[RadioTelescope] Telescope Tuesday

[Andrew Williams]

Craig Williams wrote:
> 
> As I've picked up the motors, we should probably work out how to control 
> them.
> 
> Telescopey stuff will be on tonight at UCC with the controller box and 
> the motors.

Not sure who's on the project, or who has what experience with stepper 
motors. It was a very long time ago, but I spent quite a while playing 
with stepper controllers, both commercial boards and custom made hacks, 
so I'd be happy to help (although I can't make it to UCC tonight).

In particular, you can get a lot more torque out of a stepper motor if 
you run it with a higher voltage and dropping resistors. For example, if 
it's a 12V 1A coil, the nominal coil resistance is 12 Ohm. If you run it 
with a 48V supply and a 36 Ohm drop resistor in series with each coil 
(rated to 36 Watts), you still get a nominal 12V across and 1A through 
the coil as a holding current, but you get much better torque because 
the higher voltage helps overcome the much higher coil impedance at 
turnon and increase the coil current faster.

The downsides are bulky, hot, high-power resistors, quadrupled power 
use, and the need for FETs (and/or controller boards) that can handle 
the increased voltage, but in something that's stationary and mains 
powered, it's probably worth it.

On another topic, as far as I can see from the docs you've sent out, the 
existing encoders are just potentiometers. I'm really not convinced that 
pots are a good way to get the position - even new, the linearity specs 
are probably only barely good enough, and pots are notorious for 
susceptibility to wear, moisture, dirt, etc, as far as the actual 
resistance is concerned. On top of that you've got susceptibility to 
noise on the analogue side before the ADC. Given the pots will be 
exposed to weather, and there's a lot of RFI from the steppers, I really 
don't think they are a good solution.

I'd say a better idea would first be to assume your stepper motors have 
enough grunt to not miss steps, and calculate position using the step 
counts, plus a 'home' microswitch on each axis (you'll need limit 
switches anyway, but best to add an extra switch for home sensing as 
well). If the steppers don't have enough grunt, and you find they lose 
steps under load, you can do the higher voltage trick, or simply reduce 
the slew rate. Failing that, you can make your own quadrature optical 
encoders using slot optosensors from an old ball mouse, and a position 
disk laser-printed onto an overhead transparency sheet, then cut out.

The 61cm telescope here uses a commercial stepper controller, and our 
code doesn't use any encoders at all for motion control - the steppers 
are grunty enough that they simply don't lose steps at the 
velocity/acceleration/jerk limits we use. What's more, we need position 
precision and stability down to a twentieth of an arcsecond, and overall 
pointing accuracy to a better than an arcminute, far better than a radio 
telescope. BTW, 'jerk' is the rate of change of acceleration, and it's 
something you're probably going to have to limit in the motion profile code.

Andrew