Hi. I'm currently testing the internal guider of Ekos. One thing I stumble over today is, that sometimes the guiding star SNR is zero or the value is jumping very badly. As an example please look at the attached picture:
I tried to reduce the box size but that doesnt help. If there is a value then it is in the example case around 650.
i never observed that in PHD2.
Another thing: is there a difference in calculating RMS between Ekos and PHD2? i observed that the values from PHD2 are lower tha those from Ekos. Even if I guide iwth PHD2 and look at the values in the Analyzer tab, the PHD2 values are lower. What is the reason for this?
Thank you. Thats not so good to read because I use the SNR value also for me as indication for e.g. clouds, fog e.g.... So I think I will switch back to PHD2... Would be very nice if this can be solved!
Thank you for your reply, Hy. Good to hear (and I already read in the meantime) that this is solved. Thank you for your work So I'm waiting for 3.6.1
Regarding ignoring the zero: normaly I take a quick look by eye on the SNR graph to see whats going on (fog, clouds...) but this is difficult if I have graphs like this (just to illustrate what I mean):
One additional thing I'm interesting in (its not so important for me since there are just values): as I wrote I observed that the RMS values are smaller in PHD2. Do you know if there are different maths behind that? I often compared the values given in the Analyze tab with the values I read at the same time in PHD2 and just wondering what is the reason for that.
To answer your question about how Ekos calculates RMS error vs PHD2, RMS error is the (square root of the (the sum of (the squares of the individual error terms))) all divided by the number of samples. The error in a given guider sample is probably computed the same in PHD2 and Ekos--I haven't looked at the PHD2 code for that, but I imagine it is likely to be. However, the question is: the sum over what interval and correspondingly how many samples? Is the over the last 60 seconds? The last 100 guider samples? The most recent capture? The entire guiding history? ...
I honestly have no idea what PHD2 uses for that, and Ekos itself has a few approaches. In Analyze, there are two RMS errors. They are both IIR filters designed to roughly use the last 40 or so samples, but one resets at the start of a capture, and the other doesn't. The guide tab, for historical/code reasons uses the last 50 samples unweighted (or whatever exists if there are fewer than 50). All 3 should all give similar values, but not exactly the same.
Regarding your noisy SNR graph: for some reason Ekos is having trouble detecting your guide star, though it does continue to detect enough companion stars to continue running. (If it hadn't been using MultiStar, it would have stopped early one.) Perhaps it made a bad choice of guide star, e.g. one that really two stars, or part of a galaxy or cluster? It is also possible that you have sub-optimal star detection (StellarSolver) parameters chosen for your particular setup. You might double-check your StellarSolver/guider parameters and make sure they make sense. For instance, in the "Sextractor Parameters" section, check if your "Minimum Area" too small or too large (it is given in pixels, I believe). Perhaps remove any min/max filtering in the "Star Filtering Parameters" section (by setting the values to 0), in case your guide star is being filtered out.
Thank you for the informations. I will play a little bit with the parameters you mentioned. It is on the one side much easier to work with the internal guider on the other hand I'm used to use PHD2 for nearly 2 years now So thank you for the patience with me
One last thing: is it possible to get something like the HFR value of the guiding star shown in the graphics or just as value? Also a cross section would be nice.
Yesterday I played around with the SEP parameters. It was ambigous. Most of the time it worked well, but sometimes e.g. the choice of the guiding star was questionable (at least for me). I missed to take pictures, sorry. Example: I had a field of stars which are good visible, smaller ones and bigger ones. What was choosen? A hot pixel, I had RMS 0.02 also it found many companion stars and linked them. The min area was set to 6 Pixel (even if I set the star filtering parameter "min size" to 2px)... When set min area to 10 Pixel a "normal" star was used as guiding star.
I also tried to click on a star, then this star was used as guiding star one or two steps, but then it switched back to the hot pixel...
I used the SEP profiles "Guide Default" and AllStars, where "AllStars" seem to slow down the system when there are many stars in the FOV.
I also tried to find out, what the different parameters mean and read the linked pages. But its hard to understand, at least for me. Some parameters seem to be clear but if changing them it gives weired results (like the "Hot Pixel" with min area 6 pixel).
Am I right to say that
- the Sextractor-parameters are for choosing the guiding star? Or for guiding star AND companion stars?
- the Star Filtering Parameters will search for any stars in the FOV within the parameters?
- The guiding star and the companion stars are choosen out of the "Star Filtering" stars (which should be clear I think)
I hope I'm not bothering you but I want to understand what I'm doing here
(Here and below I'm talking about the SEP MultiStar guiding algorithm). The SEP parameters are for detecting stars. The guider first detects stars then it does its thing, whether that's choosing a main guide star, choosing the references, or re-finding all those stars in subsequent images in order to guide. Without good star detection, the guider cannot work well. The star detection is done by the StellarSolver library, which uses the SEP library, which is related to the Source Extractor star-extraction software package www.researchgate.net/publication/1777454...actor_for_Dummies_v5
StellarSolver uses the "Extraction Params" to get a list of stars (i.e. positions on the image and sizes) and filters that list with the "Star Filtering Params". So, both sets of params are used to do all of the guider tasks. I don't think it's worth messing with most of the parameters. As I said in my previous email, I'd recommend that you try and get the Min Area right (so that you avoid hot pixels but keep most real stars), and perhaps Max Size to avoid very large things that aren't stars. You can set Min Size to 0, I believe, as I think it overlaps in functionality with Min Area, but I guess use it if you think it helps get hot pixels filtered out.
If you think you have an image where your Sextractor parameters are good, but it does a poor job of selecting a guide star, upload the .fits and post it somewhere where I can play with it. (That is, take an image and save it. You can do that by using the capture tab with the guidecam selected and the same gain and exposure time you're using for guiding, or you can go to KStars Settings -> Configure KStars -> Developer and (temporarily) check save guider images. You'll find them in a companion directory to where the logs are saved called "guide". E.g. on Ubuntu it would be ~/.local/share/kstars/guide
Clicking on a star won't work right with SEP MultiStar, I should somehow disable that when SEP MultiStar is chosen. It is something that was implemented long ago for previous guide algorithms.