Hello folks,
I have one of the new GM811 with the "spring-loaded" worms and tucked motors.? In this thread I will mention some of the testing I did and a few things I learned about backlash in this system.? I'm not some mechanical engineer, so a lot of things will be stuff I made up on the spot.
First thing, these are not spring-loaded worms.? The worm is held at a fixed location via that little Allen head screw, also known as the "back off bolt".? It's the same mechanism as the spring+ screw adjustment in a newtonian primary mirror, and nobody refers to these mirrors as "spring-loaded", because they don't bounce around. It is a "spring-loaded adjustment mechanism". The worm does not "float with even pressure over ring gear variations", but rather will get loose over a low spot and tight over a high spot.? Two advantages over the older system is that at least you only have to turn one screw to adjust the worm mesh, and if you adjust the worm over a low spot and slew over a high spot, then at least the worm can still lift up a little bit and you will get a "DEC heavy trace" instead of a motor stall.? One drawback is that because the worm assembly can now move around, it is a little bit loose and wobbly.
Regarding backlash, first there is the region of "loose backlash" where there is no restoring force towards center and the system can just flop around.? Next is "softly-sprung" backlash where the gears have started to push the axis in one direction, but due to flex at various points, the axis isn't moving at full speed yet.? Finally the gears get fully loaded and everything moves as expected.? Friction turns softly-spring backlash into real backlash because the gears need to "wind up" a bit first before things start moving.
To get an idea of the tolerances involved, the G8 DEC axis has a ring gear radius of 1.4".? If the worm assembly has only 0.001" of axial play, that works out to 0.001 / 1.4 * 180 / 3.14 * 3600 = 145 arc seconds.? I think sidereal rate is 15 arc-seconds per second, and PHD2 guides at 0.5x, so that comes out to about 20 "PHD seconds".
When I first got the system, there was only a tiny bit of wriggle in the DEC axis, but there was significant backlash based on holding down the hand controller and counting.? Also the Guiding Assistant backlash tests would keep failing.? It turned out that the worm cover block was actually hitting some other part of the frame and preventing the worm from fully meshing.? Oh well, this kind of thing happens all the time to me too, where I machine a piece, try to fit it on, only to realize that some part bumps into something else and I have to take it back to the shop.? Haha, fun times.? After filing down that corner, I was able to get backlash times under 10 "PHD seconds", but still not that great.
I adjusted the two silver spur gears a little closer to each other, but not enough that there was tightness between them.? I also checked that there was no looseness in the Oldham coupler by grabbing onto the worm and checking for wriggle at the upper worm gear.
To test how much backlash there is in the gears (motor + gearbox + two spur gears), I clipped a small laser pointer to the upper spur gear and pointed it at a spot on the wall a few feet away.? This is with no scope mounted, but I can still change the worm gear mesh via the back off bolt.? What I found was:
1) At least when using the hand controller, the spot still doesn't move evenly, but in these little pulses.? This is even though it already has all the gear reduction.
2) When the worm gear is loose, there is an expected "stall" of about 1.5 - 2 seconds (hand controller time) when changing directions.? This is the loose backlash in the gears.
3) When the worm mesh is tight, the laser spot has a weird behavior where it starts moving immediately upon changing directions, will go for a few seconds, then stall for about 2 seconds, then resume moving again.? If I had to make up an explanation, I would attribute it to sprung flex in the Oldham coupler plus friction at the worm.? First the axis is being pushed North with everything fully loaded and moving full speed.? Then reverse direction to South.? The worm is temporarily locked by friction and the flex in the Oldham coupler pushes the top spur gear in the South direction.? After a while, the "spring" in the Oldham coupler "runs out" and the top gear goes loose and stops moving for 2 seconds while the backlash in the gears below is taken up.? Then the top gear can start moving South again.? Like I said, I just made this up right now.
I think it would be "fair" to use TVC compensation to take out this gear lash, and I found a value of 10 helped remove most of the delay without creating initial jumps.
Question:? Since Losmandy decided to offset the motor and add these extra gears, why add backlashy gears on the slow side of the transmission instead of the fast side, where it will have less effect?
For under the stars real world testing, I use the backlash graph under the PHD2 Guiding Assistant.? I've been ignoring the number that it calls the backlash because that just states when the axis first starts moving south.? Instead, I look at the gap between the red (theoretically perfect) and white plots (real data) which gives the difference between fully-loaded going north and fully loaded going south.? Since the back off bolt adjustment is super sensitive, I added a lever on it and filed / sanded the tip of the bolt so that it is more flat / convex for more even contact.? In the attached picture, there is also a foreign object jammed between the left worm block bolt and frame.? It looks like a toenail clipper, but it's actually a calibrated wedge with the exact thickness and taper required to keep the left block from moving around.? Normally this block, because it is not tightly bolted down, can still slide around by about 0.001", which means the entire worm assembly can move also.
By looking at the backlash graphs and making tiny tweaks to the lever, I can sometimes get the backlash to a rather low level of <1.5 PHD seconds.? In my tests, each "spot" is worth 0.5 PHD seconds, so a gap of less than 3 spots is pretty good.? (This is already with a TVC value of 10)? One very good feature is that these graphs have characteristics for what a too loose vs too tight worm mesh look like.? The too-tight mesh tends to have a rising initial slope on the south side, which means that the mount tends to keep moving north despite the motors reversing.? I haven't figured out why this happens, but one of the effects is "forwards-lash", which I show an example in guiding.? On the too-loose side, there is an initial south movement, but it is not at "full speed" until the gears get fully loaded.
But now a few bad things.
One, is that if you move to a different part of the sky, the gear mesh changes and you have to do this all over again.? Because the worm mesh is not spring loaded.
Two, this is only semi-repeatable and sometimes I can get this really low backlash, but then I can't achieve it again, despite moving the adjustment back and forth.? I think I can consistently get under 4-5 PHD seconds, though.
Three, for low backlash, I need a really really light clutch pressure, like just barely hanging on, and this is even with the Michael Herman extra grippy clutch pad.? I checked the axis friction by pushing the worm out of the way and turning the axis, and it seemed fine and reasonable.? Yes, the axis friction increases a little with more clutch pressure, but that makes sense since the bearings are being preloaded more.? Oh, and I have about 40 lbs of scope on it.? But still, it shouldn't be this sensitive.
From these tests, it seems the best worm pressure is a very light one.? I also tried a configuration with a very soft spring and no back-off bolt for a truly spring-loaded worm.? With this set up it is very important to have a hard stop that prevents the worm from lifting up and skipping over the ring gear.? I found that twisting the axis one direction lifts the worm up, and the other direction pulls it down into the gears, so there must be some kind of asymmetric shape to the ring gear tooth profile.? Anyway, I could sometimes also get a very low backlash, but this system was too unreliable in terms of always getting a low backlash.? I think the low spring force makes the worm block "wobbly" and it doesn't seem to sit on the ring gear consistently.? Too bad.? And if there is too much axis friction, a turning worm gear will try to "ride up" on the ring gear, whereas in a system where the worm is locked in place, it would be forced to push the ring gear along.?
