Linear scale correction: homing to index
15 May 2014 02:26 #46938
by DaBit
Well, that's a tricky one. If you have three levels of logic and you addf them in the wrong order you suddenly have a 3-cycle latency. That could cause some surprises...
Anyway, I did a quick test:
- I removed all the offset stuff I still had in the HAL file and which is not necessary anymore since I hardwired the scale encoder index to both 7i77 encoder inputs. No difference regarding following errors
- I added the index-enable pins of the PID controllers to the xindex-enable net. No difference.
- Tried to catch the f-error using HALscope. I managed to catch it. It turns out that the f-error happens only during the first homing cycle after LinuxCNC started and when there is enough distance between the startup position and homeswitch/indexpulse. And not always, which is the weird part.
The large ferror happens for only one servo period. Could it be that the hm2 encoder position is not updated every servo cycle or is just a cycle too late depending on some internal counter when LinuxCNC starts?
- Next, I added a mux2 to mux the motor-pos-fb between motor-pos-cmd and the scale encoder depending on the index-enable. So only during the distance between home switch and index pulse I lie about the actual position.
- I had to add a single cycle of extra delay before switching back to the scale position as motor-pos-fb
This seems to work
The piece of HAL code that does the trick:
Replied by DaBit on topic Linear scale correction: homing to index
The functions run strictly in the sequence in which they are addf-ed.
Well, that's a tricky one. If you have three levels of logic and you addf them in the wrong order you suddenly have a 3-cycle latency. That could cause some surprises...
Anyway, I did a quick test:
- I removed all the offset stuff I still had in the HAL file and which is not necessary anymore since I hardwired the scale encoder index to both 7i77 encoder inputs. No difference regarding following errors
- I added the index-enable pins of the PID controllers to the xindex-enable net. No difference.
- Tried to catch the f-error using HALscope. I managed to catch it. It turns out that the f-error happens only during the first homing cycle after LinuxCNC started and when there is enough distance between the startup position and homeswitch/indexpulse. And not always, which is the weird part.
The large ferror happens for only one servo period. Could it be that the hm2 encoder position is not updated every servo cycle or is just a cycle too late depending on some internal counter when LinuxCNC starts?
- Next, I added a mux2 to mux the motor-pos-fb between motor-pos-cmd and the scale encoder depending on the index-enable. So only during the distance between home switch and index pulse I lie about the actual position.
- I had to add a single cycle of extra delay before switching back to the scale position as motor-pos-fb
This seems to work
The piece of HAL code that does the trick:
# standard components
loadrt debounce cfg=1
..
..
# index homing, position feedback and ferror suppression
net xindex_enable hm2_[HOSTMOT2](BOARD).0.encoder.03.index-enable <=> hm2_[HOSTMOT2](BOARD).0.encoder.00.index-enable <=> axis.0.index-enable <=> xservo_pid.index-enable <=> xscale_pid.index-enable => debounce.0.0.in
net x-debounced-indexenable <= debounce.0.0.out => xposfb_mux2.sel
net xscalepos <= hm2_[HOSTMOT2](BOARD).0.encoder.03.position => xscale_pid.feedback => xposfb_mux2.in0
net xpos-fb <= xposfb_mux2.out => axis.0.motor-pos-fb
net xpos-cmd axis.0.motor-pos-cmd => xscale_pid.command => xposfb_mux2.in1
..
..Please Log in or Create an account to join the conversation.
15 May 2014 03:30 #46940
by DaBit
Replied by DaBit on topic Linear scale correction: homing to index
I was happy too early...
I will try without error-previous-target set, but the problem is not in the PIDs. Something weird is happening:
encoder03 is the scale which is muxed with axis.0.motor-pos-cmd to axis.0.motor-pos-fb based on debounce.0.0.out
On the falling edge of the debounce output, when the 'switch' to motor-pos-fb is flipped from motor-pos-cmd to encoder.03.position the motor-pos-cmd changes and I get a single cycle large ferror. No matter how long I delay the index-enable to mux signal.
I will try without error-previous-target set, but the problem is not in the PIDs. Something weird is happening:
encoder03 is the scale which is muxed with axis.0.motor-pos-cmd to axis.0.motor-pos-fb based on debounce.0.0.out
On the falling edge of the debounce output, when the 'switch' to motor-pos-fb is flipped from motor-pos-cmd to encoder.03.position the motor-pos-cmd changes and I get a single cycle large ferror. No matter how long I delay the index-enable to mux signal.
Please Log in or Create an account to join the conversation.
Time to create page: 0.088 seconds