raawe i only dither small amounts and that was just based on my newer camera and not my mount. It did help settling time but my camera is low noise, has little banding and I never see any walking noise… so I did not ever need to dither for 5-10 pixel moves. I also dither every 5 frames since most of my exposures are short 2-3minutes with fast scopes.
Getting the best performance from my AM5
Been watching this thread with interest, its nice to get a level of understanding behind these mounts and try to get my head round the maths that affects their individual performance.
Thanks to all for the posts to date and particularly Chen's explanations along with the graphs and theory behind them.
Have had an AM5 for a few months now and been getting between 0.7-1.3RMS on the various nights have been able to get setup. Loads of clouds around this winter so not had more than a handful of goes yet. Will be happy if I can get to a consistent 0.7.
My report is below and after doing some rough lines / calculations I reckon mine needs around 0.2arc seconds / per second correction - am I on the right track? Assuming so then I would guess that my starting point of max moves of 150ms should be in the right ballpark to get consistent tracking.
Report:
w7ay I suspect that the worse case slope for your mount is a bit higher.
Thanks, Chen.
It's looking like Kevin's and mine are nearly twins (except for harmonic effects) since the steeper result you are seeing probably puts my slope in the 0.15-0.16 range. Looks like I'll be going back to 1 sec guide pulses which did seem to work well with it in my previous sessions.
Mike
- Edited
ShinyredAM5 I reckon mine needs around 0.2arc seconds / per second correction - am I on the right track?
Yes.
I placed two measurable lines on the graph that you'd posted.
The green one is used to determine how many screen units corresponds to 430 seconds. This is the sceen parameters:
It shows that 430 seconds of the abscissa corresponds to about 12.22-5.13 screen units. Or about 60.6 seconds per screen unit.
The red line is an eyeball of the worse case slope (and thus worse case p-p guiding). I had placed it so that the bottom is at -10 arcsec and the maximum is at +10 arcsec.
On the screen, the horizontal change is 1.47 screen units. Using the scale from the green line, the horizontal change is therefore about 89 seconds. I.e., the slope is 20 arcsec/89 secs. Or about 0.22 arcsec/sec.
If you are guiding using 1 second exposures, the RA axis will not do better than 0.22 arcsec-peak-to-peak (that sawtooth) in the high slope regions of the RA rotation.
Assuming that your declination axis is under control, you might be able to eek out a slightly better result than you were getting before (0.7" to 1.3" RMS error for both axes) by limiting the RA pulse to 0.22/7.5 of a second, if you are using 0.5x sidereal guide rate. I.e., you should not need more than 30 ms of max RA pulse (but you can give it 100 ms or so to account for mechanical flexure and King tracking rate, if your mount does not support KIng rate). That should be sufficient to correct for the worst case periodic error of the gears. You are seeing much higher RMS error that is likely coming from other sources -- like bad guide star centroid estimate, wind gusts, that has nothing to do with the gears.
Based on the above, you should be able to get away with 0.4 arcsec RMS (on just the RA axis), by using 1 second guide exposures. But if you have good enough guide scope, you can try 0.5 second exposures (you need more than 9 guide stars).
Just turn the PHD2 "aggressiveness" setting way down and not chase any disturbance that is not caused by the gears. And on top of that, limit the max RA travel per pulse also, to clamp down on any disturabce that is not caused by the gears.
Chen
Kevin_A I am not sure if the belts are set tight enough on my drives and may have some slop that is causing this… not sure.
Kevin, if the RA dithers are less than 7.5 arc seconds, there should be no backlash from the strain wave gears themselves, but it could come from other sources, like you said, a loose belt, or a cheap pulley (the ones that come with the ZWO EAF for example, are not what I'd consider good pulleys :-).
Now, if you limit each RA travel to be less than 0.75 arcsec (e.g., by setting max RA pulse to 100 ms), then a 3 arc second dither could take 4 exposure times to reach the destination, and longer, if it is not smart enough -- the ASIAIR, for example, does a slow exponential decay because it is lazily using PHD2 guiding to recover from a dither -- instead of appling a long enough pulse to directly drive the guide star to the new position and then restart guding.
Still, as you say, not a big deal when you can achieve better guiding.
Chen
w7ay as soon as the skies clear i will do some testing. I will keep my dithers at 1-2 pixels and see how it recovers. I do not forsee a problem. I did open up and look at the Dec belt and it looks fairly snug with about 2-3mm of deflection when pressed so i am happy with that as it is normal.
- Edited
Thanks for the detailed explanation Chen and the pointers for future testing.
As I suspect typically happens with new owners I started off with max dec and RA pulses of 1500ms and the aggression right up at 80% or so. Whilst the mount worked it was over-correcting and see-sawing back and forth but the average tracking was reasonably accurate.
The weather has been awful so opportunities to tweak the settings have been very limited however the last outing showed improvement with the approach discussed in this thread i.e. smaller moves. Ended up at max moves of 350ms and the aggression down at 35% on dec and 40% on RA - seemed to get down to 0.5-0.6 over the 2 hour break in clouds.
Next clear night (likely months away!) will drop the max moves from 350 with end game being 100ms, will drop in increments of 25ms and see what works.
As has been mentioned on this thread, there seems to be an opportunity for ZWO to use the report to highlight the recommended guiding settings range for each mount. Wonder if its an unwise thing to do from a retailer perspective though in this day and age of internet / mail order and the associated ease of returns.. would we all be cherry picking mounts?
ShinyredAM5 end game being 100ms, will drop in increments of 25ms and see what works.
If you are going to slice it that thinly, give it a bit of leeway if your mount is not tracking at King Rate, and your target is near the horizon.
https://canburytech.net/DriftAlign/DriftAlign_3.html
Notice in the second of two charts about 1/3 of the page down that there is a 0.03 arcsec/sec rate added to the sidereal rate near the horizon, which accounts for something like 4 ms of the max rate when autoguiding at 0.5x sidereal rate.
And, as I mentioned earlier, any mechanical flexure too.
Make sure that you don't fine tune where the periodic error curve has the minimum slope. If you do that, the pulses will not be able to keep up at the higher slope regions.
Chen
Looking at that PE trace made me think again. All the HD mount guide logs I have seen show short duration, large spikes in RA - even when guiding well. I wonder if these are actually at the point in the PE curve where the error reverses? In the case above this means the error rate goes from approx +0.23" per sec to -0.23" per sec virtually instantaneously - the peaks are sharp. Just a thought.
For anyone who is trying to get data points for these calculations from the AM5 graphs, I found this very useful - is very precise.
Jhaunton In the case above this means the error rate goes from approx +0.23" per sec to -0.23" per sec virtually instantaneously - the peaks are sharp. Just a thought.
They are probably not instantaneous -- just appears so when the curve is shrunk on a finite graph.
The way to view it is to look at the periodic error curve as a Fourier Series. For good mounts, terms after the third harmonics are very small. For worse mounts, you may see large 5th, and even 7th harmonic terms.
In any case, each of these components of the Fourier Series is a continuous function -- and since they are sinusoidal, the sign change occurs when the magnitude of the first derivative of the curve is small. I.e., you don't go from +0.23" to -0.23", you go through +0.2...+0.10, ... +0.8,... +0.1, 0, -0.1, -0.2, ... etc. And the slopes near 0 are easy to guide away.
Mathematically, the second derivative (how fast the first derivative changes) is actually quite small. I.e., no sudden change in slope.
Chen
doesnt mean a thing if you cant meridian flip using nina and ascom
biomedchad Are you saying that AM5 can't do an autoflip with NINA / ASCOM? If so, thats not correct.
- Edited
fbitto Sure.
Mine is similar. I get a max first derivative of 0.3 as/sec every 430 sec (surprise!), which lasts for about 20 sec.
Interestingly, when I did it manually, I got 0.18-0.2 as/sec, but this is obviously much more accurate.
Can this also be worked into the sub length, if we know, say, that the guiding will be under stress every 430 sec for 20 sec?
Next step is to do it on the actual graph under load (PHD2 LogViewer) and not the unloaded ZWO chart.
