Affiliation
American Association of Variable Star Observers (AAVSO)
Tue, 08/30/2022 - 13:21

I did some observations of a star last night and processed the images using Maxim DL. When I do the photometry analysis using MaximDL on a single image it gives me an error of 0.001.  The SNR of the star in the image is about 1400.  I actually took 5 images over a 2 minute period. If I process each individually and then work out the standard deviation of the 5 values of magnitude I get an SD of 0.005. This seems a more realistic value for the error but if I ask MaximDL to export a txt file for submitting as an extended format file it will obviously put the 0.001 figure in for the error.

So the question is, which of the two error values (0.001 or 0.005) would it be better to submit?

All the best

Steve

 

Affiliation
American Association of Variable Star Observers (AAVSO)
Apples and Oranges

Steve:

This is a great example of the failure of most/all photometry software to yield a 'realistic/representative' error when only one comp is used for analysis. By math, when only one comp star is available, a standard deviation cannot be calculated from the single value. Thus, most/all software reports an error calculated on the basis of only 1/SNR. Thus, you got an error of 0.001. A rounded / truncated value corresponding closely to 1/SNR (not exactly how it is calculated).

It is much better to calculate the random error / standard deviation with the use of multiple measurements/images.  In this case, you did exactly that and got a standard deviation of 0.005 from the five calculated magnitudes. The key issue here is that you must be sure that the target has not varied at all during the total duration (e.g., 2 minutes). Tough when you are working with variable stars! Of course, you could do the same with a comp star (constant?) in your 5 images.

Unfortunately, this better procedure is not easy to carry out automatically when you are conducting normal photometry with available software. At least, it is tedious/annoying enough that most observers will not do it! Neither do I. ;-)  Alternatively, I'm a believer in using multiple comps (ensemble) since the calculated error (standard deviation) is much more realistic. This continues to be a point of contention.

Ken

 

Affiliation
American Association of Variable Star Observers (AAVSO)
Thanks Ken, you've confirmed…

Thanks Ken, you've confirmed pretty much what I thought. The star I was observing has a period of about 3 days so I thought over 2 minutes the star wasn't going to change much. One problem with using multiple comp stars is that often there are only 2 or 3 available and you have to use one of those for the check star.

Thanks again

Steve

 

Affiliation
American Association of Variable Star Observers (AAVSO)
SNR over 1000

Hi Peter,

It is relatively easy to get a reported signal/noise ratio over 1000.  Remember that Poisson theory says, for an otherwise error-free signal, the random error is

SNR = signal/sqrt(signal) = sqrt(signal)

So SNR=1000 means signal = 1M.  If this were concentrated in one pixel, then you would have to have an analog/digital converter that could resolve 20 bits.  However, for photometry, you are using a circular aperture of many pixels and counting the total acquired signal.  If the aperture was about 6 pixels in diameter, it would have a total area of 100 pixels.  If the signal was uniform over the aperture area, then this would result in a signal of 10K electrons per pixel, easily resolved with the typical 16-bit ADC.

That said, my usual rule of thumb is to look at stars that approach SNR=1000 very carefully, as the signal is NOT uniformly spread over the measuring aperture.  You can pretty quickly empirically find in your images the SNR or uncertainty value that means a star is likely saturated with your equipment.  For the CCD used at USNO-Flagstaff (which typically yielded 2.5pix/fwhm), for example, things saturated when the reported error was around 0.003mag.  To get higher precision, I had to either defocus or stack images.

Arne

Affiliation
American Association of Variable Star Observers (AAVSO)
Thanks Arne. The  FWHM of…

Thanks Arne. The  FWHM of the target star is 6.47 and I'm using an aperture of 20 pixels diameter which seems OK according to the guidelines. I've looked at the graph of the star image and I'm pretty sure it's not saturated.  When observing I'm careful to increase the exposure in steps until I get a high maximum value (but less than 40000) without saturating either the target star or any of the comp/check stars. In this case the exposure was 20 seconds and the camera temperature was -20C

The intensity of the star is about 1.33x10^6

Do you think I'm doing something wrong?

Steve

 

Affiliation
American Association of Variable Star Observers (AAVSO)
The camera is a ZWO…

The camera is a ZWO ASI1600MM Pro. I'm using it Bin 1x1. The software is MaximDL 6.29

I've checked it again and what I said seems right. I've also checked the raw, unprocessed image to look for saturation. The maximum pixel value is 26160.  The saturation value is 65504. Although the camera is 12 bit the value is shifted to the left by 4 bits which is where the 65504 comes from.