Page 1 of 4 1234 LastLast
Results 1 to 10 of 34

Thread: Testing the MX-64T and MX-106T

  1. #1
    Join Date
    Jan 2008
    Location
    Norway, Stavanger
    Posts
    790
    Images
    276
    Rep Power
    79

    Testing the MX-64T and MX-106T

    Hi,

    Some months ago I bought a couple of MX-64T for my next project but also for testing before I buy (invest) more of them.
    At the age of 41 I seldom or never get the exciting Christmas feeling when getting a gift anymore. But I must admit I got the great feeling and had a big grin for a long time when I got the box of brand new MX-64T. You really get the feeling of holding a little massive "stone" in your hand, lol.

    But I've a Q about the servo horn and the Thrust washer. I couldn't find anything about it in the manual but I believe it's very obvious where to place the Thrust washer. At first I was afraid the washer made to much friction, but it seem to be ok. Looking at some pictures of the RX-64, did they use a larger white Thrust washer or no washer at all?

    Just a couple of pictures:



    Correct placement of the washer I assume?



    Like I said I'm planning to do some tests. I bet some of you have already done a lot of testing, but I like doing some for my own learning. One thing I noticed is that the gear backlash is pretty visible, but I don't think it will be a problem for my projects.

    I'm very curious about their new features, like real PID control, current sensing and goal torque. And most of all, doing some torque tests. The manual refer to stall torque. Is that holding torque or stall torque while moving? Often the holding torque is much higher than the actual torque it can deliver while rotating.

    A bit sad a fully charged 4 cell LiPo probably is to much. Maybe a 4 cell A123 would be, but the standard packages come in limited capacities (as far as I've seen).

    I'll see if I can do some testings when I'm done (or almost done) with MorpHex.
    Last edited by Zenta; 10-05-2014 at 12:47 PM.
    Kåre Halvorsen aka Zenta
    ---------------------------------
    Zenta's YouTube channel
    Zenta's Blog
    Zenta's Instagram

  2. #2
    Join Date
    Sep 2010
    Location
    ಠ_ಠ
    Posts
    2,285
    Images
    27
    Rep Power
    275

    Re: Testing the MX-64T

    The thrust washer on the MX-28 just goes behind the horn, so I expect the MX-64 is the same.

    jwatte lost several MX-64 partly from using 4S LiPo voltages, but I've run a DARwIn-OP with a 4S LiFePO4 without any obvious damage. There was one servo on the DARwIn-OP briefly 4S LiFePO4 powered that sticks a little, but not sure if that was from the battery or pre-existing damage from impacts since I did not examine it closely for a proper before and after comparison. The pack looks like a stick of butter and was made by combining two 2000mAh 2S packs together. HobbyKing has a decent range of inexpensive flat-cell packs, but shipping can be excessive if the pack is not available from a local warehouse. 3S LiPo is probably the safest choice voltage-wise and often initially cheaper than 4S LiFePO4.
    Please pardon the pedantry... and the profanity... and the convoluted speech pattern...
    "You have failed me, Brain!"
    bleh

  3. #3

    Re: Testing the MX-64T

    I would recommend 3S LiPo or 4S LiFePO4. The LiFePO4 is a safer and more robust chemistry, but slightly heavier, and more expensive.

    I think the problem with 4S LiPo on the MX64T is that the overload protection is not fast enough to turn off the servo before it damages the motor in case of stall/overload. Having stalled some motors for just a few seconds (before overload cut off,) the motors then became sticky. Then, running the sticky motors with load (but not stall-level load) caused them to generate enough spikes on the voltage and/or TTL bus to kill the circuit boards of nearby servos. This leads to a "servo vampirism" problem that requires you to fix all the servos at once, or the problem will re-occur.

    Robotis claims that the PCB damage was covered by warranty, but motor damage was not. I paid them rather than fight about it, but I think it's actually a provable design problem in the MX64T with 4S LiPo. Switching to 11.1V LiPo totally solved that problem, so I have a working bot that doesn't curl up and die when I use it now.

    If you have a sticky servo, I'd highly recommend sending it in for service. Changing the motor is < $100 which is cheaper than a new servo... Or, at the least, measure the current draw of the servo when idling; if it's more than the nominal 40 mA, you're almost guaranteed to have a bad motor and need the server. Another reason for "sticking" could be damaged gears, which would be cheaper to change, and which would not generate the same kind of servo vampirism, though, so maybe that's what's going on for you.

  4. #4
    Join Date
    Sep 2010
    Location
    ಠ_ಠ
    Posts
    2,285
    Images
    27
    Rep Power
    275

    Re: Testing the MX-64T

    There are a couple servos in the theatre group's DARwIn-OP that have slightly damaged motors, but they always run at 12V or 3S LiPo so the damage does not really propagate to other RX-28M servos on the bus (yes, most in that bot are still original RX-28M). There was definitely a batch of MX servos that had serious issues (just like the CM-900 ES and v1.01 had with the TTL buffers), but all the servos made before that suspected redesign were just fine. Not sure of long-term reliability since that batch made it out of Robotis (may have had another redesign or reverted to older design, or just been a bad batch of components), but there have been no obvious problems yet with the MX servos we got late last year.

    They have damaged several gear sets, which led to the little scuffle with Robotis support over changes to tolerances and/or coatings that prevent removal/assembly of the bearings onto the output gear. Gear damage always has a distinct, periodic resistance or clicking to it, while motor damage tends to be more of a constant resistance to rotation by hand. Impacts that cause only a small divot in the output gear are barely noticeable except for a little sticking and excess backlash in one small part of the horn's rotation (gear sets from at least two hip servos had to be replaced for that).
    Please pardon the pedantry... and the profanity... and the convoluted speech pattern...
    "You have failed me, Brain!"
    bleh

  5. #5
    Join Date
    Jan 2008
    Location
    Norway, Stavanger
    Posts
    790
    Images
    276
    Rep Power
    79

    Re: Testing the MX-64T

    Thanks for your warnings about using 4S LiPo. I did assume that would be to much. Might try the 4S LiFePO4 if I need the extra torque. When it comes to torque measuring I'm thinking of simply making a wheel with 10cm radius attached to the servo, fix a wire to the wheel and add weight to the wire. I'm still curious if the stall torque is actual stall torque while moving or just holding torque. Also want to know how much it deflects from the position under load.

    For example, the Hitec HSR 5990TG servos, are rated to about 30kg/cm. But the actual torque while moving is just about half..
    Kåre Halvorsen aka Zenta
    ---------------------------------
    Zenta's YouTube channel
    Zenta's Blog
    Zenta's Instagram

  6. #6
    Join Date
    Sep 2010
    Location
    ಠ_ಠ
    Posts
    2,285
    Images
    27
    Rep Power
    275

    Re: Testing the MX-64T

    There was a recommendation somewhere on the Robotis support site that the safe, continuous-duty, working torque for dynamixels was ~1/5 of the stall torque. Stall torque is the load where the horn will no longer be able to move to a goal position and will rapidly cause overtemperature and/or overload errors that disable torque. If the external torque or motor current spikes too quickly, the servo may not survive long enough to throw an error that disables torque, and repeated mild overload/overtemperature conditions can eventually lead to motor and H-bridge damage.

    Another precaution we took with the DARwIn-OP was changing the overheat error temperature to 55 down from the default of 80. There is no discrete temperature sensor attached to the motor casing in the MX-28, and there is a significant time lag in heat from the motor or H-bridge ICs reaching the internal temperature sensor of the STM32. Not sure if the MX-64 includes a temperature sensor outside the one integrated into the microcontroller, but overheating killed at least one MX-28's motor (damaged the bearing or brushes) so it can only hurt runtimes if you err on the side of caution.
    Last edited by tician; 05-07-2014 at 02:26 AM.
    Please pardon the pedantry... and the profanity... and the convoluted speech pattern...
    "You have failed me, Brain!"
    bleh

  7. #7
    Join Date
    Jan 2008
    Location
    Norway, Stavanger
    Posts
    790
    Images
    276
    Rep Power
    79

    Re: Testing the MX-64T

    Thanks for your heads-up and information Tician!
    Quote Originally Posted by tician View Post
    There was a recommendation somewhere on the Robotis support site that the safe, continuous-duty, working torque for dynamixels was ~1/5 of the stall torque.
    Oh. So what you are saying is that the servo can work with a constant load over a long time without heating up to much at the 1/5 of stall torque? I bet that would be a problem when using it on a robot arm then. I think it's a good idea to read the temperature while testing it then.

    I wonder if Robotis has considered to make a servo body with metal heat sink?
    Kåre Halvorsen aka Zenta
    ---------------------------------
    Zenta's YouTube channel
    Zenta's Blog
    Zenta's Instagram

  8. #8
    Join Date
    Sep 2010
    Location
    ಠ_ಠ
    Posts
    2,285
    Images
    27
    Rep Power
    275

    Re: Testing the MX-64T

    There are the Dynamixel Pro with all metal chassis, but frickin' expensive they are. The EX-106+ had an aluminium front plate for extra strength, but that appears to have been replaced with an all plastic body in the MX series. They do have vents in the sides of the servo case to help a little bit with heat, but it does not do much for the knees of a DARwIn-OP (after 20+ minutes of continuous high-speed walking and kicking) except to ease the application of CO2 duster output direct to the motor case. Really need to get work done on the gait search, because the default walking engine really pushes the knees (its basically running in place with slight changes to foot displacements from center).
    Please pardon the pedantry... and the profanity... and the convoluted speech pattern...
    "You have failed me, Brain!"
    bleh

  9. #9

    Re: Testing the MX-64T

    Also want to know how much it deflects from the position under load.
    You have control over the PID constants, so if you crank up the I gain, you get to "within backlash."
    The backlash is the worst part of these servos (apart from the servo vampirism, which I have under control.) The tip of my legs can probably move 4-5mm because of backlash in the ankle and knee joints.

    Yes, "safe continuous torque" is said to be 1/5 of stall torque. Then again, for a walker, the torque that a leg sees most of the time is a fraction of what the worst-case torque would be.

    For cooling, I came up with this arrangement:



    However, after solving the servo vampirism problem, and using 3S LiPo, I haven't needed it.

    That fan is squeezed in place perfectly by a pair of 12mm M2.5 nylon machine screws.

  10. #10
    Join Date
    Jan 2008
    Location
    Norway, Stavanger
    Posts
    790
    Images
    276
    Rep Power
    79

    Testing the MX-64T video and result

    Thanks for your input guys!

    I like your fan setup jwatte! Did it help much? Temp readings with and without?

    Another thing, before I started looking more closely to the package and servo specs I found that they use M2,5 screws. Sigh, all the time I thought it was M3. M2,5 are much harder to get.. Well, not a problem though.

    I've spent an hour doing some load/torque testing. The setup was very simple. I made a wheel of plywood with the radius of 10cm and used a lathe to make it perfect round and made a track/slot for guiding a string. By placing the wheel vertical and hooking some weight I got a good setup for testing the torque. 1 kg of load require 10kg/cm torque.

    I made a table in excel showing my findings:



    I believe my findings confirm Tician's information about 1/5 of the rated torque at constant load.
    What surprised me was that the 15kg/cm constant load made the servo shut down after just 4,5 min. @10kg/cm the servo temperature increased by about 1 deg per minute. After 18,5 min I ended the test. The motor got very hot and I didn't want to go higher..

    Measuring the stall torque while moving was interesting. 30 kg/cm was very close to the limit. @ 35kg/cm the servo didn't move at all. By using a fishing scale and placing the wheel horizontally I measured a stall torque of about 32 kg/cm @ 12,1 volt.

    Of course, the holding torque was good. The deflection got very visible at 60-70kg/cm load.

    Funny though. My findings are almost exactly like it was for the HSR-5990TG's. Actual stall torque while moving is about half of the rated torque.

    At least I know what numbers to take in consideration.

    Hereby I'm renaming the MX-64T to MX-32T, LOL.

    Here is a short unlisted video of the MX-32T under constant 15kg/cm load. 30 sec after this video the servo shut down:

    Last edited by Zenta; 05-07-2014 at 03:44 PM. Reason: typo, there are probably more too..
    Kåre Halvorsen aka Zenta
    ---------------------------------
    Zenta's YouTube channel
    Zenta's Blog
    Zenta's Instagram

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Similar Threads

  1. Mx-106t
    By HyperCube in forum Garage Sale
    Replies: 0
    Last Post: 02-18-2013, 11:25 PM
  2. AX-12 Electrial Efficiency testing
    By sarendt in forum DYNAMIXEL & Robot Actuators
    Replies: 2
    Last Post: 03-13-2011, 12:46 PM
  3. Stress testing and debugging AX-12+ servos
    By Pi Robot in forum DYNAMIXEL & Robot Actuators
    Replies: 24
    Last Post: 01-16-2011, 03:14 PM
  4. Virtual Testing
    By Firestorm65 in forum Mech Warfare
    Replies: 5
    Last Post: 02-22-2009, 12:45 AM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •