Cmos imaging sensor response

[nbo10]

Hi all,
I have a question about the response of a cmos imaging sensor. The total exposure is defined as the irradiance * pixel area * exposure time and it can be varied by changing either the irradiance or the exposure time. Would the sensor response remain constant if measurements are taken with variable irradiances and exposure times under the condition that the total exposure is constant?

My naive thinking would assume that the sensor would measure the same value at constant total exposure independent of the irradiance and exposure time values. Or are there other details, particular the circuit used convert charge to a digital number, that effect the sensor value? Thanks

 

[NobodySpecial]

My naive thinking would assume that the sensor would measure the same value at constant total exposure independent of the irradiance and exposure time values.

An ideal sensor measures the accumalated number of photons - it doesn't care about the rate they arrived.
In practice real sensors have a dark current, as signal that builds up with time even without light, and a leakage where very low level signals may be lost

 

[nbo10]

An ideal sensor measures the accumalated number of photons - t

Well this is wrong. Photons are not accumulated. There is a photon to charge conversion and charge carrier is accumulated. Now, the conversion of photons to change is linear but the conversion of charge to digital number might not be linear. The charge-to-digital number conversion is the heart of my question.

 

[NobodySpecial]

Well this is wrong. Photons are not accumulated.

Didn't say they were - I said it measured the number of photons accumulated - as opposed to measuring their rate as a photocell or photocathode might.

conversion of charge to digital number might not be linear.

The charge to voltage on the pixel unit cell is inherently not linear for a CMOS sensor - as opposed to a CCD where it almost is. But most real devices, especially consumer units, are run in a region where they are pretty linear.

The charge-to-digital number conversion is the heart of my question.

The per pixel charge->voltage conversion and the output amp are generally pretty much linear - it's too hard to do calibration otherwise.
You could make the output amp->ADC non-linear to match the response of photographic film or of an older tube TV camera but most don't.
The normal way is just ot have a programmable gain stage and switch in a different linear multiplication before the ADC.