backdrivable gears

[Steve Bailey]

This seems to be an issue of basic mechanics.

Let's say you have a servo motor/gearbox rated for 180 Lb-in at the output shaft connected to an 18 inch arm. That means the arm can impart a force of 10 pounds at its end.

When your servo drive is given a position command, it will rotate the arm to the specified position and resist any external effort to move it away from that position. If someone "backdrives" the arm by pushing on it, the servo motor will resist that force. Pushing on the arm will move it away from its desired position. This will create a position error which the servo controller will use to command the motor to move back toward the desired position. If you "backdrive" by imparting less than 10 pounds of force at the end of the arm (less than 180 Lb-in torque), the motor can overcome that force and move back to the desired position. If you "backdrive" with 20 pounds of force, that's 360 Lb-in of torque at the gearbox, twice the rating. If the gearbox is the weakest part of the drivetrain, you'll strip the gears. If the gearbox is sufficiently robust, you will move the motor far enough away from the target position to the point where the motor develops maximum torque. Motion beyond that point for too long a time generally causes a motion controller to shut down on a following error fault and the motor stops resisting the external force and simply freewheels. Perhaps the servos you burned out were set up to simply resist the externally applied force no matter what and fried because they were maintaining peak torque for too long.

 

[kamenges]

Steve-

Your description works correctly in the forward drive direction. It assumes 100% efficiency in the backdrive direction, however. Since you don't get this you will lose some of the output shaft 'input' torque in gearbox losses. Depending a several factors these losses could be significant.

student_ucsd-
Good description of the application. That will help. Until Sparkz posted about the RC servos those never even occured to me.
First of all, you said the arm only needs to rotate about 90 degrees. If you don't already do this, mechanically constrain the motion. The last thing you want is to lose control of the shoulder and have the arm whipping in circles.

I like your thought of using a DC motor as this provides for a pretty simple speed limit method. If you have more time than money making your own controls may make sense. It would be a good learning experience if nothing else. In the real world the time/money comparison most often favors money so we buy our stuff. Look here:
http://www.kbelectronics.com/
or here:
http://www.a-m-c.com/
for some commercially available drives. AMC is probably the better choice for you. If you do build your own stuff make sure you include a current limit structure. That will keep you from burning up the motor or drive, even if you backdrive it. You may also want to consider going fully digital. There are companies like these:
http://www.tern.com/portal/ or http://www.rabbitsemiconductor.com/
that make small C-programmable controllers. This will just make it easier to modify your control structure if needed.

You can use a motor with a higher voltage constant than you have available voltage in order to limit the speed. For example a motor with a K_v of 100 volts/1000 RPM will limit to 100 RPM if you apply 10 Vdc. You can use this to your advantage to guarantee that you stay below certain speeds.

I would be inclined to use as low of a gear ratio as you can get away with. This will give your system better 'feel' on backdrive. No matter what geartrain you chose the lower the ratio the easier it is to backdrive. However, there is a trade-off. Lower ratios require higher motor torque to get the same end of arm force. Unless you get a motor specifically wound for high torque/low speed operation the torque requirement may yield a system with a very big motor that take alot of current. These:
http://www.motionvillage.com/products/motors/torquers/index.html
might do the trick for you.
Keith

 

[Clay B.]

May be I am thinking to simple

From reading the description I would think a indexing clutch would be the way to go. We use them all the time in areas we think we may jam. We have a position prox that detects when we reach a position. If the position is not made during a set time we just shut down. If you are using an encoder and your position lags by to great a factor you could use it instead of a prox. Also once a colision has happened. Child grabs arm, what ever you could the return to home position, then move the arm back into position and which will reengage the clutch.

Like I said, this maybe too simple a solution.