Binning, FWHM numbers in Sharpcap and SNR?
Posted: Thu Oct 29, 2020 11:45 pm
Hi folks,
An observation - which I hope might possibly be useful to folk (or maybe is well known) and also a question or two for Robin about how the software binning and FWHM calculation works in SC - and it's dependence on SNR ?
First the observation. I recently carried out a few experiments (details below) and noticed some inconsistencies in the comparative FWHM values (~ the same in both Sharpcap and DSS) obtained when switching between 1x and 2 x binning. To cut a long story short - under~ identical conditions of focus, exposure length etc the FWHM per pixel scores did not ~ half in value when I quickly switched from 1x to 2x binning. I was puzzled about why this would be because it seemed that just changing the binning from 1 to 2 had apparently shifted the estimate of the seeing (+optics, guiding etc) from 3.6 arcsec to 5.1 arcsec ? However neither the seeing nor optical configuration had really changed.
A man who posts on the FLO Stargazer's Lounge forum under the handle "Vlaiv" offered a simple explanation. He noted that the accuracy of the FWHM calculation in DSS - and I expect also SC? - depends on the SNR because it has to approximate where the base of the Gaussian star light curve is. Therefore calculation is inherently more accurate the higher the SNR. The 2X binned images would have had better signal to noise than the 1x and therefore offered a more reliable guide to the overall arcsec resolution. He also had a lab / artificial star rig in which he could practically demonstrates that longer exposures (another route to lower SNR) also resulted in higher (and presumptively more accurate) estimates of FWHM. In practice --high SNR seems to increase the FWHM estimate.
Firstly does the above ring true? (slightly disappointing for me --my skies are generally even worse than I thought they were). Secondly (I think I have read this before on the forum and I suspect that it is an obvious point) SC carries out the software binning step first and then does the debayering (Vlaiv claimed that the converse would not be expected to work so well for SNR improvement because debayering first would cause pixels to already be correlated)?
Thanks
Tim
---------------------------------------------------------------------
Expt details.
Using an SW Skyliner 250 mm F4.7 (f 1200mm) goto Dobsonian with a Baader coma corrector and an ASI294 PRO MC camera (4.63uM pixels). Experiment with 3s subs (gain 200, offset 29) to photograph star clusters (M34 and M103) in order to compare the results unbinned and at 2x2 binning, At 3s there is relatively little added blurring due to Dob field movement. The 2X binned and 1X images were taken within minutes of each other (so no substantial change in seeing or drift in focus).
The image scale is ~ 0.8 arcsec per pixel unbinned and 1.6 when 2 x binned. Based on DSS and SC scores the FWHM -pixel of 1X frames averaged about 4.2 pixel. Thus calculated that the FWHM in arcsec was about 3.4 (~ SQR of sum of squared contributions from optics, movement and the seeing).
What puzzled me was that when I went to 2x2 binning DSS and SC reported average FWHM- pixel scores of about 3.2 pixel whereas I was expecting perhaps about 2.1 ? i.e. with 2 x 2 binning the overall FWHM in arcsec would now appear to be about 5.1 and yet, relative to 1 x1 binning nothing in the optics, field movement rate or actual seeing should have changed?
An observation - which I hope might possibly be useful to folk (or maybe is well known) and also a question or two for Robin about how the software binning and FWHM calculation works in SC - and it's dependence on SNR ?
First the observation. I recently carried out a few experiments (details below) and noticed some inconsistencies in the comparative FWHM values (~ the same in both Sharpcap and DSS) obtained when switching between 1x and 2 x binning. To cut a long story short - under~ identical conditions of focus, exposure length etc the FWHM per pixel scores did not ~ half in value when I quickly switched from 1x to 2x binning. I was puzzled about why this would be because it seemed that just changing the binning from 1 to 2 had apparently shifted the estimate of the seeing (+optics, guiding etc) from 3.6 arcsec to 5.1 arcsec ? However neither the seeing nor optical configuration had really changed.
A man who posts on the FLO Stargazer's Lounge forum under the handle "Vlaiv" offered a simple explanation. He noted that the accuracy of the FWHM calculation in DSS - and I expect also SC? - depends on the SNR because it has to approximate where the base of the Gaussian star light curve is. Therefore calculation is inherently more accurate the higher the SNR. The 2X binned images would have had better signal to noise than the 1x and therefore offered a more reliable guide to the overall arcsec resolution. He also had a lab / artificial star rig in which he could practically demonstrates that longer exposures (another route to lower SNR) also resulted in higher (and presumptively more accurate) estimates of FWHM. In practice --high SNR seems to increase the FWHM estimate.
Firstly does the above ring true? (slightly disappointing for me --my skies are generally even worse than I thought they were). Secondly (I think I have read this before on the forum and I suspect that it is an obvious point) SC carries out the software binning step first and then does the debayering (Vlaiv claimed that the converse would not be expected to work so well for SNR improvement because debayering first would cause pixels to already be correlated)?
Thanks
Tim
---------------------------------------------------------------------
Expt details.
Using an SW Skyliner 250 mm F4.7 (f 1200mm) goto Dobsonian with a Baader coma corrector and an ASI294 PRO MC camera (4.63uM pixels). Experiment with 3s subs (gain 200, offset 29) to photograph star clusters (M34 and M103) in order to compare the results unbinned and at 2x2 binning, At 3s there is relatively little added blurring due to Dob field movement. The 2X binned and 1X images were taken within minutes of each other (so no substantial change in seeing or drift in focus).
The image scale is ~ 0.8 arcsec per pixel unbinned and 1.6 when 2 x binned. Based on DSS and SC scores the FWHM -pixel of 1X frames averaged about 4.2 pixel. Thus calculated that the FWHM in arcsec was about 3.4 (~ SQR of sum of squared contributions from optics, movement and the seeing).
What puzzled me was that when I went to 2x2 binning DSS and SC reported average FWHM- pixel scores of about 3.2 pixel whereas I was expecting perhaps about 2.1 ? i.e. with 2 x 2 binning the overall FWHM in arcsec would now appear to be about 5.1 and yet, relative to 1 x1 binning nothing in the optics, field movement rate or actual seeing should have changed?