Correlated Double Sampling and Noise

[Anonymous]

Hi all,
While we are on the subject of noise, here is some related information. If you do a search with google.com on "correlated double sampling noise" you will get more information. A lot of it having to do with CCD image sensors.
I used correlated double sampling with CCD image sensors back in 1978 for Sacremento Peak Solar Observatory. Learned a bit about the transfer function this makes and other fun information. And low and behold here it is again in PI metal detectors.
Now this is a subject we can get our teeth into.
I'll put up another post with another URL.
Be sure and read the end part of this one. Has pictures of the noise transfer functions.
JC

 

[Anonymous]

Hi all,
The correlated double sampling method also has the disadvantage of aliasing white noise in the baseband. It is also aliasing external noise in the same way. This comes about because sampling at the pulse rate (say 1000 to 10,000 Hz) is undersampling if the front end has a bandwidth of 100,000 Hertz. So what this does is follow the Nyquist folding criteria of folding the higher frequency noise signals into the low frequency spectra of the output of the differential integrator, or differential sample and hold routine. Either way this increases low frequency signals which are going to look like the low frequency sweeping target frequencies.
One answer would be to reduce the bandwidth of the front end amp, but that is no fun unless looking for very large targets, as it will reduce sensitivity (voltage wise) alot. Another would be the raise the sampling frequency up, if this is possible. Or in some cases the noise would go down with increases of sampling rate (pulse rate).
Still doesn't fix the sampling frequency, beating with external signals, which can also inject large low frequency signals into the system.
JC