Clearpath Motor Tuning

Hi all - I just wanted to share a little tip that took WAY too long to figure out. I am using Clearpath servos on a custom machine that takes some design cues from the Avid CNC, and uses a few components from Avid. My gantry has a lot more material, though - probably about 400lbs total, including a 2.2kw ATC spindle.

My symptoms were a lot of machine jerk, even with lowered accelerations and constant velocity settings set pretty high.

I am using Clearpath SDSK 3421S-ELN motors on my axes and I finally settled on 150 Accel and 600 IPM velocity.

What was curious was that the jerk was worse at low speeds than at high speeds. So, eg. a rapid jog all the way to 600ipm resulted in noticeably less jerk than jogging to 20ipm. It also manifested in chatter in the cuts at reversals of direction.

The solution ended up being to adjust the RAS settings in the motor, which smooths the 3rd and 4th derivatives of the motion profile, so the torque transitions are much much smoother now. What was happening is that because the motor torque was so high, it was causing jerk.

If anyone else has a clear path motor and notices any jerk, I highly recommend turning the RAS up if you are getting any resonance, jerk, or clunking during acceleration. I settled on 99ms for my build which eliminated essentially all the bad effects I was getting.

3 Likes

Good point but didn’t the Teknic rep discover that when they manually tuned it? I have those same motors on the PRO4848 and the RAS is set to the default level. No jerk but I set the acceleration at 100 and a 900 IPM rapids.

I went through 2 tunings and we got good results during the tunes that were not great during cuts. Messing with the RAS solved my problems

How did you decide on 150 for acceleration? My decision on 100 was very subjective. The Avid default for NEMA 34 steppers was 50 and I assumed that the smoother running and tuned responses of the Clearpaths would allow at least 100. That worked well without any jerking on my reinforced frame and weighted down machine. I was afraid to go too much higher for fear of over stressing the Avid mechanics. Not very scientific.

It seems to provide the best reactivity for jogging and helps smooth the motion on complex surface profiles.

While it is counterintuitive, faster acceleration means better motion with constant velocity turned on. It will track closer to your paths.

I am jealous. I have been selling Teknic motors as an OEM for the past 20 years and they have never sent me a factory rep.

RAS (REGRESSIVE AUTO-SPLINE)
RAS is a jerk limiting, and jerk-derivative limiting feature based on
proprietary technology developed by Teknic.
The RAS software uses advanced algorithms to analyze each commanded
move and rapidly calculate and "fit" a forth-order polynomial spline to it.
This converts the sharp transitions between constant velocity and
acceleration with more gradual, rounded corners.
Controlling the rate of change of acceleration in this manner results in
moves that are lower in vibration, quieter, and easier on the machine's
mechanical parts. 

My only issue is that is something happening outside of the controllers knowledge and wasn’t accounted for by the CAM package when it calculated the tool path.

I guess what the CNC folks don’t fully appreciate is these Teknic features were developed for industrial controls. Like if I have a chicken plucker and its vibrating real bad every time the PLC commands it to pluck a feather. I use those settings in the motor because the PLC isn’t smart enough to do that itself.

So from the CAM to the machine controller is one thing. The controllers are constantly figuring out the best way to interpret the g-code. That is its function by definition.

But, when that controller is giving step and direction, it is expecting that motor to obey that command exactly. It doesn’t know that motor has sudden decided to reinterpret its own control signals and do something unexpected.

I am not opposed to function in the motor but myself, I would be worried about the acceleration settings in the controller and how this motor’s reinterpretation of the commands will interfere with each other.

I wouldn’t stick Teknic motors on an AvidCNC but if I did, I would put max acceleration to something really high in Mach4 and use RAS in the motors. I am a Teknics guy so I trust their features more than I would ESS and that Mach4 but I would NOT mix both of them!

Using RAS in a 4 axis machine would probably not give you good results.

The A axis is angular and its motion is time based and as such the fact that each of the motors is reinterpreting its own signals and calculating these splines based on their own input (without knowledge of the other motors inputs), the A axis would not not move in sync with the other axis.

How bad would the results be? Dunno. Go test it and find out for the rest of us :rofl:

So what is really high, 200, 500, 1000 ?

Great question, I love the Teknic motors but I don’t know. Essentially your motor is going to ignore the controllers step and direction stream and do its own thing. The last thing you would want is the two very similar functions (max accel and RAS) interfere with each other. The Teknic SD line are open loop servos, essentially fancy steppers with no feedback to the controller about position, power and PID settings.

So like on the dual Y motors, will the motors arrive at the same calculated spline? Dunno. Will they do it reliably? Dunno. We all know what happens with your Y motors aren’t in sync, right?

Will the XYZ motors compute similar enough splines to keep the cutter in the intended path? Dunno.

I know why I would use RAS, the PLC and the usual suspects are slamming through positions with no thought or care as to vibration. My chicken plucker sounds like a GM truck driving down a gravel road at 90 mph. Some RAS does wonders.

I just not sure if I want that in my CNC machine. Your personal factory motor tuning rep can better answer all that better than I can so consult with him.

I still don’t have a personal rep or motor tuner from Teknics. I have formally filed a complaint. :expressionless:

Here is the video explaining RAS; ClearPath's RAS: Smoother Motion Control than S-Curves - YouTube

I had forgotten this, the SD line you can only select a time period since they lack the full motion control and MCU.

Since the SD are “less smart” and kind of “hard coded” to a time you might get more repeatability from them.

Try it and see! When it breaks, ain’t my fault :rofl:

So what it comes down to is the subjective method I have already used. Teknic provides a manually tuning service for those who have a dual motor configuration and can not get the auto tune function to provide a sufficient setup. You provide two laptops or computers each connected to the USB of each of the Y axis motors, establish an internet connection with them and they manually tune the system remotely. I was asked to run the system at the maximum feed rate and acceleration I planned to use while they did the manual tune. I used a feed rate of 400 IPM and an acceleration of 100. I used auto tune on the X and Z but ask them to also to look at the tuning on them as well which they did. After the manual tuning the performance was greatly improved and did not require any changes in the RAS settings. The better performance, precision hard stop homing and HLFB that is not available in standard stepper motors is totally worth it for me. They are also a vast improvement over the old desktop rig I previously had which used Mach3, Smoothstepper and Teco servo motors.

Ok, so they didn’t turn on RAS when they set it up for you. That should tell you something right there.

I wouldn’t spend a second further chat’n with me and go back and ask then directly if they recommend use of RAS with a CNC machine and if you should disable the CNC motion planner’s anti-jerk functions.

I did speak with them about RAS, and they had it set to 34ms. Turning it to 99ms solved my particular problems with no loss of accuracy.

They key that they told me about RAS was that all the motors in the system needed to match because they are injecting a time delay.

I set all my motors to match and have had no trouble, only improvement.

Why would you lose accuracy?

34 milliseconds is a tiny amount of time, about 1 ms faster than a frame of old school TV. 99 milliseconds though is a lot. I would ask them about the anti-jerk measure in your motion planner. See what they say.

Mach 4 does not have an antijerk motion planner. Jerk is the first derivative of acceleration, and Mach 4 only has a trapezoidal motion planner so there is no anti-jerk built in. the clear path handles the anti-jerk on its own. as long as the order and duration of the signals going from the motion controller are not interrupted there is no problem.

Yes, Mach4 has a near criminal lack of basic features but I was referring to the forum’s boogieman, max acceleration.

I apologize for the poor wording. You got max acceleration values, turn them to the factory settings and jank with the RAS and tell us how it went.

Be sure to post video and photos of results. I would love to see the before and after. Cut us some pretty S and circle with a boatload of squares so we can see the ringing in the side walls and coordination in the axis.

What makes a topic in a forum get pinned is when people post video and photos with their how-to. Otherwise it’s just another trash post that gets washed down the memoryhole.

I’m not avoiding this question. I’m happy to discuss, but I don’t have a machine that is anywhere near stock. It only includes a very few parts from Avid.

Thanks fine. No pressure. But if you do get the time and can do an A-B video or photos that would be super helpful for others.