Inconsistent star FWHM values

Software Siril
1

Hello,
I have recently made some images in LRGB bands where I get inconsistent star FWHM values as follows:

  • When using the star detection feature in ASIAIR a get a average FWHM of about 2.1px

  • When using the tilt function in siril it gives a truncated mean FWHM of 2.14px

  • After registration the plot of the FWHM shows values of 5.7 - 5.9px

Its the first time I encounter such discrepancies. The difference to other images is that I am using broadband filters, so maybe it because the brightest stars are overexposed over the diffraction pattern of my Tak Epsilon 130 is becoming visible even for weaker stars and is influencing the result ?

Regards,
Goetz

2

Hello, I don’t understand what “other images” you are referring to, are they different images those with 2.14 and those with 5.8?
Registration will modify image data because of interpolation, the recent versions of Siril with the lanczos4 interpolation and its clamping will limit by about 8% the loss in FWHM that would be seen with the pixel area relation interpolation we used to use as default, but they won’t more than double like that because of registration.
Depending where you look and what information is available to Siril, FWHM may be displayed in pixels or in arcseconds, that would explain these differences.

3

Thank you Vicent,
to clarify what I mean, normally the FHWM of ASIAIR, Siril tilt FHWM values and the ones listed after registration are about the same - and are all in pixels. Typical values for good seeing around 2px for my specific setup.
Tonight and last night I did some broadband images and here I get around 2px FWHM with ASIAIR and Sirils tilt measurement, but around 6px FWHM as measured with registration. When I measure the tilt on the registered images, the mean FWHM is given as about 2px again. So only the values calculated during registration are off by about a factor 3. Pixel size is 1.8arcsec.
G.

I used the Lanczos-4 algorithm with Siril version 1.3.0-alpha-dev commit 272407be8.
Using the bicubic algorithm doesnt change the result

Will try narrowband images soon for comparision tonight.

Goetz

4

OK, with just two Halpha images just coming in I can say already that there is no discrepancy for narrowband images.

5

The way to compute FWHM does not change if your image is broadband or narrowband.

As the name says, this is a truncated mean. Spurious data are rejected.

In registration, all data are used.

6

you should look at the Dynamic PSF tool, to see mean FWHM for a selection or for the image, or information about individual stars. It would be a better comparison than the tilt tool.

1 Like
7

PSF gives values below 2.5px.

My guess is that in the broadband image more stars show the diffraction pattern of the secondary mirror holder and therefore the FWHM values of the registration (which seem to include all stars) are unreliable.

This is quite OK, it is just confusing when you first encounter it.