Measuring CCD Linearity

I had a stab - just for the exercise.

Take a series of flats at various exposures from very short exposures to long enough to fill full-well capacity (you max out the ADU value).

Because your light source can vary over time you take images in pairs:

One max exposure followed by on min exposure.
Decrease the max exposure by a fixed increment and increase the min by the same increment.

You now have two series of exposures - one increasing in exposure - and one decreasing.

If your software does image stats you can take the mean of each frame as your data (or do same from a common region you believe to be flat - I don't think flatness is critical though).

For each mean value divide by the integration time to give you the rate of increase. Plot a scatter graph (or reverse the descending order series so it increases in parallel with the ascending series and plot a regular graph of each series).

For the linear range of your CCD you should get close to a straight horizontal line.

If the two series diverge (gaps between the lines) then your light source changed brightness during the test.

I got this method from The Handbook of Astronomical Image Processing (Berry and Burnell).

I played with this a few weeks back - CCD is still sitting on my desk (I did test indoors with a white diffusing filter inserted in a spare filter ring and pointed the camera at a wall in shadow - ideally you need a low-level light source using a constant voltage source to one or more LEDS.

I'm sure there are other techniques.

Does that help?

What! no preview! All my silly errors exposed to the masses!

Mark C.

Hi Mark - I'd rather not take the CCD off the OTA.

My preferred method is to take light frames of stars and do it that way. The process is similar though - take images of a field of star, doubling the exposure time for each image (1s, 2s, 4s, 8s etc). Then you measure each image... That's where I get stuck.

Measure what? In MaximDL, I get various parameters using an aperture over the various stars. Pixel value, intensity etc etc...

Funnily enough, I just ordered the AIP book.

Thanks for the reply though - I could pssibly use the method with T-Shirt flats on a uniformly cloudy day. Should be plenty of them around...

Dave,

This is something I had to do with my ST-2000XM which is ABG and prior to the arrival of my ST-10XME which is NABG.

Anyway, the process is quite trivial and can be done during the day using your flat box!

More specifically, take three exposures at each of 1", 2", 4", 8" etc etc ... define an ROI and extract the flux from the center of each image using the ROI (ex. 100 pixels). Load these values into Excel and extract the median value for each exposure duration.

Plot these median values against exposure in Excel and where you will very easily see the behaviour of your camera. My first generation ST-2000XM is linear to 22,800 ADU and, as such, all of my photometry work was done with this upper limit in mind. In fact, as a precautionary measure, I used 20,000 ADU as a mental limit.

Anthony.

PS. I just read Mark's reply and .... I like it!

More specifically, take three exposures at each of 1", 2", 4", 8" etc etc ... define an ROI and extract the flux from the center of each image using the ROI (ex. 100 pixels). Load these values into Excel and extract the median value for each exposure duration.

OK - so what does ROI mean? I'm assuming it is something like "region of i_something_or_other".
Extracting the flux - how? If I use MaximDL, define a region and open the info window, which number should I use - the info window gives a whole series of numbers...

The median value and excel bit is grand. But if you could elucidate on the above, that'd be great.

Cheers
Dave

So long as what you measure doubles with doubling the integration (when in linear region of the CCD) then you are okay - but Anthony is right - you can leave the CCD on the scope and point it to a light box - or use a tee shirt.

The problem with doing stars is that local deviations from flat-field will throw your data as you go from linear to non-linear.

Even if you were to flat-field correct each frame your data is going to get distorted. Flat-fielding doesn't work when your data lies in (or very near) the non-linear region of the CCD. A measured value just within linear range may be scaled on flat-fielding onto a value you wouldn't get if flat-fielding wasn't needed. If you flat-field a value that is beyond linear range it isn't scaled correctly (and S/N - when imaging - gets worse).

The AIP book is excellant.

Mark C.