Oval Holes

Hi all,

I hope someone can come up with any ideas as to why I am getting oval holes on the diagonal (10.30 and 4.30 o'clock). I have managed to minimize it but still getting a 0.5mm error either side in holes 20mm in diameter. I have rack and pinion on x and y with negligible backlash.
Things I have tried include:

• Lowering feedrate,
• increased acceleration
• Exact stop mode
• Smaller tolerance settings
• Lower kernel frequency and increased buffers

I read this in another forum:

"Backlash affects you at MOTION X DIRECTION CHANGES (12, 3, 6, 9 o'clock) - the monkey motion problems at 45 deg are happening while the motors are attempting to do the most/fastest relative changes in velocity - one is speeding up, the other axis doing the opposite SIMULTANEOUSLY... A simple look and see at the motors will show/prove this
To make the motors "walk and chew gum" at the same time, they have to be timed EXACTLy the right speed relative to each other thought the motion cycle. We chased this all over hell and back until we FINALLY figured out how to properly tune for "follower error". Cutting ROUND circles is THE definitive test to assure you that you have servo follower error tuned properly.

Trust, me, it isn't back lash and it ish't backlash comp. It is a problem in the instantaneous RATE of feed change that the axis are making that is causing the 45 deg out of round problems. If the controller is NOT changing the RATE at which the step pulses are instantaneously being fed to the steppers, they won't/can't keep the poper relative speed to each other and they thus CAN'T cut round holes."

"It depends where it is out of round and by how much. Errors at the direction change points (12, 3, 6, 9 o'clock) are typically caused by backlash and take up errors in your drive'ball screws. Out of roundness at 45 degrees off the vertical or horizontal axis are due to servo tuning errors. At this point, both servor are affecting a high rate or change in opposite directions. Unless the tuning is perfect, you;ll start getting follower error and out of roundness."

I have ordered an extension so I can connect to the com port on the drivers (LCD A86Hs). Is there anyone familiar with these settings or had issues such as this before? Jazz, I know you use Lichuans, have you ever had to tune the drivers for such errors?

Attachment 29984

Quote:

---

Originally Posted by handles

I have ordered an extension so I can connect to the com port on the drivers (LCD A86Hs). Is there anyone familiar with these settings or had issues such as this before? Jazz, I know you use Lichuans, have you ever had to tune the drivers for such errors?

---

Yes, I use Lichuan closed-loop Steppers all the time, I buy them by the 100's and can honestly say I've never had any issues. However, don't confuse Closed-loop steppers with servos, they work very differently.
The Following errors described in the example you have given you won't get with a Closed-loop stepper system because they don't work like Servo's where the drive and motor work closely together for positional rotor accuracy and stability. Yes the encoder on the closed-loop stepper system checks for the position but the drive doesn't need this feedback to get into or maintain a stable position like a Servo drive does as it's still a stepping motor and just moves a set amount of steps, it's more of a positional check and then correct or fault if the error is too large.

If you don't have slop/flex or back-lash then I would look more to the motor tuning, the usual culprit is too low acceleration.

Thanks Jazz, that is very helpful.

I retrammed and recalibrated the X and Y/A axes, and could find no discernible backlash. It is basically an Avid replica with their Rack and Pinion drives. Perhaps my measuring tools are just not accurate enough, but I managed to get an acceptable result by adding 0.15 mm backlash compensation to the X with a compensatory acceleration of 20% extra (600mm/s/s/).
I set the max Velocity at 5000mm/m and Acceleration at 500mm/s/s on a small item profile.

The maximum before lost steps is actually 21000mm/m at 1000mm/s/s acceleration, but I would have to bolt the table to the floor to run at anything like those speeds with a gantry of around 80 or 90kg total. Seriously powerful motors those Lichuan Nema 23/24s. Thanks for the recommend on those btw.

The OP on that forum with the Out Of Round problem was confusing the terminology, as he linked to NEMA 34 3 Phase 12NM Closed Loop Stepper Servo Motor Driver Kit he said he was using, not true Servos. His solution was:

"I was right about it not being backlash or anything loose in the system. I went through all the settings on the drives and I found a setting for speed smoothness.. This was set at 8 on the X-axis and 2 on the Y-axis.
The consequence of this difference between the two settings was the 0.5mm OOR on all circles. When adjusted to 8 there was a pretty aggressive speed smoothing going on.
I adjusted both drives to 2 and behold, perfectly round holes and even sized diagonals in squares."

I ran the debug software on the drives and all settings are identical, just with the exception of the Current Loop Proportional and Integral Gain, which I see is set by the drive automatically according to current fluctuations I take it. The Position Loop Proportional Gain, although I take your point re servos v closed loop steppers, is set way above the default, at 2500 v default 600. The Speed Loop Feedforward is way lower than the default at 35 v 120 and Position Tracking Error Threshold is at 4000 v 400 default.

I surmise that I could play with adjusting the Speed Loop Feedforward or other settings but I don't have enough knowledge on this, and how they relate to IJK commands on steppers v servos, so I will leave that for now, after all I am not milling parts for the space shuttle.

Quote:

---

Originally Posted by handles

Thanks Jazz, that is very helpful.

I retrammed and recalibrated the X and Y/A axes, and could find no discernible backlash. It is basically an Avid replica with their Rack and Pinion drives. Perhaps my measuring tools are just not accurate enough, but I managed to get an acceptable result by adding 0.15 mm backlash compensation to the X with a compensatory acceleration of 20% extra (600mm/s/s/).
I set the max Velocity at 5000mm/m and Acceleration at 500mm/s/s on a small item profile.

The maximum before lost steps is actually 21000mm/m at 1000mm/s/s acceleration, but I would have to bolt the table to the floor to run at anything like those speeds with a gantry of around 80 or 90kg total. Seriously powerful motors those Lichuan Nema 23/24s. Thanks for the recommend on those btw.

---

Well first off 600mm/s/s is low, I'd at least double it, this will help straight away with reducing ovals if it's an IJK issue. However, R&P can be a pain in the arse regards getting the gantry square if the linearity of the racks isn't spot on, esp on a wide machine like an 8 x 4. So are you sure the gantry is spot on square.?

Quote:

---

Originally Posted by JAZZCNC

Well first off 600mm/s/s is low, I'd at least double it, this will help straight away with reducing ovals if it's an IJK issue. However, R&P can be a pain in the arse regards getting the gantry square if the linearity of the racks isn't spot on, esp on a wide machine like an 8 x 4. So are you sure the gantry is spot on square.?

---

Thanks, I will adjust the acceleration and run a few more tests. I didn't realise that this was on the low side.

I spent quite a bit of time making sure the racks were mounted square on the EN3B steel angle I used, meshing their teeth together to make sure they were symmetrical and flush to each other along their length, and then mounted them on the ally extrusion running a dial indicator off the linear bearings to ensure they were as parallel as possible to the linear rails. Similar care was taken building the sideplates and gantry itself.
I am no more than a tenth or so of a millimetre out on the hypotenuse of a 600/800/1000 mm right angle triangle. I will do another measurement tomorrow to make sure it is absolutely exact. I have a home switch both on the Y and A for self squaring but it was pretty square from the get go and I haven't had to pull it into line with the homes. Maybe it is just a little more tweaking that is needed.

Thanks for the help!

I bumped up the acceleration to 1200mm/s/s and tweaked the gantry which was out by a few tenths of a mm with the autosquare. I will be moving house in a few months so I will check then to see if there is any wear from any slight binding but it runs fine.
I dropped the backlash comp on the X to 0.1mm, and will see if I can drop it some more with a couple more tests tomorrow, but the holes are spot on now.

Thanks again for your input, much appreciated.