I had to put the 16 bit driver circuit on hold for right now, there is an over heating problem with the ADC that I cannot solve right now, I have 2 boards made up and both have the same issue.

So I have decided to go ahead and build the fast 8 bit drive instead because I already know that it works flawlessly, and can still read all 3648 pixels from the CCD, It reads 3694 pixels, but the first 30 and the last 16 are dark reference or dummy pixels and are discarded.

I have already drawn the schematic using Design Spark and will post the gerber files so you can order your own PCB and build it. I am including the BOM also.

This next set of panels are from the 16 bit driver circuit

This was the tail end of a CFL lamp (2300K/13W) but the ADC starts to overheat and you can start to see the results on screen.

Although the spectral image is inverted, this was the best initial capture of the CFL lamp, from left to right; the 2 most important calibration bands are the blue bands on the left, which should be at 432nm and to the right at the green band should be at 546nm this gives the best "fit," so to speak.

This was the first scan "dark counts."

As I stated earlier, I'll still be working on this problem, but I need my project to start moving forward and it will with the 8 bit driver and still have very good resolution because of the optics and 1800 ln/mm holographic grating I am using and Raman edge pass filter, the only difference will be digitizing a frame will take 32ms instead of 16ms.

Other than that, there isn't too much of a real difference.