Retrofit Aciera F35 CNC, Axis don't stand still

Hello everybody. I’m new here.

I need help with basic setup.

At first a bit of introducing of my project.
I have bought at the beginning of the year an old Aciera F35 CNC to make it work with LinuxCNC.
I'm be a good car mechanic but I have no experience with CNC.
Before I started I was thinking it would be possible for me to set it up, but now it shows me my limits.

Over the last months I setup the Hardware stuff.
I have install 3 Granite Devices Argon drives with new Rotary Encoder (5000) on the Servos.
The original glass scales are connected over a SIN/COS signal converter to my Mesa 7i95. The goal is to have a dual closed loop system. After some small wiring correction it looks like everything is right.

The Argon Drives are tested and tuned like it is written in the manual. I can move the axis without problems.

Over the last weeks I tried hard to get LinuxCNC work correct. I read a lot of posts and of course the LinuxCNC manual. Also have looked into some YouTube videos (thank Talla83)
to understand the hal and ini files.

My Problem:
The axis is swings back and forth. Mostly tested with the x-axis. HAL and INI is may not right for the other axis.
As I started with tuning the axis the swinging distance is rising more and more every time it change the direction till the follow error was too big.
I spend a lot of time in the PID controller tuning. My temporarily values looks like they are very high. I need lot of I and D for stop the swing up.
Now the follow error is a much smaller but still no standstill after the jog.

Please, can someone give me some advices?
I have no more idea how I can fix my problem.

Thanks in advance
Reto

Thanks tommylight.

I think the Argon Drives are well tuned. I double checked it.
At first I use the torque mode to tune and after that the velocity mode tune. I have followed the instructions from Granite Devices and EUSurplus
I don't go truth the position mode tune because I think I want go with the velocity control mode.
Right decision?

You can't know it, but my Servos are brushed DC. They are original fitted on that machine (BBC/CEM FDE T4C4B3 200V/12.8A/~2.5kw).
So no pole count necessary.
Or am I wrong?

I can do any moves out of the Granity Software. In Velocity mode and also in position mode. The axis moves nice and stops at the exact point. No overshot or swinging.

So I think my problem is in LinuxCNC.

I use step/dir from the Mesa 7i95 to the Argon Drives.
How can I do a correct setup of the step/dir time parameters in the ini?

FF1 should be exactly 1.000 if the encoder and step /dir scales are correct
You should not need any I or D for initial tuning

Hi,
your PID and F Parameter in ini are real strange. To compare, i use Delta B2 Servodrive on Z-Axis and tune from linuxcnc near zero. Because the Delta Servodrive Software are much better in tuning and adjust some more parameter. I think in your case same situation.
Second looks your encoder section in hal also very strange. X and Y axis are confused for me. I add my hal encodersection to.
Try this Paramter for your card and it should be working.

My ini

[JOINT_3]
TYPE = LINEAR
HOME = 0.0
FERROR = 10.0
MIN_FERROR = 1.0
MAX_VELOCITY = 200.0
MAX_ACCELERATION = 400
# The values below should be 25% larger than MAX_VELOCITY and MAX_ACCELERATION
# If using BACKLASH compensation STEPGEN_MAXACCEL should be 100% larger.
STEPGEN_MAXVEL = 230
STEPGEN_MAXACCEL = 500
P = 5
I = 0
D = 0
FF0 = 0
FF1 = 1
FF2 = 0
BIAS = 0
DEADBAND = 0
MAX_OUTPUT = 0
ENCODER_SCALE = -400
# these are in nanoseconds
DIRSETUP = 1000
DIRHOLD = 1000
STEPLEN = 1000
STEPSPACE = 1000
STEP_SCALE = 800
MIN_LIMIT = -400.0
MAX_LIMIT = 1.0
HOME_OFFSET = 0.000000
HOME_SEARCH_VEL = 5.000000
HOME_LATCH_VEL = 1.00000
HOME_FINAL_VEL = 0.000000
HOME_USE_INDEX = NO
HOME_SEQUENCE = -1


My hal
# ---Encoder feedback signals/setup---

setp hm2_7i76e.0.encoder.00.counter-mode 0
setp hm2_7i76e.0.encoder.00.filter 1
setp hm2_7i76e.0.encoder.00.index-invert 0
setp hm2_7i76e.0.encoder.00.index-mask 0
setp hm2_7i76e.0.encoder.00.index-mask-invert 0
setp hm2_7i76e.0.encoder.00.scale [JOINT_3]ENCODER_SCALE

net z-pos-fb <= hm2_7i76e.0.encoder.00.position
net z-vel-fb <= hm2_7i76e.0.encoder.00.velocity
net z-pos-fb => joint.3.motor-pos-fb
net z-index-enable joint.3.index-enable <=> hm2_7i76e.0.encoder.00.index-enable
net z-pos-rawcounts <= hm2_7i76e.0.encoder.00.rawcounts

A big step forward, thank you tommylight

I tested only for short but it looks promising.
My workshop is a bit cold at the moment.

I opened the loop with changed the position feedback in hal to stepgen.00.feedback-fb. Also changed the step_scale to -4000 to have it in the right scale. It is now possible to make jog without any problem. The positioning at the end of the jog is good and no swinging. Naturally I need to retune the Argon and LinuCNC. The servo does not sound smooth.

I will trie to reconnect the glass scales later.

Did I need the fb/cmd_deriv signal in my hal?



Do I understand it right, step/dir is only for position control (stepper system) and the velocity control is only with analog signal useable? Is it a advantage for me to have it at the end in velocity mode? Should I change to a analog card like Mesa 7i77?
I have reading that velocity control is the best way to do a dual closed loop system.

Thanks for the additional information about the Argon Drives.
At stationary the current is around 0.5A.

Thanks PCW for the notice.

Thanks chris@cnc. I will compare it.

Also, if you look closely at your plot, there is almost a 100 millisecond
delay from the commanded motion to the encoder feedback, indicating
actual motion. I suspect that this is some delay or filtering enabled
in the drive setup. You will not be successful with tuning if the drive
does not respond _much_ faster that this to external commands.

Now I made my first chips
Thanks a lot, guys!
It was a tuning problem.
Now it is mostly fine. I needed to add some more deadband like 0.001mm and more. It is, I think, needed by the 0.005 mm resolution of the glass scales is not fine enough. Specially the Y-Axis moves in a micron-range at still.
I added my last hal and ini.