Subject:
|
Re: 0.5 degree resolution with rotation sensor
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Sat, 3 Jun 2000 22:39:49 GMT
|
Viewed:
|
594 times
|
| |
|
|
In lugnet.robotics, Fredrik Glöckner writes:
> "Moleque BR" <molequebr@cs.com> writes:
>
> > In order to get a 1 degree precision, a 22.5:1 gear train would be
> > necessary, but then, why not a 45:1 gear train? That causes the
> > input to turn 720 times for each turn on the output, which gives a
> > 0.5 degrees precision.
>
> Aren't you fooling yourself slightly here? Just because the input
> turns 45 times for each turn of the output does not mean that the
> precision of the output will be 45 times as good. What happens, for
> example, if you reverse the input axle? The angle sensor will start
> registering that the input axle turns the other way at once, while the
> output will lag behind because of the gear train.
Actually, I have built such a beast. Or rather, an even more precise beast-
don't remember the exact gear setup, but it was roughly in the 625:1 range (4
sets of 8:40(?) gear combinations.) Slop was minimal- a harsh stop and reverse
might lose 20 or so counts out of 1,000. However, there were other problems:
1) I used legOS. The generic RCX firmware does not sample the rotation sensor
as frequently, so if you intend to use such a huge geartrain at high speed,
you'll lose counts left and right. If you can guarantee that you'll have <
1count/3ms, then you should be OK with the standard firmware.
2) It took a long time to construct the train, because even very slight
deviations in axle alignment made the plastic bind terribly. This requirement
also made the entire structure bulky and unwieldy. The practical application I
intended it for failed as a result :( Don't know what your intent would be, so
I don't know if that will be a consideration.
3) Remember that as you increase gearing, the necessary forces change. You'd
be surprised at how much it takes to overcome just friction once you reach
that point.
Anyway, just my observations...
Luis (RCX 3.0 had better have Lego ball-bearings ;)
|
|
Message is in Reply To:
 | | Re: 0.5 degree resolution with rotation sensor
|
| (...) Aren't you fooling yourself slightly here? Just because the input turns 45 times for each turn of the output does not mean that the precision of the output will be 45 times as good. What happens, for example, if you reverse the input axle? The (...) (24 years ago, 3-Jun-00, to lugnet.robotics)
|
6 Messages in This Thread:
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
