Recording PE with PemPro

[scagman]

Hi,

When recording PE with Pempro does it matter on the time between images.

Eg. If I use my QHY8L it takes apporx 7sec to download each image so it would be about 10-12 secs between measurement, or would I be better using my Orion SSAG which is much quicker to download, giving a lot more data points per worm cycle.
Also would using a 2x powermate give better accuracy?

Thanks for any advice

Regards

[AndrewJ]

Gday John

For "analysis" purposes, faster framerates are always better.
For creating a PEC "model", the limit depends more on your firmware.
PEMPro uses Fourier analysis to break down the raw data into fundamental frequencies and then reconstitutes it using the integer fundamentals only. The accuracy of the FFT analysis is only useable to what is called the Nyquist freq, which is effectively 1/2 your sampling frequency.
ie if you sample at 10 seconds, the best errors you can detect and model will be ones with a 20 second period or greater.
Sooo, if you know how long each PEC bin in your particular handbox takes to run, you would sample at a rate at least half of that time ( or faster )

A power mate may be helpful if you use a short focal length scope, but again, under normal conditions with good seeing, the lower the arcsec/pixel you can record at the better.

Andrew

[scagman]

Hi Andrew,

Thanks for the explanation. I Don't suppose you now the Nyquist Freg. for a G11 w/G2 to is?:)

I'm assuming it would be quick so a faster frame rate wuld be better.

Cheers

[alistairsam]

Hi John

I have no idea about the sampling relationships but I think of it in terms of accuracy or resolution of the curve.
If you have more samples, your curve would be finer, less samples more coarser the curve. So I'd go with samples taken closer. 1 sec is usually adequate. But you'd need to set your image scale.

Cheers
Alistair

[AndrewJ]

Gday Alistair



Yep, but if you cant actually use the resolution, its wasted, hence why i mentioned faster is better for "testing", but may be overkill for creating a model. ( It cant hurt to go faster, but if your camera cant do it, there is no need to go faster than required )



Nope:D
The G11 has a 360 tooth worm, hence one rev of the worm will take ( close as dammit ) 240 clock seconds.
What you need to know is how many segments, one turn of the worm is broken into for the PEC model ( the Bincount ), and i have never seen that mentioned for the G2 firmware. ( I havent looked real hard tho :P)
If you can find that data ( or someone here knows it ), then your minimum framerate would be ( 240 / 2 ) / BinCount seconds per frame ( or faster ).
As a practical limit, 2 seconds would probably work without problems,
as PE isnt always periodic, so its a rough fit at the best of times.
To see this effect, use PEMPro to collect say 10 revs of the worm in one hit. Then in the PEMPro display window, look at how well each pass correlates with the others. You should see its not as "periodic" as you think.

Andrew

[scagman]

If the weather actually stays clear like it says it will, I'll do some runs tonight. I'll use the Orion which I should be able to use 1-2 sec exp with about 1.19 acreseconds and see how the graph looks.

Thanks for the help.

Cheers

[AndrewJ]

Gday John

If it does stay clear :rofl::rofl::rofl::rofl:
and you do end up doing several runs,
save the log files and post em here.
I have a little app that allows data from multiple runs to
be plotted at the same time, using the same scales,
so you can see how variable the raw tracking data is.

Andrew

[PRejto]

I don't know about the QHY8L, but when I use my Moravian G2-8300 camera (also on the slow side to download images) it certainly doesn't take long at all to download images using PEMpro. I'm fairly certain that this is because PEMpro uses a subframe and it is fairly small and easy to download. If you have not tried your camera I certainly would! I would do this if only because PEMpro will have a much easier time finding a star (you don't get to select it, PEMpro does!). And, also the wide field will make it easier to do the required camera calibration before you start measuring PEC.

I agree that the exposures should be as short as possible and at a high resolution (barlow!), but you need a good steady night and some patience! Also, pretty good polar alignment will help too and keep the star more easily centered during the measurements.

Peter

[jase]

Between 10 and 20 worm cycles, more the better. Like most statistical data, the more samples the better the analysis. Sampling is typically every few seconds. Keep in mind that we are not talking about full frame downloads but a sub frame hence your download speeds will be fast. Experiment with the drift fitting algorithm, don't just accept the default. You may find a better (read tighter) fit with a different algorithm. There is good information in the Pempro manual. Study it and understand frequency waveform fundamentals to get the most from the tool. Finally make sure you measure your results post application of PEC. If all things have gone to plan you should end up with a relatively flat line.

[AndrewJ]

Gday Jase



This is why, as part of understanding the characteristics of the mount, i also like to compare long datasets from different runs from different true sky positions. The "drift" fitting is not only dealing with drift, it also deals with tooth spacing errors etc on the final wormwheel.
Running the PEMPro display in "non overlapped" mode really highlights this well, and displays why PE aint always PE.
ie the PE model will work well in some places but not so well in others.

Andrew

[scagman]

Its not looking good here at the moment, it was looking promising around 7 with some stars poking their heads out thru the clouds but they've pulled'em back in again.
Yep i will post the logs,
Thanks



Hi Peter,
I think it does support subframes. Thanks for pointing that out. I probably wouldn't have thought of doing that.



Hi jason,

yeah I will do a heap of cycles. Been going thru the manual a bit this afternoon to try and work out whats what.

Thanks

[PRejto]

Just one more thing! Ray Garalak poimted out on his forum at CCDWare that doing more than 5 cycles of PE measurement can actually be detrimental in that it can make it more difficult for the algorithm to work properly.

[AndrewJ]

Gday Peter

Yes and no :P
I mentioned doing long duration plots was very useful for getting a good idea of how consistent your mount is, esp if run at several different HA's.
ie you get a really good overview of the system.
If your mount is extremely consistent, then lots of data shouldnt matter.
If your mount has a lot of "non periodic" errors, then the algorithms may not work as well as they could, but by the same token, a single model cant be generated that fits all locations.
"Long data" at the beginning of the process merely gives you a different view of how much benefit a PEC model may give "your" mount based on imaging location.

edit
Just dug out part of a long duration data grab for a ZEQ25, that appears to have a glitch in its wormwheel.
If you did a PEC model over 3 turns in each marked zone, you will get a different answer.
A PEC model that works well in zone 1 will not work as well in zone 2, but by looking at the overal data, it is moderately clear that the zone 1 model would be the best overall.

Andrew

[White Rabbit]

I'm no expert on the matter but I would think that when measure PE you want your mount to be in as close a state (balance wise) as it is when you will be imaging. Meaning that if you measure pe you want to use the imager through the scope that you will be using to image with. Someone correct me if I'm wrong.
Also Im pretty sure you can bin the image with Pem pro.

[PRejto]

Andrew,

Thanks for that explanation! I'd also wondered about the "assumption" that the gear was "perfect" and the worm always the culpable party when it comes to PE. I can see how this would be an issue with a more modest mount than I have taken for granted.

Peter