LOL - and I thought 5-year-olds could only embarrass us oldies with their computer skills
I think that your damper solution is absolutely valid - to fix the gimbal, not the gimbal controller. As Denny says - if hard-mounting the RSGS is passing vibration into the sensors then it makes sense to prevent that first, then see what can be improved on the gimbal itself. Denny's solution needs some work to establish the right level of isolation to overcome the problem without introducing new ones - one of the keys there is a lead washer to create inertia and mass: that solution may not appeal to all.
In an ideal gimbal, the mechanical arrangement of each axis must be "stiff" (in the engineering sense), i.e. rigid and without any radial or axial play whilst still having very low friction. Any friction present must be overcome by the servo before any actual movement can take place. This means 1) a delay and 2) a jolt in the system as the friction is overcome and the rotation accelerates more than was originally intended. Any play in the system (a.k.a. looseness, slop, backlash, etc.) must also be "taken up" before any actual movement of the axis can take place. Unfortunately the PH gimbals are poorly designed in this respect - as far as I know, neither side of the pitch axis (for example) has any pre-loading on the bearings or special bearing arrangement to limit/remove radial and axial play. The servos have (unavoidable) backlash in their gears and the belt drive has (unavoidable) backlash. The net effect of all of this is that the camera is not rigidly fixed in its orientation - it can "wander about" either side of its nominal position by an amount equal to the sum of all the backlash in all the components. And it's easy to see - with the servo holding any particular position, try moving the camera very gently - it will be able to move (in just about every direction, but rotation will be the most obvious). This unconstrained movement will show up in video, particularly at longer focal lengths. It's one of the big advantages of the Zenmuse, which (by its mechanical design and the absence of any gearing between the motor and the camera) will have (almost) zero backlash: it is a "stiff" structure end-to-end.
The primary reason for the appearance of belt drive on this type of gimbal is because it will tend to mask the backlash in the servo (which is far and away the worst offender). The belt tension creates a radial bias force on the axis - in other words, a frictional force that needs to be overcome before that axis can rotate. This will mask, to some extent, the backlash in the servo but it won't reduce it in any way. Note, too, that its effectiveness at masking the backlash will vary according to how tight it is fitted. Furthermore, of course, it also introduces its own backlash to the system. As with any gearing, the belt drive also has the simultaneous effect of reducing the drive speed and increasing the torque by the same proportion. Increased torque is helpful but reduced speed isn't!
This backlash has another nasty trick up its sleeve - if your axis measurement system (in this case the RSGS) is able to "see" that unconstrained movement, i.e. it senses the axis/camera has rotated when actually it hasn't (it's just "flopped" from one side of the backlash to the other) then it thinks "ooh! I must correct that!" and tries to do so. Through the miracle of mathematics, this can quickly become a meaningless oscillation as the system bounces from one end of the backlash to the other (and beyond, thanks to momentum) and the controller battles to "correct it". High speed oscillation... a.k.a. jitter. The only way to stop the maths misbehaving like this would be to filter out tiny movements, in other words, reduce the resolution of the system. The maths can't tell the difference between backlash and rotation of the frame in the air.
Fitting a rotary damper is supplementing the "masking" of backlash on that axis - as such it will be effective in preventing backlash-induced jitter (and should allow you to have the gain higher so you have a more responsive axis). You want a low resistance, just enough to stop the axis "flopping" and not enough to hold the servo back too much when it does try to rotate the axis. As you say: more controllable and constant than a friction washer (or adjusting the tension of the belt). There are adjustable bi-directional rotary dampers available, of course - that would be ideal to fine tune the effect. But, at the end of the day, the backlash will still be there - a better mechanical design of the gimbal as a whole would remove the need (for the RSGS or any other controller).
Jeremy as you know I went through the whole servo thing a while back, I stopped using gear/belt drives in favour of direct drive servos such as MKS 787 at 8.4 volts due to the speed of response, when that was exhausted I saw that I reached the end of the road. It was just not fast enough. Enter the torque motor and 24 volts. The interest in finding a sensible IMU to go with this has always been a major consideration. Good to see that OP has got into the MPU600. James knows what he is doing alright. Now that it has the license to use the built-in sensor fusion algorithm it has freeded up a lot of space to do some interesting things. The 500 dg/sec setting is very good but at 250 deg/sec with the 16 bit ADC we are looking at .0038 deg accuracy. I'll take that any day.
Last edited by a moderator:
