Flexicam ATC router retrofit with Kflop/Kanalog - Page 3


Page 3 of 3 FirstFirst 123
Results 25 to 34 of 34

Thread: Flexicam ATC router retrofit with Kflop/Kanalog

  1. #25
    Gold Member TomKerekes's Avatar
    Join Date
    May 2006
    Location
    USA
    Posts
    2998
    Downloads
    0
    Uploads
    0

    Default Re: Flexicam ATC router retrofit with Kflop/Kanalog

    Hi Eug,

    Making progress. But you still seem to have huge error (2000 counts) and may be missing some concepts.

    Thanks for including the data. Here is a plot of Error: Error is what you should really be plotting.

    Flexicam ATC router retrofit with Kflop/Kanalog-poserror-png

    Note that the Error is like a mirror image of Output. That is because your P gain is really the only thing doing anything. So Output is simply Proportional to Error.

    The Integrator is not really doing anything because the gain of 1e-7 is so small and it is still limited to 200 DAC counts. It seems you might not have read the link provided in my last email.

    In the Video you seem to be zooming in and focusing on any tiny (~2 counts) error after the move finishes. More importantly you should focus on errors (~2000 counts) while the machine is moving (assuming you want accuracy while moving and not just when moving point-to-point like drilling holes or something).

    Something else you might be confusing is that as you increase the I gain, and you observe some overshoot after the move finishes, you consider the increase in I gain to be bad. But this is the nature of how I gain works - which is to force the average error to be zero. Because your system has thousands of counts of undershoot during the move, there needs to be some overshoot to balance out the errors. Because the I gain is so small it causes the Integrator to react so slowly that it does almost nothing to reduce the lag/undershoot during the move, but still observes it so then overshoots by a tiny amount for a very long time (minutes?). The goal should be to get the Integrator to make corrections as fast as possible. If you are lucky it will make multiple overshoot/undershoot corrections throughout the move and after the move completes there will not be a huge undershoot to need to be corrected. Hope this makes sense.

    Regards



  2. #26
    Registered xyz-cnc's Avatar
    Join Date
    Jun 2011
    Location
    Australia
    Posts
    37
    Downloads
    0
    Uploads
    0

    Default Re: Flexicam ATC router retrofit with Kflop/Kanalog

    Hi Tom

    Apologies for my frustrations. I guess the purpose of the video was to show you that I'm not sure how I can increase the P & D gains much higher than what they are now. Unfortunately I didn't make that clear, and confused things by zooming in on the small error at the end of the move. My thinking was that if these are rapid moves than yes, I have a large error while moving, but at least I arrive at my commanded destination.

    But as you can see and more importantly hear, (if not, please turn your volume up), when I increased P to 0.8 (at 0.26secs into the vid), I got severe instability. So I dropped it back down to 0.7 and moved on to D. When D got to 3, (at 1.05secs into the vid) again, severe instability. And increasing I wasn't helping at all....

    So at P 0.7 and D at 2, my error was about ~2000 counts which is huge, but at least the axis moved smoothly. In hindsight, I think my mistake was thinking I needed to increase P. I realise all motors/amplifiers behave differently, but what I saw and read in the posts that you linked to (e.g in post #9 above, you refer to rbickles tuned plots which have P at 8 and D at 100 on the X axis). So I assumed that mine were much too low and needed to be increased to reduce the error.

    I went back and forth, a little more P (going up in .05 steps), a bit more D, (yes, I read to increase D by x2 or x3 times, got it to 12) a fraction of I and the vibrations were so bad I though pieces of the machine would fall off (unlikely as the thinnest part is the cable chain support at 6mm thick) but sounded really bad - surely can't be good for it). And even if I backed them off a touch I could do a 5" move, but a 10" move would cause instability. I tried so many combinations but as per my post above, the only numbers that would allow smooth motion over any commanded distance were P 0.5, I 0.000001 and D 0.25 (with a large error). It would be an understatement to say I was frustrated, spending a lot of time and making little progress.....

    Oh and things were no different with lower velocity and acceleration, in fact they were worse. I looked up the specs for this machine and the current model says speeds up to 60m/min (about 2300IPM) and I'm nowhere near there. I calculated top V was 500000 (37.6m/min or 1483ipm) and have been doing the tests at V 400000. With J at 20000000 it is very smooth.

    Anyway, today I took a different approach. I started with my stable combination and made the decision to leave P at 0.5 and only change D and I. This proved successful, and I managed to get the error quite low with D and I. I tried increasing P throughout, but as I suspected, it would only make things worse. After some time, I ended up with what I think are reasonable results for Y axis - smooth motion over any commanded distance with a relatively small error. I then moved to Z, where I was able to get P a touch higher to 0.65. Z took more "work" but I think it also turned out OK with a relatively small error.

    I did this on both axis at a relatively low V of about 100000 and A of 1000000. I think the Jerk was x100 higher. I think these were the screenshots below. But interestingly enough (for me) increasing V and A to 2, 3, 4 and even 5 times seemed to have little effect on the error. It may have increased a little but that is irrelevant as I wouldn't be cutting at those speeds. And I reduced the Max Following Error on both axis to 300 with no problems.

    Please don't tell me I'm still missing some concepts

    Attached Thumbnails Attached Thumbnails Flexicam ATC router retrofit with Kflop/Kanalog-y-axis-png   Flexicam ATC router retrofit with Kflop/Kanalog-y-axis-error-png   Flexicam ATC router retrofit with Kflop/Kanalog-z-axis-png   Flexicam ATC router retrofit with Kflop/Kanalog-z-axis-error-png  

    Attached Files Attached Files
    Last edited by xyz-cnc; 09-16-2018 at 11:08 AM.
    Eug


  3. #27
    Gold Member TomKerekes's Avatar
    Join Date
    May 2006
    Location
    USA
    Posts
    2998
    Downloads
    0
    Uploads
    0

    Default Re: Flexicam ATC router retrofit with Kflop/Kanalog

    HI Eug,

    Looking much better! 40X less error and 100000X more I Gain. I think you get everything

    Although looking at the error plots there is some oscillation so I suspect the systems are borderline unstable. Its hard to tell if it is servo instability or something mechanical like gear noise or such. If it something mechanical there is usually little that can be doe about it (except fix the mechanics). Sometime a clue will be that something mechanical will be related to speed and servo instability will be unrelated to speed. You might back off on all the gains something like 25% and see if the oscillations go away. Its up to you to decide how aggressive you wish to be to run right on the edge for max performance or to back off some for a good margin of stability and smoothness.

    Y axis seems to be ~ 40Hz

    Flexicam ATC router retrofit with Kflop/Kanalog-osc40hz-png

    I think the next step would be to include feedforward. Up to now, with only feedback, it is like driving a car while staring straight down at the center line of the road. When we come to a curve we don't do anything until after we start running off the road to realize we need to make a correction. Feed forward looks ahead, anticipates a needed correction, and applies the expected correction without waiting for or considering any error. Here is further description:

    Step Response Screen

    Here is the point in a Thread where PeterTheWolf performs his Feedforward adjustments:
    https://www.cnczone.com/forums/dynom...ml#post2132108

    I like his video called:
    Here is the Kmotion videos of the results with I-Gain I=0.0001 (ZIP file 42mb)

    Note as discussed in the Thread it is sometimes advantageous to lower PID feedback gains while tuning feedforward. This is sort of like closing your eyes while making a turn to learn how to make good turns open loop without any corrections. If you can do that well then making a turn with your eyes open should be very easy.

    HTH
    Regards



  4. #28
    Registered xyz-cnc's Avatar
    Join Date
    Jun 2011
    Location
    Australia
    Posts
    37
    Downloads
    0
    Uploads
    0

    Default Re: Flexicam ATC router retrofit with Kflop/Kanalog

    Hi Tom

    Feels good to be making progress.

    In regards to oscillation, I did as you said and reduced all gains by 25% but couldn't see any improvement, if anything the error became larger. I also increased V but couldn't see much change and oscillation remained at 40Hz. Realistically, this is a solid 2.5 ton machine. The linear bearing blocks are high preload and everything is made from heavy steel box sections. As far as I can tell, the only "adjustment" is the pinion mesh into the rack. This is not easy to access on the Y axis but I'll do it if I have to.

    Otherwise, moving on to Feed Forward proved successful. Only small amounts were required, but the improvement was impressive. I tried different size moves at different speeds and the errors remained quite low. However while adjusting the FF, I found that reducing I gain to 0 was detrimental and I had to increase the Max following error to be able to get any movement. So I put I back to 0.01 and went from there. Did I do this correctly?

    Following Peter's thread, it would appear the next step is to determine my max cutting speed and error size, and then come up with a cutting profile and a rapid profile.

    BTW, my servos are still a little noisy. Better than before, but can they ever be completely silent?

    Attached Thumbnails Attached Thumbnails Flexicam ATC router retrofit with Kflop/Kanalog-y-error-reduced-75-pid-png   Flexicam ATC router retrofit with Kflop/Kanalog-y-error-base-pid-png   Flexicam ATC router retrofit with Kflop/Kanalog-y-error-ff-png  
    Attached Files Attached Files
    Last edited by xyz-cnc; 09-16-2018 at 11:38 PM.
    Eug


  5. #29
    Gold Member TomKerekes's Avatar
    Join Date
    May 2006
    Location
    USA
    Posts
    2998
    Downloads
    0
    Uploads
    0

    Default Re: Flexicam ATC router retrofit with Kflop/Kanalog

    Hi Eug,

    Great. That all sounds correct.

    I assume you mean "Noisy" as in dithering while still? Please read and try this wiki article.

    Regards



  6. #30
    Gold Member TomKerekes's Avatar
    Join Date
    May 2006
    Location
    USA
    Posts
    2998
    Downloads
    0
    Uploads
    0

    Default Re: Flexicam ATC router retrofit with Kflop/Kanalog

    Hi Eug,

    In regards to oscillation, I did as you said and reduced all gains by 25% but couldn't see much improvement, if anything the error became larger.
    I wanted to point out that I think the lower gains did make a significant improvement in stability. Here are two comparable regions in your two plots. Can you see the oscillation dampens out much quicker with the lower gains?

    Flexicam ATC router retrofit with Kflop/Kanalog-hilowgains-png

    However you are correct lowering the gains caused the errors to increase. That is to be expected and is the trade-off. Higher gains tend to make bigger corrections faster and provide better performance/accuracy but tend to make the system unstable. I think either set of Gains may work fine. Just wanted to point this out so you know what to do if you later run into stability problems.

    I found that reducing I gain to 0 was detrimental and I had to increase the Max following error to be able to get any movement. So I put I back to 0.01 and went from there. Did I do this correctly?
    If you reduce I Gain to a low value or turn it off and don't yet have feed forward set correctly then as earlier we would expect huge errors. So I should have mentioned to temporarily set a large Max Following Error so you can still make moves and observe large or small errors in order to tune the Feed Forward. But regardless It seems you achieved the correct result. Sometimes it can be confusing if the Integrator is making corrections while trying to tune feed forward.

    Regards



  7. #31
    Registered xyz-cnc's Avatar
    Join Date
    Jun 2011
    Location
    Australia
    Posts
    37
    Downloads
    0
    Uploads
    0

    Default Re: Flexicam ATC router retrofit with Kflop/Kanalog

    Hi Tom

    Thank you again for taking the time to spell things out. If I see/hear any instability I'll drop the gains.

    I assume you mean "Noisy" as in dithering while still?
    No, I read that article, mine are sitting still. Occasionally 1 count of movement can be seen on 2 of the channels. I mean you can hear the servo motors. Hard to describe the sound but it's like all the servos are both groaning and hissing, although some are worse than others. It's not a problem but as I have no experience with servo motors I don't know what is "normal". If there is a way to quiet them down I'd like to do so.

    I quickly tuned the Z axis and I also disengaged the Xa and Xb motors from the racks and tuned them individually. It appears very similar PID gains work with all of them. Regarding the noise, looks like Z has 50Hz and both Xa & Xb are 70Hz. As the Z is ballscrew and both X were disengaged, I'm thinking it's not mechanical and caused by something else?

    What do you think may be the cause of the spike seen at the first sign of movement on the Xa servo? Both Xa and Xb are rigidly mounted to the machine, they have just been disengaged from the rack. I was very surprised to see the plots look so different. I spent some time making small adjustments to PID and FF (didn't just copy/paste the numbers).

    Attached Thumbnails Attached Thumbnails Flexicam ATC router retrofit with Kflop/Kanalog-xa-ff-png   Flexicam ATC router retrofit with Kflop/Kanalog-xb-ff-png   Flexicam ATC router retrofit with Kflop/Kanalog-z-ff-png  
    Attached Files Attached Files
    Last edited by xyz-cnc; 09-17-2018 at 12:57 AM.
    Eug


  8. #32
    Gold Member TomKerekes's Avatar
    Join Date
    May 2006
    Location
    USA
    Posts
    2998
    Downloads
    0
    Uploads
    0

    Default Re: Flexicam ATC router retrofit with Kflop/Kanalog

    Hi Eug,

    I mean you can hear the servo motors. Hard to describe the sound but it's like all the servos are both groaning and hissing, although some are worse than others. It's not a problem but as I have no experience with servo motors I don't know what is "normal". If there is a way to quiet them down I'd like to do so.
    That noise may be just the nature of the amplifiers. The amplifiers switch the power supply on and off to regulate current. Newer amplifiers switch at 20KHz or higher so you can't hear it. Sometimes the motor cabling length, shielding, proximity to other cables can influence it. To prove it isn't being caused by KFLOP/Kanalog you might disconnect and GND the input to the Amplifier and check if there is still the noise. It may be something you have to live with or replace the amplifiers.

    I quickly tuned the Z axis and I also disengaged the Xa and Xb motors from the racks and tuned them individually. It appears very similar PID gains work with all of them and the 40Hz noise appears to be present on all. As the Z is ballscrew and both X were disengaged, I'm thinking it's not mechanical and caused by something else?
    Its probably the tuning combined with the other components in the system, motors, encoder, amplifiers, amplifier tuning, coupling, compliance, mounting, etc... The bandwidth of your system is probably around 40Hz. Which is a place it may then oscillate if not enough gain/phase margin. You would probably need to do frequency domain analysis and Bode Plots to understand this. Things are likely to change when the rest of the mechanical components and loads are added. I'd recommend moving on and revisit this later if necessary. I think these errors are quite small for a router.

    What do you think may be the cause of the spike seen at the first sign of movement on the Xa servo? Both Xa and Xb are rigidly mounted to the machine, they have just been disengaged from the rack. I was very surprised to see the plots look so different. I spent some time making small adjustments to PID and FF (didn't just copy/paste the numbers).
    Actually I see Xb has a spike at the first point of movement not Xa. Xa has a spike when the axis is enabled at time=0. That axis might have more amplifier offset or something. I wouldn't worry about a tiny spike at the beginning of the day when the system is first enabled. I see you have FFA set completely different. Do you have the filters set the same for both axes? I also don't think they are actually much different. Just plotted at different scales due to the spike on enable.

    Regards



  9. #33
    Registered xyz-cnc's Avatar
    Join Date
    Jun 2011
    Location
    Australia
    Posts
    37
    Downloads
    0
    Uploads
    0

    Default Re: Flexicam ATC router retrofit with Kflop/Kanalog

    Hi Tom

    I pushed on each axis lightly and the noise stopped, so you were right, the noise I could hear was indeed dithering. It was noticeable due to the silence, but when I plug the spindle cable back in and the spindle fan comes on I can't even hear it. Apologies for bringing it up.

    Also, in regards to the 40Hz noise, I'm not concerned about it at all. If you didn't point it out to me I wouldn't have thought anything of it.

    Regarding the Xa spike, it's there on each motor move, not just at the beginning of the day when the system is first enabled. And it's there regardless of the size or velocity of the move. But no matter what I did, it was still there so I'll just move on and see if it causes any issues after I start cutting.

    Having completed the servo tuning, I have moved to the cutting and rapid profiles. The rapid profile is pretty straight forward, but in regards to the cutting one, I have a question... (probably still missing another concept) LOL

    Doing some (metric) maths, I have approx 800 counts per mm of movement which obviously means 80 counts per .1mm and 8 counts for 0.01mm. For a large bed router, that is more than enough precision, so I'll be happy to keep my errors under 20 counts,

    When I was checking the errors, I was doing it at about 75000 c/s and the errors were quite low, about 10 counts. I found that I could easily increase the velocity to 200000 c/s and the errors were still under 20 counts. Even at 300000 c/s the errors were relatively small. In the past, depending on the job, I typically would cut at under 140,000 c/s but with this machine I'm happy to cut faster if the results are adequate. However I'm sure there will be times when I can't cut at that speed and will be cutting much slower. Do I have just the one cutting profile, and if so, what happens when I cut slower? e.g I choose V 200000 as my cutting profile with 20 count errors. I then cut a job at 75000 c/s. (at 75000 c/s, errors are 10 counts). Am I likely to get 10 count errors? (this is theoretical, don't worry about all the other mechanical errors for this question). If that is the case, why not just set a cutting profile near to max velocity?

    When this machine was working, everything was done using a hand held pendant. Even if I could get it to work with Fflop I wouldn't bother as I like having the PC with screen and keyboard. I got my hands on a Moneygram kiosk, gutted it and today installed the Dell PC that I have been up until now testing with. I put the kiosk on wheels so I can easily position it where it suits best and move it if necessary, but I'm wondering if there is a limit on the length of the USB cable that goes from PC to Kflop? Is 5m OK?

    Eug


  10. #34
    Gold Member TomKerekes's Avatar
    Join Date
    May 2006
    Location
    USA
    Posts
    2998
    Downloads
    0
    Uploads
    0

    Default Re: Flexicam ATC router retrofit with Kflop/Kanalog

    Hi Eug,

    I pushed on each axis lightly and the noise stopped, so you were right, the noise I could hear was indeed dithering. It was noticeable due to the silence, but when I plug the spindle cable back in and the spindle fan comes on I can't even hear it. Apologies for bringing it up.
    No problem. I'm not certain that proves it is dithering however. The Servo should still be dithering with you pushing on it.

    Regarding the Xa spike, it's there on each motor move, not just at the beginning of the day when the system is first enabled. And it's there regardless of the size or velocity of the move. But no matter what I did, it was still there so I'll just move on and see if it causes any issues after I start cutting.
    Well every time you do a Step Response Test the axes are disabled, parameters updated, and re-enabled. Normally the axes would just be enabled once and then you might run Jobs all day is what I meant.

    Having completed the servo tuning, I have moved to the cutting and rapid profiles. The rapid profile is pretty straight forward, but in regards to the cutting one, I have a question... (probably still missing another concept) LOL

    Doing some (metric) maths, I have approx 800 counts per mm of movement which obviously means 80 counts per .1mm and 8 counts for 0.01mm. For a large bed router, that is more than enough precision, so I'll be happy to keep my errors under 20 counts,

    When I was checking the errors, I was doing it at about 75000 c/s and the errors were quite low, about 10 counts. I found that I could easily increase the velocity to 200000 c/s and the errors were still under 20 counts. Even at 300000 c/s the errors were relatively small. In the past, depending on the job, I typically would cut at under 140,000 c/s but with this machine I'm happy to cut faster if the results are adequate. However I'm sure there will be times when I can't cut at that speed and will be cutting much slower. Do I have just the one cutting profile, and if so, what happens when I cut slower? e.g I choose V 200000 as my cutting profile with 20 count errors. I then cut a job at 75000 c/s. (at 75000 c/s, errors are 10 counts). Am I likely to get 10 count errors? (this is theoretical, don't worry about all the other mechanical errors for this question). If that is the case, why not just set a cutting profile near to max velocity?
    Yes that is all correct. The main thing to determine is the acceleration that still gives acceptable accuracy at your highest cutting rate and with "infinite" Jerk. The Velocity will normally be determined in the GCode by the specified feed rate anyway. The machine should cut reasonably accurately at all speeds less than the maximum. Even if you specify a certain feed rate when cutting something like a circle the axis will be required to move at an entire range of speeds form the specified feedrate down to zero.

    For Rapids the criteria is quite different. All you really care about is not shocking the system so badly that it is still shaking and rocking after the Rapid completes and cutting begins. You should be able to go much higher Velocity and Acceleration because you don't really care about accuracy and you can specify limited Jerk to dramatically reduce shocking the system.

    I'm wondering if there is a limit on the length of the USB cable that goes from PC to Kflop? Is 5m OK?
    The specification for USB cable length is 5m maximum. But try to keep the cable length as short as possible and away from noisy signals. Use a good quality shielded cable with ferrite rings on both ends. You might check the shield continuity from shell to shell as there are lots of cables that don't meet specification. The shield resistance is should be < 0.6 ohms.

    HTH
    Regards



Page 3 of 3 FirstFirst 123

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  


About CNCzone.com

    We are the largest and most active discussion forum for manufacturing industry. The site is 100% free to join and use, so join today!

Follow us on


Our Brands

Flexicam ATC router retrofit with Kflop/Kanalog

Flexicam ATC router retrofit with Kflop/Kanalog