This Centroid setup is wrecking my machine!

[ashesman]

So, after a few years of running the Centroid Oak on my Mazak, I have always had an annoying banging sound at the start of movement. It has been there since the Centroid setup went on, was not there on the OEM Mazak controller.

I had got to the point I believed it was mechanical. I even re-balled all the linear rails and pre-loaded ball screw bearings. I wondered if it was caused by the soft plastic insert motor couplers so I changed to some disc type couplers that are true zero backlash (the machine originally had solid couplers). Unfortunately, now the knocking is terrible, it seems the couplers were helping to damp it. I spent a ton of time trying to diagnose it and finally think the issue is caused by the Oak controller! Hear me out...

I will preface this by saying I have not proved for sure the Oak is the problem as that would require scoping the signals to the drive. I will do it if I have to. I am basing my assumption that the Oak is the issue as the problem has happened on several motors and drives. I am basing my findings on the signals that the Delta ASD drives see from the oak.

So, the issue is that sometimes, almost randomly, the movement command to the drive is not consistent. It has some stalls in it. These cause the drive to accelerate/decelerate in an oscillation that vibrates right through the machine as a "bang". Refer to the picture below:

The white line is the drives command signal when I issue a G01 X50 F1500 (starting from X0) (metric). It is zoomed in to show the moment movement starts. You can see the white line has two flat spots in it. Where commanded position stops changing a couple of times. You can see the motor RPM (orange line). The motor accelerates then decelerates, accelerates, decelerates. In this capture there are three flat spots in the command. These flat spots are only 15ms apart. it is these acceleration cycles that are rattling the machine to death.

For comparison, here is an example when the machine doesn't bang and moves quietly. When the stepping is not in the position command signal, movement is great.

You can see the motor accelerate smoothly up to a steady 150 RPM.

Here are some more findings:
- Happens on all three axes, probably one out of every four MDI commands, kind of random.
- Initially I though it was related to backlash comp, but happens if you keep moving in the same direction, G01 X50 F1500; X100; X150 (bang), and sometimes when changing direction.
- The difference in number of steps at flat spot from resting position is not consistent, and not the same as the backlash value. This setup has 30000 steps per rev.
- Some times get one flat spot, two or three.
- Tested with backlash off (in motor settings) and with backlash acceleration at 0.001 (Param 208-215)
- Fresh report file, Delta drive configuration file and more pics attached

Here is another example from the Z axis. I can hear the brake inside the motor rattle when this happens!

In this example the flat spot happens after 978 steps movement (0.326 axis movement). Backlash comp is 0.0025mm (7.5 steps).

[ashesman]

I did another test tonight. I changed the pulse/rev numbers to match Centroids recommendation for Delta ASD drives. So changed pulse count to 32768 in Oak and 8192 in drives. Recycled power to everything, homed and checked movement was correct distance after change. Unfortunately it made no difference, still has steps in the pulse output.

As far as I can tell, many people use Delta ASDA2 drives. So, this cant be a new problem. I'm only asking it to move at 1500mm/min (60 inch/min).

I am going to need some help with this one, not sure I know where to go next. I cant keep running the machine like this!

Picture of movement at 32768 steps/rev:

[centroid467]

Hello Ashley,

I will be looking into this for you. I am not sure what could be causing this so I will need to do some digging and asking around.
Would you say that the average duration of one of the flat spots is around 10 ms? Do they only happen under acceleration/deceleration or do they also happen during constant velocity motion?

[ashesman]

Thanks for looking into this issue. I have only seen this happen on acceleration. But I havent tried to capture it on deceleration. I dont believe it happens at constant speed. The machine moves smoothly at rapid of 15000 to 24000 mm/min. It seems to only be at cutting speeds I see this problem. 1500 mm/min is what I have used to reproduce this issue.

Not sure the exact width of the flat spots. You can get an idea from the time scale on the traces. Probably less than 5ms I would have said.

[centroid467]

Alright, I am going to try something here with similar settings to what you have and see if I can get any signs of similar behavior.

In the meantime, could you please run a PID gather on CNC12 with your current settings to replicate the banging and post the resulting scope-gather-xx.txt file? You can find that utility in F1 Setup -> F3 Config -> F4 PID -> F1 PID Config.

Also, have you tried increasing the axis acceleration time? 0.2 is pretty fast for a mill axis. I'd suggest testing the defaults of 0.5 and see if it makes any difference.

[ashesman]

I will do the PID test tomorrow night and send results.

As for accel times, I want it to go faster not slower! Since the Centrod (and Delta motors/drive) conversion, I have never been able to get the machine to move like it did on the old Mazak controller. Mainly due to banging. It used to rocket along. See this video.

[centroid467]

I understand the desire for more acceleration. I'm mainly asking you to try slowing it down to see if it reduces the banging. Might provide a hint as to what is causing those flat spots.

[ashesman]

OK, so I did another round of testing tonight. I am confident now the issue is caused by the Oak. Actually I have two issues now!

I tested with acceleration at 0.5s, 1.0s and 0.2s (previously 0.35s). In all cases the same stepping flat spots happened. I have traces and screen shots if required.

I tried to use the PID tuning screen to create a log of the problem, but it wouldn't knock! After some going round in circles, I noticed that the PID tuning test file had a "G64 off" in it. And after running the file I was not able to make it knock ever again. After shut down and reboot the problem was back. Anyways to cut a long story short, the stepping pulse output does not have flat spots after G64 off, and does after G64 on. Completely reproducible.

So, you might say just leave smoothing off then. Well, that brings me to my second problem. I was surprised to find that with smoothing off, movement is much quieter and nicer (MDI commanded movements at cutting feed rates). When smoothing is on, the start end end of motion is harsher, the opposite to what you would expect. It is definitely noticeable.

The reason smoothing is turned on is because without it this machine knocks a lot while doing cutting movements. Smoothing really calms it down, not completely but I cant really run without it. You should be able to bring a tool down in Z, then go across in X (square corner), but if I don't have a radius in there, thunk! But if I increase feed rate, it goes down and across quietly.

I fluffed around for hours with little test programs and smoothing on/off. Basically it seems to come down to slow speed movements thump when they start/stop (say < 4000mm/min), where fast movements (say > 5000mm/min) are mint. I ran my test program that did noting but different distance square steps between various axis at different feed rates. At anything above cutting speeds, the machine moves so quietly and smoothly, no knocks and bangs. 5000 to 24000 mm/min (200 to 950 inch/min) is so nice. But down around the 1000 to 2500 mm/min (40-100 inch/min) every start/stop movement is harsh.

I ran a gcode file with real world profiling operations. Without smoothing on it was harsh, much better with smoothing on. I would rather not have smoothing on as it ultimately loses precision I guess.

If I could find a way that those cutting feedrate movements weren't so harsh, I probably wouldn't need smoothing on. I tried mucking around with the deadstart and delta values, but didn't really notice any difference. Of course slowing the acceleration right down smooths things out, but why should I have to! It can do fast movements nicely at higher accelerations, why not slow movements? Also, slow accelerations are hard on tools.

So, I am still quite worried about the stepping flat spots. But also after advice on how to tune this machine for better low speed performance.

Two other issues I noted tonight:
- On the bar of enabled G codes, G64 dos not show as on/off. But smoothing is definitely turning on/off. See attached picture.
- Two finger "pinch" zoom on the 3d view is backwards. Pinching your fingers together should zoom out, but instead zooms in!

[ashesman]

Just to add a bit more to this. To convince myself I wasn't dreaming, I looked at the acceleration curves when smoothing is on and off. Having smoothing on definitely reduces the acceleration time and sharpens the rate of motor speed increase from stationary.

Both pictures are the exact same movement. From X0, G01 X50 F1500. The first picture is smoothing off (G64 off). You can see the smooth transition of the white (commanded position) line. And gradual increase in motor RPM (orange line - goes from 0 to -ve RPM). It takes 70ms for the motor to reach full -150 RPM).

When smoothing is turned on (G64 on), the sharpness of the transition from stationary to moving increases significantly. You can also see how much faster the motor accelerates initially. Which is why the machine seems more harsh. it takes 50ms for the motor to reach full -150 RPM.

Here is a close up picture of the stepping. Time axis is 5ms per division. Sometimes the steps aren't completely flat as per this example.

[Houseman303]

It is normal that if you let the machine run completely stopped with G61, it will be very jerky and it will "bang". You can only work with high dynamics with G64. However, you can set the desired precision. I had the problem with my Oak that if the line for the plugged-in 3D probe (wireless) was active with G64, the Gcode would be very jerky to process, despite G64 being on. I now use a switch to disconnect the plugged-in line for the 3D probe and it runs without jerking or banging.