Why do my stacked Seestar S50 M57 histograms show such a sharp lower cutoff before stretching?

ZWO Seestar S50 Siril Team Siril Stacking Stretching Deep Sky M57
3 replies•58 views
Claudio Pedrazzi avatar
Claudio Pedrazzi avatar

I have uploaded two versions of M57 taken with my newly acquired Seestar S50: one based on 20 s IRCUT subs and one based on 60 s LP-filter subs. Both were stacked in Siril with drizzle 2×.

📷 M57 (Ring Nebula) with Seestar S50, balcony observatory :-), no LP filter, 20 sec subframesM57 (Ring Nebula) with Seestar S50, balcony observatory :-), no LP filter, 20 sec subframeshttps://app.astrobin.com/i/t1sqz5/

📷 M57 (Ring Nebula) with Seestar S50, balcony observatory :-) , LP filter & 60 sec subframesM57 (Ring Nebula) with Seestar S50, balcony observatory :-) , LP filter & 60 sec subframes

https://app.astrobin.com/i/2kpf0v/

I am trying to understand something I see in the pre-stretch histograms of both stacked images.

I attach three Siril screenshots:

  1. one original, non-debayered 20 s IRCUT light;

  2. the 20 s IRCUT drizzle stack;

  3. the 60 s LP drizzle stack.

📷 Light_M57_20s_IRCUT_202606013-014400.pngLight_M57_20s_IRCUT_202606013-014400.png📷 r_20s_IRCUT_drizzle_stacked.pngr_20s_IRCUT_drizzle_stacked.png📷 r_60s_LP_drizzle_stacked_ps.pngr_60s_LP_drizzle_stacked_ps.pngThe screenshots include both the histogram and the image statistics.

The single original light looks quite normal to me: the background distribution is broad, with a tail toward low values, and the minimum is 0 in all CFA channels.

In the stacked images, however, the RGB histograms show a very sharp lower cutoff, even in logarithmic scale. There seem to be essentially no pixels below a certain threshold. At the same time, the statistics do not seem to indicate a simple fixed zero floor: for example, in the 20 s IRCUT stack the minimum values are about 4405, 5110 and 4995 ADU in R, G and B, while the median is around 5790 ADU. In the 60 s LP stack the minima are different again, about 187, 2835 and 0 ADU, with a median around 4350 ADU.

My current interpretation is that the sharp cutoff is mostly the expected narrowing of the background distribution after stacking many calibrated frames, perhaps combined with Siril normalization, drizzle and rejection, rather than simple clipping to zero.

Does this interpretation sound reasonable? Or is there something specific in the Seestar FITS / Siril drizzle stacking workflow that could produce this kind of sharp lower histogram edge?

I also notice that both the single and stacked frames contain saturated pixels near 65535, so the 60 s subs may be locally saturating some stars. Any comments on that point are also welcome.

Well written Engaging
Jeff Reitzel avatar
Jeff Reitzel avatar

There is a known issue with dark calibration in recent software updates for the Seestars. It over aggressively clips the blacks. This may be what you are seeing if you allowed the internal calibration to occur. A possible workaround could be to save all the individual raw files and create new dark frames to save as well. Then calibrate and stack in other processing software. I do see that you were already using external software to stack your images.

CS,

Jeff

Well written Helpful Respectful Concise Engaging Supportive
Willem Jan Drijfhout avatar
Willem Jan Drijfhout avatar

Looks like a pedestal value is added during the calibration/stacking process. Is very normal to prevent any zero-values.

Helpful Concise Supportive
Claudio Pedrazzi avatar
Claudio Pedrazzi avatar

Thank you both, this is very helpful.

@Jeff Reitzel , I was not aware of the recent Seestar dark-calibration issue, but it could indeed be relevant here. As far as I know, however, the Seestar does not give direct access to truly raw sensor files. My understanding is that the FITS files saved by the Seestar are already internally calibrated by the device.

If I am wrong about this, I would be very interested to know how to access the actual raw files and how to save matching dark frames from the Seestar. That would certainly be useful for testing this more rigorously.

@Willem Jan Drijfhout , the pedestal explanation also makes sense to me. In fact, in my stacked files the minimum values are not simply zero in all channels, which may support the idea that some kind of offset/pedestal and/or normalization is involved during calibration or stacking.

So at the moment my tentative interpretation is that I may be seeing a combination of several effects: Seestar internal calibration, possible black clipping or pedestal handling, and then additional narrowing of the background distribution due to Siril stacking, drizzle and rejection.

Thanks again — this helps me understand the behaviour much better.

CS,
Claudio

Well written Respectful Concise Engaging Supportive