Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: Will SV II software be updated to automatically calibrate primaries in sRGB mode  (Read 1712 times)
WombatHorror
Sr. Member
****
Offline Offline

Posts: 299


« on: January 10, 2012, 01:50:47 PM »
ReplyReply

I was wondering if SV II will be updated so that instead of just calibrating the white point it would also automatically calibrate the primaries when in sRGB emulation mode.

Currently you either:
1. select sRGB mode and then leave things as is and use the original factory settings for all time

2. have to mess around doing it yourself by:
a. opening SV II probe window and start creating a custom mode with the multiprofiler sofware
b. setting sample color to 255,255,255 in SV II probe window and setting it to continuous read
c. adjust the xy coords of the white point in MP until the readout read 0.313,0.329 then going back to the gamut screen in MP
d. then setting SV II probe window color to 255,0,0 and then adjusting xy primary location for red in MP until you hit the 0.640,.330 and then write the values you had to set in MP to create that read out in SV II
e. repeat the above only for 0,255,0 green this time (matching to .300,.600)
f. and then for blue 0,0,255 (matching to .150,.060 only the PA screens often can't hit .150, many or most hit down to .152 so if you try moving the x coord below and nothing gets it lower than the default reading then just forget about it and leave the x coord untouched for blue)
6. set preferences in SV II so it does .icc profiles uses probe measurements for primaries (otherwise it would end up using the custom values you enter which is NOT what you would want)
7. in the broadcast or sRGB calibration instead of leaving it at the default sRGB gamut you need to select custom gamut and then enter the xy values for R and G and the y value for B that you found in the earlier steps)
8. then proceed as normal

It seems to me the software should be doing all of that for you, as an option, since that is what software is for.

For something like a spyder3 that might not read the primaries all that accurately it is probably left best as is in all the defaults for everything but with something like their custom i1D2 or an i1Pro and perhaps even the new i1 Display Pro (although Ihaven't checked the latter yet) then it seems like it might help compared to sticking with factory defaults, especially once the monitor has broken in??
Logged
WillH
Full Member
***
Offline Offline

Posts: 138


WWW
« Reply #1 on: January 11, 2012, 06:04:17 PM »
ReplyReply

We have put a lot of thought and research into this idea. The lab grade equipment used during the factory characterization is almost certainly more accurate than consumer level devices are at measuring the RGB primaries. While we could certainly modify the software to adjust the gamut to what your sensor thinks is the correct RGB primaries, chances are the factory calibration is more accurate, even after the monitor has been in use for thousands of hours.
Logged

Will Hollingworth
Senior Manager, Product Development
NEC Display Solutions of America, Inc.
WombatHorror
Sr. Member
****
Offline Offline

Posts: 299


« Reply #2 on: January 11, 2012, 10:44:14 PM »
ReplyReply

We have put a lot of thought and research into this idea. The lab grade equipment used during the factory characterization is almost certainly more accurate than consumer level devices are at measuring the RGB primaries. While we could certainly modify the software to adjust the gamut to what your sensor thinks is the correct RGB primaries, chances are the factory calibration is more accurate, even after the monitor has been in use for thousands of hours.

Thanks.

I was wondering a bit about that.

sRGB gamut:

One curious thing is that here and the readouts from an i1 Pro and a NEC i1D2 from sRGB mode:
and comparing sRGB emulation mode which I custom tuned according to what my i1 Pro said:

i1pro:
.639, .330
.301, .599
.152 , .061
whle i1d2 said:
.642, .328
.303, .600
.152,.061

for blue, both probes read back the exact same results
for green there is a .002 diff for x and for y they basically give exactly the same results, just .001 off
now for red things are a bit more in disagreement at .003 and .002

but it seems surprising they would both agree almost completely for the G and B primaries if the probes were so poor at reading the primaries (although it could just be chance, of course)

to get these values I had to switch from factory programming of .640,.400   .300,.600   .150,.060 to:
.638,.333
.291,.602
.150,.057

So the i1 pro and i1d2 read the blue the same but differ by .003 in y compard to the factory value.
For green they read nearly the same .002,.001 delta compared to each other, but delta to factory values of .010,.003 and .012,.002
And for red they differ by .003,.002 and compared to factory by .001,.003 and .004,.005.

So the i1 pro reads a bit closer to the factory settings than the NEC i1D2, the difference only being for red really. It reads closer a bit closer to the i1D2 than to the factory settings overall, mostly due to the x-coord of green (differing .002 with respect to each other but by .010 and .012 to the factory setting), so does that imply that even tuned consumer probes simply tend to mess up reading the x-coord of green primaries badly and in the same exact way or since they both miss it in almost the exact same way that the set does drift over time more than thought or that my set had a mess up when they measured the x-coord of green in the factory?

I guess it's hard to say with so few data points. I will measure it with an i1 Display Pro later this week and see what it says. Not that it will answer the question, but it would be interesting.


Native Gamut:

Another odd thing I noticed is that when in native gamut mode, while the factory settings and i1 pro and NEC i1D2 were at least in some ballpark for most values, for some reason they measured the x of G wayyy differently and, again, in the exact same way. Is it likely that both would read so far off in the exact same way?

i1pro:
R .679, .309
G .214, .690
B .152, .056
NEC i1D2:
R .683, .306
G .214, .691
B .152, .057

so we have delta .004,.003 for R and .000,.001 for G and .000,.001 for B so they are in remarkable agreement with each other, even for the native gamut coordinates, for G and B, if starting to differ somewhat on their assessment of  R.

the factory settings give:
R .678,.312
G .200,.694
B .152,.054

The deltas to the i1pro and i1d2: .001,.003 and .005,.006 for R; .014,.004 and .014,.003 for G; .000,.002 and .000,.003 for B.

So once again the i1pro reads closer to the factory settings and it agrees reasonably closely for B and R but once again the x coord of G is different, way different this time but both probes are in exact agreement as to the degree of difference.  (When the sets and probes were newer I though the i1pro and i1d2 read a little more closely on avg so perhaps my i1d2 is starting to drift a touch.) Here the i1pro agree more with the factory than the i1d2 when it comes to R but more with the i1D2 for B and much moreso for G.

So does this mean the factory setting for the x-coord of G got mis-measured on my set at the factory?
That the set does drift over time, noticeably for at least some components of some primaries?
That the consumer probes are really poor at reading the x-coordinate of the green primaries (and in this case, if they all have a bias for over-reading the x component of G then why do the manufacturers not apply a negative scalar to how they read x of green [well I suppose the probe would need lots of logic and some guessing as to what it was measuring then otherwise it would then toss off all of the other x measurements, they could add driver code to be informed when they were trying to read a green primary perhaps or maybe automatically scale more and more the close the measured location go out towards the greens?])? I wonder if I can find my old data from when i first got the set. I wonder if the factory readings for x of G were closer then or if it was the same story.

I wonder if the closer agreement between the probes than either to the factory settings means that, the i1 pro at least if maybe few others, does read the locations, other than perhaps the x-coord of G, better than what the factory settings are, at least after much usage??

Anyway I obvious have very few data points to go on and I'm doing a lot of perhaps rather wild, if at least a trace educated, speculating here.
Have you noticed any tendency for probes to mess up the x coord of Green primary readings? Any drift in the coordinates as the monitor ages?

(I actually would expect the native primary locations, at the least, to remain rather constant over time, but I don't really know much about the backlights and how their spectrum changes over time or whether the colored filters over the subpixels in sets change much over time.)

Once again thanks and sorry for all the babbling here.
Logged
Tim Lookingbill
Sr. Member
****
Offline Offline

Posts: 1146



WWW
« Reply #3 on: January 12, 2012, 02:18:07 AM »
ReplyReply

Do you see any differences to color managed images with this level of precision scrutiny with XY coordinates?

Just curious.
Logged
WombatHorror
Sr. Member
****
Offline Offline

Posts: 299


« Reply #4 on: January 12, 2012, 12:36:40 PM »
ReplyReply

Do you see any differences to color managed images with this level of precision scrutiny with XY coordinates?

Just curious.

I don't actually know hah. You may have a point there.

I never looked into that hah. It's probably not as easy as it sounds since ideally you'd want an instant flip quick A/B comparison but most software makes you have to do a few steps and often re-start each time you swap profiles and it's hard to remember subtle color differences over time. I might eventually be able to do that with some software I'm about to start writing. I guess I could, fo rnow, try to load up a photo and then go into multiprofiler and move the coordinates of a primary around and see how much change made it look obviously different. I think that tends to make the monitor kinda flicker in a sense though so even that might not be the easiest A/B comparison. I'm sure some photos would make things more obvious than others.

EDIT: or I guess I could calculate the dE and that might give some rough idea
« Last Edit: January 12, 2012, 12:52:00 PM by LarryBaum » Logged
WombatHorror
Sr. Member
****
Offline Offline

Posts: 299


« Reply #5 on: January 12, 2012, 05:25:18 PM »
ReplyReply

Some interesting findings, just tried an i1 Display Pro on it. Quick story is that it seems that this probe is closer to factory measurements than i1 Pro or NEC i1D2 (although I didn't test this one warmed up yet will update soon) or a spectrcal calibrated DTP94b or an older ColorEyes DTP94b. It is not shock that it would be closer than the DTP94b since SV II applies no compensation for that probe.

It seems that the greatest differences between the sRGB probes like DTP94b and the others are in the x value (of xyY) of the white point and the xy values of the green primary. The DTP94b actually didn't measure all that far off for the red or blue primary compared to the wide gamut/wide gamut compensated probes or all that far off on the y of the white point (although it was starting become a bit more different there). But the x of the WP and the xy of green primary were well off by .08 to .18 so that's pretty far off on the other values it often was just .000-.003 different though.

Anyway my i1 pro and nec i1d2 disagreed quite a bit with the x value for the green primary compared to the factory measurement. The DTP94b had crazy trouble with this so it seems this may be the single trickiest value for a probe to measure properly at least the spectral spikes on green on the PA series sets. Both the i1pro and NEC id2 agree with each other but not with the factory measurement. So which was truer? The factory measurement after thousands of hours of the two probes in agreement?

Well the i1 Display Pro didn't match either the factory or the i1pro/nec id2 but it crept much closer to the factory measurement and was closer to the factory than two the other two probes. So my wild guess, based on not enough information, is that the spectral spikes for green on this wide gamut and perhaps most others have a lot of energy that gets missed by the wide spectra sample size of the i1 pro and that something about it may prove troublesome even for a tuned colorimeter such as the NEC i1D2. The i1 Display Pro colorimeter is spectrally tuned to the PA series and this perhaps lets it get a better reading.  So my wild guess is that maybe most of these probes do have trouble measuring the green primary, at least the x coord, and that even after a lot of drift due to thousands of hours of usage that the factory measured x value for green may be more accurate. OTOH the other probes I think agree more with each other overall for the other primary values than with the factory settings so perhaps the better probes could measure them about as well or better, after time, than the original factory settings. Perhaps they do drift a bit though and the i1 Display Pro has the x of green a bit better than the factory settings and perhaps with this probe it would make sense to use it to reset all primary locations even not necessarily with the other probes, at least not for some of the values.

only tested srgb emulation not native gamut yet though

anwyay sorry for sloppy rambling post, in a rush

here are prelim data taken with broadcast, gamma 2.2, 93 brightness mode, comp level 1:

the custom tuning that made the calibration (i.e. what the factory settings think it should be reading):
WP .307,.327
R .638,.333
G .291,.602
B .150 (treat this as .152),.057

i1 Display Pro:
WP .311,.329,92.6
R .640,.329
G .295,.603
B .152,.059

i1pro:
WP .316,.334,95
R .642,.330
G .301,.601
B .152,.062

NEC i1D2 (not warmed up, so not really valid, will post warmed values later):
WP .319,.335,94.8
R .645,.327
G .303,.600
B .152,.063

DTP94b 1 Coloreyes:
WP .331,.333,92
R .642,.330
G .318,.589
B .155,.063

DTP94b 2 from Spectracal (they claim they calibrate ones with their HDTV kit):
WP .329,.331,90.5
R .642,.331
G .315,.592
B .153,.062

For the WP luminance:
The DTPs report lower, perhaps because they filter too much of the wider gamut out.
The i1 Pro and NEC i1D2 (not warmed) report almost the exactly same luminance.
The i1 Display Pro actually comes out closer to the DTP, especially one of them.
Which is more correct, who knows? I doubt the one 90 reading from the one DTP.
Why the other DTP and i1 Display Pro get 92/92.5 and the i1pro and NEC i1d2 get 95 I don't know. I guess a 2-3% difference isn't huge but I thought this the one thing that you always claims about how just about any probe measures this perfectly....
Of course it is true that spot to spot monitors vary, especially in luminance so the shapes of the probes perhaps led me to place certain probes slightly over compared to others. I should try to make sure to align the measurement spot and probe eye very carefully and redo the luminance stuff. 2-3% could easily be placement spot.

Really hard to say with such small sample size, but it does seem to hint that having SV II calibrate the primaries might not be worth it after all unless perhaps you use an i1 Display Pro.  You might not necessarily do worse overall with the NEC i1D2 or i1 Pro but it does seem that whether you would do any better overall is in doubt, although, again, can't say much with so few probes sampled. But yeah NEC may be right to say why bother when their is a good chance it might just be a wash at best.
It is still possible you might do a trace better on average, but who knows, maybe you'd do a bit worse with an avg copy of those probes.

There does seem to be some possibility that calibrating the primaries for sRGB actually might help if you use the i1 Display Pro though. Perhaps.
« Last Edit: January 12, 2012, 05:46:45 PM by LarryBaum » Logged
WombatHorror
Sr. Member
****
Offline Offline

Posts: 299


« Reply #6 on: January 12, 2012, 07:14:54 PM »
ReplyReply

OK I let the NEC i1D2 sit on the screen for about twenty minutes and then I let it read continuously for about 25minutes. Of course about 90 had passed since I first measured so the monitor itself may have drifted a bit, anyway:

earlier, pre-probe warmup/after the probe warmed up (and monitor had been on longer):
.319,.335,94.8 / .316,.331,94.45
.645,.327 / .644,.327
.303,.600 / .3025,.600
.152,.063 /.152,.062

So the primary measurements really didn't drift at all with the monitor having been on longer of the probe having gotten to warm up but the white point measurement very much did change and it seems the NEC i1D2 measured it quite a bit more accurately after being warmed up (the continuous reads seemed to warm it better than sitting on the screen alone.

I quickly re-did it again with the i1 Pro ["( )" for the earlier measured values]:
.315,.3325,95.4 (.316,.334,95) so maybe the monitor drifted down .001 or so over the time?
.640,.329,20.43 (.642,.330)
.302,.600,68.8 (.301,.601)
.152,.062,7.1 (.152,.062)
not really much difference, maybe the monitor drifted .001 on WP over time taking this with the above

recalling the factory settings claim it should read:
.307,.327
.638,.333
.291,.602
.150 (.152),.057

If we assume that the truth would be getting close to the factory settings for the primaries and neither what the probes say nor the factory settings for the WP but when probes differ assuming that whatever read closer to the factory settings is more likely to be true (biasing in favor of factory settings having been relatively true when monitor was fresh because well just because plus the well re-regarded and brand new i1DPro tends to be a bit closer to them) then it seems taht:

1. my i1Pro may read the white point a bit more accurately than my year plus old NEC i1D2.
2. my i1Pro reads the red primary a trace more accurately than " "
3. they read the green primary identically accurately
4. they more or less read the blue primary identically accurately
5. the i1 pro reads brighter than any of the other probes (lack of filters make it more accurate? or is is less accurate than the rest when it comes to this?)

So it seems my year plus old i1 Pro calibrates this set, most likely, a bit more accurately than my year plus old NEC i1D2 other than perhaps for luminance. They both may suffer a bit in accurately reading the x coordinate of the green primary and to a lesser extent the y coordinate of the red and blue primaries.

Now as for the i1 Display Pro:
at first I got:
.311,.329,92.6 (later after quick measurement I got 93.5, the 92.6 might have been a weird spot)
.640,.329,19.8
.296,.603,67.4
.152,.059,6.72
after letting it read continuously, the second time, for maybe 10+ minutes:
.3095,.327,92.9
.639,.329,19.72
.2955,.6015,66.96
.152,.058,6.73

again recalling what the factory settings claim it should be:
recalling the factory settings claim it should read:
.307,.327
.638,.333
.291,.602
.150 (.152),.057

The after continuous read warm up of 10+ minutes is slightly closer to the factory settings.

The brand new i1 Display pro reads a bit closer to the factory settings than the year plus old i1 Pro did and closer than the year plus old NEC i1D2 did.

It mainly differed from the factory settings on how it read the y component of the red primary (all three probes, each of very different type did as well, so is it the probes better than the factory measurement at this time after much monitor usage or just that all of them struggle with that a touch in exactly the same way despite the different technologies and calibrations?) and a bit on how it read the x component of the green primary (although it was far closer to factory setting than any of the other probes, hinting that it or factory settings are most likely to be true I guess and that all the other probes probably do struggle with that a lot, it would be nice to test it with a second i1 Display Pro though and see if that agrees with this copy or with the other probes). And it was a tiny bit off in the x of the white point but that might be set drift. Is the fact that it was closer to the factory implying it is more accurate or that the set has drifted a lot and it and the factory settings are well off and the other two probes got it better?

I guess I will make a very tentative guess and suggest that:

1. my i1 Display Pro is measuring more accurately overall than either my i1 Pro or NEC i1D2 are at this point and that it might be more reliable than the factory settings at this point for all measurements. Although it might potentially be slightly under reading luminances. Brightness measurements themselves aren't actually that important anyway so long as they are within reason, who cares about 100 cd/m^2 vs 106 cd/m^2, doesn't matter at all. You set brightness by what is comfortable for the situation, you pick it yourself. Wht does matter is how it TRACKS brightness, since if it doesn't do that well than it will measure the tone response wrong. That I did not look into at all.

2. while the other better probes might possibly get a few bits of the primaries measured better than a well used set's factory values they probably get one or two much more off and that overall it might NOT be worth adjusting them (as NEC suggested above) and that if you do, to largely forget about adjusting the x coord of green, at the very least


anyway a lot of perhaps, maybe, etc. it's hard to be sure with only one copy of each probe, only one copy of the monitor and no $30,000 reference measurement equipment on hand

The new i1 Display Pro is demonstrating some potentially impressive initial results though, very much so for the price being only $200-250 for it.

« Last Edit: January 12, 2012, 07:55:31 PM by LarryBaum » Logged
WombatHorror
Sr. Member
****
Offline Offline

Posts: 299


« Reply #7 on: January 12, 2012, 07:21:26 PM »
ReplyReply

To sum up:

The i1 Display Pro shows tentative initial signs of being a very impressive unit and remarkable for the $200-250 price, potentially the best there is unless you go to like $1500+ class.

It seems, as suggested by NEC, that overall, it would be kind of a wash, at best, to use most probes to reset the sRGB primary locations. Perhaps if you only use them to reset certain values but not others (such as x of green), it might help a tiny bit, but who knows, it may not be worth it and it may be too tricky to figure out what would be ideal one's own.

Maybe it might be worth resetting the primaries if you use the i1 Display Pro though, if not any of the other probes out there. Remains to be seen for sure though.

Playing around with settings a bit, my eye seemed quite sensitive to changes in white balance with even just .002 difference being very readily distinguishable but that primary xy differences might be a little bit trickier to spot until they become a good bit larger.

I could probably say a lot more if I only had access to one or two more copies of each of the i1 Pro, NEC i1D2 and i1 Display Pro probes. Without that it is hard to say things without a lot of maybes and perhapses.
« Last Edit: January 12, 2012, 08:04:03 PM by LarryBaum » Logged
WombatHorror
Sr. Member
****
Offline Offline

Posts: 299


« Reply #8 on: January 12, 2012, 09:06:40 PM »
ReplyReply

And for native gamut (not that there are anything you tried to adjust to target gamut since there is no target but whatever it does naturally, but just to show how my copies of these probes measured the native primaries at this point vs. what the factory settings said) I got with deltas to factory settings in "( )":

With factory settings:
R .678,.312
G .200,.694
B .152,.054

With i1 Pro:
R .679,.309 (.001,.003)
G .214,.690 (.014,.004)
B .152,.056 (.000,.002)

With NEC i1D2:
R .683,.306 (.005,.006)
G .214,.691 (.014,.003)
B .152,.057 (.000,.003)

With i1 Display Pro:
R .679,.308 (.001,.004)
G .207,.694 (.007,.000)
B .152,.054 (.000,.000)

So the i1 Display Pro registered the same or super close the factory settings other than for y of R where it was .004 delta and x of G where it was a large .008 delta. It measured closer than the other two probes.

i1 Pro measure the next closest and the NEC i1D2 the least close (at this point in time)

« Last Edit: January 12, 2012, 09:17:01 PM by LarryBaum » Logged
WillH
Full Member
***
Offline Offline

Posts: 138


WWW
« Reply #9 on: January 12, 2012, 09:58:55 PM »
ReplyReply

Another odd thing I noticed is that when in native gamut mode, while the factory settings and i1 pro and NEC i1D2 were at least in some ballpark for most values, for some reason they measured the x of G wayyy differently and, again, in the exact same way. Is it likely that both would read so far off in the exact same way?

Keep in mind that you are looking at the numerical values in CIE xy space, with has the drawback of overemphasizing differences in green compared to red and blue. A seemingly large difference in xy in the green area may not be visually significant. You would get a more meaningful comparison by evaluating in CIE u'v' space than CIE xy. The aim of u'v' is to try and make color differences more visually equal over the colorspace area.
Logged

Will Hollingworth
Senior Manager, Product Development
NEC Display Solutions of America, Inc.
WombatHorror
Sr. Member
****
Offline Offline

Posts: 299


« Reply #10 on: January 12, 2012, 10:15:28 PM »
ReplyReply

Keep in mind that you are looking at the numerical values in CIE xy space, with has the drawback of overemphasizing differences in green compared to red and blue. A seemingly large difference in xy in the green area may not be visually significant. You would get a more meaningful comparison by evaluating in CIE u'v' space than CIE xy. The aim of u'v' is to try and make color differences more visually equal over the colorspace area.

Good point.

I also found that the i1 Display Pro registered values closer to the factory settings than the other two. I'd need more probes to say, but maybe it does hint that the i1 pro and so on are having a tiny bit of problem reading it. Anyway you are probably right that there isn't too much point to calibrating sRGB primaries with by far most probes (with the new i1 Display Pro or some super high end stuff, maybe, exceptions).
Logged
Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad