View Full Version : Dynamixel Limitations?

11-29-2010, 05:13 PM

I've been doing some reading trying to determine the actuators I should use on my project. The end goal is to use ROS with an on board PC (PC is still undecided... Roboard, Gumstix, Fit PC etc.) on a humanoid. I'm hoping to be able to build a humanoid capable of dynamic balancing, but I probably won't use that capability at first.

I need to have a better understanding of the dynamixel's capabilities and limits before I make a purchase. The AX-12's might work for the arms or a few low load joints, but in general I'll have to use the $200 to $500 versions (probably separate power supplies), which is why I'm not just buying them to see what they can do. (I haven't done torque calculations yet to see how much I need - will be done shortly).

Now the questions.

1. What is the backlash and real world precision of the output angle. Are the encoders mounted on the output shaft or on the motor shaft?
2. What is the max update rate of a series of 20 to 24 dynamixels connected together. I've read as low as 10 Hz on some forums. I saw a post saying the roboard can communicate with one dynamixel with a delay of 1.2 ms, but that's not so useful. What are you guys seeing in your robots?
3. Do the motor torque outputs relatively closely match the specs? Does anyone have a torque RPM curve, or is it close enough to linear to just use the basic specs and interpolate? I've read of a bug that causes the servos to shut off if the load is only about 40% of max. Has anyone encountered this?
4. The kit brackets are set up to use the output shaft as the joint axis. Is the servo output axis really designed to take the load directly? Has anyone broken a dynamixel shaft/gearbox?

Thanks in advance.

11-30-2010, 08:36 PM
Well I found one of the answers in this teardown http://www.hizook.com/blog/2010/03/14/robotis-dynamixel-servos-overview-applications-tear-down-and-open-source-software

It says the encoder is mounted on the output shaft (which is good). I wonder if backlash compensation is automatically handled by the dynamixel then.

I'm still looking for answers to the other questions if anyone has them.

2. Max update rate of 20 to 24 servos connected together
3. Does anyone have a torque rpm curve? Are you seeing torque values matching the specs?
4. Has anyone broken a dynamixel with normal use?

11-30-2010, 08:40 PM
4. Has anyone broken a dynamixel with normal use?

I've only broken 2 AX-12s -- and my servos get a lot of abuse. Haven't had the bigger ones long enough to have any feedback. The first one, I turned off overload detection on Issy during RG'09 and stripped the gears -- $3 in parts later, it was working again. The second was on a robot arm that had RX-64s for shoulders -- I forgot to snip a strap holding the arms together for transport and ripped the side-mounted bracket off the servo (along with 4 of the mounting holes) -- really, fairly minor case damage.