@Dominic Chan

Continuing from the latest discussion, could you expand further on your current thinking of what diffuse white and peak brightness to try aim for for best of possible HDR experience with a projector?

My current understanding is that diffuse white is 100 nits, anything above this are spectral highlights. Naturally projectors are not as bright as screens that can hit 100's to 1000's of nits so we have a limited range for these.

Currently on my 5040 I'm getting the following using the 50%/100% Masciola pattern...
BC HDR1 med = 21 nits
BC HDR1 high = 27 nits
Dynamic HDR1 med = 39 nits
Dynamic HDR1 high = 54 nits

And the with 100%/100% pattern...
BC HDR1 Med = 116 nits
BC HDR1 high = 150 nits
Dynamic HDR1 med = 188 nits
Dynamic HDR1 high = 248 nits



Am I right in saying you are recommending for diffuse white to be 50 nits therefore leaving headroom for highlights? What I'm struggling with is how does the projector know that 50 is the diffuse white and maps from that?

 

@Dominic Chan

Continuing from the latest discussion, could you expand further on your current thinking of what diffuse white and peak brightness to try aim for for best of possible HDR experience with a projector?


Am I right in saying you are recommending for diffuse white to be 50 nits therefore leaving headroom for highlights? What I'm struggling with is how does the projector know that 50 is the diffuse white and maps from that?

There is no HDR standard that applies to projectors, due to the much more limited peak luminance. What I suggested is to use half of the luminance, consistent with the SDR practice of calibrating to 50 nits white instead of 100 nits white.
This approach is similar to what Steve Shaw refers to as the "multifplier":
https://www.lightillusion.com/uhdtv.html

 

There have been some very good conversations past week or so on various other threads regards calibrating the Epson 5040/6040 for HDR.

These threads were...
http://www.avsforum.com/forum/68-di...rs-under-3-000-usd-msrp/2563857-official-epson-5040ub-6040ub-owners-thread.html
http://www.avsforum.com/forum/139-d...alibration/1393853-hcfr-open-source-projector-display-calibration-software.html
http://www.avsforum.com/forum/139-display-calibration/2248138-official-chromapure-3-thread.html

In an effort not to swamp those threads and possibly take them off topic, I have created this one.

That's a good idea, although the links can be more specific to a page or post where this topic was discussed. HCFR thread has close to 10,000 post!

 

The 'Multiplier' approach is basically what Dolby use for their HDR projection based cinema, with their 103 nits projection configurations (I think it is 103 nits, without checking my data...).

To allow for different peak luminance we have a user definable multiplier within [Redacted] to allow for user preferences.

(https://www.[Redacted]/lut_manual.html#destination_colour_space)

We also have a user definable tone-map, as varying the rolls-off is key for getting the best final result, rather than just relying on the fixes time mapping of BT2390.

As they say, every little helps!

[Redacted]

 

Sorry folks, I'm still struggling with the concept, maybe just one of those days/weeks! Also can't see mention of multiplier in the https://www.lightillusion.com/uhdtv.html article.

Maybe I just need a better understanding of how the HDR works. If a UHD is created against 1000 nits, how does a projector with say 100 to 150 peak nits use this data? Lets say theres a scene where a bunch of pixels need 500 nits, does the projector just use max, so in essence for a scene with various high nit pixels eg 300, 400, and 500, a projector with 150 will just show max for all of them so there is no definition between them if that is the right word. I'm an engineer by trade and I can't let things go until I understand how they work, to my own detriment! Thanks for your patience.

 

Sorry folks, I'm still struggling with the concept, maybe just one of those days/weeks! Also can't see mention of multiplier in the https://www.lightillusion.com/uhdtv.html article.

https://www.lightillusion.com/hdr_calibration.html

 

The curve has a different shape for 141 nits max white compared to 300 max so I don't think that linear scaling will work.

 

Yes, I understand that. I've explained my rationale in a previous post
http://www.avsforum.com/forum/139-d...9-display-calibration/2248138-official-chromapure-3-thread-44.html#post55174284

BTW, when I said "too bright", that is for projector in a completely dark room, where people calibrate to 50 nits rather than 100 nits SDR.
Hence I asked for a 300 nits curve rathan than a 141 nits curve.

I believe the JVC projector users are doing something similar for ST2084, although they don't seem to be aware of the tone mapping standard in BT.2390. LightSpace is also using something similar, in terms of aiming for a lower luminance.

 

ok, you can compare to this curve where I have shifted the knee parameter to give you ~50 nits at 50% stimulus. (actual 59 nits for 1000 nit master, 41 nits 4000 nit master)

 

Thanks for this Zoyd.

For some of us, hitting 59 nits at 50% stimulus won't be possible. It's a challenge in my case to even get to 50 nits. Is it possible to force both the 1,000 and 4,000 curves to hit 47 nits more precisely while maintaining the max luminance target of 141?

 

Just noticed my black is poor eg .11 nits. My env is a dark room with no ambient light. I see others have much better "black" eg Dominic has 0.019. What could be causing this? Possible light leakage within the projector or ? Because of this I have v poor contrast ratio, approx 1:1000.

Edit: Think that was with iris off. Will have to check what it is with iris on. Am I right in saying all calibration should be done with iris off, and then when calibration is complete turn it on? Dominic, you getting 0.019 with iris off?

 

Think that was with iris off. Will have to check what it is with iris on. Am I right in saying all calibration should be done with iris off, and then when calibration is complete turn it on? Dominic, you getting 0.019 with iris off?

Yes and yes.
Try measuring with the meter facing the projector lens, diffuser on. This provides a more accurate measure of the contrast ratio, based on which you can deduce the black level nits.

 

I just finished the fixes when using Ryan's disk and I'll add the tone curve stuff next. Probably release a test version in a day or two.

 

Great thread

Thanks for this thread

 

Quick update, reset my BC profile to defaults, iris off, high lamp, left contrast at 50, brightness to 52, dialled in grayscale (peak of 148 nits), used Dominics method above to dial in luminance, then did my 50/50 sweep and calibration across the 6 colours using auto generated patterns casted via nvidia sheild. When done the picture looked terrible! Most colours blown. Spun up Ryans disk and put a 50/50 red pattern on the screen and it was seen that the colours calibrated using the generated patterns were miles off. Eg ryans 50/50 pattern was landing on the 75! HCFR was set to 2020/P3. Should HCFR (3.4.5.1) be sending the same patterns as Ryans eg when colour space is set to 2020/P3 eg 50 sat free measure and generated patterns is sending a 50/50 or actually is sending a 50/75? I'm thinking the later the way the colours are lining up.

Currently redoing the colours using Ryans disk. While I can dial in the hue and sat using the CIE digram, how will I dial in the 3rd dimension, brightness, as appears HCFR is off showing the delta luminance for that sat?

 

Currently redoing the colours using Ryans disk. While I can dial in the hue and sat using the CIE digram, how will I dial in the 3rd dimension, brightness, as appears HCFR is off showing the delta luminance for that sat?

Hard to tell what you are doing but a couple of points with HDR calibration and HCFR:

1. You have to first calibrate to the PQ transfer function up to your clipping point, if you have done some sort of tone mapping, color calibration with a subset of BT.2020 won't work. All targets assume a raw PQ transfer function.

2. The sequences on the HDR disk that are currently supported at less than fully saturated BT.2020 primaries are in the HCFR flow section labeled:

a. P3 in BT2020 Color Sweep (pseudo-primaries scan at fully saturated P3 locations and 50% stimulus): Use the primaries/secondaries page for targets
b. P3 in BT2020 Color Sat Sweep (50% stimulus at P3 saturations from 0% (white) to 100% in 25% steps): Use the saturation sweeps pages for targets

Choose Color space BT.2020/P3 and check SMTPE HDR transfer function

The internal generator should give you the same results within 1 dE00 due to some rounding issues I'm currently fixing and assuming no additional errors coming from your video card vs. the Blu-ray player.

 

Sorry I was actually doing that, in that, adjusting the gamma on the projector but checking it against the luminance view in HCFR.

 

Sounds right, if you could attach a zipped .chc file from a session with the disk that has a grayscale run and a BT2020/P3 run (using sequence 2a from above), I'll take a look.

 

So just to spell it out essentially, the order for HDR is...

brightness, contrast, 2 point grayscale, luminance (adjusting gamma controls on projector), then disable gamma changes, primary and secondary saturations, re-enable gamma changes?

 

I'm new to HDR calibration myself, but the way I understand zoyd's procedure is that you do not disable the gamma changes prior to adjusting primary and secondary satuations; on the contrary, you should adjust the EOTF to match ST.2084 (the first column in zoyd's tables) as closely as possible, adjust the colours, then re-adjust the tone curve according to the second column (for 1000-nit master) or third column (for 4000-bit master).

 

Whichever one you are trying to make sense out of is fine with me.

 

Color Uniformity

5040UB owners, this might be a bit off topic but I am finding it important as I try to tackle HDR calibration.

When I did my SDR calibration, I noticed that the color uniformity was a bit off (greener on the right side and redder on the left side). I was able to use the color uniformity adjustments and my meter to get it reasonably balanced. I have read about people returning their projector 3-4 times with this issue so I decided to make it work. Overall my SDR calibration went very well despite this issue.

Now as I pursue HDR calibration, I am noticing that as I use brighter modes, the uniformity issue is more pronounced. And I cannot have two saved adjustment settings. So I am considering asking for a replacement.

This is my first projector, so I am not sure what I should be expecting. When you put up a full white screen, do you get a high degree of uniformity? If I don't, is my unit faulty or is this to be expected?

And by the way, the adjustment allows one to adjust the entire panel. So if you are getting an overall green skew you can use this to reduce the skew. I think this should be used as a last resort after adjusting gain / offset etc. but it could come in handy.

 

Color Uniformity

5040UB owners, this might be a bit off topic but I am finding it important as I try to tackle HDR calibration.

When I did my SDR calibration, I noticed that the color uniformity was a bit off (greener on the right side and redder on the left side). I was able to use the color uniformity adjustments and my meter to get it reasonably balanced. I have read about people returning their projector 3-4 times with this issue so I decided to make it work. Overall my SDR calibration went very well despite this issue.

Now as I pursue HDR calibration, I am noticing that as I use brighter modes, the uniformity issue is more pronounced. And I cannot have two saved adjustment settings. So I am considering asking for a replacement.

This is my first projector, so I am not sure what I should be expecting. When you put up a full white screen, do you get a high degree of uniformity? If I don't, is my unit faulty or is this to be expected?

And by the way, the adjustment allows one to adjust the entire panel. So if you are getting an overall green skew you can use this to reduce the skew. I think this should be used as a last resort after adjusting gain / offset etc. but it could come in handy.

The refurb game with this Epson projector is sooooo not worth it. Keep what you got IMO.

My experience going through four refurbs is post 355. Look at those images. The refurbs can be (are?) atrocious!

Refurbished Epson 5040UB - $1,151.00

Pretty insane deal for a 5040UB. Since it's refurbished, I'm not certain of the warranty that's included on these, but I would think it would have at least some period of Warranty. Someone else mentioned that it may have the full regular warranty, but buyer beware, because I'm not sure...

www.avsforum.com

 

correct, adjust first to ST.2084 and use that function as a baseline for color adjustments, as this is the target that all patterns use for HDR10 calibration. After that apply any tone mapping to try and capture additional highlights (at the expense of absolute color accuracy if your adjustments affect diffuse white and below range).

 

correct, adjust first to ST.2084 and use that function as a baseline for color adjustments, as this is the target that all patterns use for HDR10 calibration. After that apply any tone mapping to move to try and capture additional highlights (at the expense of absolute color accuracy if your adjustments affect diffuse white and below range).

Dominic, would you mind putting that in Epson 5040 terms? Adjusting the epson gamma controls while checking against the luminance graph is that "adjusting to ST.2084"? If so, what is then tone mapping?

 

Yes, technically to calibrate you would set your contrast so that you see no flashing levels above 539 for 140 nits max luminance and adjust your 10 or 20 pt controls (if you have them) to match the shape of the white dotted line. The material was mastered to this line. If you can't clip at 539 then I would get 50% (502) as close to 92 nits as possible. Then you would adjust your color controls using patterns that are designed for 50% levels (92 nits). At that point you would back off on contrast and use 10/20 point controls to match one of modified curves (what we've been calling tone mapping), to bring back some of the highlights.

 

Unfortunately with the Epson 5040 we are very limited in our controls, we only have two point control of offset and gain for RGB. For gamma we have 9, known as colour tone 1-9. Then for colour we have hue, sat, and brightness for each of the RGBCYM colours

 

Something else I'm struggling with is the concept of the "multiplier" as discussed here https://www.lightillusion.com/hdr_calibration.html

If its scaled down and a curve produced for that and calibrated to, how does the projector know this eg there is no way to tell the projector that your peak is x and therefore treat this peak as level 940 and scale all content to this level. Eg content is mastered to 1000 nits, a bunch of pixels are at level 609 which is 300 nits, the projector cannot produce that, where as a scale value would be 110 nits which the projector can.

 

That multiplier is similar to your contrast control which you can set such that an incoming 1000 nit code (723) is converted to your peak luminance of 140 nits. All codes under 723 will then also produce luminance levels your projector can display and the recommendation for best perceptual mapping of these codes is to use a curve based on the BT.2390 formulas.

 

OK, I think I'm beginning to see the light, parden the pun!!

 

White is no different in this respect than any other color in HDR.

 

@zoyd

Any ETA on the next HCFR build as I'm going to give calibration another go once available rather than asking for a curve to be manually created for me. Eg will adjust screen size till 50% pattern produces 92 nits, then see what 100% produces and create a curve from that.

 

I might get something out for testing later today.

 

FYI, any one using Cosmos Laundromat or any other HDR clip, do a reality check by just playing it as SDR. I taught it looked great on the "calibrated" HDR until I watched it on SDR and while it doesn't have highlights it was much better looking. I guess its easy to get so tied up with HDR to see the highlights that it can cause the rest of the picture to suffer.

Here is a HDR copy (~6G) of it https://drive.google.com/uc?id=0B7Gr_0zXIXEjSWhhUzBmVmNZWFU&export=download

 

What was the luminance for diffuse white when you did the comparison? Assuming it's equal to the SDR white, I don't expect it to look worse.

 

Bright Cinema reset to defaults and using auto bright (HDR1) I zoomed in till 50% grayscale was 92 nits, this results in a 62" image compared to my 106" normal size. Contrast was at 50. White clipping showed up to level 642 flashing. 100% grayscale was giving 323 nits.

Use the 300 curve or wait till new HCFR drops?

Is the contrast ok or should it be adjusted?

 

Here is a test version of HCFR which has both a diffuse white setting for scaling the PQ curve and BT.2390 tone mapping. You may encounter some bugs but the main luminance plots and Y targets should be fine as well as BT.2020/P3 primaries and saturations using the shifted diffuse white (don't know what happens with these with tone mapping) In the plots below the upper curve is following raw PQ, the middle one is with diffuse white at 47 nits, and the bottom one is diffuse white 47 nits with tone mapping.

Unzip this and drop into your HCFR folder. You should probably delete the .ini file prior to starting.

 

Here is a test version of HCFR which has both a diffuse white setting for scaling the PQ curve and BT.2390 tone mapping. You may encounter some bugs but the main luminance plots and Y targets should be fine as well as BT.2020/P3 primaries and saturations using the shifted diffuse white (don't know what happens with these with tone mapping) In the plots below the upper curve is following raw PQ, the middle one is with diffuse white at 47 nits, and the bottom one is diffuse white 47 nits with tone mapping.

Thanks zoyd.
I just did a quick simulated run and the result matches my manually scaled tone mapping curve. However, I noticed that the min Y target seems to remain at 0.000, even when I specific 0.05 as Master MinL and Target MinL.
Also, as I previously asked, what is the difference between the Y target values in the grid, and the Reference values in the Luminance graph? Is one the raw ST.2084 and the other tone-mapped BT.2390? (Both have 0 nits at 0% input).