LEARNED IT · 09/26
A lower resolution is not a region of interest — the SV205 scales, it doesn't crop
Dropping AstroDMx from 3264x2160 to 800x600 shrank the files fifteen-fold, shrank Venus threefold, and left the frame rate exactly where it was. The camera downscales the whole field instead of reading out part of the sensor, so the mode throws away the resolution you are trying to record and buys nothing back.
Planetary capture wants a sensor crop: read out a small window of pixels, keep the image scale exactly as it was, and get a much higher frame rate because there is far less data per frame. That is what "ROI" means on a dedicated planetary camera.
The resolution dropdown on a UVC webcam is not that. Tested it on 12 September, same target, three minutes apart:
| 3264×2160 | 800×600 | |
|---|---|---|
| file size, 1000 frames | 21.15 GB | 1.44 GB |
| frame rate | 15.12 fps | 15.12 fps |
| Venus across the disc | 166 px | 53 px |
| image scale | 0.222"/px | ~0.91"/px |
The planet shrank by the same factor as the frame. 3264 / 800 is 4.1; the disc went from 166 px to 53 px, which is 3.1. That is downscaling, not cropping. A real crop would have left Venus at 166 px sitting inside a smaller frame.
And the frame rate did not move at all. Not by a single frame per second. The camera streams at a fixed 15.12 fps and the mode change did nothing to it. The only thing all night that changed the rate was a long exposure — 87 ms dropped it to 11.5, because at that point the exposure itself is the limit.
So the trade is: fifteen times less disk, three times less resolution, and no speed. For a target that is 160 px across on a 3264 px sensor, giving up resolution is exactly the wrong currency to spend.
How to tell which one you've got
Shoot the same object at both settings and measure it in pixels.
- Same size in pixels → real crop. Take it.
- Smaller in pixels → scaling. The mode is only saving disk.
The frame rate tells you too: a genuine crop should go substantially faster because there is less to move over the wire. If the rate is identical, nothing was cropped.
What this means for the workflow
Full sensor, and accept 21 GB per thousand frames. Ten runs filled 161 GB in twenty-three minutes on 12 September. The lever is therefore run length, not resolution: shorter runs, or fewer of them, or prune the same night rather than next week.
Worth remembering the drift number alongside this. Venus moved 41 arcseconds a minute, so a genuinely cropped field would need re-centring every couple of minutes anyway. The full frame holds it for a quarter of an hour.
Generalises past this camera: on a UVC device the resolution list is a list of output modes, and the sensor is free to get there however it likes. Assume scaling until a measurement says otherwise. Same shape of mistake as [[usb3-or-no-max-resolution]] — the mode list describes what comes out, not what the sensor is doing.