Klaus Karcher wrote:
There seems to be a second (dark) calibration at the beginning of each strip, maybe with different integration times (Graeme mentioned something like this when we met in Munich, but I can't remember the details).
For reflective strip readings, that is the dark reference. It ensures the same integration time, but I guess it also improves thermal tracking, and may compensate somewhat for mechanical effects as well (ie. if the guide spaces the instrument slightly from the media, the resulting light leakage will get subtracted out.)
When I compare the dark current values obtained from the initial calibration with the (assumed) dark readings at the beginning of the strip (213...233), band 3 catches my eye as it seems to be an extreme outlier (a "hot pixel"?) -- see<http://digitalproof.info/argyll/eye-one_IR_calibration/offsets.pdf>
Hmm. Perhaps the sensors are only qualified over the official wavelength range.
... but in the strip dark calibration, bands 1...4 get out of line:<http://digitalproof.info/argyll/eye-one_IR_calibration/boxplot_strip_dark_cal.pdf>
More conformation that they shouldn't be used I think.
- my failed attempt to interpret the raw data correctly (10 bands out of 127, strip calibration removed):<http://digitalproof.info/argyll/eye-one_IR_calibration/10_bands.pdf>
I'm not sure. All I can suggest is adding some printf's to output the raw floating point values from within the driver. Graeme Gill.