Joint following error
06 Apr 2011 08:36 #8631
by wwr
Replied by wwr on topic Re:Joint following error
I've put the old harddrive back about an hour ago. The machine is still running without any errors. I started a test ngc file and let it run for 15 minutes.
It is definitely a software issue. I'll try the latest version now.
It is definitely a software issue. I'll try the latest version now.
Please Log in or Create an account to join the conversation.
06 Apr 2011 09:31 #8633
by wwr
Replied by wwr on topic Re:Joint following error
The update didn't solve it, however now it took 12 minutes to produce the error.
Please Log in or Create an account to join the conversation.
06 Apr 2011 10:43 #8635
by andypugh
Replied by andypugh on topic Re:Joint following error
This is pretty weird.
Quadrature encoders should be very resilient to noise in the wires (as unless the noise is exactly right in both lines, it just causes a count-up followed by a count-down.)
How many encoder counts does the sudden offset correspond to? (encoder scale x axis offfset).
Quadrature encoders should be very resilient to noise in the wires (as unless the noise is exactly right in both lines, it just causes a count-up followed by a count-down.)
How many encoder counts does the sudden offset correspond to? (encoder scale x axis offfset).
Please Log in or Create an account to join the conversation.
06 Apr 2011 11:05 #8636
by wwr
Replied by wwr on topic Re:Joint following error
The scale is 2000 and the coords jump about 300000 mms so it is 600M in encoder counts.
This is impossible to be electrical noise.
This is impossible to be electrical noise.
Please Log in or Create an account to join the conversation.
06 Apr 2011 12:49 - 06 Apr 2011 12:57 #8643
by andypugh
Replied by andypugh on topic Re:Joint following error
Yes, definitely not noise I would say.
Also, not an obvious counter-wrap (65535 counts, 32768, 2147483647....)
I wonder if there is a glitch in your motenc firmware?
The history of the emc2 motenc driver is here:
git.linuxcnc.org/gitweb?p=emc2.git;a=his...deb3db5a5a368f9d82a1
Nothing I have looked at so far looks like a likely culprit. (do you know the release date of your working emc2 version?)
Also, not an obvious counter-wrap (65535 counts, 32768, 2147483647....)
I wonder if there is a glitch in your motenc firmware?
The history of the emc2 motenc driver is here:
git.linuxcnc.org/gitweb?p=emc2.git;a=his...deb3db5a5a368f9d82a1
Nothing I have looked at so far looks like a likely culprit. (do you know the release date of your working emc2 version?)
Last edit: 06 Apr 2011 12:57 by andypugh.
Please Log in or Create an account to join the conversation.
07 Apr 2011 13:24 #8676
by wwr
Replied by wwr on topic Re:Joint following error
Sorry but i don't remember the release date.
I checked the motenc driver's history and i don't see anithing suspicious.
I noticed that the rate of the sudden offset is always different. From 60 to 300000.
I checked the motenc driver's history and i don't see anithing suspicious.
I noticed that the rate of the sudden offset is always different. From 60 to 300000.
Please Log in or Create an account to join the conversation.
07 Apr 2011 13:49 #8678
by andypugh
Replied by andypugh on topic Re:Joint following error
Can you set up a halscope in single shot, triggered off of the Motenc encoder output value (as "close" as possible to the driver). It might be instructive to see if the value jumps to the new value, or ramps there.
Actually, to rewind a bit, can you trace back the offset in Machine->show-hal-config? You should be able to go back up the HAL chain from axis.N.position and see where the offset appears. I have been assuming that it is in the
motenc.<board>.enc-<channel>-count and motenc.<board>.enc-<channel>-position values. But we have no evidence of that yet.
Actually, to rewind a bit, can you trace back the offset in Machine->show-hal-config? You should be able to go back up the HAL chain from axis.N.position and see where the offset appears. I have been assuming that it is in the
motenc.<board>.enc-<channel>-count and motenc.<board>.enc-<channel>-position values. But we have no evidence of that yet.
Please Log in or Create an account to join the conversation.
18 Apr 2011 10:02 #9033
by wwr
Replied by wwr on topic Re:Joint following error
Sorry for the long delay.
I managed to catch the error with the halscope, although it was quite hard because the scope triggered off on any count.
You can see a screenshot here: sztaki.hu/~paniti/Screenshot-1.png
After all it does offset to a counter-wrap, but it wasn't obvious for me at first sight. Today it was 131072, 524288 and 32768.
I managed to catch the error with the halscope, although it was quite hard because the scope triggered off on any count.
You can see a screenshot here: sztaki.hu/~paniti/Screenshot-1.png
After all it does offset to a counter-wrap, but it wasn't obvious for me at first sight. Today it was 131072, 524288 and 32768.
Please Log in or Create an account to join the conversation.
18 Apr 2011 11:15 #9040
by andypugh
Replied by andypugh on topic Re:Joint following error
I am actually rather confused, as the wrap you see is a 16-bit wrap, and the motenc encoders (and the HAL driver) are all 32 bits wide.
If you re-plot that Halscope trace so that the step-change is visible (with the slider on the right) it probably ought to be submitted to the bug-tracker at sourceforge.net/tracker/?atid=106744&group_id=6744&func=browse
If the step is as instant as it looks on that trace then it has to be a software problem, but whether it is with emc2 or with motenc is not clear.
The HAL driver seems to export the encoder value read from the Motenc card pretty much unmodified, so I do actually suspect the Motenc. (Also, if they all did this, surely we would have heard by now?)
It might be worth contacting Motenc, the one time I emailed them they were pretty responsive.
If you re-plot that Halscope trace so that the step-change is visible (with the slider on the right) it probably ought to be submitted to the bug-tracker at sourceforge.net/tracker/?atid=106744&group_id=6744&func=browse
If the step is as instant as it looks on that trace then it has to be a software problem, but whether it is with emc2 or with motenc is not clear.
The HAL driver seems to export the encoder value read from the Motenc card pretty much unmodified, so I do actually suspect the Motenc. (Also, if they all did this, surely we would have heard by now?)
It might be worth contacting Motenc, the one time I emailed them they were pretty responsive.
Please Log in or Create an account to join the conversation.
Time to create page: 0.076 seconds