[thoth]

Is the conversion from 8-bit to 10-bit in Calman likely to make any substantive difference to the readings?

Converting 8-bit RGB Limited to 10-bit RGB Limited is simply "multiply by 4". Converting from RGB to YCbCr is more complicated, and is different for BT.709 vs BT.2020. I haven't heard of anyone who's done digital capture of G1 output to see if it's accurate. (I have the means, but no interest in spending money on a G1.) Assuming accurate output, I wouldn't expect that large of a difference, but I haven't tried that same pair of measurements. You could try cross-checking by making a similar pair of measurements with ZRO and PGenerator.

 

[thoth]

From your previous measurments ITP was more lenient on dE calculation compared to 2000.

Chuckle. Picking a pair of measurements at random:
xyY: 0.29762647982055,0.319425287164715,0.178747163376
xyY: 0.297126363178279,0.320098721507976,0.17543285532

dEITP: .466
dE2000: .038 (assuming the measured White luminance)

There is no actual standard for the use of dE2000 with displays; it's specified for physical objects viewed under 1000 lx simulated D65 illumination. It has farcical results when applied to HDR. dEITP is an actual standard expressly designed and specified for use with displays.

 

[sjy1969]

Converting 8-bit RGB Limited to 10-bit RGB Limited is simply "multiply by 4". Converting from RGB to YCbCr is more complicated, and is different for BT.709 vs BT.2020. I haven't heard of anyone who's done digital capture of G1 output to see if it's accurate. (I have the means, but no interest in spending money on a G1.) Assuming accurate output, I wouldn't expect that large of a difference, but I haven't tried that same pair of measurements. You could try cross-checking by making a similar pair of measurements with ZRO and PGenerator.

I didn’t cross check between software but did do so within Calman. RGB 8-bit and YCbCr 4:4:4 10-bit read almost identically.

 

[sjy1969]

The A95L pumps; HDR luminance goes up the longer the pattern stays up, until it goes down again. Try .5s.


From what I see, CalMAN is still only operating in 8-bit, and the G1 is simply converting to 10-bit for output.

I settled on delay 0.5s as you suggested plus 15s black frame insertion between each reading in autocal and 10s between each reading during post calibration measurement. It provides rock solid consistency in readings with peak luminance high and gave a good calibration result. Takes a while to do it that way but the result looks worth it.

I also used YCbCr as that’s what Sony prefers.

 

[Rob2la]

Delete

 

[thirdkind]

I finally took some time over the weekend to calibrate my A95L using the direct 20-pt method for each format (SDR 2.4, HDR10, and DV) rather than relying upon Sony's math to translate SDR gamma 2.2 calibrations to SDR 2.4/HDR10/DV. The basic process I used is detailed by @thoth in his post, along with this sage advice from @titleexaminer92 for patch sorting to achieve stability on high luminance displays. Here are my results. Any tips/recommendations are welcome.

SDR (Expert 1)

I don't think the difference between Sony's method and the direct method is meaningful for SDR. My results are similar for both (generally pretty good). I was able to get the gamma curve slightly closer to ideal, but we're talking nitpicking here. I still have issues in the low end in SDR 2.4 that I can't calibrate out.

HDR10 (Expert 2)

Relying upon Sony's method, I had good grayscale in HDR10, but slight EOTF undertracking. By targeting HDR10 directly, I was able to adjust luminance where necessary and achieve both excellent grayscale and EOTF tracking.

Dolby Vision (Warm)

DV grayscale with the 2-pt Warm method was decent, but had a higher dE than I would've liked and a slightly S-shaped EOTF curve, with undertracking at the bottom of the slope and overtracking near the top. Direct 20-pt calibration with DV Verify in ColourSpace resulted in a much more acceptable dE and EOTF tracking that's pretty spot-on. For some reason, DV clips peak white unless I lower Contrast considerably (which I didn't ultimately).

This is using Gradation Preferred for tone mapping. I might try again with tone mapping disabled, but the overtracking is severe and I imagine the 20-pt adjustments would have to be fairly large to compensate.

 

[7685949]

I finally took some time over the weekend to calibrate my A95L using the direct 20-pt method for each format (SDR 2.4, HDR10, and DV) rather than relying upon Sony's math to translate SDR gamma 2.2 calibrations to SDR 2.4/HDR10/DV. The basic process I used is detailed by @thoth in his post, along with this sage advice from @titleexaminer92 for patch sorting to achieve stability on high luminance displays. Here are my results. Any tips/recommendations are welcome.

SDR (Expert 1)

I don't think the difference between Sony's method and the direct method is meaningful for SDR. My results are similar for both (generally pretty good). I was able to get the gamma curve slightly closer to ideal, but we're talking nitpicking here. I still have issues in the low end in SDR 2.4 that I can't calibrate out.

View attachment 3669655
View attachment 3669657


HDR10 (Expert 2)

Relying upon Sony's method, I had good grayscale in HDR10, but slight EOTF undertracking. By targeting HDR10 directly, I was able to adjust luminance where necessary and achieve both excellent grayscale and EOTF tracking.

View attachment 3669661
View attachment 3669662


Dolby Vision (Warm)

DV grayscale with the 2-pt Warm method was decent, but had a higher dE than I would've liked and a slightly S-shaped EOTF curve, with undertracking at the bottom of the slope and overtracking near the top. Direct 20-pt calibration with DV Verify in ColourSpace resulted in a much more acceptable dE and EOTF tracking that's pretty spot-on. For some reason, DV clips peak white unless I lower Contrast considerably (which I didn't ultimately).

This is using Gradation Preferred for tone mapping. I might try again with tone mapping disabled, but the overtracking is severe and I imagine the 20-pt adjustments would have to be fairly large to compensate.

View attachment 3669663
View attachment 3669664

Have you tried DV with TM off by doing sdr with the dv offset WP and targetting 2.2 while setting sdr gamma to +1?

 

[Masterwarzombie]

I finally took some time over the weekend to calibrate my A95L using the direct 20-pt method for each format (SDR 2.4, HDR10, and DV) rather than relying upon Sony's math to translate SDR gamma 2.2 calibrations to SDR 2.4/HDR10/DV. The basic process I used is detailed by @thoth in his post, along with this sage advice from @titleexaminer92 for patch sorting to achieve stability on high luminance displays. Here are my results. Any tips/recommendations are welcome.

SDR (Expert 1)

I don't think the difference between Sony's method and the direct method is meaningful for SDR. My results are similar for both (generally pretty good). I was able to get the gamma curve slightly closer to ideal, but we're talking nitpicking here. I still have issues in the low end in SDR 2.4 that I can't calibrate out.

View attachment 3669655
View attachment 3669657


HDR10 (Expert 2)

Relying upon Sony's method, I had good grayscale in HDR10, but slight EOTF undertracking. By targeting HDR10 directly, I was able to adjust luminance where necessary and achieve both excellent grayscale and EOTF tracking.

View attachment 3669661
View attachment 3669662


Dolby Vision (Warm)

DV grayscale with the 2-pt Warm method was decent, but had a higher dE than I would've liked and a slightly S-shaped EOTF curve, with undertracking at the bottom of the slope and overtracking near the top. Direct 20-pt calibration with DV Verify in ColourSpace resulted in a much more acceptable dE and EOTF tracking that's pretty spot-on. For some reason, DV clips peak white unless I lower Contrast considerably (which I didn't ultimately).

This is using Gradation Preferred for tone mapping. I might try again with tone mapping disabled, but the overtracking is severe and I imagine the 20-pt adjustments would have to be fairly large to compensate.

View attachment 3669663
View attachment 3669664

what colorspace stabilisation did you use in SDR and HDR?

 

[thoth]

I don't think the difference between Sony's method and the direct method is meaningful for SDR.

Agreed.

I might try again with tone mapping disabled, but the overtracking is severe and I imagine the 20-pt adjustments would have to be fairly large to compensate.

I got half way through doing that one day, and abandoned it. The adjustments were indeed large.

 

[thirdkind]

Have you tried DV with TM off by doing sdr with the dv offset WP and targetting 2.2 while setting sdr gamma to +1?

I tried the SDR gamma method previously and it was so-so. More experimentation might have yielded better results. Calibrating DV directly is definitely easier, and based on the opinions of others who are far more knowledgeable than me, enabling tone mapping shouldn't be destructive.

what colorspace stabilisation did you use in SDR and HDR?

5 seconds with a ~25% grey patch. Sorting the patch set to alternate from light to dark really helped stabilize the measurements.

 

[7685949]

I tried the SDR gamma method previously and it was so-so. More experimentation might have yielded better results. Calibrating DV directly is definitely easier, and based on the opinions of others who are far more knowledgeable than me, enabling tone mapping shouldn't be destructive.



5 seconds with a ~25% grey patch. Sorting the patch set to alternate from light to dark really helped stabilize the measurements.

I have a c6 hdr5000 and g1 tpg arriving to me soon and will be experimenting more once I get them. I tried the sdr gamma method for dv and my results were ok I think, with just a little overtracking. I did aim luminance lower than the target throughtout the grayscale but my i1d3 is giving me fits.

 

[Masterwarzombie]

I tried the SDR gamma method previously and it was so-so. More experimentation might have yielded better results. Calibrating DV directly is definitely easier, and based on the opinions of others who are far more knowledgeable than me, enabling tone mapping shouldn't be destructive.



5 seconds with a ~25% grey patch. Sorting the patch set to alternate from light to dark really helped stabilize the measurements.

as recommended, I want to sort the hdr and DV patches. On your screenshot and from the patches provided, where do you find patch 21? what does it correspond to

 

[thirdkind]

as recommended, I want to sort the hdr and DV patches. On your screenshot and from the patches provided, where do you find patch 21? what does it correspond to

I added that patch myself because I wanted to measure it. There's no control for that point in the Sony 20-pt menu.

 

[thoth]

I still have issues in the low end in SDR 2.4 that I can't calibrate out.

Although it can't be fixed, have you tried setting the Bias controls to -1 to see if it improves?

 

[thoth]

Having discovered that PGenerator 4:2:2 output is incorrect, I now think the A95L measurement differences between 4:4:4 and 4:2:2 are because the A95L processes 4:2:2 as 12-bit, and hence gets inaccurate triplets. For example, all PGenerator 4:2:2 grayscale triplets are inaccurate as 12-bit. So I think it's better to calibrate using 10-bit 4:4:4 instead. I redid SDR that way last night; will redo HDR when I can get to it.

 

[thirdkind]

Although it can't be fixed, have you tried setting the Bias controls to -1 to see if it improves?

I've always followed the advice not to touch bias on Sony TVs because it's traditionally caused issues, but if you're saying it doesn't, I'll give it a try.

Having discovered that PGenerator 4:2:2 output is incorrect, I now think the A95L measurement differences between 4:4:4 and 4:2:2 are because the A95L processes 4:2:2 as 12-bit, and hence gets inaccurate triplets. For example, all PGenerator 4:2:2 grayscale triplets are inaccurate as 12-bit. So I think it's better to calibrate using 10-bit 4:4:4 instead. I redid SDR that way last night; will redo HDR when I can get to it.

I'd like to see Ted's response to your last post. He's always said YCbCr 4:2:2 on the PGen is only accurate at 10-bit, so it sounds like there's a debate to be had about whether the PGen's output is the problem or the A95L's handling of it. If the HDMI spec says the source should assume the sink will process 4:2:2 10-bit as 12-bit and pad the output appropriately, then it's PGen that needs to be fixed.

How does this relate (if at all) to the conventional wisdom that Sony handles 4:4:4 incorrectly and 4:2:2 is preferable? Wouldn't that also affect the results? How do we know which is better for calibration purposes?

 

[8521564]

@thirdkind If you’re lowering RGB evenly by one click there is no harm - just don’t adjust the RGB Bias channels independently, like R+2 G-2 B-3 would be bad.

 

[FromPlasmatOled]

@thirdkind If you’re lowering RGB evenly by one click there is no harm - just don’t adjust the RGB Bias channels independently, like R+2 G-2 B-3 would be bad.

By inserting a 0.1% pattern of the Ted disk and observing it in a dark room, you notice that you actually do not lose information as long as you move the bias to - 1, probably the situation has also improved thanks to the hours of use of the panel...
All Bias -2 the pattern merges with black and in case of positive shifts a strong dithering is generated.

Having exceeded 2000 (as soon as I have time) I will check that the old calibrations are still valid as a reference and will possibly do a new calibration by trying this approach on the biases.

 

[thoth]

I've always followed the advice not to touch bias on Sony TVs because it's traditionally caused issues, but if you're saying it doesn't, I'll give it a try.

I haven't seen any adverse outcome from lowering all three to -1, and from my measurements of +2 grayscale steps below the first adjustment point, it's an overall improvement.

If the HDMI spec says the source should assume the sink will process 4:2:2 10-bit as 12-bit and pad the output appropriately, then it's PGen that needs to be fixed.

The spec is crystal clear that zero padding "shall" always be used. The spec doesn't say the sink will process as 12-bit. There is no information in the HDMI stream to tell the sink how many bits the source used, so the sink doesn't know if the source used 8 or 10 or 12 bits. The sink is free to use however many bits it wants. This makes 4:2:2, in general, problematic for perfect accuracy, unless you know how both source and sink work. You'll never have perfect accuracy for all colors if the sink uses more bits than the source used, even with zero padding.

How does this relate (if at all) to the conventional wisdom that Sony handles 4:4:4 incorrectly and 4:2:2 is preferable?

Can you point at any specific claims about this? I could imagine differences in upscaling and/or chroma resampling, for example, but neither should impact single-color fields for calibration.

I'm going to try putting together my own 4:2:2 pattern generator with full 12-bit output, so that I can measure the difference between 10-bit 4:4:4 and correct 12-bit 4:2:2.

 

[Masterwarzombie]

Regarding the choice of resolution, does the frequency 60 Hz instead of 24 Hz influence the measurements?