qtdragon - probing issues
13 May 2023 23:37 #271321
by Roguish
Replied by Roguish on topic qtdragon - probing issues
Chris,
I have seen one error.
when I go to Status to see what's going on, and click on Clear, it throws an error.
don't have an image. i just shut down........
I have seen one error.
when I go to Status to see what's going on, and click on Clear, it throws an error.
don't have an image. i just shut down........
Please Log in or Create an account to join the conversation.
14 May 2023 02:26 - 14 May 2023 02:37 #271326
by cmorley
Replied by cmorley on topic qtdragon - probing issues
ok that bug is fixed. didn't get to looking at the probe routines yet
Last edit: 14 May 2023 02:37 by cmorley.
Please Log in or Create an account to join the conversation.
14 May 2023 02:41 - 14 May 2023 02:44 #271329
by rodw
Replied by rodw on topic qtdragon - probing issues
I had an idea while reading this thread to speed up probing and to make it more efficient and accurate.
forum.linuxcnc.org/show-your-stuff/49100...routines-for-probing
Here all probing is done at F200mm and only one pass is done.
I then realized the maximum accuracy is determined by how far the probe can move in 1 servo thread cycle.
Rather than using a G1 move to back off after the initial touch, you could use a G38.4 move which would probably move no more than 0.2mm away from the material before the probe cleared. . Then a really slow probe move back to the material would enhance accuracy
I'm always looking for efficiency!
Anyway, shelve that idea for now until the bugs are fixed...
forum.linuxcnc.org/show-your-stuff/49100...routines-for-probing
Here all probing is done at F200mm and only one pass is done.
I then realized the maximum accuracy is determined by how far the probe can move in 1 servo thread cycle.
Rather than using a G1 move to back off after the initial touch, you could use a G38.4 move which would probably move no more than 0.2mm away from the material before the probe cleared. . Then a really slow probe move back to the material would enhance accuracy
I'm always looking for efficiency!
Anyway, shelve that idea for now until the bugs are fixed...
Last edit: 14 May 2023 02:44 by rodw.
The following user(s) said Thank You: Roguish
Please Log in or Create an account to join the conversation.
14 May 2023 06:22 #271332
by cmorley
check the length status after probing the boss - it didn't seem right. I don't think it compensated for the probe diameter properly.
Chris
Replied by cmorley on topic qtdragon - probing issues
Ok I think those are fixed now for bosses - yes backwards.Chris,
well, updated. testing
a couple of small things.
for the 'Outside Boss' (center pick on the Outside page) the X Neg and X Pos results are backwards, as are the Y Neg and Ypos.
I think the assignments in probe_routine.py (line 658, etc) are backwards. see the image attached.
also the move from the -x probe completion over to the +x probe is too big.
check the length status after probing the boss - it didn't seem right. I don't think it compensated for the probe diameter properly.
Chris
The following user(s) said Thank You: Roguish
Please Log in or Create an account to join the conversation.
14 May 2023 15:31 #271346
by Roguish
Replied by Roguish on topic qtdragon - probing issues
rodw,
Very interesting to look at the entire string of actions involved with probing, including all the times. all aimed of course toward accuracy.
I can't and don't do any super accurate work, like sub 0.0001 inch (0.00254 mm).
There is lots going on to consider. following error, accel and decel, program time, computer speed, servo speed (I use real servos, not steppers).
I finally got one of my probes centered to within 0.0002 in with very careful adjustment and a lot of luck. For some reason I started getting a few errors of probe tips during non-probing moves. I previously had a debouce in the hal circuit, but had commented it out. Well, I put it back in with a timing of 2 and the false trips dropped a lot. Just to see how the debounce time affected the routine, I increased the time to 5, and then to 50. The change to 5 made no real noticable difference in the probing results (I am testing on an internal pocket, physically it's the inside of a good condition bearing, 30mm ID, 62mm OD), but the 50 made a significant difference of like .005 inch in the diameter. Well, tells me that time does matter. I have a servo period of 1,000,000, and the default Trajectory period of 100,000.
So if one created a fully detailed timeline of all the stuff going on in the probing routine, I bet it'd be very enlightening. I'm not sure how to do that 'cause there is way too much going on inside the code for me to follow.
I'm trying to figure out why the diameters I measure are off by several thousands of an inch. I am confident the bearing is damn close to exact, and I've measured the probe tip diameters the best I can and they are definitely within a tenth or so (thousands of an inch). I use 5 and 2 inches per min. to probe, fast and slow, respectively. The axes following errors at those speeds are way small, about 0.0001 inches or less. So even adding up all these inaccuracies doesn't even come close to the approx 0.005 inch diameter error. I'm stumped. I did find that if I modify the probe diameter by like 0.004 I can get the correct results.
Anyway, this is all pretty interesting. And I'm confident that 'tween all of us working on it, it'll get much, much better. Thanks
Very interesting to look at the entire string of actions involved with probing, including all the times. all aimed of course toward accuracy.
I can't and don't do any super accurate work, like sub 0.0001 inch (0.00254 mm).
There is lots going on to consider. following error, accel and decel, program time, computer speed, servo speed (I use real servos, not steppers).
I finally got one of my probes centered to within 0.0002 in with very careful adjustment and a lot of luck. For some reason I started getting a few errors of probe tips during non-probing moves. I previously had a debouce in the hal circuit, but had commented it out. Well, I put it back in with a timing of 2 and the false trips dropped a lot. Just to see how the debounce time affected the routine, I increased the time to 5, and then to 50. The change to 5 made no real noticable difference in the probing results (I am testing on an internal pocket, physically it's the inside of a good condition bearing, 30mm ID, 62mm OD), but the 50 made a significant difference of like .005 inch in the diameter. Well, tells me that time does matter. I have a servo period of 1,000,000, and the default Trajectory period of 100,000.
So if one created a fully detailed timeline of all the stuff going on in the probing routine, I bet it'd be very enlightening. I'm not sure how to do that 'cause there is way too much going on inside the code for me to follow.
I'm trying to figure out why the diameters I measure are off by several thousands of an inch. I am confident the bearing is damn close to exact, and I've measured the probe tip diameters the best I can and they are definitely within a tenth or so (thousands of an inch). I use 5 and 2 inches per min. to probe, fast and slow, respectively. The axes following errors at those speeds are way small, about 0.0001 inches or less. So even adding up all these inaccuracies doesn't even come close to the approx 0.005 inch diameter error. I'm stumped. I did find that if I modify the probe diameter by like 0.004 I can get the correct results.
Anyway, this is all pretty interesting. And I'm confident that 'tween all of us working on it, it'll get much, much better. Thanks
The following user(s) said Thank You: rodw
Please Log in or Create an account to join the conversation.
15 May 2023 00:35 #271380
by cmorley
Replied by cmorley on topic qtdragon - probing issues
I would think that the physical latch distance must be accounted for.
When the probe touches the material, it doesn't trip right away.
You must move enough to break the contact.
I believe this is why you must calibrate probes, if you want to get high precision measurement.
For comparison measuring (ie find the center) it doesn't matter as much. Though asymmetrical trip points will come into affect.
When the probe touches the material, it doesn't trip right away.
You must move enough to break the contact.
I believe this is why you must calibrate probes, if you want to get high precision measurement.
For comparison measuring (ie find the center) it doesn't matter as much. Though asymmetrical trip points will come into affect.
The following user(s) said Thank You: Roguish
Please Log in or Create an account to join the conversation.
15 May 2023 13:36 #271399
by andypugh
It certainly claims to restore on exit. Whether it does on error exit is something I haven't tried experimenting with.
Yes, and going back to the previous units is better than what I implemented in the probescreen_ng code.
Replied by andypugh on topic qtdragon - probing issues
There is m70/m72 that can do the recording/restoring, not sure how it works with errors.
It certainly claims to restore on exit. Whether it does on error exit is something I haven't tried experimenting with.
Does this seem right/reasonable to you?
Yes, and going back to the previous units is better than what I implemented in the probescreen_ng code.
Please Log in or Create an account to join the conversation.
15 May 2023 14:36 #271403
by Roguish
Replied by Roguish on topic qtdragon - probing issues
By physical latch distance, are you referring to the small distance the probe internal mechanics must move to make/brake the contact? sounds right.
well, I have 3 different probes. the one I referred to above is a Tormach brand a couple years old (no longer sold). I have one from Drewtronics, a small US manufacture, and an inexpensive Chinese one.
I'll try each and watch the results very carefully.
As for calibration, qtdragon VersaProbe does not have a routine for that. Probe Basic does, but I'm not sure how it operates, or if it is correct or not. I have done some searching for calibration routines and have found a few. So far they are primarily accounting for off-centerness.
Wish I had one of those super good Renishaws.
The mechanical latch distance (and associated time) goes into what Rodw was speaking about.
And it's good to see AndyP joining in.
Let's keep at it.
well, I have 3 different probes. the one I referred to above is a Tormach brand a couple years old (no longer sold). I have one from Drewtronics, a small US manufacture, and an inexpensive Chinese one.
I'll try each and watch the results very carefully.
As for calibration, qtdragon VersaProbe does not have a routine for that. Probe Basic does, but I'm not sure how it operates, or if it is correct or not. I have done some searching for calibration routines and have found a few. So far they are primarily accounting for off-centerness.
Wish I had one of those super good Renishaws.
The mechanical latch distance (and associated time) goes into what Rodw was speaking about.
And it's good to see AndyP joining in.
Let's keep at it.
Please Log in or Create an account to join the conversation.
15 May 2023 15:16 #271405
by andypugh
Replied by andypugh on topic qtdragon - probing issues
I think that calibration is generally looking at the lobularity of the detecting circle, as the three-bar contacts mean that the probes trigger fractionally earlier in some directions that others.
When I made my probes I planned to write a calibration scheme that used the spindle encoder. (as my tooling only fits into the spindle in a fixed orientation). The idea was to probe a test ring, then turn the spindle approximately 90 degrees and probe again. From the exact angle turned and the difference in X and Y centres of the two measurments you can derive a correction fr any spindle angle (as read by the encoder). .
This might be a bit specific to my machine, as there is an absolute spindle encoder and tooling that only fits in one orientation.
When I made my probes I planned to write a calibration scheme that used the spindle encoder. (as my tooling only fits into the spindle in a fixed orientation). The idea was to probe a test ring, then turn the spindle approximately 90 degrees and probe again. From the exact angle turned and the difference in X and Y centres of the two measurments you can derive a correction fr any spindle angle (as read by the encoder). .
This might be a bit specific to my machine, as there is an absolute spindle encoder and tooling that only fits in one orientation.
The following user(s) said Thank You: Roguish
Please Log in or Create an account to join the conversation.
15 May 2023 19:09 #271420
by Roguish
Replied by Roguish on topic qtdragon - probing issues
well, interesting word that is. but you're definitely correct. since the vast majority of probes are based on a 3 legged stool concept, there is most certainly a small difference in activation.
I try to install my probes into the spindle in the same direction such as the connecting cable always out the left side (-y direction). if I probes diameters only in the 3 contact directions, it might help. but then what about corner probing.... or arbitrary directions like with your perimeter probing routine...... oh well.
just one more nuance to add to the bag.
I try to install my probes into the spindle in the same direction such as the connecting cable always out the left side (-y direction). if I probes diameters only in the 3 contact directions, it might help. but then what about corner probing.... or arbitrary directions like with your perimeter probing routine...... oh well.
just one more nuance to add to the bag.
Please Log in or Create an account to join the conversation.
Moderators: cmorley
Time to create page: 0.108 seconds