ASIMount@ZWO It seems most of the slopes are as same as mikenoname's AM5 and yours got much low numbers of Periodic error.

The amplitude of the Periodic error is not important for auto-guiding (it is only important for non-autoguided visual). The first derivative of the Periodic Error is the limiting number for auto-guiding. Large, smooth periodic errors can be easily guided away (take for example the Avalon mounts), rough slopes are what makes it harder to guide.

Taking your rescaled overlays:

I see a factor of 1.6 between the worse case derivative for Mike's curve compared to the worse case for Mutt's (worse) curve. For auto-guiding, Mutt's mount is worse than Mike's mount, even though Mutt's periodic error has much lower ampitude.

Mutt's curve has very high fifth order harmonics. From the white paper that I am writing, the first derivative of the N-th term of a Fourier Series looks like:

Notice the "N" that is inside the argument of the sinusoid pops outside (using the Chain Rule in calculus) as a direct multiplier of the amplitude of the derivative. A large 5th harmonic therefore contibutes to a much larger error.

Mutt's high derivative regions also occur 5 times more often during a long exposure than Mike's mount.

Chen

tempus Where do you see this in the plot?

Count the cycles per period. Use Mike's graph for the period, since Mutt's graph is too irregular.

Chen

The 2 first point are straight from Chen's playbook. I don't think any of the rest will make that much of a difference with such an erratic guiding unless you are way out on some. Get a hold of your retailer. Hopefully he's a good one. I had an issue with a camera at one point. ZWO was ignoring me. I spoke with the retailer and he contacted them. Things got things moving in a matter of days and I got a resolution. Unfortunately, getting a good retailer is also a matter of luck or careful selection. Keep hope!

tempus The AM5 is the biggest lottery of the astro world at this point.

Nah, the USB hardware in ZWO products beats the AM5 by a mile :-).

The AM5 simply is QC'ing the wrong thing. They are QC'ing the amplitude of the periodic error (and still boasting about it for the AM3), while the parameter to QC is the first derivative of the curve.

They should trash the gears with the large slopes instead of passing it to the customer as a usable product. (Note that large harmonic distortion means large slope by the chain rule of calculus. So, when you see lots of wiggles in the curve, run, don't walk, to return the stuff.)

And if I am not wrong, they have even reduced the period of the AM3 from 430.82 seconds down to 288 seconds. That would instantly bump the already large slopes of the strain wave gears up by another 50% (again, from the chain rule). What are they thinking?

It is not clear, but I saw recent spec for the AM5 to also have a 288 second period. It is either a documentation error, or that mount has been revised too to the shorter period.

Chen

tempus So without having access to the proper data or any data at all prior to purchase, which harmonic drive would be the safest to buy?

I already have two RST-135 that I have been using for a few years now (the first was from RainbowAstro's first production run in 2019), so I am not really in the market for another one.

I was recently tempted to order the Nyx-101 just in case I need more payload ability when a Q3 manufacturing run of the Mewlon 180C is expected (paid for already; just now collecting other accessories I will need -- reducers, FTF, adapters from PrecisionParts, etc). But it looks like (check out their PHD2 screen captures), they too, like ZWO, have no understanding of strain wave gear characteristics before actually selling a product that is based on it. Whatever happened to good old fashioned engineering anyway.

My next strain wave geared mount could be the RST-300 if the Mewlon proves too much for the RST-135. The RST-300 is a little heavy, but I have been leaving the RST-135 outdoors 24x7x365 with no ill effect for 3 years now, so the plan is to leave the RST-300 outdoors too and don't have to carry it in and out.

When I had only one RST-135, back in 2019, I had babied it because I couldn't imagined having to go back to my Tak EM-11 if the RST-135 had failed. But after I bought the second one in 2020, I just left one outdoors all the time. And I still have both (similarly equipped with identical dual saddle mount, USB and power hubs, etc).

Now, if RainbowAstro were to announce something in between the RST-135 and the RST-300, I would buy it in an augenblick. You should see how clean their mount protocol is, a software engineer's dream. It takes after Niklaus Wirth's philosophy -- Pascal was designed with no operation that is not absolutely neccessary.

Chen

@KC_Astro_Mutt Best guiding came at 65% agg (both axes), 500ms durations, 0.5 second exposures.

If you have no other choice, try (both at the same time) (1) reducing the max pulse durations to 100 milliseconds, and (2) reducing the RA aggressiveness to less than 50%. Try 35% or less.

As bad as the harmonic distortions are for your mount (many large wiggles per primary period - so much so that you can't obviously see the fundamental frequency), you should still be able to restrict the correction pulses to no longer than 70 ms or so.

Allowing the autoguider to issue larger corrections than what is absolutely needed to correct for tracking error will cause problems (notice that the ASIAIR recommends 2000 milliseconds for ZWO's own mount, which it totally bogus). If there is a centroid estimation error (and there will be, even when you use optimal guide scope and cameras) large correction pulses (overshoot) will be issued (and exacerbated by large aggressiveness factor; thus, keep it low) which in turn produces a negative pulse later (undershoots) when the centroid is more correctly measured.

Now, the only problem when you reduce the pulse widths and aggressiveness, dither recovery will also be very slow (since we are only trying to move the mount by 0.1 to 0.5 arcsec/sec). But that is cause by how the current guiding software treats dither (it uses the same pulse width and loop gain). If you wite your own software, you can use different parmeters for dither recovery vs guiding, and the problem goes away. In the meantiome, if you don't write software, is to use as small dither as possible. 2 pixels worth will work with most monochome camera, and 3 will work with any Bayer camera. And if your plate scale is around 2 arcsec per pixel, the dither is limited to 4 to 6 arc seconds, and you can recover in 5 to 10 seconds.

Chen