Originally Posted by Joe Johnson (Post 913204)

...Some say I am too conservative but my experience with FIRST and with automotive actuators tells me that these numbers are not far from the right ones.

Originally Posted by Mike Betts (Post 913140)

Chris,

While I am humbled among the likes of Al and the Good Doctor, I'll give you my 2 cents...

A motor at stall is delivering no mechanical power. All of the electrical power, even constant current, is converted to heat. Most external cooling that teams implement is over the case and not forced into the motor. The heat will build up in accordance with the thermal resistance of the internals of the motor.

This will likely happen very fast.

That little fan inside the FP is directly cooling the motor windings. If that little fan stops, my experience is that failure will occur very quickly.

In my opinion, the PTC added to the FP in recent years is to attempt to save a child's extremity or to mitigate an actual fire in the toy for which the motor was designed. It may not be fast enough to limit damage to the motor.

I would not design a system where the motor could stall.

JMHO,

Mike

Originally Posted by Joe Johnson (Post 913204)

Yes, efficiency is a huge factor and very important for sizing gearboxes and designing mechanisms.

I use the following for designing using torque*:

• Straight Spur Gear with good bearing condition 95%
• Straight Spur Gear with funky bearing conditions 90% (look at the final stage of the FP transmission for an example of a funky bearing condition)
• Planetary
  • Low ratio (<5:1) 85-90% (depends on how good the bearings are, the grade of the gears, the size of the planets w.r.t. their axles... things like that)
  • High ratio (>7:1) 50%
  • Very High Ratio (>20:1 -- "hunting tooth" stages) 15% <<these are great for speed reduction by division, but lousy for torque increases by multiplication
• Helical Gear, parallel axis 80%
• Helical Gear, cross axis see Worm Gear
• Worm Gear -- totally depends on lead angle
  • Best case 50-60% (high lead angles of 40 deg, good bearings, etc.),
  • Worst case 5-15% (lead angles of 10 deg, bad thrust management, etc.)
• Conical Gears / Bevel Gears depends on bear arrangement and alignment 60-90%
• Chain 90% (assuming good alignment and tension)

This is PER STAGE.

Example: If you have a 4 stage 4:1 per stage spur gear gearbox with good bearings it would be .95^4 = 81% efficient. So... ...instead of getting a ratio of 256:1 your "effective ratio" (from a torque point of view) would be 207:1.

Continuing with the example, if you put a FP in with a stall torque of .45N-m then you would get 93N-m out of this gearbox, not 115N-m. Now suppose you are trying to lift your robot with this gearbox and you have the output connected to a .17m arm (and assume your robot weight is 600N, then you need 100N-m to lift your robot.

NOTE: You are not going to lift that robot, all you are going to do is turn a lot of electrons into heat.

Continuing, if you put a 3:1 chain stage between the arm and the gearbox, the effective ratio would be 560:1 (207X3*.9). You could put 250N-m of torque on your arm. Now your motor would be loaded at 40% of its stall during your lift (and the motor would be running at 60% of its free speed or the arm would be turning under load conditions at 12RPM = 16,000RPM *.6/(256*3)<<Note: Actual Ratio used for SPEED, Effective Ratio used for Torque).

Now you'd lift in a heartbeat (1/2 turn in 6 seconds -- well... ...kind of a LONG heartbeat ;-) and you have extra torque should another robot get in your way on the way up.

Life is good.... ...always.

Joe J.


*Some say I am too conservative but my experience with FIRST and with automotive actuators tells me that these numbers are not far from the right ones.

Originally Posted by billbo911 (Post 913500)

I do have have one question though. The following comment lost me, I would like to know where you came up with the 100N/m value?

I'm sure the answer is obvious to some, but for some reason I just can't seem to see how you came up with this value.

Originally Posted by Vikesrock (Post 913508)

The assumed the use of a .17m arm is the connection between the two numbers.

600N * .17m ~= 100 N-m

Originally Posted by vamfun (Post 913440)

Yes, we are all humbled :)

I think from the discussion, that the FP motor will trip the PTC if it is stalled, although, I have yet to see any test data. I would propose a test where the motor is stalled and voltage increased until the PTC trips to see what the max continuous current is (at least for the lab conditions) , except I don't want to risk a motor for the test. My hair dryer has tripped a few times when the dust builds up over the fan exhaust, but it has always recovered. I would hope the FP would recover too. Has anyone actually tripped an PTC yet?

I really am not sure how to design a fixed roller that sucks up a ball to ensure a motor will not stall. It would seem that this requires limiting the normal force on the roller which is difficult when a ball is squished between a wall and a robot. Even with a frictionless back bar, it seems the ball can deform and push hard against the roller. If the roller is at all sticky getting a normal force equal to the output torque/radius doesn't seem too difficult.
We are using a FP with a banebot 16:1, a 1.6 in dia roller. This can deliver about 30 lb of tangential force to the ball and getting a normal force 30/u_roller seems plausible for any reasonable u_roller.

Squirrel and others seem to be able to spin the ball on the rug which implies u_roller>u_rug. Using a slippery roller can keep from stalling probably , but degrades ball magnet performance.

So, we limit the normal force on the roller and actually shut down the motor when the normal force lifts up the roller. I still would be very happy to know if we do stall that the motor is not damaged.

Originally Posted by Joe Johnson (Post 913555)

If you can use a Jaguar and use CAN, you can easily know the current. There are other ways to measure or infer current (sensors, shut resistors, measure speed & calculate, etc.). Once you have current you can easily build a "how close to tripping the FP bi-metal breaker am I?" model. It will not be too hard. If you are using Labview, I think it may be trivial.

Originally Posted by vamfun (Post 913687)

As a controls engineer, this was my first thought but from this thread it seems the problem requires motor speed to be of any use. When Al and I were pondering the Jaguar linearity vs frequency phenomenon, I had proposed a BackEMF motor speed project but never got around to doing it last summer. I wish I had so we could just use the CAN bus and motor voltage to create a current trip schedule without the use of an encoder.<snip>

Originally Posted by Joe Johnson (Post 913700)

Now we have gotten far afield but if you want a project to implement on the Jaguar, consider the following:

This paper (Closed-Loop Motion Control for Mobile Robotics -- by Rich LeGrand available for $1.50) makes reference to the idea of using Back EMF to control a motor. The great thing about it is that you don't need to http://www.chiefdelphi.com/forums/ne...ly&p=913687put any extra stuff on the motor to get PID control. FYI, Rich LeGrand is from Charmed Labs which hosts this example video that dropped my jaw.


If you want to learn more about how it works, look at this link from another cool robot website, Acroname.com. I think they really do a pretty good job of making the concepts understandable. Actually, Acroname.com is part of the way there already in that they offer an H-bridge with velocity feedback built in.

So... I want to see this built in as a feature of the Jaguar. You could command motor speed for any motor on the robot! Think of the well behaved robots you'd see if the default code implemented velocity control as the standard method move a drive wheel? Need a beater bar to move a a standard speed without having the fuss an muss of adding a sensor and implementing a feedback loop? no prob.

This would be a very big deal for FIRST robots.

For your consideration.

Joe J.

Originally Posted by vamfun (Post 913444)

What would you use for a Banebot 16:1 P60 gearbox? Banebot would not fess up to any number so I've been using 40%.

Do you use this factor to modify the free current (free speed) on the output of the gear box? I generally don't, but I know the friction load will slow the free speed down.