Why is PHD2 tracking on the faintest star it can find?

Sequence Generator
21

Where will I find this Debug Log Andy ?

Cheers
Paul

22

Noticed something which MAY be relevant, or may not be !

If I have the ‘Camera Exposure Duration’ set to 2 secs then quite often I see a dim star auto selected

If it is set to 1 sec then all auto star selections seem to be on a reasonably bright star

???

Regards
Paul

23

I have not done any math on this factor, but my gut feeling is it will not make any measurable difference to the guiding results. Rotation, at least with my rig, has always been close to zero. I have quite good polar alignment, and my rotation is a small fraction of the polar movement between images. If the rotation is actually significant then your images all have circular star trails anyway. I’ll bet if you actually tried it you would get excellent results.
You are absolutely correct, in theory.

24

Same folder as the Guide log you posted earlier.

More info available in the Getting Help page on the PHD2 web site, and the Problem Reporting section of the help file included with PHD2.

Andy

25

@pscammp I have been using 1 sec and I am still getting dim stars selected.

@Andy I have attached a zip with several fits images and PHD2 logs I took previously. The fits images were taken when the guide star had been lost. I don’t have the SGP logs from that night.

Here are PHD2 logs and the SGP log from last night when I think the guide star stayed lost and SGP had to do a recenter process, about 4 times during the night. Since it was a beautiful clear night with better than average seeing, it is a mystery why it would lose the guide star.

Also last night I tried to take an image when the star was lost. When I clicked Save, PHD2 instantly terminated. I opened it manually and started the guiding again. SGP didn’t seem to care and just kept taking the image it was on.

26

Andy,

Here it is, sorry about that:

https://dl.dropboxusercontent.com/u/4938813/PHD2_DebugLog_2016-01-27_162545.txt

Too big for an upload so it’s on my Dropbox, added it to my original post too.

Regards
Paul

27

jmacon, thanks for uploading the logs, I will take a look later tonight.

Could you please upload the PHD2 Debug log for that incident. Thanks.

Andy

28

Thanks, Paul. Looking at that log the poor star selection appears to be caused by the PHD2 camera simulator not rendering realistic star profiles. Nearly all the stars had peaks near the saturation limit, so PHD2 chose the best star that was not past the saturation limit. I would consider this to be more of a limitation of the star rendering of the simulator than a problem with the star finder itself. It was a useful exercise to look at it though and we’ll keep that behavior in mind when we look at some actual camera frames like the ones provided by jmacon.

Andy

29

It’s the log attached to my prior post.

30

Here’s more. Hot off the press.

This is this evening up until now.

I started taking images about half an hour ago, every time it lost the guide star. Has been happening frequently tonight, I think because the seeing is average or worse.

Well I got 3 images, then on the 4th PHD2 closed when I clicked the Save Image button, which caused the first set of logs to be written.

So I started it up again. SGP did not notice and kept right on chugging along.
Took 3 more images, then I closed PHD2 so it would write the logs, the second set in the zip.

None of these lost stars when I took the images were permanent forcing a recenter. However, already in the first 2 hours of imaging I have had 3 that I know of, maybe more, complete failures where SGP was forced to go through the centering process. The first set of attached logs cover those times earlier this evening, but I did not take any images then. However the ones I took later are very indicative of the star choices PHD2 is making.

I you want more images I can easily generate them. Only takes a couple of minutes for a lost star message to appear.

Surely there is some setting in PHD2 that will allow it to guide continuously without these constant interruptions. There are not even any clouds tonight. Beautiful clear night, just not the best seeing.

31

jmacon,

I have been looking at the logs you sent earlier and I have come up with a theory of what might be happening in your case. Those most recent logs will also be useful, thanks for uploading them.

I’d like to get your help diagnosing a possible fix if you’re willing. If you have a chance, could you save your dark files to a temporary folder, then rebuild your darks? To save the darks, locate the folder C:\Users\YOUR_USERNAME\AppData\Local\phd2\darks_defects\ then copy all the files named PHD2_dark_lib_*.fit to a temporary location. After rebuilding the darks the problem should be mitigated; if so, this will help confirm the theory.

The problem I believe has to do with how PHD2 adds a pedestal during dark subtraction to avoid negative pixel values. The pedestal is not accounted for in the saturation detection and that is a bug; the result is stars incorrectly being identified as saturated and incorrectly rejected by the star finder. The bug would only affect cases where the darks had pixels brighter than the lights and hence requiring the pedestal. I’m working on a fix for the problem right now. When the fix is ready I’d like to have you test it using your original darks. Sound ok?

Andy

32

Sure Andy, be happy to.

OK, that’s done. Now we will see what changes may be. Rig set up to image all night, but some clouds have appeared.

33

@jmacon Just a thought and you’ve probably already done so but have you tried / considered changing the exposure length from 1 to say 3 seconds?

34

Ordinarily that would be a good approach to make the dimmer stars better guide star candidates. But because of the pedestal in jmacon’s guide images, increasing the exposure duration would just cause the next tier of too-dim stars to be selected.

There are a couple workarounds to the bug: rebuild the dark library (so no more pedestal), or use a bad-pixel map (no dark subtraction so no pedestal.) Anyway, I should have a fix available shortly so we won’t need workarounds.

Andy

35

Here are last night’s logs.

I installed the version you sent me. I could not use the old dark libs because it said they did not match my camera, so I made new ones.

I put this in the SGP hang thread:
I should mention that to start the evening I installed a trial version of PHD2 that Andy sent me to try to improve the dim guide star problem, something about a pedestal adjustment. I am suspect of that version in that it was very sluggish in posting new guide commands. I had it set variously from 1 to 4 seconds, but it often only updated every 10 to 30 seconds. I have never seen this behavior before. I frequently got its message that the image had not competed in 17 seconds, and that it was going to disconnect all the equipment (which I don’t think it ever did). None of these caused any problem for SGP.

The timeout was set to 15 seconds so I changed it to 45. that mostly kept the message from appearing. I tried both 2 sec and 4 sec and sure enough, I got stronger stars, sometimes even the one PHD2 chose was saturated.
However, this release did not seem to help my lost star problem. Recenter processes still occurred, and it was another nice clear night for this time period. Cloud came in around 11:00 so I quit. I would think that recreating my dark lib would give the best results. Not sure why you wanted me to use the older one, which I was not able to use.

The sluggish response I was getting last night was quite new. May have been some new and problematic interaction with SGP. To try to test this out and if it has anything to do with the test version, I reinstalled 2.6.0, but by that time the clouds had shut me down. Sorry.
I may get some early clear weather tonight to have another go at it. After that a major winter storm will shut me down for 4 days.

36

jmacon,

I just sent you a PM with a link to an updated build. I think the build I sent you yesterday may have been a bad build because I saw some inexplicable behavior with it too-- the simple pedestal calculation change cannot have had anything to do with camera timeouts!

Andy

37

Finally got a chance to try this version Andy. Worked great all night long. I did not notice any re-centers triggered by guide failures. The guide stars selected had SNR values between 40 and 60, whereas with earlier releases it was around 20 to 30. Appeared to make a big improvement in ability to keep hold of the guide star. I did move the time from 1 sec to 2 sec, but in the past that has only made a small difference. The star profiles mostly looked a bit saturated, but the guiding seemed to be as good as it ever was. Ranged from .7" to 1.5" over the night, most of the time around 1", which is typical for my site.
Great. Thanks.

38

jmacon,

Excellent, thanks for the update. We will publish a new PHD2 version with this change. Feel free to continue using the test version in the meantime.

Andy

39

Hi Andy,

When will this new version of PHD2 available?

Thanks,
Yizhou

40

Yizhou,

Version 2.6.1dev1 is available now on the PHD2 development builds page. v2.6.1dev1 contains the fix that jmacon tested for us.

Andy