KPF has two 16Mpx CCDs for its detectors for each red and blue arm of the spectrograph. The setting of their operating points and waveforms, and their characterization was performed at Caltech. I was the lead of the analysis of the detector characterization data to ensure the detectors are ready for precise RV measurements.
An image of one of the KPF detectors.
Some key things to check for with CCDs are low read noise and good charge transfer efficiency. When charge is read in a CCD, it is moved to a capacitor whose voltage is read before and after the charge arrives, then the difference in the two voltages gives the amount of charge. This is called correlated double sampling (CDS). The measurement of the capacitor charge doesn’t, and shouldn’t happen once. While the charge is there, it is read out continuously for some amount of time in what is called the ‘video signal’. Transients in this signal in time due to, for example, oscillations in the signal from the reset pulse, can be excluded from the average signal to improve the read noise. In KPF we unfortunately saw that the read out electronics were introducing an oscillation in the video signal that we could not remove without serious intervention. Luckily we were able to tune the read out parameters to still meet the 4 e- read noise requirement. Other things that can destroy your read noise include not properly grounding your read out electronics, but that’s a fixable problem.
Example video signal showing a read out of two pixels. The difference of the reset level and the data level give the charge in a pixel. Source.
One cool thing we did with the KPF detectors was hook them up to the JPL spot projection laboratory (which is actually at Caltech) to measure the intrapixel sensitivities (just like how there are pixel to pixel variations in sensitvity, a pixel itself can have different sensitivities within itself). The setup basically allows us to project onto our detector an image of a mask that has thousands of tiny pinholes cut into it. This mask is illuminated with an LED of our choice and can be scanned very precisely. By shifting these tiny spots around, we can probe how sensitive a pixel is in different regions inside the pixel. Below is a gif that shows a zoom in of some sub-pixel sized spots moving across the detector in small subpixel steps.
Subpixel sized spots moving around one of the KPF detectors.