Brother TC-225 / TC-229 adventure!

soooo, i guess noone found any glaring errors in my hal and ini files?

I haven't had time to do any further tuning on my axis. Your P value seems high and there is no I value. My machine is metric (should be 1/25.4 of your values), but anything more than 1.5 for P in X,Y gave resonances in my case.

OK, after one evening of tuning the X axis I think it's the servo driver that needs additional tuning. No matter how I tune the position loop, there are always oscillations which I am unable to get rid of. Probably the jerk limit in the original controller allowed a stiffer tuning of the velocity loop.

Its possible but if you have oscillations even at low position loop gains, that does suggest the the drives velocity loop is not stable, regardless of any jerk limiting the velocity loop should be stable because even if the trajectory contains no high frequencies that excite oscillations, cutting forces (disturb forces) can have high frequencies

sounds like mine. sigh.

pcw, remember that these drives in their current tuning state are running 100% perfectly under the stock control, at the same speed and acceleration.

I also did try tweaking the gain on one axis and it had minimal effect. the only time my machine behaves smooth is in a g64 corner, which is effectively jerk control on each axis as it rounds the corner - even if the rounding is 0.005" at 787ipm 2.5G. so I would not be shocked if jerk control is a major factor here - though I cant believe linuxcnc simply cant handle it correctly.


that said - could someone with a velocity controlled drive with high performance (500+ipm, 0.25G) show me their results? both the scope and the machine moving. cause, to be completely honest I have never actually seen a properly tuned velocity drive at high speed and acceleration. ive seen some low speed ones. and low acceleration, and some torque controlled ones at what seem to be moderate performance, but not one single example similar to my machine.

the only high performance linuxcnc machines I have ever seen are step and direction..

if we have to just say linuxcnc cant handle velocity drives well enough, that's at least a resolution. this wild goose chase for some hardware issue that seems to not actually exist is wearing me down.

@pcw sure, jerk is not the only source of high frequencies the velocity loop gets to see.
I was a little imprecise with my description: I am seeing ripples and overshoot. It's not unstable. It's just, that the velocity ramp has some pronounced frequency. My guess is, that this frequency component is weaker in the jerk limited control and hence more tolerable.
I will do some screenshots next time I'm in the workshop. I think it's possible to solve this problem with tuning the servo.

Which brings me to the next problem: I don't have a manual for the Toshiba RA Drive.
There are 6 potentiometers on the front. I, SZ, P, SPD and WZ, UZ
I and P are clear,
SPD = speed?
SZ = ?
WZ = ?
UZ = ?

@ihavenofish
Do you know the stock values of the acceleration? Were does this information of jerk limit in the brother control come from? Could the jerk control be an analog low pass filter?.

Another reason why I suspect the velocity loop is: Depending on the tuning I can have ripple in ferror which stays positive during the whole ramp. So there is no sign reverse in the position loop which could cause the oscillation.

ihavenofish wrote: that said - could someone with a velocity controlled drive with high performance (500+ipm, 0.25G) show me their results? both the scope and the machine moving. cause, to be completely honest I have never actually seen a properly tuned velocity drive at high speed and acceleration. ive seen some low speed ones. and low acceleration, and some torque controlled ones at what seem to be moderate performance, but not one single example similar to my machine.
the only high performance linuxcnc machines I have ever seen are step and direction..
if we have to just say linuxcnc cant handle velocity drives well enough, that's at least a resolution. this wild goose chase for some hardware issue that seems to not actually exist is wearing me down.

velocity mode
ferror 0.05mm, no ripple nor overshoot
g0 = 72m/min = 2800ipm

MAX_VELOCITY = 1200
MAX_ACCELERATION = 3200.0
P = 51
I = 0
D = 0
FF0 = 0
FF1 = 0.97
FF2 = 0.0013
BIAS = 0
DEADBAND = 0
MAX_OUTPUT = 1200

well that works
thanks!

you don't happen to have a hal scope pic while its doing that?

your f error is high - is that error during acceleration, or through the whole move? mine is about 0.025mm +- spike during acceleration, then only about 0.005mm or so during the steady motion.

only other thing I notice is you have deadband at 0. I cant remember what happened when I tried 0, or if I tried 0 at all.

also what drives and motors and encoder counts are on that machine?

Depending on your Gcode you may always have to have G64 PN QM in force to do smooth motion.
Are these set to reasonable values?

Also its easy to differentiate between tuning and jerk limiting issues with LinuxCNC by doing a motion that is inherently jerk limited (like a G2/G3 helical move) Once up to speed, commanded motion will be jerk limited