[argyllcms] Re: [Possible bug in Spyd2.c] Inconsistant readings from Spyder3

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Sat, 07 Mar 2009 02:50:32 +1100

howdy555@xxxxxxxxx wrote:
The spyder 3 produces values of exactly 0.0 all the time for black on my CRT.
So even if there is a problem of erroneous readings of something close
to zero, it would be indistinguishable from valid measurement...

I figure it's showing this problem all the time on my
display, because the brightness is below the quantization
threshold, even with an integration count of 32e6.

Could I ask you to try to measure very dark grey (I tried using XYZ: 0.196480
0.200497 0.384248) instead. Maybe you would be able to reproduce the problem.
Please make at least 20 measurements before giving up :)

If I choose and R=G=B on my display to give a Y of about 0.2, this is what
I get:

Reading X   Y   Z
1   0.000000    0.175008    0.124630
2   0.000000    0.176625    0.126857
3   0.000000    0.177723    0.127206
4   0.106032    0.187275    0.159078
5   0.106109    0.187766    0.162659
6   0.000000    0.122717    0.193857
7   0.000000    0.180656    0.130376
8   0.107101    0.190537    0.163706
9   0.000000    0.182667    0.132824
10  0.109261    0.191781    0.165216
11  0.156973    0.134002    0.232604
12  0.000000    0.182397    0.132627
13  0.109114    0.191958    0.165993
14  0.109026    0.191553    0.165962
15  0.158718    0.134652    0.232149

so the X value seems to be quantized below about 0.1.
Given that typical LCDs have a white of about 400, this is
0.025%, not unreasonable in many ways. Measuring my
CRT it's not so good, since it's brightness is about 50
(i.e. this then amounts to 0.2% for my CRT).

Also, could you make the same test but with full white and about 20
measurements? Does the result decrease, increase or stays more or less
the same?

It depends on the stability of the display of course. Many LCD
displays drift markedly on warmup, sometimes for extended
periods of time (20 minutes ?) - I know my MacBook display does.

On my CRT it bounces around a bit (possibly because the Spyder 3
doesn't seem to have the same smarts as the Spyder 2 to synchronize
it's readings to the refresh rate), but there were no obvious
trends.

[For comparison the i1pro produces XYZ values of 0.048 0.051 0.064,
and the Spyder 2 values of 0.035 0.042 0.043 ]
So the error is 0.064 -> perfectly acceptable for me :) I was
referring to errors of about 0.200 when measuring a value of about
0.200 :).

Right, but what's your white level ?

It's interesting that the specifications for the Spyder 3 seem to
be less comprehensive than the Spyder 2.

Spyder 3:

        Accuracy (x,y typical)        0.0025

Spyder 2:

        Accuracy: Chroma x,y:         0.0020 (CRT), 0.0035 (LCD)
        Luminance:                    Y Typical +/- 1.5%
        Dynamic Range:                Black Point: 0.02 cd/m2,
        White Point:                  > 5,000 cd/m2

Graeme Gill.

Other related posts: