PHD2 DEC guiding mystery

[Tommohawk]

Hoping someone might be able to solve this!

PHD2 guiding in DEC has been very erratic - large deflections for no obvious reason and then slow to apply correction. So, I discover that the "resist switch" algorithm may be the issue, and was confident that "hysteresis" option would be beter.

But - I still got significant unexplained deflection - OK, the correcting pulses cut in sooner, but still followed by overcorrection, and similarly slow to correct. Seems to be periodic - so OK for a bit and then off it goes.

Last night I did good polar align, as usual so thought I'd switch DEC guiding off completely. And - big surprise, the large deflections stop!. OK, there is a very slow drift, but the big deflections disappear. It seems that the defelction are actually caused by PHD2 - howe can that possibly be??

Unlsees somehow the RA corrections are being applied to the DEC?? I calibrated using the settings calculated within PHD2.

One other thing - the last couple of DSOs have been imaged at the meridian, and I haven't flipped as its gone through - I just disable the mount limits in EQMOD - and run on for about 45mins. I dont see hoe this should make a difference unless PHD2 automatically reverses the correction direction? But if that happened it just wouldnt correct at all presumably.

Bit desperate for a solution - any ideas??

[Pompey Monkey]

You might be experiencing a combination of a very finely balanced 'scope coupled with some backlash in the dec axis: small corrections push the 'scope and it just keeps going until the slop runs out.

Try slightly unbalancing the rig so that it is a touch front heavy and see what happens.

[Tommohawk]

Thanks for the idea - but I considered that and its weighted to one side enough to stop any backlash.

The puzzle is - why is there little or no deviation with DEC off, but marked deviations with it on? What is causing the deviation? It doesn't seem to make any sense!! I cant find a way of replicating the screen plot from the logs which doesnt help.

[Owmuchonomy]

This is indicative of Dec backlash.  Make the setup 'Camera heavy'.  Weighting to one side will affect your RA axis.  Once you are set up switch to EITHER North or South correction only, not  Auto.  Decide which option suits you set up best and leave it at that.

[Tommohawk]

51 minutes ago, Owmuchonomy said:

This is indicative of Dec backlash.  Make the setup 'Camera heavy'.  Weighting to one side will affect your RA axis.  Once you are set up switch to EITHER North or South correction only, not  Auto.  Decide which option suits you set up best and leave it at that.

Thanks for that Chris - and I did find that watching the drift and setting N or S only was better, but no better than just switching DEC off. The set up is camera heavy for sure.

Also I dont think it answers the basic question:

When Dec is off there are no sudden deviations, when Dec is on there are sudden deviations. But why? What causes the initial error?

[Uranium235]

There are no deviations when dec is off becuase... dec is off! (its not guiding on that axis, hence not being used)

The dramatic shifts in direction in dec is the mount "catching up" due to backlash, as mentioned above - the best way to overcome this is use either north or south only. If the graph drifts up, then has a sudden correction downwards it means you need to switch it to North only, and vice versa for if the graph drifts down then jumps up all of a sudden.

Be aware that you need to switch it round should you need to do a meridian flip (if it was north before the flip, it will drift south after).

[Owmuchonomy]

As Rob says. You can also try this: Next time you switch Dec to OFF make sure the option "Trendlines" is checked on. Then you will see where the drift is. You can then analyse that and choose North or South corrections to compensate.

[Tommohawk]

Well... one thing this confirms is just how frustrated I get when I cant explain something, or cant understand something, or possibly both! 

Let me try another tack. 

When I watch the graph with DEC guiding off, I get small erratic deviations either side, caused by vibration etc. Also a very slow overall trend in one direction, probably due to imperfect PA. Guiding is doing nothing - all I'm seeing is the deviations from the axis. I do understand the option for dealing with this: either leave it if its small, use N or S only, using trendlines maybe, or the default which is resist switch or alternatively hystereses which works the same as RA. I also understand that if the DEC is finely balanced I may get sudden movement due to backlash. 

The point is this: If I switch to auto (hystereses or resist switch) without changing anything else, I suddenly get larger  initial deviations - and importantly this occurs before any guiding pulses are sent. Nothing else has changed - all I've said to PHD is "if you see a deflection, correct it using XYZ algorithm."  So what causes the increase in initial deviations?

Any offers?