For those who are not familiar with mounts that use Strain Wave gears (trademarked as Harmonic Drive[tm] by the original patent holder), they are not new. Hobym (Korea) sold one of the first commercial hobby mounts based on strain wave gears, RainbowRobotics (manufacturer of Robotics arms) created a subsidiary called RainbowAstro, just to sell mounts that are based on strain wave gears, SS One (Japan) had a low volume production mount, etc.

The strain wave gears have been used in many applications that benefit from high torque. The NASA Mars Rover wheels are driven by motors which are connected with Strain Wave gears.

https://en.wikipedia.org/wiki/Strain_wave_gearing

The new comers to using Stain Wave gears (the ads call them called "Harmonic Drive" even though they may not be built with gears manufactured by the US company -- we don't know yet) are SharpStar (Sharpstar Mark III) and ZWO (AM5).

In the past, the gears have been very expensive since only Harmonic Drive (and a Japanese licensee) could sell them. The gears for a single axis alone in the RainbowAstro RST-135 is close to about $1000 US.

The main attractiveness of strain wave gearing is that the gear ratios are quite high (even reaching 200:1) which makes then a good replacement for worm gears. Worm gears, even with the large gear ratio, have terrible torque since there is only a single tooth of the worm that is in touch with the spur gear. With a strain wave gear, you typically have a dozen teeth that are meshed.

As a result, you can drive very heavy payloads without the use of a counterweight. As such, they are perfect for travel, since not only do you not have to carry a counterweight and counterweight shaft, the mount itself is small and light. The RainbowAstro RST-135 itself weighs 3.3 kg (not a typo; three point three kilos), while being able to toss around a 13.5 kg payload like it is a rag doll. The larger RainbowAstro RST-300 weighs 8.5 kg, and it can carry a payload of 30 kg. The weight ratio of the ZWO mount is not as good (they probably use rougher manufacturing).

If you add a small counterweight, the strain wave gear mounts can take even higher payloads. The payload capacity of the RST-135 for example goes up from 13.5 kg to 18 kg when you add a small counterweight. Most of the time, the counterweight is necessary not because the high torque gears cannot handle it, but because the tripod would topple over without the counterweights!

Again, for people who are not familiar with strain wave geared mounts, not only do you not have to balance the mount, you cannot balance the mount. The RST-135 uses a 100:1 ratio gear and it is impossible to move the RA and declination axis by hand. The only way to do it is to apply power and control the motors electronically.

Another good part with having all that torque is you can easily use a dual saddle plate (I do) and not have to worry about the "3rd axis" balance problem.

The bad part (for astronomy use) is that they have very large period error (they were not invented for that purpose). The worst case for the RST-135 is about 70 arc seconds peak to peak -- most people see 40 to 60 peak to peak arc seconds on their mounts (while even a small legacy German mount as the Takahashi EM-11 has a peak-to-peak error of just 7 arc seconds). Looking at the specs, the ZWO AM5 also has a high PE. The graph shown in ZWOs product page claims 40 arc seconds peak-to-peak (they call it +/- 20 arc seconds, but most mounts specify peak-to-peak, since the PE may not be symmetrical.

In fact, the error graph shown in the ZWO product page shows a highly non-sinudoidal curve, together with a slower drift term. The drift term probably make it impossible to have a Periodic Error Correction -- PEC. We shall see if I am right, i.e., if ZWO can create a PEC for their mount.

(It looks like the ZWO AM5 may need to use very fast autoguiding feedback loops to counter the periodic high slopes, perhaps even faster than the 2 FPS that I use for my RST-135.)

The periodic error of the RST-135 is very sinudoidal, with a small 3rd harmonic term. The ZWO is quite different, I include ZWO's curve from the AM5 product page here:

The reason I bring this up is that one of the ability to guide well is not really the error, but the slope (first derivative) of the error. That is what guide pulses will need to keep up with. When a curve is not sinusoidal, the slope of parts of the curve is unnecessarily large. The reason why an Avalon M-series mount can be guided so well even though it too has a large error is that the curve is very smooth (i.e., small slope) (The Avalons do not use strain wave gears.)

Already, you can see from the above graph that the three cycles of the periodic error are not the same even though they immediately follow one another. To do good PEC, all cycles (except for noise term) will need to be identical.

I currently use two RainbowAstro RST-135. Bought my first one (serial number 13 from the first production run) in the Spring of 2019. Both are in use. In fact, one of them is outdoors all the time under just dry bags, in rain and snow. There is no grease to freeze or melt -- I don't know if ZWO's strain wave gears need grease.

If you just have a home observatory, with no need to travel, it may not be worth the addition cost of the strain wave gears. However, once you are spoilt by the light weight and the lack of need to balance, you will never want to use the legacy German mounts again. I bought a second RST-135 not just because I wanted two mounts, but I could not see myself using my EM-11 again, in case the first RST-135 fails.

Chen

wvreeven How is the RMS of guiding in general?

As you can see, the large periodic error of all strain wave geared mounts need to be guided.

Guiding will depend, as I mentioned, on the worst case slope of the periodic error.

The RST-135 has huge periodic error, but curve is very sinusoidal (so the slope is also sinusoidal (d/dt(sin(t/T)) = cos(t)/T). Because T (period) is moderately large, and the slope is bounded by the magnitude of the error itself (cosine is just shifted by 2.pi) divided by T (chain rule of derivative), the slope that needs to be guided away is pretty small -- but you still need to guide with high cadence. Guiding with 1 second frame rate is not fast enough -- I needed to use 0.5 second exposure and 2xBin to get ASIAIR to guide at 2 FPS, for example. The only reason I use 2x binning on ASIAIR is to get the frame rate high enough.

I recently (when v1.8 beta allowed 2xbinning) on October 30, posted my result here:

https://bbs.astronomy-imaging-camera.com/d/12100-asiair-pro-and-guide-camera-binning/21

We'll have to see if the ZWO AM5's more peaky periodic error can be guided away like the RST-135. The slope looks to be higher than the RST-135, even though the latter has a larger peak error. The derivative is the important part, not the curve itself.

The actual duration of the RST-135 and AM5 periods are the same, 430.82 seconds -- 1 sidereal day divided by 200.

Chen

EDIT: I threw pretty much everything at autoguiding in that post. The guide scope was a Borg 55FL (55mm, 250mm focal length) fluorite APO scope, with a 685nm IR-pass filter and ASI462 camera for near-IR guiding, and gain adjusted to not clip and get 12 guide stars, and focused with Bahtinov mask. I recently got a Askar 230 to use as a guide scope at 275mm focal length, but have not had a chance to try it yet. 250mm focal length with 2.9 µm pixels is a little on the short side when binned by 2.

w7ay @asiair@zwo

Chen, ZWO,
Could you clarify if your RST, or any RST mount uses a synchronous belt, as ZWO is advertising for the AM5?
Would you know where to find out more on this belt and how it "help"?

Thanks.

sebriviere Could you clarify if your RST, or any RST mount uses a synchronous belt, as ZWO is advertising for the AM5?

As far as I know, the RST-135 does not use belts. So I have no idea about using belts combined with strain wave gears.

RainbowRobotics (the parent of RainbowAstro) had been being doing research as part of the University of Nevada (Las Vegas) Robotic apartment for a while now. Their CEO is an amateur astronomer, so is one of their R&D managers (B.J. Jeong, xuranus on Cloud Nights). Even the command set (they support both their own command set and the LX-200 command set) is very streamlined and a pleasant to write code to support. So, if they don't use belts, I take them at their face value. @Byrdsfan1948 is a retired MIT Mechanical Engineer, so he probably can tell you loads [sic] more. I am just a retired Stanford EE (coupled with a couple of summers at the National Radio Astronomy Observatory).

Chen, ZWO,

Just to be clear, I have nothing to do with ZWO.

To keep my independent stance, I have even refused a free ASIAIR "Plus." I also don't want to have to work around hardware bugs too. They actually wrote me again after my first refusal, that it would be sent to me at no cost, and I told them I fully understand, and still don't want one. (heck, taxes on gifts from a foreign country cost more time to resolve than the cost of the device).

I am in frequent touch with a couple of ZWO's folks, all the way to the top. So I knew they were going the way of the Compute Module 4 as a Raspberry Pi 4 replacement a year before they released it for beta testing by the current testers. I also knew about the strain wave gear mount development from about a year ago (the info came from the top). And a couple of other unannounced projects, which may require more precision than their manufacturing capability.

Stain wave gears make it easier to design a mount than traditional German mounts, which require precision bearings and tons of hands on experience. That allows people like SharpStar and ZWO, who have no prior experience with mounts, to enter the market once the patent for Harmonic Drives [tm] expired, and cheap knock-off gears can be sourced.

The inconsistency of the successive cycles of the Periodic Error curve in the AM5 could become the "tall tent pole," and we shall see how well those can be guided away. The problem with knock offs is they can produce something that superficially behave the same, but may miss some key ingredients (many examples of that, leaky tantalum capacitors, overheating Li Ion batteries etc, etc). The instantaneous bumps and the slow drift of error appear to hint at less precise manufacturing, but we shall see if the high dx/dt can be guided away. The very slow super-period on top of the 420 second periods (their first graph) is another indication of something imprecise, but that should be able to be guided away. But it means that even if they added a shaft encoder (price range probably outside ZWO's target) that error component may remain, depending on the location of the encoder.

From just casually looking at the curves, it will need fast (perhaps even faster than the RST-135) guide sampling. It would be nice if ZWO publish CSV data so we can do Fourier Transforms to look at it more carefully; but I have little interest in buying ZWO's mount, so I will leave it to others to pursue ZWO for the data.

The periodic error curves are really less helpful than a guided curve, since you really can't use a strain wave gear mount for DSO instrumentation anyway without guiding (unless for casual visual work). We will also need to see how the AM5 performs with unguided planetary; but since Sam's interest was in planetary photography, perhaps that may be OK.

Chen