Measuring CCD Linearity

Hi Dave,

To achieve linearity result depends on your camera type. Is the temperature relative or absolute. This is useful to know before attempting the linear test below.


In order to check the linearity of the response of your CCD, you could carry out measurements using the camera with a lens and LEDs. For example you could place eight LEDs in a sphere. The sphere is matt black inside. Attach the camera lens to the CCD and locate the CCD at small hole of the sphere. You then could image inside of the sphere for various exposure time ranging 5 to 80 seconds. For each exposure time, take 5 frames. The measurements could be done by using slow readout mode, and the dark component is subtracted. You could derive a mean count of 100 ◊ 100 pixels near the centre of the field against exposure time.

You would end up with a very accurate measurement with a mean deviation from the range of ADU values. Your non linearity degradation should show up from the ADU count you achieve at 65,000 ADU.

I hope that helps.

Eamonn A

The intensity *could* be in ADUs - my understanding of the word flux is that it is the total light gathered from a star. That is the sum of all pixels under the central aperture compensated for the background - which is where the annulus comes in (outer ring).

The sum of multiple pixels could easily exceed 65535.

However, I really don't think that this Intensity value is what you want for the task at hand.

You can try using a star - in which case use the maximum valued pixel within the aperture - but I feel that using low-light flats and a region of interest (and average an area) is the better way to go.

MaximDL can do such a 100x100 area - it calls it Area mode (and use region mode to mean something to do with examining RGB colour planes).

This page helps a little:


www.cyanogen.com/help/maximdl/HID_VIEW_INFORMATION.htm

Mark C.

mjc wrote: The sum of multiple pixels could easily exceed 65535.

Dang - now you say it, of course it could. I guess I've been getting hung up on the number 65535 - but that applies at the pixel level. Sure, sum all the pixels in the target aperture, take away the sky annulus and hey presto, you've have a number much larger than 65535... I've just had a Doh! moment.

I have trouble with taking any measurement of a single pixel value. That seems to me to be asking for trouble. If it happens to be a hot pixel on a dim star it would skew the results. And with sensor having 1048576 usable pixels (being a CCD it has an overscan region - which is also very useful), if you took say 10 samples, that less than 10ppm...

I think I'll go with the PTC and DTC curves anyhow and forget the whole star idea. Unless I get tied in knots.

Eamonn - the ProLine gives an absolute temperature - from ambient to 70 c below ambient. I've heard tell that LED's can be problematic - but I'm chasing the folks on the CCD-Tech forum to get a definitive answer. There was one mention of the sphere idea - but that was highlighted as the "holy grail" method.

By the way, if anyone is interested, here is a link to both the PTC and DTC methods (thet're closely related anyhow - one sues darks, the other uses lights):

www.narrowbandimaging.com/images/ptc_tal...nal_comments_web.pdf

Dave

Dave,

You had me shaking my head with the message above (actually a couple of messages up) and thanks to Mark for reminding you that we are summing pixel values and whose total easily exceeds the 65,535 max.

You also mention hot pixels ... my mistake for not mentioning that we also apply bias and dark frames as part of the process!

Anthony.

PS. When checking the linearity of the sensor, we are literally doing photometry (!) and, as such, all of the associated rules for photometry (vis a vis reduction) apply.

mjc wrote: However, I really don't think that this Intensity value is what you want for the task at hand.

Mark,

It is PRECISELY what we desire since the intensity tells us the behaviour of the 100x100 pixels as a group (or the star if using this route). Now, double the exposure and the individual as well as the collective behaviour of the 100x100 pixels (or star) should also double etc etc etc etc.

Anthony.