Problem with Flats

Hey Mark

Open discussion

I sorted the initial "FLAT Problem" but then the topic of Characterisation reared its head, I've learned a lot already particularly from Dave, things I had never even thought of before and it all makes sense, but it's getting the right data in the right place to get it right

One thing for certain, Astronomy and Astrophotography is a constant learning curve

Simon

Hi Mark - well said. PTC and DTC curves give certainty about the various parameters of your CCD rather than you relying on either the manufacturer's spec (they use terms like 'typical') or what other owners have calculated. I certainly learned a lot about my own imager - much I have yet to use but is necessary for producing good photometry. I also have to be honest in that whilst I characterised the CCD, it was just after that that other parts of life took over and astronomy had to take a back seat. So this is a good exercise for me in remembering what I did and also more precisely about what the numbers mean and how they can be used.

Some of the data may be superfluous to your kind of imaging Simon - but some of it will be useful, particularly for flats. Certainly, I have some software packages that request gain and full-well - so they can calculate an optimum exposure for a given purpose (e.g. pointing exposures). Knowing precisely what my gain and full-well is, is very useful.

So, onto your data... And before I kick off, I think part of the problem is that I didn't read the Stark material. I've done that now and see that he doesn't mention overscan at all and the methodology is different to Crisp's (who bases his methodology I think mostly on Janesick but Howell might be in there too). So we're a little at crossed purposes. But no matter for now...

I had a muck around - I'm struggling a bit to figure out what to do about the lack of overscan data - and more importantly, what impact it has on interpreting the charts. The data I'm interpreting at the moment (this is real rough and taken must be taken with a very large pinch of salt) is that Full Well is coming in at around 15000 DN, I can't get the gain but it seems to be very low. Read noise seems to be about 200 DN which is preposterous. Hence the large pinch of salt. Your maxing out at about 82 sec mark with the lights (with however many sheets of paper you used).

Linearity is OK from about 20K to about 58K DN - mine was similar first time round. Second set of images gave me time round, I got a near perfect straight line response across 90% of DN. That was down to a better measuring technique. You may see exactly the same second time round...

This is what would be helpful:
1. See if you can get details of the how to get into the overscan region of the KAF8300 (email Atik)
2. If you get it, before you shoot lights/darks again, let me know. I'd like to see more data for the lights, particularly around the mid to lower end. Darks seem to be fine but with the overscan (if) I'd shoot everything again. Hopefully you can remember roughly how many sheets of paper you used - if no, no harm. Next time round, I'd go for maybe a couple of extra sheets to lengthen exposure times before you max out - say around the 90 to 100 sec mark. instead of low 80's.
3. If you don't have any luck with overscan, I dunno...

Do you have the manufacturer's (or others) data re gain, read noise and full-well?

One final thing - is it OK if I send Mark on the spreadsheet with your data? He might see a lot of stuff I can't.

Thanks
Dave

Hi Dave, sure send on the info.

I have the sheets of paper in a document wallet, so I will re-use the same number of sheets

According to Atik
Very low read-noise, 7 electrons, on a par with many Sony sensors, enabling faint details to be detected.
Great linearity, with a regression correlation coefficient of R = 0.9998 over the range 1,000 to 64,000, making the camera suitable for sensitive photometric measurements.

Thanks - I got the spreadsheet.

The thing about inheriting spreadsheets is that they always appear complex for the recipient - so I've taken some essential data and reworked in a new sheet - instructions below.

To get linearity and gain:

1. I started a new sheet
2. Copy Cols A and B from old sheet to Col A and B in new sheet
3. Copy Cols I and J from old sheet to Cols C and D - but paste special - as values (no formulae)
4. Mark all table (exclude headers) sort by col A in ascending order
5. Delete all rows that start with "A"
6. Delete last row (for exp 2s - Average ADU - after Bias correction is bad)
7. Delete row for exp 82s - ADU's unrealistically low - out of general trend
8. Plot Mean ADU vs Exp Time (s) - I used scatter graph with joining line.
Visually appears linear up to around 60,000 ADUs - possibly as high as about 62,000.
One could apply a regression line (trend-line) display graph and R value and start deleting the highest data-points in sequence on R^2 is maximised and that might be a little more precise. But I reckon visual identification should suffice.

So we can estimate (pretty well) the full well capacity and the gain.
I'll have to do a bit more playing to see what else can be determined from this data.


9. Add a new coloumn (E) - title Gain (e-)
and populate cells with formula =c5/d5^2 for row 5 (An average ADU for a region divided by the square of the std dev for that region yields the gain of the CCD)
copy and paste into remaining cells of column and C5/D5 should increment for each row
Top three values and bottom two values are out of general value of 0.3 something
So we take an average of E8:E19
I did this by placing the formula = AVERAGE(E8:E19) in cell F19
The average is 0.35 (to two decimal places) - this is the gain of the CCD - and is close to values of around 0.4 that I see when trawling the web.
That means you get one ADU (or DN) for every 0.35 photo electrons.

What I have done avoids using the overscan region - it uses data values less "signal offset" - which is the bias.
As far as I know the overscan region is preferred because it is a better source to determine what the bias is than using "real" pixels - I don't know why that is - but I'd like to find out.

By plotting a PTC (or DTC) one has another method of ascertaining the gain.
I've tried a PTC on the data - but I too am rusty and was getting VERY low values for the gain - will have another go later.

Dave's suggestion for more data points is good - if there are more values in the middle and low end the graphical method may yet deliver results.

Hope this helps - I threw this text together a little more hastily than I would have liked and I hope it serves rather than hinders.

Mark C.

Hi Mark

One question, because the "Original" spreadsheet contains values in column E and F, these are having an effect on columns I and J which you want me to copy, so surely I should Zero columns E and F as I have no "Overscan" data, would that be correct? My Flats had the corresponding "Dark" subtracted, so BIAS was removed also

Simon

Also Column N "Offset Correction" has a value that is affecting I and J also, so should that be 0 for the same reason?