PDA

View Full Version : How to performance test servos?



billyzelsnack
03-16-2010, 10:43 PM
Over in the humanoids forum I've been blathering on about testing the Turnigy/MKS HV-300 servo. It makes much more sense for me to post over here in actuators.

This is where I am at at this point..
http://billyzelsnack-robotics.blogspot.com/2010/03/initial-hv-300-servo-test-rig.html

I'm most interested is seeing the relationship between the various types of torque ( dynamic, stall, and holding. ) I'm especially interested in seeing the dynamic torque on this particular servo as I know the dynamic torque of the AX-12 has been reported to be less than half its stall torque.

If anyone has suggestions or ideas on the best way to go about this I would love to talk about it. Once I have the method worked out I'll test out some other servos as well.

Also.. If anyone has any data on big boy servos such as the RX-64 ( or anything else ) I'd definitely appreciate seeing that too.

darkback2
03-17-2010, 12:15 AM
OK...

I think the first couple of things is to form a standard jig in which to measure various servos. I saw someone on the forum using a shop vice to hold the servo, and then using KG weights at the end of two back to back SES long servo brackets. That way they could vary the amount of load at a known torque arm length and test performance. I suppose you could mount a protractor behind the servo and measure specific angle movements to see how well it is hitting the mark, and use a stopwatch to test how fast it is moving both without, and under load.

Finally, the test you did already isn't bad, but I would like to see you try it with smoother movements programmed. Right now it seams to move well, but I'm not sure how much of the jerkyness of the movement is the servo, and how much of that is the programming.

Thank you for doing the testing you have done. I'm pretty stoked about these servos, and may pick up a couple.

DB

billyzelsnack
03-17-2010, 01:15 AM
My software is plain dumb at the moment. I just have 5 keys to set various positions. When it goes up in the video it's going from say 0 to 180. When it goes down I step it a little so the weight does not slam down as hard.

This does bring up an issue. Should I be doing these tests at full speed? ( speed in the sense of interpolating target positions ) I guess I should do tests at a variety of different speeds.

btw. Here's the test rig that vstone uses for their high-end servos..
http://www.vstone.co.jp/top/products/robot/v2/vservo/50kgfcm.wmv ( 5MB )

sam
03-17-2010, 03:59 PM
That servo looks really good for the price!

For your rig, I would also suggest clamping down or screwing the servo to something to make it stand still. It depends a lot on the precision you want to get from these tests. The bigger the wheel, I believe the better the precision you could get (let's say you have something like this (http://www.sparkfun.com/commerce/product_info.php?products_id=67))

Your % of error would probably go down since the length is bigger. Get good weights and a way to hold them. I would also (if you can) suggest doing everything in metric (Kg, centimeters and newtons) then you can convert.

If you have a SSC-32 you could program the same movement but slower and faster to see the difference. I'm not sure of your setting, but there could be a way of doing that.

This is interesting, keep us posted!

EDIT : Oh, do you know if the work with the SES?

Sam

billyzelsnack
03-17-2010, 07:30 PM
I'm planning on doing a setup that is solid enough to be repeatable over various speeds. I just threw together an Arduino so I'll be able to do the testing at various speeds. I'll probably make some graphs if the data looks interesting.

I think you are right that the bigger wheel will be more precise. I'll see what I can find. I also will be changing my setup so the load is supported on both sides of the pulley to get rid of some of the side-load.

It's funny.. I used to convert everything into ozf-in to so that I'd have some sort of mental concept of the torque values. No I find myself converting to kgf-cm. I feel much better now. haha.

What's all this SES business people are talking about? I know it's Servo Erector Set, but I don't know how it's related. Do they have a bunch of pulleys available or something?

lnxfergy
03-17-2010, 07:39 PM
What's all this SES business people are talking about? I know it's Servo Erector Set, but I don't know how it's related. Do they have a bunch of pulleys available or something?

I think the question is: does this servo conform to standard sizes well enough to use SES parts on it (I may be wrong, but that's the typical question I see for non-hitec/futaba servos)

-Fergs

Quantum
03-17-2010, 08:00 PM
Regarding the SES compatiabily there already is a new hitec that recently was released that is a little to big for the SES. There is a thread over at Lynxmotion that covers this.

My guess is that the mounting pattern will never change on the servos. But the hieght size of the case might start to change. In time as performance is reguired as RC models get faster the servos that control them need to get stronger as well. The design of the gears will get beefier.

If they released one there will be plenty more.

zoomkat
03-17-2010, 08:50 PM
For testing the servo, I'd mount it on the edge of a table or shelf and attach the center of a long arm to the servo horn (equal am lengths on each side of the servo horn to balance out the arm weight). On the end of one end of the arm attach weights via a string. Start the servo with the arm straight down and let it move upward. When it finally stalls, measure the distance horozontally from the verticle centerline of the servo horn out to the centerline of the string. Multiply this distance by the weight of the weights to see the max torque being produced. Move the arm up to a more horozontal position and release. The arm should go down and stop. This should be the ~holding torque.

sam
03-17-2010, 09:01 PM
That's a nice idea to get nice holding torque and holding torque!

It could be a bit tricky to calculate the horizontal distance precisely thought, even if you decide to measure the angle and use trigonometry. But the setup could be easier.

Oh and I'm pretty sure that servo won't fit with the SES.

zoomkat
03-17-2010, 09:25 PM
It could be a bit tricky to calculate the horizontal distance precisely thought, even if you decide to measure the angle and use trigonometry. But the setup could be easier.

Use a plumbob setup centerded at the center of the servo horn and hanging down. Then measure the horozontal distance between the two strings.

MikeG
03-18-2010, 09:06 AM
You might want to keep track of current and voltage too.

bonmot
03-22-2010, 11:24 AM
Hey Billy, I got my HV-300 2 minutes ago. Wow what a beautiful thing!

billyzelsnack
03-22-2010, 03:14 PM
What are your initial plans for it?

bonmot
03-22-2010, 03:27 PM
My plan was to get the first one tested, and then order more to make two legs for balancing test.
Right now I am working on a two-wheel project (inverted pendulum) to test out the gyro (LPY510AL) and PID logic.
Lot to learn from this project. motor noise affecting the gyro output, integral gyro output to get the absolute position, PID response time, car body flex will afftect the gyro, DC motor PWM, etc ...
So I will finish the two-wheel first. (I will post the project here for sure).
Then start designing a simplest two-leg model (no feet, no knees, trying to avoid kinematics for now) using HV-300.

billyzelsnack
03-23-2010, 10:55 AM
I updated my servo torque to weight ratio chart..

http://billyzelsnack-robotics.blogspot.com/2010/03/servo-torque-to-weight-ratios.html

darkback2
03-23-2010, 11:50 AM
The turnigy servos say "for high speed cyclic servo use only." Is this because they run hot and burn up if you run them for too long? Seams like that would be a big check in the minus column for me.

Might be able to fix that by swapping out the cases or filling them with some sort of ceramic heat sink gel.

Neat list though.

DB

billyzelsnack
03-23-2010, 12:11 PM
That's one of the reasons I want to do the tests before I buy a bunch of servos. My guess is that the electronics don't really like to be at stall current and cyclic uses really won't be an issue there. If that's the reason I hope it's the case over fragile gears. I can deal with stall current ( either by fuse or replacing electronics. )

billyzelsnack
03-23-2010, 02:54 PM
Here's how the invenscience guy tests his big servos..

YouTube- 12 VDC Super High Torque Servo Torxis Cycle Testing

I'm liking the simplicity and it is also testing how well the servo handles the side-loading.

billyzelsnack
03-23-2010, 03:00 PM
Here's another tester I found..

YouTube- Sanwa High Torque Servos Crazy Test

Though.. It's kinda sketchy when they don't stay the coefficient of friction ( static or dynamic ) between the two surfaces.

billyzelsnack
03-28-2010, 04:02 PM
http://www.youtube.com/watch?v=S6eDEW_MepQ

darkback2
03-28-2010, 05:25 PM
Nice testing jig. Do the servo horns meet either futuba or hitech spechs? cause they make metal horns for both...might help in the slip department.

Thanks for the video.

DB

billyzelsnack
03-28-2010, 05:43 PM
They are Futaba splines. I'll be ordering a metal one soon.

I'm very interested in seeing how this works out. Stall torque for this lever length should be about 15lbs. The plastic horn stripped at 7.5lbs, but you can see it slowing down. It'd be cool if I put on an external pot and graph the speeds for given weights.

Another thing that is interesting is that peak current is already at 2.25 amps. With that I kinda have the feeling that I'm gonna be blowing the FET's before I get to full weight.

darkback2
03-28-2010, 07:05 PM
I guess that is why its meant for momentary use.

DB

RobotAtlas
04-02-2010, 07:31 PM
Billy, what are you going to do once you find a perfect motor (for your bigger biped?)?
Are you going to turn it into a servo using something like what's discussed over here:
http://forums.trossenrobotics.com/showthread.php?p=39394

billyzelsnack
04-02-2010, 11:00 PM
I actually have two "bigger" bipeds in mind. One that's about 50 or 60cm and one that is Asimo-ish size of 120cm. A few years ago I bought a bunch of surplus paper shredder gearboxes. They are compact, solid, not too heavy, nice gear ratio ( ends up at 54rpm ) , but the motor it comes with has an insane 90amp stall current! The theoretical torque on this puppy is something ridiculous like 8000oz/in! I'm not sure I even believe my own numbers, but a boy can dream.

Anyway. I have a 30amp bidirectional ESC that I've intended to fake test with ( with a fuse ) for awhile. Now that I have a proper servo test rig I might actually get around to doing it.

RobotAtlas
04-03-2010, 10:25 AM
At 8000oz/in, how heavy should your robot be to stall the motor? :)

Billy, on a separate note, have you thought about finding partner(s)/sponsors to go to RoboCup, so you don't have to do/pay everything yourself? I know going to RoboCup has been your dream for a while (4 - 5 years or more?).

billyzelsnack
04-03-2010, 11:43 AM
The big gearboxes with motor weigh 365g so a full robot would probably not get near stall torque, but it'll probably routinely reach in the tens of amps. That's one of the things I'll find out once I finally test.

It's funny. I either have the time or the money, not ever both at once! haha. I've made a lot of progress in the last 6 months now that I have more time, but money is tighter. I've considered sponsorships, but for the moment I'm not comfortable with where I am at to approach people.

I'm patient though. I'll eventually make it to robocup and I think I'll be pretty competitive after a couple competitions.

bonmot
04-08-2010, 09:56 PM
Hi Billy, what's your next step on these HV-300s? Are you going to do the electronics? I just finished my "helloworld" on my GYRO test.

http://forums.trossenrobotics.com/showthread.php?t=4032

It performed as I expected. And works pretty stable with the body self resonance and electrical noise from the motor PWM control. The next thing maybe an acceleration sensor to detect linear movement. (right now I just use the motor PWM signal to estimate the moving distance)
Then I am ready to start my "two leg" self balancing platform.
Rightnow I am thinking about some smaller servos to get the sensors and software going. After it can self balancing itself, I will scale it up.

billyzelsnack
04-09-2010, 10:56 AM
I'm moving to Iowa in 2 weeks and my stuff is packed up already. My house here in Wisconsin is really small and it looks too scattered for open houses with my work desk setup.

My next step was/is to get an aluminum servo horn and continue with my torque tests. I want to finish up my series of tests, but mainly figure out what the dynamic torque is and if it over heats quickly. When it does warm up I'd like to know if it shuts itself down or blows itself up.

If the gears slip or blow up that's bad news. If the motor blows up that's bad news unless I can find another motor. If the electronics blow up I'm not really concerned. If the servo manages to survive the tests I'll probably immediately order another one and replace the knee servos in my AX-12 bot and see how well the HV-300 do in a real bot. My custom dynamixel bus interface supports PWM servos directly so I'd not even need to replace the servo electronics at first.

bonmot
03-23-2011, 11:43 AM
Hi Bill

How are you. My robotics project stopped for almost a year. Busy making $$$ :-D
Even though, I just ordered a desktop CNC to make plastic parts without any plan yet :-D.

But how about this servo story? The one I bought still sitting in my tool box :veryhappy:

Thanks

billyzelsnack
03-23-2011, 12:45 PM
The servo still remains untested. Doh.

bonmot
03-23-2011, 01:16 PM
LOL, my servo still remains unboxed.

bonmot
03-25-2011, 10:23 AM
Hi Bill
Where did you get your metal horn for the HV300 you have just sold?