Measuring CCD Linearity

My flats were taken as follows:

I made a translucent filter from two layers of the white plastic sheet that separates layers of Serena ham (it might be three - I made it a while ago) and used an empty 1 1/4" filter ring.

Then placed it on the nose of the CCD and placed the CCD on a shelf with the filter ring about 10cm from part of the wall appeared to have uniform shadow.

This is not an ideal solution - nor particularly repeatable as brightness of ambient light can vary over time. I used ambient electric light and avoided daylight hours.

For linearity test concurrently walked exposures up from min to max and also from max to min. Differences between the two curves can reveal that ambient brightness changed.

I did my stats on full frames. This is a weakness of my measurement. Are my flats as flat as they could be and if they are not then doing full frames probably isn't wise.

If I were to examine my flats and find a 100 x 100 pixel (or so) region that I can confirm is flat then measurements based on that ROI would probably give me more accurate data.

Ideally one would use a low-level light source as per the AIP (Berry and Burnell) and still search out flat sub-regions.

The device is described on pg 4 ff in

www.jlab.org/~salgado/notes/CCDReport.pdf

And in case that link doesn't work here it is as just text (I think)
www.jlab.org/~salgado/notes/CCDReport.pdf

CCD is cooled - single stage peltier - brings chip down to about 20 - 25C below ambient.

Dave you said: "but I'd imagine as long as temperature conditions vary by only a small amount over the imaging time that the data is still good"

A very good point. My CCD appears to have very low dark current. and small changes of temperature by 1 - 2 degrees isn't likely to affect things by much. Because I used relatively bright ambient light a lot of my exposures were less than a few seconds.

If I got my physics right you'd have to look at how the temperature changes on the absolute Kelvin scale (so zero Celsius is 273K - one or two degrees kelvin isn't much) - especially when dark current is so low and my exposure time so short.

I think my chip has an overscan but no software/firmware to utilise it.

Its all interesting stuff...

Mark C.

mjc wrote: ...two layers of the white plastic sheet that separates layers of Serena ham

Presumably without the ham?

That sounds like a fairly good process. I've heard tell that incandescent lamps are better than the fluorescent (inlcuding "green") bulbs/tubes can be problematic. I also saw that two folks simply use sheets of white paper - the number of sheets varied until approx a 60 sec exposure gave max ADU's. I also followed the same process with simple laser printer paper.

My ROI was chosen by taking a very long dark (4 hours long!) and then examining the image and avoiding obviously regions containing lots of hot pixels. I narrowed it down to a region of 140 x 140 pixels - it just happened to very close to the center of the chip - though I don't think that mattered too much.

The practicalities were carried out by:

1. I got a cardboard box and lined it with black card
2. Imager went into the box
3. A lint-free cloth was placed tightly over the nosecone of the imager to prevent any paper particles falling onto the CCD window when the shutter was opened
4. Sheets of paper (I think about 15) were placed over the lint free cloth
5. I have a light-box - this was placed over the top of the cardboard box and swithced on. I let it "warm up" for about half an hour
5. Images were taken in my study with all other light sources off

Since my max exposure was about 80 or 90 seconds, it didn't take long to iterate through two exposures at each exposure duration (I started at about 0.1 sec). It took about an hour or so.
Images were taken at -20 deg c.

I took corresponding darks for flat calibration.

The ROI was cropped as well as the corresponding overscan region. The cropped areas and the overscan regions were then analysed and the data thrown into the spreadsaheet. Then I did the pixel-math with the duplicate ROI's - data again thrown into the spreadheet.

That's pretty much the process I wanted to write-up but haven't had the time yet...

Darks were easier but the duration far longer. They were taken at room temp but went up to 4hrs duration (and they are duplicated - so that was the imager out of use for about 16 hours).

Its a shame you have no software facility to overscan - perhaps an email to the manufacturer?

Anyhow - hope the above helps in some way to you and others following this thread...

Dave

Mark,

I've read Howell's book and have tried out some of the tests. However, Howell only briefly mentions Janesick method. There is a lack of detailed equations on linearity, gain, readout noise and dark current. Does Janesick book show detailed equations for example for gain and readout noise? Or does he have some published papers that show his methods for the above. I have tried on the internet, but had no luck.

Eamonn A

Eamonn

I haven't got the Janesick book yet (next few days) so I can't really answer your question. It is my expectation that it does go detailed and hope that it is covered. I have tried to rationalise this equation in my mind and can't honestly say that I understand how it comes about.

A quick scan of the internet got:
ftp://tesis.bbtk.ull.es/ccppytec/cp163.pdf
And on page 121 indicates that Janesick's method heralds from Janesick,
Evans, & Elliott 1987 - but I can't find that referenced document.

If I find some supporting logic/maths for the method (either in the Jansick book or otherwise) -I will post.

Mark C.

Eamonn - I used Richard Crisp's methodology. It takes a bit of navigating but all the data and a partial methodolgy is on this site:
www.narrowbandimaging.com/home_page.htm

I say partial in as much as I had to contact Richard (and Tim Khan) to expand on a few points I didn't understand. They (and others) can be found on the CCD-Tech Yahoo group. Richard frequently references Janesick.

I have the Howell book and it really is only an overview. I took delivery of the AIP book last week - its seems a lot more comprehensive but I haven't had a chance to read it in detail yet.

Hope that helps.

I got the Janesick book yesterday.

In quick response to Eamonn's question (if I got that question correctly)

The equation giving gain as the sum of two flats minus the sum of two bias all over sigma^2 difference in flats minus sigma^2 difference in the two bias does not appear to be in the book.

Janesick seems to favour the PTC as the method of choice for determining gain. There is, however, a reference / description of some method (and maybe related to the PTC) involving histograms but I know nothing, at the moment, of that technique.

The book is basically an aggregation of material published in SPIE papers over the years as well as some lectures at UCLA.

Equations presented are explained but, and I have had limited time looking at it, only to the degree that justifies them without going through tedious intermediate steps. I haven't worked any of these through myself yet.

My first impression was - I'm well out of my depth.

In the preface is the following sentence:-

"This book is written for a wide audience - from the novice to the advanced CCD user."

I'd love to know what Janesick regards as a "novice".

However, from experience, I know that things can look daunting and intimidating but perseverance usually pays off - if its something one really wants. However, it is encyclopedic and there will be stuff that will just have to be parked.

From the back cover:-
"James Janesick has authored more than 75 publications on CCDs, has contributed to 17 NASA Tech Briefs, and holds 12 patents for various CCD innovations. At the Jet Propulsion Laboratory, his work focused on scientific CCDs and support electronics utilised in many CCD NASA spaceborne imaging systems, including the Hubble Space Telescope, and the Galileo and Cassini projects. Janesick received NASA medals for Exceptional Engineering Achievement in 1982 and 1992."

From scanning the book one gets insights into certain technical specs of those missions - such as full well capacity etc. - and these little snippets also have their value. It never dawned on me, for example (a historical reference in the book) that Voyager was pre-CCD roll-out.

Bottom line - I'm glad I bought the book - but it's a book that one studies more than reads - and will take time to measure what net value I get out of it.

Mark C.

[Edit: When I say "encyclopedic" what I mean is it goes to depths and detail that one needs (in my case) to refer to the glossary because it's discussing unfamiliar ground and terminology]