Luminous Landscape Forum
Raw & Post Processing, Printing => Colour Management => Topic started by: GWGill on February 16, 2016, 07:19:35 pm
-
I've been tending to use my i1Pro Rev E (i.e. i1Pro2) for random spot
measurements a lot. Recently it's accuracy had been niggling at me.
What I noticed was this: From cold, do a calibrate.
Measure spots. The errors would rise (literally the L* was rising).
The closer the measurements together, the faster the errors would rise.
Let it cool, and the errors would drop. Recalibrate hot, and the error
would increase as it cooled down.
At its worst, I was seeing errors of 2.0 Delta E between cold and hot!
[ It's pretty easy to replicate this experiment - ArgyllCMS spotread
makes it straightforward, but you could do something similar with
ColorPort etc., repeatedly measuring the white calibration tile.
I noticed this type of thing before with one of the i1pro Rev A's I've
got, but it wasn't as extreme, and I put it down to being an old lamp
in a well used instrument. ]
Then it got weirder. This characteristic didn't seem stable. I started
looking at how it was affecting strip measurements, and it kept changing.
Sometimes it would get better. Then it would get worse again.
Then the light went on (so to speak), and I arrived at a theory.
What if the lamp is a miniature Quartz Halogen lamp ?
This would make sense in terms of how long the lamp should last,
QH lamps typically last 2-5 times longer than conventional
incandescent lamps, and would also be helpful in ensuring constant light
output over the lamp life, something that is rather desirable for
an instrument light source. I measured a color temperature of
just over 3000K, which is consistent with a hotter QH lamp.
QH lamps (https://en.wikipedia.org/wiki/Halogen_lamp) work on the basis of re-circulating evaporated
tungsten back onto the filament. To do that they have to get hot enough to
evaporate any that gets deposited on the inside of the glass, and even hotter
to disassociate the resulting halide back into tungsten at the filament.
If the spot measurement gets the bulb hot enough to evaporate
tungsten from the filament and the glass, but not quite hot enough
to deposit it back at the same rate onto the filament, then the glass
will build up a layer of tungsten on it. This reduces light output.
As it warms up a bit, the tungsten gets evaporated from the glass,
increasing light output. When it gets cold again the tungsten
gets deposited back on the glass. Hence my increasing L* error
as the instrument warms up, and hot/cold behavior.
But if the bulb gets warmer still, say by being used for strip
readings, then the tungsten gets re-deposited on the filament,
and the bulb gets "cleaned". So after some extended strip readings
(i.e. several measurements in a row, made with the button down
and the lamp on for as long as possible), the instrument no longer
displays such dramatic error increases with temperature. In fact
after "cleaning", the worst I could get was 0.08 Delta E, a 20x reduction,
and back in line with instrument specifications.
If you predominantly use your i1Pro for reading strips, you
probably haven't noticed this effect at all, because your lamp
isn't getting dirty.
I've seen similar effects on all my i1Pro's, and also Spectrolino's,
although none as dramatic as the Rev E was.
So if you do lots of spot measurements with your i1Pro's, and get the
impression that your accuracy and consistency of measurement is getting
squirrelly, you may be right. But there is a way of fixing it!
[ Also posted to the ColorSync mailing list. ]
-
What is the way of fixing it? Paying X-Rite a fortune of money for "service"? I wonder how prevalent this problem is - i.e. whether generic or specific to your device.
-
So if you do lots of spot measurements with your i1Pro's, and get the
impression that your accuracy and consistency of measurement is getting
squirrelly, you may be right. But there is a way of fixing it!
so for example if I take my i1Pro2 (relatively new and never used to read strips/charts - only spot measurements and probably less than 5000 clicks total so far), calibrate it once ("cold") and then do for example 100 sequential spot measurements of a single patch (granted spots will fluctuate because of the manual placing) with say 1-2 sec delay between measurement I shall see some clearly detectable L* drifting ?
-
What is the way of fixing it?
firmware can (at the expense of the lamp life ?) just detect that you are not readings strips and force lamp heating to keep tungsten balance.
-
so for example if I take my i1Pro2 (relatively new and never used to read strips/charts - only spot measurements and probably less than 5000 clicks total so far), calibrate it once ("cold") and then do for example 100 sequential spot measurements of a single patch (granted spots will fluctuate because of the manual placing) with say 1-2 sec delay between measurement I shall see some clearly detectable L* drifting ?
Use the calibration tile if you want to measure the D.E. change.
-
Use the calibration tile if you want to measure the D.E. change.
you mean use the same tile that is used for instrument calibration for the spot measurements ? but if there is an error because of this tungsten vapor flow then it shall be clearly seen on any patch as well (granted brighter patches are better), no ?
-
What is the way of fixing it? Paying X-Rite a fortune of money for "service"? I wonder how prevalent this problem is - i.e. whether generic or specific to your device.
1) I suspect only a few people will have seen this problem.
2) If you discover you are one of these people, I described exactly how to fix it.
-
you mean use the same tile that is used for instrument calibration for the spot measurements ? but if there is an error because of this tungsten vapor flow then it shall be clearly seen on any patch as well (granted brighter patches are better), no ?
Yes, but measuring random patches is no way to diagnose problems!
You diagnose accuracy problems by measuring something that is known and consistent, like the calibration tile.
-
1) I suspect only a few people will have seen this problem.
2) If you discover you are one of these people, I described exactly how to fix it.
Where did you describe exactly how to fix it? Maybe I'm missing something, but I re-read your opening post twice and I found ample description of the problem and your hypothesis about why it occurs - most interesting - but I did not see how to fix it.
-
But there is a way of fixing it!
Please kindly enlighten! A friend has done a similar experiment and shown the i1Pro to be unstable in scan mode for the first 16 passes or so. The L* drops steadily, instead of rises, from the first measurement pass until around the 16th pass, then holds fairly stable from that point onwards. That's a brilliant theory about why we are noticing the inconsistencies in measurements! I think it very well may be the truth. So how should we go about fixing the problem? Depress the i1Pro button to activate the lamp continuously for a certain amount of time to heat the lamp sufficiently until it's 'cleaned'? Then begin measurements and do not pause for too long between measurements?
-
Where did you describe exactly how to fix it?
"So after some extended strip readings
(i.e. several measurements in a row, made with the button down
and the lamp on for as long as possible), the instrument no longer
displays such dramatic error increases with temperature."
60 seconds accumulated "on" time seems a good place to start.
I wouldn't recommend doing this unless you need it, and
can verify that it helps - there is no point in running down the
lamp life for no reason.
-
"So after some extended strip readings
(i.e. several measurements in a row, made with the button down
and the lamp on for as long as possible), the instrument no longer
displays such dramatic error increases with temperature."
60 seconds accumulated "on" time seems a good place to start.
I wouldn't recommend doing this unless you need it, and
can verify that it helps - there is no point in running down the
lamp life for no reason.
Ah, OK, thanks.
-
A friend has done a similar experiment and shown the i1Pro to be unstable in scan mode for the first 16 passes or so. The L* drops steadily, instead of rises, from the first measurement pass until around the 16th pass, then holds fairly stable from that point onwards.
A steady cadence is a workaround for the problem, but you have to do dummy reads until it stabilizes.
I notice the L* dropping with successive measurements when it is "dirtying" the lamp (assuming my theory is correct), which seems to happen more dramatically for measurements longer than normal spot readings. I've only tested 1 - 5 seconds as doing this, and I'm not sure at what length measurement it stops "dirtying" and start "cleaning".
Depress the i1Pro button to activate the lamp continuously for a certain amount of time to heat the lamp sufficiently until it's 'cleaned'?
I wouldn't use lamp life up without good reason.
Then begin measurements and do not pause for too long between measurements?
Maybe that can help.
-
Or maybe press X-Rite to make an instrument that works properly?
-
Or maybe use the i1Diagnostics?
i1Diagnostics when checking the instrument also set some parameters if necessary.
Good luck.
-
I have 3 I1 spectros. One, an M0, is almost 10 years old. I have a uv Cut that is about 6 years old and an I1P2 about a year old. On a colorchecker the oldest has average dEs of about .9 compared to the I1P2 and the uV cut is about .5 dEs different.
-
I tested my i1Pro 2 Rev E (3 months old) in cold and hot without recalibration, also I retested without recalibration after measured 540 patches, I compared the measurements from the previous day with the current in every way.
Measurements were made with 10 single point measurements and analyzed with i1Profiler
Always against the tile.
In all comparisons the Delta E 2000 is always less than 0.01.
-
In all comparisons the Delta E 2000 is always less than 0.01.
Maximum or average ?
None of my 4 i1Pro's are that consistent with regard to warming up. At their best I get DE76 of 0.07.
My Spectrolino is about the same or slightly worse.
With carefully controlled cadence, I can get consecutive measurements that are consistent to 0.003 - 0.008 DE76, but the change with consecutive measurements causing warming up is always worse than that.
[ Others have reported similar observations to mine on the ColorSync list. ]
-
First of all, on all comparisons the Delta E 2000 is always less than 0.1, not 0,01, sorry for the mistake.
I sent a copy of the i1Profiler's screen few minutes ago, in which ten measurements made yesterday after 75 minutes of heating and 544 patches measured are compared with ten measurements made today after only one minute of heating, with totally cold instrument.
Always against the tile.
Two questions arise:
** Have you checked your i1Pro with i1Diagnostics?, this software when checking the instrument also modify some parameters if necessary: http://www.xrite.com/product_overview.aspx?Action=support&ID=766
** All my measurements were performed with i1Profiler software. Perhaps the software performs any correction ?. I suggest that you take the same test with i1Profiler
Last, my i1Pro 2 today have a lamp burning time of 8.762.2 seconds.
Good luck!
-
With carefully controlled cadence, I can get consecutive measurements that are consistent to 0.003 - 0.008 DE76, but the change with consecutive measurements causing warming up is always worse than that.
[ Others have reported similar observations to mine on the ColorSync list. ]
Surely you don't mean 0.003 - 0.008? If I can get 0.03 - 0.08 I would be delighted. I'm mostly seeing 0.1 deltaE2ks with my i1Pro 2 Rev E.
-
after 75 minutes of heating
how did you manage to keep the lamp burning for 75 minutes uninterrupted ?
-
If my lamp, today have an usage time of 8.762,2 second, how can you think that before my test I heated the lamp during 4500 seconds (75 minutes)?
Hi...Hi.
Good luck.
-
If my lamp, today have an usage time of 8.762,2 second, how can you think that before my test I heated the lamp during 4500 seconds (75 minutes)?
then what do mean exactly by "after 75 minutes of heating" which was "made yesterday" ?
-
So if you do lots of spot measurements with your i1Pro's, and get the
impression that your accuracy and consistency of measurement is getting
squirrelly, you may be right.
I tested my i1Pro2 using BabelColor PatchTool (non XRGA mode) - one calibration and then imitating manual spot reading patch by patch on a calibration tile with ~2 sec delay between readouts
this is the file with the 300 sequential spot readouts = attached
L* =
(http://s26.postimg.org/d7kaczdrb/L_300.jpg)
a* =
(http://s26.postimg.org/90zfxne5j/a_300.jpg)
b* =
(http://s26.postimg.org/leca4k3tz/b_300.jpg)
-
Could there be the possibility of manufacturing differences between halogen bulbs to account for why some don't have this warming problem? I don't know how quality control is managed in QH lighting.
-
The I1 is much more stable reading the same spot but I've noticed variations averaging about .10 dE going up to about .5 dE on more saturated colors in different positions of the same printed patch. Also seems to vary based on paper texture (glossy, pearl, semi, matte...).
Anyone know of products that make v4.3 profiles using the new Float32Number based LUTs? They would seem to offer much less inversion error in BtoA -> AtoB and enhance proofing accuracy. OTOH we also need products that can utilize these new types.
-
now do the same again __BUT__ with BabelColor PatchTool "lamp restore" done once just before the series of 300 manual, patch by patch measurements (of the same calibration tile - click were by mouse, not through button on the device - so position on tile was pretty the same, just in case)
the measurements attached
L* = the same scale as above - see the difference :
(http://s26.postimg.org/44vt5ye07/L_lr_300.jpg)
a* = the same scale as above - see the difference :
(http://s26.postimg.org/hzu3of8fb/a_lr_300.jpg)
b* = the same scale as above - see the difference
(http://s26.postimg.org/4wyh55i7b/b_lr_300.jpg)
funny that a* stays OK, but b* still drifts...
-
quote from PatchTool manual
(http://s26.postimg.org/uts5hrlux/BPT.jpg)
-
now with Spectrolino + argyll spotread, 400 readouts off the calibration tile (a different one from i1Pro2 naturally)
Argyll 'V1.8.3' Build 'MSWin 64 bit' System 'Windows V10.0 SP 0'
Device: Spectrolino
Serial No: 17984
Part No: 36.55.52
Prod Date: 9/8/2004
SW Version: 1.68
the data attached
L* =
(http://s26.postimg.org/bjpqc2p9j/L_spectrolino_17984_1_400.jpg)
a* =
(http://s26.postimg.org/cn9uo19wn/a_spectrolino_17984_1_400.jpg)
b* =
(http://s26.postimg.org/9hp8xtraf/b_spectrolino_17984_1_400.jpg)
-
FWIW, I saw a .1 dE76 shift on my 10 y/o I1 after doing the bulb renew. Considering how old the thing is I was half expecting more. They are solid instruments.
-
the same Spectrolino, after some rest, reset, couple of recalibrations
the data attached
L* =
(http://s26.postimg.org/ds3wtewdj/L_spectrolino_17984_2_400.jpg)
a* =
(http://s26.postimg.org/sckzo8rc7/a_spectrolino_17984_2_400.jpg)
b* =
(http://s26.postimg.org/4a8r7402v/b_spectrolino_17984_2_400.jpg)
-
Attached a screenshot of i1Profiler comparing two sets of 30 measurements:
Measurement A: 10 patches measured patch by patch, from cold after calibrating, plus 10 patches measured 10 minutes after without recalibration, plus 10 patches measured 45 minutes after without recalibration. Total 30 patches.
Measurement B: 30 patches continuously measured, patch by patch after calibration, the instrument measured from cold and followed with 2 second pauses between measurements.
All against the tile.
De 2000 standard error: 0,11
De 2000 maximum error: 0,33
-
First of all, on all comparisons the Delta E 2000 is always less than 0.1, not 0,01, sorry for the mistake.
Right - that matches my experience of an instrument that is in good shape, and is within the manufacturers short term repeatability specifications.
The phenomenon I'm referring to takes the instruments outside their specifications. Subsequent experimentation with my i1Pro2 indicates that something like short strip reads (1-5 seconds lamp on time) with about 5 seconds off time between strips is quite effective in developing this behavior.
** Have you checked your i1Pro with i1Diagnostics?, this software when checking the instrument also modify some parameters if necessary: http://www.xrite.com/product_overview.aspx?Action=support&ID=766
As far as I can see (looking at how i1Diagnostics works, and what operations it performs over USB), it doesn't do anything special - just operates the instrument like any application does. I presume it analyzes the resulting measurements in a way that normal applications do not, but it doesn't appear to modify the instrument in any way.
The report "Reflectance drift test: Successful" is intriguing - perhaps this test is looking for exactly the type of thing I've noticed >
So I tried "dirtying up" my i1Pro2 - 50 x 2 second measures with 5 secs between increased the thermal drift from 0.08 DE to 0.18 DE, and then ran i1Diagnostics on it.
But it still reported "Reflectance drift test: Successful". So either it's not testing for this, or it's error threshold is quite high.
After 60 seconds of "cleaning", it was back to 0.07 DE thermal drift.
[ Interestingly, I tried "dirtying up" one of my i1Pro RevA's, and couldn't budge it - drift stayed at 0.06 - 0.07. Either the effect is very device instance specific, or the i1Pro2 has a different type of lamp in it ? ]
** All my measurements were performed with i1Profiler software. Perhaps the software performs any correction ?. I suggest that you take the same test with i1Profiler
I've cross checked the behavior using X-Rite driver output, as well as my own driver. The driver is what feeds the measurements to applications such as i1profiler, and yes, I've checked that i1profiler doesn't manipulate the numbers it is fed by the driver. So it is a characteristic of the device, not the software (although how it is driven certainly influences the behavior.)
-
Surely you don't mean 0.003 - 0.008? If I can get 0.03 - 0.08 I would be delighted. I'm mostly seeing 0.1 deltaE2ks with my i1Pro 2 Rev E.
Yes, that is correct. "With carefully controlled cadence" and without moving the instrument, on the white calibration tile.
i.e. not normal measurement conditions.
-
I tested my i1Pro2 using BabelColor PatchTool (non XRGA mode) - one calibration and then imitating manual spot reading patch by patch on a calibration tile with ~2 sec delay between readouts
I calculate a shift of 0.26 DE. So outside spec., but not in the direction I would expect if it was due to the effect I'm talking about.
-
Could there be the possibility of manufacturing differences between halogen bulbs to account for why some don't have this warming problem? I don't know how quality control is managed in QH lighting.
There's certainly variation - one of the i1Pro RevA's I've got doesn't seem to show this effect at all, whereas the i1Pro2 I have is quite repeatable. Whether the variation is systematic is another question.
-
I calculate a shift of 0.26 DE. So outside spec., but not in the direction I would expect if it was due to the effect I'm talking about.
"burning" the lamp with a special option in PatchTool in the 2nd test of i1Pro2 makes it work better... now if you can add something to argyll utilities like spotread to burn Spectrolino lamp instead of a user trying to burn "manually" it though measurements ?
-
one of the i1Pro RevA's I've got doesn't seem to show this effect at all
may be it's lamp so old that it reached the state of spectrometer's nirvana ;D
-
quote from PatchTool manual
Right - so X-Rite know all about this, but just haven't told anyone but SDK licensees ?
-
may be it's lamp so old that it reached the state of spectrometer's nirvana ;D
Maybe :-)
It's got about 9 hours lamp use on it, but that isn't much, even for a pure incandescent lamp.
If it's actually QH, it might be expected to last 1000Hours, although the constant on/off
works against that. On the other hand, I'm reasonably sure the i1 drivers it with a "soft start",
so the lamp life shouldn't be excessively degraded by constant on/off.
-
"burning" the lamp with a special option in PatchTool in the 2nd test of i1Pro2 makes it work better... now if you can add something to argyll utilities like spotread to burn Spectrolino lamp instead of a user trying to burn "manually" it though measurements ?
I'm thinking about what adding a "check for drift and try to fix" would involve.
-
GWGill,
In one of my test with i1Diagnostics it says aproximately "adjusting the instrument, this is done occasionally", and for a minute or two he worked.....he did?, I do not know.
At that time unaware that i1Diagnostics creates a log with quite a bit information.
-
In one of my test with i1Diagnostics it says aproximately "adjusting the instrument, this is done occasionally", and for a minute or two he worked.....he did?, I do not know.
This didn't trigger in my testing, so I'm not sure what it would be doing.
-
I'm thinking about what adding a "check for drift and try to fix" would involve.
It will be the "-Yl" to test, and "-YL" to test and fix options in the next release of ArgyllCMS/spotread.
-
It will be the "-Yl" to test, and "-YL" to test and fix options in the next release of ArgyllCMS/spotread.
great, can't to see how it works with Spectrolino !
-
great, can't to see how it works with Spectrolino !
Sorry, i1Pro only - I don't know of any way of turning the Spectrolino lamp on for a length of time.
-
Sorry, i1Pro only
oops :(