> A question for those who live in the world of GMP/GLP.
>
> We have a number of imaging systems for gels, including densitometers, CCD
> Cameras etc. We have tried to address a system suitability test
> (performance qulaification) for these. While we have processes in place, I
> am never satisified that we have totally solved the issue.
>From Rob Watson of Alpha Innotech Corp, a leading maker of gel imaging
systems:
One needs to specify exactly what is to be calibrated. Are the measurements
meant to be absolute? ( Very tough ) Relative measurements against known
standards are the norm as they are easy and traceable. For the most
stringent and demanding use of CCD's one must look to the astronomer. All
calibration is done relative to known stars of specific spectral classes
for colorimetric measurements. Intensity measurements are also specified
relative to certain wavelengths from specific stellar objects.
Measurements of step wedges are subject to many variables and were
originally defined arbitrarily by Kodak based on performance of specific
lots of film stock. The variation in illumination must be corrected. The
glossy finish of the step wedge causes reflective distortion that can be
corrected. A deviation from flatness or excessive texture causes
distortion. The CCD has an intrinsic variation of ~1% across the surface in
QE that can be corrected.
The closest to a gold standard that any aspect of CCD performance has been
rated is the charge transfer efficiency. A radioactive isotope of Iron
emits a specific energy gamma ray that deposits a known quantity of energy
into a single pixel. This energy generates free electrons that are then
transfered from the target pixel out to the readout node. The number of
transfers is known and the resluting loss of charge then determines the per
pixel transfer efficiency.
Ultimately, absolute measurements and calibration may be impossible at
worst, exceedingly tedious and unstable at best. Relative measurements are
much more robust and traceablity is based on the standards used.