Loading:

  • Deep Sky Colors
  • astrophotography
  • stars
  • tools
  • books

Forum

DSPX 1.0.0 Beta 6 –...
 
Notifications
Clear all

DSPX 1.0.0 Beta 6 – test report (IC 5070, OSC, CFA drizzle): weights, Noise Reduction clamping, Deep Star Removal at 2× scale

1 Posts
1 Users
1 Reactions
92 Views
(@sergiusz)
Eminent Member
Joined: 2 months ago
Posts: 12
Topic starter   [#3367]

Hi Rogelio,

I processed IC 5070 end to end in DSPX: 62 × 60 s subs, Askar 400 at 280 mm, Player One Ares-C PRO (OSC, RGGB), first without drizzle and then with CFA drizzle 2×. Overall the program works very well. CFA drizzle, plate solving, the Gaia color calibration, Star Recombination, Pixel Math and the detailed logs and reports are excellent. Below are the issues I found, in order of importance.

Pre-processing

Default integration weights. With weights = auto DSPX uses SNR estimate. The worst frame of the session (FWHM 1.93 px vs ~1.36 median, PSF Signal Weight 0.018, the lowest) got weight 1.000, the highest. Switching to PSF Signal Weight fixed it. I would suggest PSF Signal Weight as the default, as in WBPP.
The report says “59 frames within the limits” but integrates all 62 and does not say which 3 frames are outside the limits.
The registration reference is a trail-flagged frame (0048), while the normalization reference explicitly skips flagged frames (0061). That is inconsistent.
After Analyze builds the calibration masters, the Run report says “0 built, 3 reused (built earlier)”, although they were built minutes before in the same session. It looks as if nothing was rebuilt.
The estimate in the status bar (“62 light frames”) is not refreshed after frames are left out by hand.

Processing

Noise Reduction does not clamp its output to [0,1]. I got negative values: down to −0.016 on the first run, and 2 pixels at −0.003 on the drizzle image.
Curves keeps the previous curves of all channels when it is opened again or used on another image. It is easy to apply an old RGB/K curve a second time by mistake. Resetting on a new target, or a visible warning, would help.
Background Extraction re-attaches to the image if you click the image after Apply and Save. The following dialog (Apply / Discard / Keep working) risks a second subtraction.
Deep Star Removal fails at drizzle 2× scale. It leaves thousands of small stars in the starless image, and the brightest star becomes a grey disc (heal diameter 40) or a bright ring (heal diameter 80). On the same image downsampled 2× it works perfectly. My workaround was: Integer Resample −2 → DSR → Resample ×2 → stars computed in Pixel Math. A scale-aware model, or automatic internal downsampling, would solve this. Also, the “Passes” field turned out to mean healing passes; the label could say so.
Processed XISF files carry no history of the processing steps. They also still have IMAGETYP = 'Master Light' and the NOISE keywords of the linear master after non-linear processing.

I can send logs, run reports or sample files if they would help.

Best regards,
Sergiusz Gobiecki



   
RBA reacted
Quote
Share:
This website uses cookies to improve your experience. Cookie Policy