Skip to main content

Exercise #4 Calibration Proceedure

2 posts / 0 new
Last post
nlx
nlx's picture
Exercise #4 Calibration Proceedure

My typical calibration routine begins on a cloudy night before an observing night.

I build up libraries of dark and bias frames for different chip temperatures.  I thought initially that I should aim for the lowest chip temperature I could for each imaging night.  So, one night I might achieve a chip temperature of -26C, and another night -24C and soon.  When cloud was about, I would set about doing bias/darks at those temperatures, for exposures typically used during the imaging night.  It was soon realised that this was a very laboroius task.

After some thought and discussion with others, it was decided to bulld a library at set temperatures and set exposures.  For example, bias/darks might be done at -5C, -10C, -15C, -20C and -25C.  On the night, if the ambient temperature was higher, the camera might only reach -17C with a 90% cooler applied to the camera (powers over 90% were not used as the set temperature might not be stable).  The camera would be set at -15C, and whilst this might increase the SN slightly, the Peltiers would be running well in their capacity and their would be a stable chip temperature.

On the cloudy night the library was being built, the chip temperature might be set to -15C for example.  A number of bias frames would be taken, usually 100. Then a series of darks would be done.  Typically 100 of 1s, 5s, 10s, 50 of 30s, 60 and 100s, 40 of 200s and 25 of 300s.  At the completion of the run, median combined masters would be made of each series. These darks have had the master bias removed from each of them.  This is necessary as the darks will have to be scaled - a 40s exposure might be used for a science frame and a 60s dark would have to be scaled. Median combining eliminates cosmic ray hits, and though the the median combine process has a higher SN than averaging, the number of frames taken dramatically lowers the SN ratio.

At my observatory, sky flats are performed. I have the camera cooling before sunset, then open the dome and point the telescope to near the zenith.  When I am about to commence the exposures, I ensure that the scope is about 10 degrees east of the meridian as I don't want the scope to do a pier flip.  I have a milk perspex cover that goes over the front of the telescope to help obtain a better flat.  The filter sequence used is typically BIVRg'r' - a rough order of increasing sensitivity to the sky.  The biggest dilema I find is when to commence exposures, and what exposure length to use.  A long exposure of say 10 seconds will virtually eliminate any shutter effects, but in the time the exposures take for a given filter, the average ADU might drop by more than 50%.  I have arbitrarily settled on an exposure of 2-3 seconds to try and get a comprimise. Usually 11-15 exposures are done for each filter.

The temperatire used to obtain the flats is usually different to that used during the night, thus a chip temperature of -10C might be used at dusk, but an temperature of -15C during the night.  Care has to be taken to use darks of the same temperature as the flats.  Master flats are made for each filter.

Each science frame is then calibrated with the master bias, appropriate dark and flat frames.

The master dark and bias frames are done about every 3 months, and the flats usually each week.

After completing this course, I am interested in finding out how much the SN of the darks change with temperature.  There has been an assumption that the difference in SN between 5 degree increments in the master darks is not that significant.  This assumption should be assessed with a view to reducing the incremental step in the temperatures the darks are taken at.

I have never examinied the statistics of the darks from each run to see if there are significant differences between them.  It might be necessary to increase the frequency at which the calibration libraries are done.

Reading the article on flats by Arne, it has been realised that more QA needs to be done on them.  I have not critically examined the quality of the flats taken using a shorter exposure vs using a longer exposure and subsequent demination of the average ADU count.  Sky flats need to be taken in the evening as well as morning to look for gradients when one set is divided by the other.

Peter Nelson

HQA
HQA's picture
darks

[quote=nlx]

...Median combining eliminates cosmic ray hits, and though the the median combine process has a higher SN than averaging, the number of frames taken dramatically lowers the SN ratio.

After completing this course, I am interested in finding out how much the SN of the darks change with temperature.  There has been an assumption that the difference in SN between 5 degree increments in the master darks is not that significant.  This assumption should be assessed with a view to reducing the incremental step in the temperatures the darks are taken at.

I have never examinied the statistics of the darks from each run to see if there are significant differences between them.  It might be necessary to increase the frequency at which the calibration libraries are done.

Reading the article on flats by Arne, it has been realised that more QA needs to be done on them.  I have not critically examined the quality of the flats taken using a shorter exposure vs using a longer exposure and subsequent demination of the average ADU count.  Sky flats need to be taken in the evening as well as morning to look for gradients when one set is divided by the other.

Peter Nelson

[/quote]

Not quite sure what you mean by SN - is this signa/noise ratio, or SNR?  Median combine gives poorer SNR than average combine, but as you say, eliminates cosmic rays.  You always pay a penalty when rejecting images.  When I'm trying to image as faint as I can, for example, I combine the science frames with averaging/summing whenever possible.  Sometimes you can't avoid a cosmic ray near your object of interest without median combine or some sort of pixel rejection, but I always try to maximize my signal/noise.

Dark current doubles for every 6C increase in temperature for the typical CCD.  However, most CCDs have extremely low dark current in the first place, so imaging at -10C or 0C makes little difference - for the average pixel.  What kills you are the "hot" pixels, where they are affected nonlinearly with temperature.  It is not uncommon for a hot pixel to saturate at 0C while the rest of the array is still at 1 ADU mean.  So running as cold as you can, within reasonable limits with your TEC and humidity, is always a good thing to do, as the hot pixels decrease rapidly.

The main point is to experiment, and then to compare everything.  Plot the mean dark current as a function of temperature.  Histogram to find the hot pixels; extremely bad ones you might want to document and avoid.  Ratio flats to see differences.  Learn how changes affect your calibration.

Arne

Log in to post comments
AAVSO 49 Bay State Rd. Cambridge, MA 02138 aavso@aavso.org 617-354-0484