hi,
I dont know if indi does it already but I guess it doesnt. In the v4l2ccd tab there is an option for stacking. If I click on it while having the live stream window open It seems to stack nothing. In (open)PHD the v4l video gets stacked by an adjustable integration time. Do I misunderstand the function in ekos or is this only made for cameras that do internal integration?

[edit] and would'nt the stacking of an bad v4l signal help the autoguide process to reduce noise and some seeing?

Hi,
I addded stacking while porting the existing V4L2 driver to the INDI CCD framework, but never had time to go further. I wanted to test it with a CCTV camera but as my grabber is not in the kernel, I have to modify and recompile a module for it to work. So I left it for future work.
At the moment it is only effective when you take an exposure, not when you do live streaming. It simply adds the frames received for the duration of the exposure. The values are truncated to 8 bits at every addition so this is not stacking as usual. Actually I don't know if this is really useful. I should use a temporary 16bpp image when adding frames and then divide or truncate the final frame. Notice that there is an error with the measurement of the integration time: it starts as soon as you start the exposure, but it should start when the first frame is received (to take into account the time needed to start the device itself). So the number of stacked frames will be (Exposure time - Startup time) * Framerate. Startup time may be 1 second or more, depending on the driver.
I will have a look to how openPHD works. In the case of Ekos, I wonder if stacking while streaming would be useful as I think that Ekos continuously takes exposures.

Jean-Luc.

Ok thank you for the insight. You're right that stacking while streaming might not make sense. It would be maybe just an 'eyecandy' to see an stacked video from the v4l stream. It would maybe make the already cpu-heavy live-steam eating more cpu ressources. My thoughts where basicaly it would make not so good low light cameras usable to find a proper guide star and improve autoguide. Thats my guess why openPHD implemented it.

Hi,
I've just commited a version which performs some stacking for 8 bits GRAY Images by summing luminance after linearization. Here is a simple how-to:
- connect cam and choose capture format (use YUV formats if any)
- be sure to use GRAY images (linearization is not handled and crashes when used with COLOR images)
- you may start streaming to adjust focus (I usually augment brightness to 200 here). You may check quantization (dealing with limited range colorspace quantization) and linearization buttons in the Capture tab to see the effects of those processings.
- Readjust your brightness to a middle value (I use default 128) and take a series of stacked exposures to determine the best exposure time which will suit your needs. Before using a stack mode, be sure to CHECK linearization in the Capture Tab. It is made internally but as linear/stacked exposures are 16bit depth, checking this button sets buffer sizes as required(not yet automatic). Using my setup I use 1.8second exposure (yielding in ~15 stacked frames, startup time is quite high).
- You may take a dark frame now using the corresponding stack mode and your desired exposure time. Don't forget to close your tube, select Take dark and start an exposure. This will create a dark frame which will be subtracted to all subsequent stacked exposures. You may reset this dark frame or take another one as needed.
- I then use additive stack mode to perform subsequent exposures. I attach an example image I took yesterday evening using a 80x600 guide refractor from inside home. This is HD49166 (near Sirius), and these 5 five stars are between 8.9 and 9.4 magnitude. They were low in altitude (less than 30 deg) so there should be some dispersion, not forgetting the full moon. I use a cctv camera (one of those sold as CCTV Sony 700TVL) with an old videograbber (based on EM28xx chip).

I will make other experiments but for now, I did not succedd to solve any images using astrometry.net (well I made only 3 tries). Problem first comes from edges (which could be cropped), and from some background pixels which are detected as bright sources. But maybe that could be enough to perform guiding.

Jean-Luc.

Great! I will definitely test it when I am back from vacation.