Gday Peter
Quote:
I'm having a brainfart, how would the encoder on the worm work?
|
Same as per usual, but the position of the encoder determines how much "guessing" the system has to do.
There are also 2 scenarios here
a) Use an external Hi accuracy encoder to give tracking corrections via the guider port.
b) Use a mid accuracy encoder on the worm to directly replace the feed from the existing encoder on the motor.
I am familiar with the Meades ( which already use encoders ) so would choose option b) and explain it the following way.
With an LX200, the encoder is mounted to the motor.
There is then a 50:1 gearbox then a 180:1 worm drive.
Assuming we had an absolute encoder on the final axis,
to get say 1 arcsec resolution, it needs 1,296,000 counts per rev
If we put the encoder on the worm, it needs 7200 counts per rev
If its on the motor, we only need 144 counts per rev.
If you put the encoder on the output axle, then if it moves, the axis is moving, ie there are no backlash or PE errors between the encoder and the OTA.
If its on the worm, ( and the worm is springloaded so always in contact ), there is virtually no backlash, but there is still the worms PE.
If its on the motor, you have all the lash and PE in the gearbox plus the PE in the worm.
When tracking, the system needs to know the PE, and when guiding ( esp in DEC ), the lash needs to be constant ( and its not )
The Meade LX200 gearbox requires three revs of the output axle before all the gears resynchronise ( hence the three turn PEC model they use )
The LX90s etc require 128 turns or the worm for the gearbox to resynch.
As such "modelling" PEC when it includes a gearbox like this is a non trivial problem, esp if the gearbox PE is large relative to the worms PE.
An encoder on the output axle is optimum, but horribly expensive ( at present ).
A mid range encoder on the worm is a much cheaper proposition, but still needs a PEC model to be used to get good tracking. However, as the encoder is on the worm now, no PE effects from the gearbox contaminate the model, hence an accurate model can be generated.
So its just looking at it on a cost benefit analysis of what method gives the best bang for the buck.
Andrew