Deep dive into DNG settings and Pixel Binning capabilities of Q3…

[LarsOtium]

Advertisement (gone after registration)

 

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[John A Fleming]

My head was spinning - I’m not a statistician! But I believe I understood Mathphotographer’s findings. 

• Do the pertain to the M11 since the sensor is similar? The next questions pertain to the M11.
• M-DNG is better than H-DNG = 1 more stop of dynamic range = better hi ISO performance (less noise)?
• If 2 is true, it contradicts other reviewers, including Mathphotographer, statements that there’s no difference in photo quality between H-DNG (60mp) and the other two (M-DNG, L-DNG). What caused the change?
• The best usable ISO is 64 to 25K?

Thanks,

John Fleming

[Miltz]

No one else tested it like this that’s why those reviewers saw no change. Most videos or articles were just promotional pieces for Leica anyway. Furthermore you need very specific conditions to see the one stop difference. Glad he tested it and now we know. 

[Qwertynm]

Regarding dynamic range. Whether true or not, has anyone any images to share where you can see that one stop difference?

I quickly shot 2 images out the window and see no discernible difference. The histogram is slightly different but then again I didn’t use a tripod so probably just got more of a dark area in one over the other. 

[SrMi]

A couple of comments:

- There is no pixel binning in M11 or Q3; raw size reduction is made with smart firmware algorithms (Leica's PR department messed it up initially but removed later all references to pixel binning from their websites).

- When looking at Lightroom and whether any modifications are done, one should also look in the Detail panel as NR and sharpening is often applied automatically.

- We do not know when DCG (dual conversion gain) kicks in with Q3. If it is the same as M11, it would be ISO 200. Most new cameras are pretty ISO invariant above the DCG point. I wonder what the DCG point of Q3 is. Note that Q3's base ISO is 100, but the video tested only the range from 200 upwards. Therefore the video does not show that the Q3's sensor is ISO invariant. However, I think it is practical to have DCG point early at ISO 200 (larger ISO invariant range).

- Leica has said that in M11, lower resolutions have the same DR as the large resolution once the raw images have been reduced to the same size in the post. I assume that is the case with Q3 as well. You should see the noise difference only if you compare at different resolutions. That means that the benefit of smaller resolution is not DR (you can replicate it by downsizing) but the smaller raw file only.

- The video apparently looks at the noise of JPEGs, not at DNGs. JPEGS include any NR in the camera and any default NR in LrC. 

I did not watch it to the end.

 

Edited June 20, 2023 by SrMi

[clifak]

22 minutes ago, SrMi said:

A couple of comments:

- There is no pixel binning in M11 or Q3; raw size reduction is made with smart firmware algorithms (Leica's PR department messed it up initially but removed later all references to pixel binning from their websites).

- When looking at Lightroom and whether any modifications are done, one should also look in the Detail panel as NR and sharpening is often applied automatically.

- We do not know when DCG (dual conversion gain) kicks in with Q3. If it is the same as M11, it would be ISO 200. Most new cameras are pretty ISO invariant above the DCG point. I wonder what the DCG point of Q3 is. Note that Q3's base ISO is 100, but the video tested only the range from 200 upwards. Therefore the video does not show that the Q3's sensor is ISO invariant. However, I think it is practical to have DCG point early at ISO 200 (larger ISO invariant range).

- Leica has said that in M11, lower resolutions have the same DR as the large resolution once the raw images have been reduced to the same size in the post. I assume that is the case with Q3 as well. You should see the noise difference only if you compare at different resolutions. That means that the benefit of smaller resolution is not DR (you can replicate it by downsizing) but the smaller raw file only.

- The video apparently looks at the noise of JPEGs, not at DNGs. JPEGS include any NR in the camera and any default NR in LrC. 

I did not watch it to the end.

 

Mathphotographer used RAW DNG files out of the camera without any post processing for is in-depth SNR tests. He details that and the rest of the methodology in the video. Initially when he compared 3200 and 200 ISO images in PixInsight he used a JPG that was converted from the RAW DNG from the camera but that was part of his initial testing, not the detailed testing when he later used for S-DNG, M-DNG, and L-DNG images straight from the camera at both 3200 and 200 ISO.

[beewee]

Advertisement (gone after registration)

53 minutes ago, SrMi said:

- We do not know when DCG (dual conversion gain) kicks in with Q3. If it is the same as M11, it would be ISO 200. Most new cameras are pretty ISO invariant above the DCG point. I wonder what the DCG point of Q3 is. Note that Q3's base ISO is 100, but the video tested only the range from 200 upwards. Therefore the video does not show that the Q3's sensor is ISO invariant. However, I think it is practical to have DCG point early at ISO 200 (larger ISO invariant range).

High gain should kick in at ISO320, assuming the Q3 shares the same base sensor as the A7Riv and A7Rv, of which I’m fairly certain of.

[SrMi]

1 minute ago, clifak said:

Mathphotographer used RAW DNG files out of the camera without any post processing for is in-depth SNR tests. He details that and the rest of the methodology in the video. Initially when he compared 3200 and 200 ISO images in PixInsight he used a JPG that was converted from the RAW DNG from the camera but that was part of his initial testing, not the detailed testing when he later used for S-DNG, M-DNG, and L-DNG images straight from the camera at both 3200 and 200 ISO.

Thanks for the update. Do you know whether he resized the results (like P2P and DxO) or whether he compared at different resolutions.

 

 

[SrMi]

15 minutes ago, beewee said:

High gain should kick in at ISO320, assuming the Q3 shares the same base sensor as the A7Riv and A7Rv, of which I’m fairly certain of.

Assuming Q3’s sensor is similar to M11’s sensor, DCG kicks in at ISO 200. If it kicks in later, the video would show a difference between ISO 200 and 3200. 

[beewee]

5 minutes ago, SrMi said:

Assuming Q3’s sensor is similar to M11’s sensor, DCG kicks in at ISO 200. If it kicks in later, the video would show a difference between ISO 200 and 3200. 

Definitely not the same sensor as M11. The M11 sensor can’t do PDAF.

[SrMi]

7 minutes ago, beewee said:

Definitely not the same sensor as M11. The M11 sensor can’t do PDAF.

Yes, certainly not the same sensor.

[Qwertynm]

vor 5 Stunden schrieb SrMi:

The video apparently looks at the noise of JPEGs, not at DNGs. JPEGS include any NR in the camera and any default NR in LrC.

I think it’s fair to criticize an online video but apparently you didn’t watch but only skim trough it. I‘m not affiliated with the guy (although he‘s also from Zürich) but he put a lot of effort into the video explaining what he’s doing and how he reached his conclusions based on the data he gathered. Before you make claims about his methodology maybe watch the video first? All images were shot in dng and converted to jpg using a python script and analyzed using astro software on the pixel level. 
 

Whether or not what he claims is true (more DR at mdng) is another matter imho. I haven’t seen any real life difference from the brief tests I did. 1 stop more DR should show up in 1:1 tests

[SrMi]

6 hours ago, Qwertynm said:

I think it’s fair to criticize an online video but apparently you didn’t watch but only skim trough it. I‘m not affiliated with the guy (although he‘s also from Zürich) but he put a lot of effort into the video explaining what he’s doing and how he reached his conclusions based on the data he gathered. Before you make claims about his methodology maybe watch the video first? All images were shot in dng and converted to jpg using a python script and analyzed using astro software on the pixel level. 
 

Whether or not what he claims is true (more DR at mdng) is another matter imho. I haven’t seen any real life difference from the brief tests I did. 1 stop more DR should show up in 1:1 tests

Yes, I wrote that I did not finish watching it. I stopped watching it because I did not want my post to become too big 😄. The JPEG was only one of several issues I mentioned.

[nicci78]

11 hours ago, beewee said:

Definitely not the same sensor as M11. The M11 sensor can’t do PDAF.

It is certainly the same, Sony Semicon can manufacture your sensor with or without sacrificed pixels for PDAF. 
For economy of scale reason it has to be the same. M11 volume is too low Leica needed a 60MP Q3 to sweeten the deal with Sony Semicon. 

SL3 will certainly sport the same 60MP with PDAF for the same economic reasons. 

Sony Semicon is also happy to sell it to Leica and Sigma, because Nikon did not want it at any cost. Stating that their 45MP (in house design but manufactured by Sony) is better. 

 

However M11 is using the old Maestro III whereas Q3 is using the new L2 technologies Maestro IV processor. So Q3 may have an edge in image processing. 

Edited June 20, 2023 by nicci78

[nicci78]

12 hours ago, SrMi said:

- We do not know when DCG (dual conversion gain) kicks in with Q3. If it is the same as M11, it would be ISO 200. Most new cameras are pretty ISO invariant above the DCG point. I wonder what the DCG point of Q3 is. Note that Q3's base ISO is 100, but the video tested only the range from 200 upwards. Therefore the video does not show that the Q3's sensor is ISO invariant. However, I think it is practical to have DCG point early at ISO 200 (larger ISO invariant range).

 

https://www.dxomark.com/Cameras/Compare/Side-by-side/Sony-A7R-IV-versus-Leica-M11-versus-Sony-A7R-V___1326_1379_1385

Looking at DXOmark results for M11, A7R IV & V, dual gain did kick in at 200 ISO, but its effect is very very subtle. Seems like a real ISO invariant sensor

[SrMi]

20 minutes ago, nicci78 said:

https://www.dxomark.com/Cameras/Compare/Side-by-side/Sony-A7R-IV-versus-Leica-M11-versus-Sony-A7R-V___1326_1379_1385

Looking at DXOmark results for M11, A7R IV & V, dual gain did kick in at 200 ISO, but its effect is very very subtle. Seems like a real ISO invariant sensor

The effect of DCG often seems subtle with Leica cameras as they do not have intermediate ISOs (the graphs are smoother).

P2P shows quite a drop in read noise after DCG for M11 and a7rV (similar amount), though Sony seems to be more ISO invariant than M11 after DCG.

I doubt that M11 is invariant from ISO 100.

[BrianCO]

On 6/19/2023 at 7:09 PM, SrMi said:

- The video apparently looks at the noise of JPEGs, not at DNGs. JPEGS include any NR in the camera and any default NR in LrC. 

 

Also, the pixel values in JPEG aren't linear, they work with a gamma to convert to viewable values in a non-linear way. The analysis probably doesn't make sense due to this. The ideas of the analysis seem ok but doing it on the JPEG doesn't make any sense. It needs to be done on the 16 bit DNG files. I look forward to seeing the analysis here, which uses the raw data: https://www.photonstophotos.net/Charts/PDR.htm.

[pcgarner]

All very interesting, but I think Cartier-Bresson would turn his back on the discussion, pick up his 1930s Leica and be taking photos, grain and all.

Edited June 26, 2023 by pcgarner

[Advertisement]

Hallo,

Look here: Leica Q