periodic error

[orly_andico]

Measuring periodic error is pretty much the only thing you can do when it's hazy and only the brightest stars are visible.

The idea is you train PHD, after the training, turn off guiding, click the "Brain," check the "disable guide output" checkbox, then start guiding again. Since the guide output is disabled, the star will go all over the place. PHD will measure the deviations, and from that you can measure the periodic error.

A software called PECPrep can be used to analyze the PHD log file. It is important to check the compensate for star declination checkbox when importing the file, and enter the star's declination in the little box, otherwise the periodic error reported will be falsely small.

So here's the periodic error of my CGEM. The CGEM has 180 teeth, so one period of the worm is (86164 seconds / 180) = 479 seconds. 86164 seconds is 1 sidereal day (which is less than the standard day of 86400 seconds)



Under FFT Smoothing I have unchecked all the components except the two major ones. As we can see the biggest component (magnitude 100%) is at a period of 482 seconds. This is the worm fundamental frequency (since it matches the 479 seconds worm period).

The second biggest component (55%) is at 182 seconds. This is the famous CGEM 8/3 component due to the poor-quality gear in the motor. 8/3 = 2.6667, and 479 (fundamental) / 2.6667 = 180 seconds.

The total periodic error is +22 / -20 arc-seconds, or 42" peak-to-peak. Which is huge.

If we un-check the 8/3 component, leaving only the worm fundamental:



we see that the fundamental periodic error (which is solely due to the worm gear) is +14/-15 arc-seconds, , or 29" p-p. A large figure. For reference the typical Mach1 is 7" uncorrected, and the typical AP900 is 3" to 4" uncorrected.

I was able to get an Aeroquest worm from Aeroquest Machining. This is analogous to the Ovision worm for G-11's. Here is the periodic error:



As we can see - the largest component is now the 182-second component!!! the total PE is +21 / -19 arc-seconds, or 40" p-p. Not much of an improvement.

If we un-check the 182-second component, leaving only the 482-second component (which is the worm fundamental) we get this -



The PE is now +8 / -8 arc-seconds or 16" p-p. So in summary the Aeroquest worm is twice as smooth as the stock CGEM worm.

Moral of the story: a good worm cannot overcome the bad 8/3 gear inside the CGEM.

All of this is moot if you auto-guide fast enough. Both the fundamental (479 seconds) and 8/3 (180 seconds) are slow-moving terms, and can be guided out.

If we look at the first screenshot, the "maxRate" is 0.48 arc-seconds / second.

If we want to maintain guiding within 1" (normally our image scale is about 2" to 3") then this means we must guide at a rate no slower than 1 correction every 2 seconds (thus allowing an accumulated drift of 0.48 x 2 = 0.96")

Periodic error correction (PEC) can reduce the fundamental term only - because the 8/3 term is non-integral (2.6667) and not in phase with the fundamental. So for this mount, a good PEC training should reduce the PE to about 22" or so - as the 8/3 motor term is 20". That would halve the PE and allow guiding corrections every 4 seconds, instead of every 2.

[Gary]

Thanks for sharing the info! Most of it sounded greek to me but if I ever pick up serious AP in the future, will definitely re-read this again and try to understand it better.

[kimo]

Thanks for your nice graph. very much educational.

[kimo]

I don't know why mine is like this. I followed your procedure. This is GP2 mount.

[orly_andico]

There's definitely something wrong there, because PECPrep thinks the period is 214 seconds.

The fundamental period of the Vixen GP2 is 598 seconds.

The wrong worm period is because you chose the wrong mount (notice under "Statistics" it says the worm period is 638 seconds). This is why I thought you have an HEQ5, with the 135-tooth worm. You have to choose the correct mount (under the "Mount" dropdown menu) which is Vixen Great Polaris or Super Polaris or CG-5 (they all have 144-tooth worms).

You don't have enough samples. Typically you need at least 3 worm periods worth of data, which would mean you need at least (598 x 3) = 1794 seconds worth. Since your sampling interval is 1 second, you need about half-an-hour of data. You only have 247 samples, so you are only seeing about 1/2 of the worm cycle.

Notice that in all my screenshots there are many cycles - the more cycles captured, the better the noise / seeing / etc. get averaged out. Also obviously if you only see 1/2 of one cycle, you are only seeing 1/2 of the periodic error at best. It could be much worse - you see, if you are at the peak of the sine wave, the slope is minimal there, so if that's the part you recorded, the periodic error will look amazingly low.

Also you need to ensure that the focal length, size of the pixels on your camera, and the declination are all properly set when you import the PHD log.

Not to diss the GP2 - I had a GP until some time ago - I thought mine had +/- 10 arc-seconds of PE, and I was very happy. Turned out it had +/- 40" - and the 10 was because I was guiding at the time. The Vixen GP/GP2 has very smooth PE, not like the lumpy CGEM PE - but it's definitely not +/- 8", that is Takahashi EM-11/EM-200 territory.

Again, so long as the PE is smooth, it can be guided out. Smooth and large is better than small and lumpy.

[kimo]

I had selected as you mensition Mount select, focal length, size of the pixels of camera, and the declination are all properly set.
here is the result. i am writing from dark room .

[kimo]

today i cleaned and regresed my mount and adjusted the gear. its working much much btr then last time. i can understand this from its tracking accurecy.

[orly_andico]

This illustration should make it clear:


The entire worm cycle (here, simplified to just the fundamental) is 598 seconds. Since your data above is only 247 seconds, you are not seeing the whole worm cycle. Hence, you are not seeing all of the periodic error. You need to get at least 1200 samples (2 worm cycles) to get a meaningful result.

[orly_andico]

BTW.

If it was your post on the astronomyforums.net asking about converting a GP2 to GoTo -

the easiest solution is to buy the EQ5 Synscan GoTo conversion kit. Can't find it anywhere except the UK, and it is 300 GBP (less 20% VAT, then convert and ship to SG).

That would give you EQMOD control (actually it's the only solution that would give you EQMOD control).

There are several other options, but none of them are particularly cost effective. For example, the German Littlefoot Elegance Photo is incredibly advanced (though not EQMOD compatible). This is what I use on my AP600 -

http://orlygoingthirty.blogspot.sg/2012 ... rsion.html

it is not a cheap solution (the GoTo controller alone is over $1100 SGD) but it has a lot of features -

1) encoder support (you can unlock the clutches, manually point the mount, and it won't lose GoTo alignment

2) Robofocus controller

3) DSLR control

4) SD card for catalogs

5) Ethernet port

6) encoder-assisted PEC (I don't have this because I don't have the $800-plus Gurley or Renishaw encoder)

There's a reason I got rid of my Great Polaris... most or all of the GoTo conversion kits are too expensive to spend on something like a GP - you'd be better off buying a used Celestron CG-5 ASGT for the same price as those kits.

In my case... well the AP mount is worth more, so it was cost-effective to convert it. Still won't match a Mach1 performance-wise, but a lot cheaper.

[PETER LOO]

As last night i we eager to test out this GP2 head newly acquired from our friend here despite the very cloudy condition, result has been satisfactory (just for that little opening time). Orly you are right, need to accumulate more samples though (can only do it next round, when the sky is better ..)