method, already been used in smartphones for some time. One exposure, two readouts, one from high well capacity mode, another from low capacity mode. Since there's still only one circuit, one ADC, the readout time doubles.

GH7 and G9II on the other hand, are using DAG-HDR. One exposure, two parallel analog amp circuits. Readout time stays the same, but the bandwidth doubles.

I don't have a technical background. This is just a guess according to a whole day of searching. But no matter what, I found it hard to believe it's only due to NR.

Google or AI are not the truth

Phones have quad bayer cells and can do what you suggest but this camera doesn’t have quad bayer sensor so it cant

Who said it has to use quad bayer cells to achieve DCG-HDR?

 

[Interceptor121]

It might be DCG-HDR in the ISO100~400 range, which is a single-frame HDR method, already been used in smartphones for some time. One exposure, two readouts, one from high well capacity mode, another from low capacity mode. Since there's still only one circuit, one ADC, the readout time doubles.

GH7 and G9II on the other hand, are using DAG-HDR. One exposure, two parallel analog amp circuits. Readout time stays the same, but the bandwidth doubles.

I don't have a technical background. This is just a guess according to a whole day of searching. But no matter what, I found it hard to believe it's only due to NR.

Google or AI are not the truth

Phones have quad bayer cells and can do what you suggest but this camera doesn’t have quad bayer sensor so it cant

Who said it has to use quad bayer cells to achieve DCG-HDR?

It’s the only way to do that with a single shot unless you use dgo which has already been ruled out

 

[FOTONOTO]

It might be DCG-HDR in the ISO100~400 range, which is a single-frame HDR method, already been used in smartphones for some time. One exposure, two readouts, one from high well capacity mode, another from low capacity mode. Since there's still only one circuit, one ADC, the readout time doubles.

GH7 and G9II on the other hand, are using DAG-HDR. One exposure, two parallel analog amp circuits. Readout time stays the same, but the bandwidth doubles.

I don't have a technical background. This is just a guess according to a whole day of searching. But no matter what, I found it hard to believe it's only due to NR.

Google or AI are not the truth

Phones have quad bayer cells and can do what you suggest but this camera doesn’t have quad bayer sensor so it cant

Who said it has to use quad bayer cells to achieve DCG-HDR?

It’s the only way to do that with a single shot unless you use dgo which has already been ruled out

I thought the same before my searching. I thought DCG-HDR means that, in each pixel group of a quad bayer sensor, different pixels are using different conversion gain, so it sacrifices resolution in exchange for DR. But turns out it's just one way to do DCG-HDR. The other way is that you can actually read out twice sequentially with different conversion gain from every single pixel of a single exposure. And all of those happen before entering the analog circuit. So compared to DAG-HDR (which needs two parallel analog circuits, two ADCs), it only needs one analog circuit, but the price is longer readout time. I don't understand how it's done though.

Again, I don't have a technical background, so it could be all wrong. But by "searching" I don't mean just asking AI or reading the 1st page of google results. That doesn't need a day.

 

[Interceptor121]

It might be DCG-HDR in the ISO100~400 range, which is a single-frame HDR method, already been used in smartphones for some time. One exposure, two readouts, one from high well capacity mode, another from low capacity mode. Since there's still only one circuit, one ADC, the readout time doubles.

GH7 and G9II on the other hand, are using DAG-HDR. One exposure, two parallel analog amp circuits. Readout time stays the same, but the bandwidth doubles.

I don't have a technical background. This is just a guess according to a whole day of searching. But no matter what, I found it hard to believe it's only due to NR.

Google or AI are not the truth

Phones have quad bayer cells and can do what you suggest but this camera doesn’t have quad bayer sensor so it cant

Who said it has to use quad bayer cells to achieve DCG-HDR?

It’s the only way to do that with a single shot unless you use dgo which has already been ruled out

I thought the same before my searching. I thought DCG-HDR means that, in each pixel group of a quad bayer sensor, different pixels are using different conversion gain, so it sacrifices resolution in exchange for DR. But turns out it's just one way to do DCG-HDR. The other way is that you can actually read out twice sequentially with different conversion gain from every single pixel of a single exposure. And all of those happen before entering the analog circuit. So compared to DAG-HDR (which needs two parallel analog circuits, two ADCs), it only needs one analog circuit, but the price is longer readout time. I don't understand how it's done though.

Again, I don't have a technical background, so it could be all wrong. But by "searching" I don't mean just asking AI or reading the 1st page of google results. That doesn't need a day.

Yes you don’t have a technical background and it shows

 

[ATP62]

Pretty impressive :



_______

https://www.dpreview.com/reviews/image-comparison/fullscreen?attr134_0=nikon_z6iii&attr134_1=panasonic_dcs1ii&attr134_2=panasonic_dcs1r&attr134_3=canon_eosr5ii&attr136_0=1&attr136_1=1&attr136_2=1&attr136_3=1&attr176_3=efc&attr403_0=1&attr403_1=1&attr403_2=1&attr403_3=1&normalization=print&widget=962&x=-0.9699355476699227&y=-0.6810968541718139