my real motor question

[Hank Fletcher]

I'm a bit bummed out, because after a few days of thinking a drill motor would be the solution I want I've realized that the backlash in a drill is far too great for what I need.

What I want is 600rpm and about 1.4Nm torque (absolute max), with a max amount of backlash of 1/2 a degree. I've been looking for this for about a year now! I want to use the motor to drive a wheel of about 60cm circumference, and be able to stop it to within a millimetre (hence the need for low backlash). Am I nuts, or is this possible? Bear in mind that like most hobbyists I'm, ahem, thrifty. Is making my own gearbox a solution, and if so does anyone know where I might get the parts I would need?

 

[3v0]

Is there any chance you could provide more details about the motion of this wheel.

 

[mneary]

Anticipate your stopping point, and stopping at (backlash) earlier?

 

[3v0]

There are games that one can play to make mechanical things preform better then spec. To play them you need to know the machine and what you need to do with it.

Time for 20 questions:

Can you add drag to reduce the backlash (make the machine more predictable)?

 

[Hank Fletcher]

Is there any chance you could provide more details about the motion of this wheel.

I want to be able to turn the wheel from a centre axle, but not continuously in one direction. Just back and forth just under 360 degrees. So it should be able to go from the extremes in each direction in about 1/10 of a second. I'd like to be able to stop the wheel to the precision of about half a degree, or if you're going by the circumference of the wheel, about one millimetre. I think I can sort out the feedback for that okay, so what I'm looking for help here is a motor/gear combo that can deliver these mechanical characteristics.

 

[Hank Fletcher]

Can you add drag to reduce the backlash (make the machine more predictable)?

Anticipate your stopping point, and stopping at (backlash) earlier?

I don't think I can. I've tried this for prototypes and I need something better. The objective is 1mm precision positioning over 60cm, and it's got to be fast, e.g. any of 600mm in less than 1/10s. I'm not sure fudging for backlash will cut it.

 

[mneary]

You can perhaps get position indication from the detents in the trombone? They would cause very predictable blips in current consumption.

 

[mneary]

oops, wrong thread? Which is the real one?

 

[Hank Fletcher]

You can perhaps get position indication from the detents in the trombone? They would cause very predictable blips in current consumption.

There's no such thing. It's all done by feel, ear, and a whole lot of experience.

 

[Hero999]

Please avoid making duplicate threads, it's very confusing and annoying too.

 

[Cabwood]

Backlash can be completely eliminated with feedback. What you should look at is servo positioning. Use a position sensor (like a potentiometer, or shaft encoder) to provide a signal which represents the current wheel position, from which you subtract a signal representing desired position. The result is amplified and is used to power the motor. The advantages are numerous - linear response, accuracy, control of overshoot, ability to maintain position under load etc etc.

 

[Hank Fletcher]

Backlash can be completely eliminated with feedback. What you should look at is servo positioning. Use a position sensor (like a potentiometer, or shaft encoder) to provide a signal which represents the current wheel position, from which you subtract a signal representing desired position. The result is amplified and is used to power the motor. The advantages are numerous - linear response, accuracy, control of overshoot, ability to maintain position under load etc etc.

Thanks. I still have a lot to learn about feedback and servos. My interest is in keeping noise to a minimum. My biggest fear is that following overshoot, the motor whacks back and forth trying to get into the desired position. From what I've learned so far about available rotary position sensors, one of my obstacles is that I'm trying to sense 1/600 of a revolution, and that my wheel is only going to turn just less than one revolution. The best single turn, rotary sensor I've been able to find is only accurate to 0.5%fs (or 1/200 of a rev).

It might be my newbieness that's feeding my skepticism. If a servomotor has a 1000ppr encoder, is it reasonable to assume that the servo will, via feedback, be able to position itself at any 1/1000 of a rev (and in a timely fashion), regardless of the backlash in the servo? Can it be done, and if so, what's the compromise (or is there one)?

 

[Hank Fletcher]

Just a hypothetical situation to illustrate my concerns in my last reply:
Suppose to move my axle I use a cordless drill motor, which with its transmission has a lot of backlash. And for feedback control I use the 0.5%fs sensor, because it's the best I can find/afford. I send a signal of the desired position to the motor controller which it compares to the current position. The motor controller runs the motor, but for whatever reason the motor slightly overshoots the desired position. The controller senses this difference in the feedback loop, so it moves the motor the difference in the opposite direction in order to correct the overshoot. But the backlash in the drill transmission interferes with there being any power put into the axle at all, because it's all taken up in the slop of the backlash.

So what's the best solution, then? To anticipate the amount of backlash in the feedback control, and presume it'll be constant? The timing of the motor control is also crucial to my application, so while the motion of the drill is compensating for the backlash (i.e. not yet moving axle) I suppose I'll have to anticipate/calculate how much time that will/has take(n)?

Sorry if it was naive of me to presume that zero backlash would be an option in any situation - the more I think about it, the more I realize that there will always be a mechanical difference between what you want and what you can make happen. I guess it's a judgment call from application to application as to how much backlash (and its inherent noise, delays, etc) will be acceptable?

As a novice to the field, I just want to make sure I'm on the right track so that further down the line I don't have to face criticism of "Why the heck did you do it that way?!" Please don't be afraid to share you opinion or state the obvious - it's all good!

 

[Cabwood]

The first point I'd raise is that any gearing system will have slack (backlash as you put it), although with care I think it should be possible to reduce it to below 0.5 degrees. Even then, though, in a servo system there will be oscillation because of that. My experience (very little, I admit) is that in servo systems pulleys tend to out-perform gears for precision (because there's no slack), but it's a trade-off against power/speed. If I was to build this myself, I'd almost definitely use a pulley or pulleys, because servos handle pulley slippage and elasticity much better than they handle slackness in gears.

My initial thoughts on your need for precision are about momentum. The 60cm diameter wheel surely has quite some mass, and together with the motor spindle, this in my opinion is the dominant factor in your design. So I believe you are right to question whether a simple first order servo will do the trick, because of momentum.

If you can derive (or find on the web) some second order expression describing the acceleration and deceleration required to move as quickly as possible quickly from one position to another, then you are well on your way to solving this. This would involve feedback of the second order, with all the associated headaches. What you have here is the same problem that has had robotics people tearing their hair out for decades.

Failing that, instead of using the motor to brake the wheel, how about a separate braking system to halt and hold the wheel at the correct position? At least the maths would be simplified.

The resolution of your position sensor is not such a major concern. You could consider monitoring the motor shaft instead of the wheel axle, but that might require a zero-position calibration each time the system is powered up. Position sensor accuracy, in my opinion, is an issue to be addressed after momentum.

Lastly, the "why the heck did you" criticism is something I've always tried to avoid, but been slapped with many times. Somebody will always have a "better" solution. Don't worry about it.

 

[Hank Fletcher]

Hey, you call it slack, too?! I called backlash that once to a U.S. motor supplier, and he practically tore me a new one, so I've been avoiding it! Must be one of those colloquial vernacular things...

Thanks for the input on the pulley vs. gear. I'd given up on the ideas of using a pulley, but your argument has given me reason to look into it again.

The main wheel is made of a thin rim of aluminum, with a simple two spoke axle to the centre. Even then, I also had concerns about the inertia of the wheel and factored them into my calculation that I would need 1.4Nm to move the whole thing from top speed in one direction to top speed in the other.

I've thought about a seperate brake, too. I'm trying to avoid this if possible in the interest of limiting the noise of the thing, or if necessary keep all possibilities for noise isolation open.

Thanks for the notes on criticism. I'm always forgetting that criticism is a dime a dozen, whereas people who actually try to do something is rare.

 

[Sceadwian]

There's slack in the geartrain yes. But they call them anti-backlash gears not anti-slack gears =) The slack in the geartrain when the motor stops cause the driven load to backlash

 

[Hank Fletcher]

^^^ That's both a concise explanation of the terms, and illustrates my concern for the wheel position. If it's just bouncing around after the motor stops, how can I control where it is? I guess the answer is a seperate brake?

 

[Cabwood]

Or pulley + feedback! Sorry. My coat is around here somewhere.

 

[Hank Fletcher]

Or pulley + feedback! Sorry. My coat is around here somewhere.

It's all good! I keep forgetting about the pulley idea, thanks for the reminder. Or antibacklash gears, of course. There's always options...

 

[mneary]

Hmmm. When the motor is slowing down the slide, aren't the gears always meshed in a predictable way? It seems to me that the only time you don't know where the output shaft is, is at the moment when you transition from acceleration to deceleration.

If there's overshoot, or if there isn't enough friction, then there might be some uncertainty.