linearising: qspline not producing good curve
2007-07-07 by David Burren
I'm still having the odd issue with linearising. One example is a target that on first scan produced non-linear L* values, with QTR- Linearize-Data reporting: > The Lab values are not in order. > Cannot be linearized. Perfectly understandable, and a fresh scan (after cleaning a couple of dust motes off the print) produced linear values that QTR- Linearize-Data _was_ happy with. It produced a LINEARIZE string which I then fed into quadprofile, which then reports: > Invalid linearize curve -- not constantly increasing Huh? QTR-Linearize-Data is happy, but quadprofile isn't... It turns out that when fed the following data the qspline program doesn't produce a good curve! LINEARIZE_CURVE="0,0 3.61,2 7.24,4 11.1,6 14.88,8 18.56,10 22.33,12 26.46,14 30.89,16 35.86,18 41.73,20 45.52,22 50.19,24 53.85,26 55.79,28 58.68,30 61.29,32 63,34 65.73,36 68.29,38 69.55,40 72.04,42 74.12,44 75.25,46 77.63,48 79.38,50 80.23,52 82.1,54 84.04,56 84.5,58 86.4,60 87.72,62 88.09,64 89.55,66 90.35,68 90.81,70 92.13,72 93.21,74 93.29,76 94.63,78 95.47,80 95.72,82 96.8,84 97.48,86 97.73,88 98.72,90 99.11,92 99.26,94 99.74,96 99.94,98 100,100" ... 35279 35378 35382 35366 <<< the offending data 35452 ... With the change in slope of the curve at that point, the algorithm is producing a tiny bump in the spline curve, and the whole thing falls apart. Has anyone else run into issues like this? Is there any consensus on how to deal with it? Should I specifically NOT be trying to use 51-step targets for linearisation? I could use 21 steps for linearisation and 51 for profiling, but that seems like a bit of a hack workaround... The 51-step target used in this instance was the 51x4-random one I posted a link to yesterday, so most scanning errors should have been averaged out. Thanks for any advice __ David Burren