Official JVC RS20 / HD750 Calibration and CMS thread (NEW FIRMWARE V1.1) - Page 34

leDahu and all - what an incredible tool this spreadsheet is! I cannot thank you enough for the time you have spent creating this, translating it, and sharing it with all of us.

Normally I put off calibrating the gamma because it is such a long and frustrating ordeal. But motivated by your spreadsheet I decided to try it out last night and the results were nothing short of fantastic.

To start, let me say that I was able to accomplish an ideal custom gamma curve in 15 minutes using your spreadsheet (see pic below). This used to take me several hours (6+!) using my manual method, and with the spreadsheet the results are much better as well.

Previous gamma calibration
The first thing I did last night was to measure how my custom gamma was holding up which I created probably about 150 bulb hours ago. I measured the current settings and the curve was very poor.

Average gamma was 2.07 and WAY off on the top end (5-55% was in the 2.27 - 2.11 range, then things dropped pretty sharply, with 60-80% in the 2.04 - 2.0 range, 85% at 1.93, 90% at 1.9, and 95% at 1.58!).

I know that with this prior gamma calibration I couldn't get the top-end too well when I finished, although it was significantly better than this. At any rate, clearly with the top end gamma so poor there will be a much improved picture.

Latest gamma calibration
So with leDahu's spreadsheet in hand I set out to resolve the poor gamma curve. I was really pleasantly surprised to see how well it came out, especially considering how easy it was to follow his procedure compared to the manual way I was doing this before.

Here is the graph with my latest results:


I started with the gamma correction value at the default 2.2 and followed leDahu's excellent instructions. The first attempt only took 15 minutes including time to run the grayscale measurements and the results were very good (despite a spreadsheet issue with #N/A being returned which I detail in the notes below).

The top end needed quite a bit more work though, along with some tweaks in the range overall, so I ran through the refinement process several times. Probably about 10 refinement cycles overall, but it was well worth it.

For reference here are the values I ended up with. These are for 5, 10, 15, 20 - 90, and 95% respectively:
50, 93, 136, 177, 258, 350, 448, 556, 668, 776, 897, and 956.

By the time I finished up last night it was too late to watch any material, but at a very quick glance the picture definitely now has more pop and depth to it. I look forward to checking out the results more fully tonight.

Here are a few points in particular:

- I encountered a small issue with the spreadsheet. If you enter a value less than 1.8 in any of the cells in row 28, then the corresponding cells in rows 29 and 31 show as "#N/A". I guess this is because the lookup table is set to work only with values up to 1.8? In my case I had three values in row 28 (the measured gammas for 85%, 90%, and 95%) under 1.8. So to work around this I used 1.8 for the first pass and then refined it with the subsequent measurement. Hopefully leDahu can update this to support values below 1.8, perhaps as low as 1.4 (or lower?) to handle this situation.

- I took leDahu's advice and worked from the top down. This is really critical to getting the top end as perfect as possible. As he mentions it is very sensitive at the top end.

- I found it very interesting that when looking at the individual R, G and B lines on the gamma graph, these are perfectly converged from 5-50%. Then starting at 55% they start to diverge, first gradually and then continually, even more so as you move to 95%. HOWEVER, this does not occur on the RGB levels graph (which remains flat across 5-100%) and the dE is all < 3 through the range. For instance at 80% R, G and B gamma is 2.32, 2.21, and 2.13 respectively. And at 95% its the widest, with R, G and B measuring 2.7, 2.4, and 2.09. Do you find this odd? Some of this may be to custom gamma tweaks I made to RGB to get the grayscale flat, come to think of it...

- There were some slight dips at 75% (2.19) and 85% (2.16). But of course no controls are available for there and I suppose this is close enough. I guess all things considered it is a testament to JVC's excellent algorithm that all the other half steps (25%, 35%, 55% etc) track so well.

- I have heard (mostly) good things about the RSx series and higher gammas and have always wanted to try 2.4 or 2.5. However the calibration process was way too much work to go through, especially considering I may have found that I didn't like the results. Therefore I never bothered and stuck with 2.2. Now thanks to this spreadsheet I should be able to dial in a very nice 2.4 or 2.5 curve in less than an hour. I am really looking forward to experimenting with that and seeing how I like these alternate gammas.

- Per the above note I plan to try a 2.4 and 2.5 gamma curve, but with 5% set to use a gamma point based on a 2.2 curve (to try and improve shadow details with these higher gammas, per other things I have read). I would love to hear your feedback on how you think 2.4 and 2.5 looks on the RS20.

Again I want to thank leDahu for his excellent spreadsheet and others here who have contributed to the excellent discussion. This spreadsheet is my new favorite tool (if you haven't been able to tell yet ) and I look forward to more experimentation now with 2.4 and 2.5 curves.

As always your comments and feedback are encouraged and welcome!

Yes.
In case of compromise to be taken in some circonstances, the decision for rounding is left to the user. Sometime you have to force a different value


For my white gamma I started with a corrected value of 107=>2.13 , then 97=> 2.15 and finished up with 89=> 2.22

In the example, the current gamma table is initially preset to 2.4 in order to get an average value of about 2.22
So the gamma table show 81 at 10 IRE.
This 81 gradation does produce the actual gamma 2.0
The new gradation is calculated, taking into account the actual gamma, the target, the current gradation and some data taken in the moving table.
This 55 gradation value (step 7 line 31 ) is then entered in the JVC and replicated for a new cycle (step 4 line 22).
The new gamma produced with this 55 gradation will be read in HCFR and entered on line 28

Gamma values are referenced to 100 IRE.
If R/G/B have the same gamma value at X IRE then their representation on the luminance graph is the same point. As a consequence dE=0 at X IRE.
Achieving the same gammas for R, G and B at all the IRE levels leads to achieve also the grayscale adjustment.

The correction of the white gamma gradation gives to R, G, and B gamma gradations the same value, destoying the current particular R G B values. Therefore the grayscale returns to its initial state. The previous work is lost.

Working first on the white gamma has the advantage of preserving the grayscale and to move the 3 gammas close from each others. We do 3 in one go.

The green will be very close to the White one.
Red and blue will be close enough to allow a quick adjustment.

Working directly on R G B, forgetting W and using RGB data from HCFR is the royal road.
Wait and see...
Something faster and much better.

This can happen if the next point is to far from what JVC is expecting.
Sometime, the gradation is accepted but not really saved.

Give a 'better' value to the next point to be processed and come back to the one to be worked on.

The projector also works on its side...

Any ideas about what is causing the gamma drift?

This is a pretty big issue, especially for those who don't have the ability to fine tune their own gamma. Doing an a/b comparison between my old settings and new ones shows a very dramatic difference in dimensionality. This gamma change is probably the single biggest change that takes place over time. IMO, it has a bigger effect than greyscale or color drift, both of which are pretty modest on my unit.

I have to post my values, but I want to take another go at it before I do that.

Thank you very much for your incredible efforts, leDahu!
I tried your method last night and it worked outstanding. I have managed to get a straight 2,22 gamma without any bumps. All within some 30 minutes. Your excell sheet and rules/tips are fantastic!

Last night I dialed in a 2.4 gamma on the RS20. It took about 30 minutes to get a decent rough cut. Then I spent the next hour running several iterations to refine it. It came out nice including the top end which can swing wildly but ultimately has an exact setting that when set right gets you right at or near the target.

Also note that I purposely added a bump to give a gamma of 2.2 at 5% and 2.3 at 10% in order to avoid possible black crush and lose of fine details etc.

For reference here are the numbers I wound up with for 5%, 10, 15, 20 - 90 and 95%:

50, 85, 119, 160, 238, 328, 425, 535, 644, 757, 885, 950

Note that these numbers are highly bulb dependent so I include them for reference only (though feel free to plug them in if you want). I haven't looked but I think I am around 400-500 hours on the bulb.

For reference, here are the default numbers from the RS20's 2.4 gamma setting (which on my unit yields a gamma much lower and very uneven):

36, 81, 128, 176, 272, 375, 481, 584, 693, 799, 911, 967. For comparison purposes you should ignore the first two numbers (for 5% and 10%) because as I noted above mine are purposely bumped up to achieve the lower gamma at 5 and 10%.

I did some quick A/B testing courtesy of the preset User 1/2. I cannot really decide on whether I like 2.2 or 2.4 better. 2.2 seems to have a bit more "pop" to the image because it is brighter, while 2.4 seems to have a bit more dramatic effect and perhaps depth. I was nice to see that with the extra bump at 5 and 10% that there was no noticeable black crush or loss of detail with the 2.4 gamma (although in fairness I did not bother to see if such lose would be present without the bump).

I know some folks like Greg are big proponents of 2.4 or 2.5 gammas on machines like the RS20 that have the CR to pull it off. So I will keep experimenting so I can decide which suits my taste.

It seems some types of material (sports, animation) lend itself well to 2.2, and other material such as dark movies may be better at 2.4. I did some reading on AVS and it seems all discussions on which gamma is better are mostly talk on technical standards and why one is "correct" or not.

I was looking for more of a discussion on what movies/materials tend to look better at 2.2 vs 2.4 or 2.5. If you have any thoughts on this please let me know.

Gamma Tweaking - New version Available

Create your own custom gamma profile more easily

Special gamma procedure: no change here

 

gamma HD750_eng_2.zip 18.7578125k . file

The new version handles this overflow.
Nevertheless any lower value is accepted but not used. It is replaced by 1.8 instead.
Since it is an iterative process this will lead to an additional cycle.


At 90%, the gradation increase by 1 per 0.01 gamma.
At 95%, the gradation increase by 0.5 per 0.01 gamma.

Depending on the case, changing the target from 2.22 to 2.21 or 2.23 would potentially make the adjustment possible for the White .


W adjustment should be run first. Then G R B.


Yes.
It seems it's an effect of the spline if the leading points are not perfect.


So the new version is for you

I bought an I1D2 meter and CalMAN about 6 weeks ago and have been calibrating my RS20....but have run into one key problem I hope someone may be able to help me with.

I can not get the blue CIE point for REC709 or SMPTC-E to reach the blue target no matter what value of Saturation or Hue I use and with the meter facing the screen or the projector. The blue point can only be moved by the CMS to just above or to the right of the target -- I can not go into or below the blue target.

I have the exact same problem with a Sony XBR4 LCD 52 inch set, so it is not the RS20. I can increase the RS20 CMS saturation from say 0 to + 60 and clearly see the blue change in Color Bars and real images -- and yet the measured blue point reaches a limit in movement on the CIE chart and will not go down into the blue target zone.

Do I have a meter problem with the blue measurement is my key question?.

If I drop Manni's blue CMS settings into my other new R,G,C,M,Y settings for SMPTE-C and REC709 the image looks great and except for blue the 1976 Delta E values are all under 3. But the meter's blue output appears to reach a limit when the blue point gets near the blue target (if I lower Blue Saturation the blue point withdraws correctly and moves easily toward the D65 point) .

Thanks for any thoughts on this I1D2 problem.

KT

I am very curious about why you say this. Is it because you think it is easier to reach the final result that way? Or is it because doing R G B first and then W would cause incorrect results?


Thanks. I really like how you can now set the gamma target at each % level! I have a question about the defaults you are using however.

I see that you started at 2.22 for 5% and then moved SLOWLY up to 2.4 at 50%. Whereas when I calibrated it I went with 2.22 at 5%, 2.3 at 10% and then right to 2.4 for 15% and up.

Is there a specific reason you chose to demonstrate it this way? Do you think that is better than moving it up much more quickly like I did?

The reason I did it this way is because in the various places I have read about this "gamma bump" trick it has mentioned a bump at 5% to help prevent the higher gamma from crushing blacks. And with the current way I am using it certainly seems to prevent crushing blacks.

However if you think there are adverse affects from what I did and I would have better results using a more gradual approach like you showed, I would certainly then want to try that.

Thanks again for such terrific work on this. For the folks here that have the calibration equipment I highly recommend fine turning your gamma, even if it is just to adjust to 2.22.

I'm looking foward to trying this new, and super cool tool.

Have you guys seen the need to go through CMS after making your gamma changes? Last time I calibrated gama, I had to go back to the CMS and tweak a bit.

I did not have a chance to go back and remeasure the gamut. However I don't think it changed much if at all, just by A/B testing by-eye between my old untweaked gamma and my new one (both my new 2.2 and 2.4 curves). For instance color bars looked identical as I switched.

Hi Manni

Looks like a very strange problem to me. No one else on this thread or the calibration threads so far as I have seen has indicated a problem like this.

As I said, your blue settings seem to work well on my projector. I am judging that by comparing the blue visual behavior of the THX mode to your settings--and with a good variety of bright - almost electric blue HD sources I have collected off the air.

So this raises the question -- how did you arrive at your blue settings -- if you could not calibrate your projector with your i1Pro meter?

KT

Blue is a dim color. Check out its "Y" in relation to the other colors. This makes it hard to measure. I had this issue with blue until I zoomed the picture until it was as small as I could get it. This increases the overall brightness, making it easier to measure. I know this did not solve Manni's problem. His screen has less gain than mine so I am thinking that this could be a factor. I also don't recall what Manni's throw distance is. That could also be a factor because it would affect how much he could zoom in.

Hello Manni,

have you tried that?

FWIW the secondaries we adjusted before the primaries which he was advised to do by JVC Japan.

http://www.avforums.com/forums/proje...ml#post9305403

Did you notice that the secondaries don't move at all when you adjust the primaries with the CMS? Very strange world when we thought that R + G = Y.
It seems that the RGBYMC domains are physicaly separated.

Ddvd, LG, leDahu: thank you guys for all this useful info! I've been following your work only loosely since I've been distracted on audio matters in recent weeks (just got a Mark Seaton SubMersive and have been engrossed in sub eq matters. This is the 'Lumis' of subs, thought not QUITE as expensive!)

Tonight I've just replaced the lamp in my RS20, so as soon as it ages a bit I will be studying all of your recent reports very carefully (and probably have lots of idiotic questions). The original lamp was at 1360 hrs, and in High lamp mode and Iris wide open (0), it wasn't unwatchable by any means, but I want to be sure to get my 2 Mack Camera lamps within the 3 yr period. But WOW, does the new lamp bring back the dynamic pic! I'm now back to 'Normal' (i.e., low) lamp, and with Iris = -6 the pic is so much more striking. If LED illumination is able to eliminate the lamp in 2 more years, say, and give a pic as good as I'm getting right now, I will be willing to upgrade. If the lamp stayed as it is, I would see no need to upgrade.

BTW, are you guys using the HDMI 'Enhanced' input? Initially it was recommended to use 'Standard', but I think the last discussions I heard recommended 'Enhanced' (with Brightness and Contrast of course re-adjusted appropriately).

Standard clips btb and wtw.

Try opening the iris an extra 3 or 4 stops with the 2.4 gamma curve. This will keep the strong contrast of the 2.4 curve, but give you more equal light output levels at the mid-IREs.

I have a 2.3 and 2.5 setup right now, and did an A - B with my wife (no videofile) beteen 2.3 with a -10 iris and 2.5 with a -6 iris. She consistently prefered the 2.5 gamma and actually used the word 'pop' unprompted by me. Hardly scientific, but I find 2.5, with the iris open a little wider, to be excellent -- a ton of depth in the image.

Hey millerwill - any chance that you could measure the gamma with your new lamp at default settings? I would be interested if the new lamp restores proper gamma values toward the high end. My guess is that it does and that the better PQ that you are seeing is a result of the gamma curve being restored.

I have as many hours on mine as you did on your old lamp and fixing the gamma curve really made a huge difference.

LeDahu -

Thanks for sharing your work sheet. Yesterday I spent a few hours with my i1pro and Calman. Using your spread sheet enabled me to get gamma calibrated without throwing things across the room for the first time. I used the curve you set up too. I like it. After running W gamma, I was able to bump RBG gamma curves to get a pretty flat grey scale as well.

While I had the meter out, I went through CMS. I must have either been drunk or just stupid the last time I calibrated the gamut. Every one was totally screwed up. I think I failed to calibrate the Calman decoder before starting. I don't smoke pot, so I just can't think of anything else that would have caused these being off so far.

The end result of my calibrating efforts was a very nice looking picture. So good in fact, I'm looking forward to watching my BR collection again.

I should get the chance this weekend to start playing with video again. I now have an I1Pro, and did learn how to do Grayscale calibration with it, but haven't yet tried a custom gamma. It sounds like leDahu's work above will be very helpful; just have to get started learning about it.

Guys - is it just me or is there no way to temporarily hide (or substantially minimize) the custom gamma controls menu when calibration?

For instance, let's say I have a 5% or 10% full field up and trying to calibrate W on the gamma curve. Having the RS20 gamma calibration menu up seems to add additional light, which then throws off the measured gamma.

Other controls such as grayscale and CMS have the ability to be minimized. For instance if you are working on RED you can click it and the whole menu disappears except for the Red control which is then centered on the bottom.

Basically I am looking to do the same thing within the custom gamma menu. This way I can move the setting up or down using just a small portion of the menu/screen and minimize the affect this menu has on the overall light output.

Because there doesn't seem to be a way to minimize the custom gamma menu I have to save and close the entire gamma menu, tweak the setting, measure and navigate back to the gamma menu and drill back down to the W % level I am working on, and repeat the process each time.

Am I missing something or is there no way to minimize or temporarily hide the custom gamma menu? Thanks.

Good point
It can be hopefully a good suggestion for the next generation RS series and add it in their firmware release
JVC... Take a note, please !

I wrote:
rule #1- The action on a point of the curve always causes one modification of both lateral points and of these two points only.

This is true as long as the amplitude of the leading point is moderate.
If not, outer points might also move a bit: because the internal algorithm tends to smooth the strong variations.
This is why I would recommand to start with a raw gamma curve close enough to the target (the average value). If not, the next cycle will correct that.

I also realize that I missed to detail some other rules .

#- After resetting or presetting the custom gamma, the WRGB gamma curves are straight and therefore identical.
#- Playing with W ALONE, will change R G and B as well. RGB gammas will follow exactly. WRGB stay identical. As long as you do not touch RGB separatly.

That way, you actual gamma curves will change, keeping their relative distance -about-

#- Now, changing one RGB gamma alone will also affect W according to the equation Y709 = 0,2125R + 0.7154G + 0.0721B.
As from now you loose part of the control over the process because the internal logic wokrs after you.
You are obliged to go to the end of the entire process RGB to get what you want.
That way you finish up with nice gammas and if skill enough with improved grayscale (RGB graph) or a good compromise of gammas and RVB

#- Coming back to W will affect RGB which will follow according to the above equation. But you have not the control of them.
Your preceeding accurate efforts for adjustments might be ruined.
_______________________________

Working on RGB's first is appropriate for adjusting grayscale.
But the algorithm of the spreadsheet is not designed for that.
Because of that this process will also affect the relative positions of RGB gammas.
Working W at the end, will produce a very nice W gamma (as you did) but at the expenses of RGB gammas. Grayscale might change a bit.
Nevertheless the quality will be improved drastically.
_______________________________

In the current situation, the most critical problem is gamma.
So I do prefer the first approach

To solve the problem and get rid of any compromise, I will design the spreadsheet in a much better way


There is no rule.
Playing with gamma is particular.
Mainly for a general purpose.

I'm using the gamma control on some of my photos; it has been a long time to understand.
Of course the DLighting control is the Rolls Royce, but not always.

The gamma control is one way for normalyzing the dynamic of lighting.
What is better for one scene will be worse for the other.
The first rule is to have moderate actions.

As a general purpose, increasing gamma reveals the details in the mid range, where the gradient of the graph has the stronger variations.
So, going slowly, I take care not to add a seconf degree of variation.
Another point: when you look at the RGB histograms you can often see that they are not superposed. I mean that is the reason why sometimes when the gamma change is too abrupt, the colors turn un-natural.
The spline line could also produce some waves.
The second rule is to make the change smooth.

15% is still a dark shadow.
In photoshop or equivalent, moving the mouse over the image gives you on the histogram the brightness level. So you can experience where and what you should correct and improve.
Sometimes I adjust the black level in addition to the gamma.
I do not know how to transpose the photoshop results in the projector.

All is question of taste, feeling and testing.

Hi leDahu,

Thanks for the detailed response as always! I am having a little bit of trouble understanding parts of the post.

It seems like you are saying that using the custom gamma to calibrate RGB runs somewhat against using the custom gamma to calibrate W. And that you are updating your spreadsheet to provide an improved approach that handles the compromises involved better? I think this may be more clear once I see the updated spreadsheet.

The way I have gone about this is that I first used the custom gamma to get the grayscale flat. Then I used the W point to get the gamma correct. After changing the W point considerably, the RGB points needed some work to fine tune the grayscale again.

However, once I went back and touched up the grayscale again by tweaking RGB in the custom gamma, I was surprised to see that the gamma still measured the same. Perhaps this is because I was just making a couple clicks here and there in Red and Blue only. Had I been making changes with Green I suppose the gamma measurements may have been affected.

Based on this, the next time I do this I think I will do the W adjustment first. Then I will go back and do RGB. Then I will recheck to make sure that did not mess up the gamma, and tweak W again if necessary. Then again, perhaps your spreadsheet will lay out a different procedure.

Yes it's my opinion.

Wait until the next release.
The new approach will not handle the compromises.
It should avoid them -if I succeed-