To stack or not to stack, which is best?

[John Coffin]

Hi, I am learning how to do photometry in order to flux calibrate my spectra. I am using an 80 mm refractor and an Atik 460 with a field flattener.

I have been using ASTAP to plate solve and calibrate the individual images. I wanted more control over the process so changed from using ASTAP’s photometry facility and am now using Astroimage J.

I can’t image for more than 10 seconds before my target is saturated. I collected 20 frames. I found that processing the frames individually or after stacking them made very little difference to the result.

A typical spectroscopy session lasts for an hour, and I aim to measure the magnitude of the target during this time.

How many frames should I collect to get an accurate result? I suppose the advantage of processing the frames separately is that I could spot any major fluctuations in the target’s magnitude.

Any thoughts?

[Robin Leadbeater]

The nice thing with the individual measurements (both in photometry and spectroscopy) is you can use them to make a true estimate of the stochastic uncertainty (though not any systematic errors)

[Andy Wilson]

Hi John,
In theory stacking the frames or combining the photometric measurements should give similar results as you have found.
With short exposures of 10 seconds, you need to be careful of atmospheric seeing. You might see changes due to atmospheric turbulence from frame to frame.
One option would be to slightly defocus the camera. This will spread the light over more pixels and lower the peak, so you should be able to expose for longer. Obviously you need to avoid blending the light from nearby stars, so you only want a slight defocus.
Andy

[John Coffin]

Great, thanks for your advice Robin and Andy. I assume you use the propagation of errors formula Δz = √((Δx)² + (Δy)² etc ) to work out the total uncertainty.

[Andy Wilson]

Hi John,
If you are combining errors/uncertainties from individual images then you also need to divide by the square root of the number of images.
A useful check is to calculate the standard deviation on the magnitudes derived from the individual images. This won’t be identical but should not be vastly different. The standard deviation with give you a direct estimate of the random error, and large differences could indicate systematics that are over or under estimated in the formula used to calculate the individual magnitude uncertainties.
Best wishes,
Andy

• This reply was modified 2 weeks ago by Andy Wilson.

[John Coffin]

Thanks Andy. I forgot about the square root of the number of images which in fact Google AI explained to me. AI is not all bad!
Best wishes, John

[Author]

Posts