EMC2 + Steppers + Encoders = following error detection (so far) :)

[samco]

There is this guy - you see - that has encoders hooked up to the back of his stepping motors. He thought 'Hmmm I would like to see if I can set emc2 up so that if I start missing steps, Emc will throw a following error - and estop'. He is reading the encoder back thru the printer port. You can see how flexable emc2 is. I think this is the first time it has been tested. (It has been talked about a lot - there may still be bugs in the theory) But is seems to be working. Here is an edited IRC chat with the emc gurus from here.

Gives an idea of the thought proccess.
http://www.linuxcnc.org/irc/irc.free...2007-09-12.txt

18:47:13 alex joni: i have an encoder on my z axis and its hooked up to my breakout board with a and b channel,is there a place on the wiki that deals with adding an encoder 18:47:29 wanting to get ferror 18:48:02 jlmjvm: search the configs for the component encoder 18:48:17 you need to loadrt encoder num_chan=1 (if my memory serves me right) 18:48:32 then add the 2 functions of encoder to base_thread and to servo_thread 18:48:40 you need to get the scale right 18:48:59 and the output from encoder needs to get connected to the axis feedback position 18:51:37 configs?do you mean where the hal and ini file is?dont have encoder component in my configs, 18:52:15 may not be looking in the right place 18:54:31 jlmjvm: look at the etch-servo hal for a starting point for setting up the encoder module 18:54:43 will do 18:57:50 looks like what i need to get started,thanks 19:24:25 heres my hal file,crashes when emc starts 19:24:30 http://pastebin.ca/694392 19:25:33 im only using 1 encoder right now,and the a and b pins are right 19:25:43 uh what's the error? 19:26:03 HAL:20:error message here 19:26:06 (20 is the line number) 19:26:30 want me to paste error msg 19:26:36 encoder.2 is wrong 19:26:42 you need encoder.0 19:26:55 and loadrt encoder num_chan=1 is enough for now 19:27:24 if you want encoder.2.* then you need num_chan=3 (0,1 and 2) 19:27:59 my z axis is number 2 19:28:09 encoder on z axis 19:31:43 jlmjvm: 'Z axis' doesn't mean you automatically put .2. everywhere 19:32:07 jlmjvm: in this case, this is the first and only thing you are hooking up to 'encoder', so you use encoder.0 19:32:34 just tried it with 0 19:33:01 still no go 19:33:13 yeah but the error is almost certainly different now :-P 19:33:27 prolly so 19:33:42 want me to post it 19:34:26 well that's the only way anybody here is going to be able to help you 19:47:37 http://pastebin.ca/694421,thats the changes i made 19:47:59 how do you send an error,cant copy 19:53:59 jlmjvm: start emc from a terminal 19:54:06 then you can select and copy the error 19:54:36 jlmjvm: but the error is probably : 19:54:42 newsig Zpos-fb float 19:54:46 ahhh 19:54:47 that one should already be defined 19:55:04 jlmjvm: the linksp should be enough 19:55:36 adam2: http://groups.yahoo.com/group/CAD_CAM_EDM_DRO/ 20:04:51 http://pastebin.ca/694451 20:05:01 finally got the error msg 20:05:31 jlmjvm: don't see the actual error 20:06:22 k,on the way,that was the dmesg 20:09:04 http://pastebin.ca/694459 20:09:15 that should be the error 20:09:46 jlmjvm: I already told you that you need to remove the newsig 20:10:18 gives another error when i did 20:10:39 jlmjvm: what's the error? 20:10:42 already has a writer? 20:11:17 lemme remove the newsig and resend error 20:11:26 jlmjvm: no need 20:11:31 remove the newsig 20:11:38 then open core_stepper.hal in the same dir 20:11:46 and look for the following line 'linksp Zpos-fb <= stepgen.2.position-fb 20:11:53 and comment that line out 20:13:36 k,will give it a try 20:15:06 your going to have to scale the encoder module correctly to work correctly with you input scale 20:16:01 I think I could have squeezed another 'correctly' into that sentence 20:16:40 that worked 20:17:20 emc came on and if i turn the motor the numbers change 20:17:28 Sweet 20:17:31 motor poer is off 20:17:36 power 20:17:54 where do i set the cpr amount at? 20:18:18 1024cpr quadrature 20:18:27 20480 counts per inch 20:21:33 scale I think 20:22:09 it's up to you actually 20:22:10 setp encoder.0.position-scale 20:22:21 you can stick a line like that in the .hal file 20:22:22 setp encoder.0.position-scale 20480 20:22:40 you can use any name you want in the ini file, and reference that in the .hal file 20:22:44 I ment scale pin on the encoder. 20:22:48 yes 20:23:01 so setp encoder...scale 20480 is one way 20:23:03 SWPadnos: he is doing steppers with encoders. 20:23:08 rigth 20:23:10 right 20:23:19 right 20:23:30 so he has the scale in the ini for the stepper which will be different than his encoder scale. 20:23:42 * skunkworks is thinking out loud 20:23:45 you can add any variables you want to the ini, and reference tem in the HAL file, so it really depends on how you want to set it up 20:23:55 really - cool 20:24:03 so you can add a section called [MY_STEPPER_ENCODER_FEEDBACK] 20:24:04 really? - cool 20:24:29 and a variable in it called THE_NUMBER_OF_COUNTS_PER_INCH_ON_THE_X_AXIS 20:24:45 so add that setp line to my hal file? 20:25:01 and you can set that by using "setp encoder.0.scale [MY_STEPPER_ENCODER_FEEDBACK]THE_NUMBER_OF_COUNTS_PER_INCH_ON_THE_X_AXIS" in the hal file 20:25:08 setp encoder.0.position-scale 20480 20:25:21 that'll work too, of course 20:30:16 hmm - wasn't there an issue with the stepgen with using encoders as feedback - like it doesn't reset it position to the 'encoders' position after an estop or something like that? 20:30:22 changed the hal file,still have joint error,was i supposed to put something in the ini file also? 20:30:48 what is the error? following error? 20:30:55 yep 20:31:19 encoder is doing its job already 20:32:18 with it off - does the readout count correctly? like 20 turns equal an inch on the readout? 20:32:35 are they going in the same direction? 20:32:39 i can check 20:33:40 SWPadnos: did you read what I wrote above about stepgen? 20:33:58 yes. that could be a problem, I don't know 20:38:34 counting correctly 20:41:26 jlmjvm: so the second you turn the machine on - you get a following error - or is it when you start to move? 20:42:05 when i start to move 20:42:18 so when you start to jog - it will error? 20:42:24 yes 20:42:27 I would negate the scale on the encoder 20:42:51 jlmjvm, you may need to fiddle with the stepgen accel and vel limits 20:43:11 setp encoder.0.position-scale -20480 - or does it count the right direction? 20:43:11 or use halscope to display what is happening with commanded & feedback positions during the short time you try to move -- that will give you a better idea what is going on 20:43:40 jepler: what? systematic aproach? bah 20:43:41 yes 20:43:41 20:44:40 jlmjvm: yes what? 20:44:57 yes its reading the right way 20:45:25 ah - ok. Jepler really has the correct way to see what is going on. use halscope. 20:45:32 and my rapids are slow right now,60 ipm 20:45:47 try moving much much slower than that 20:45:49 what if you jog slow - like 1ipm or such 20:45:56 same thing 20:46:04 tried that 20:46:13 use MDI to do something like G1X1F10 20:46:46 was doing this same thing yesterday on mach and it could read at 100 ipm 20:46:48 jlmjvm: what is your following errir set to in the ini? 20:46:57 lemmee see 20:47:56 ferror=.050 minferror=.010 20:48:19 I would think that would be more than enough. 20:48:28 ferror that is. 20:48:37 jlmjvm, you have a setup error, and you'll need to use the diagnostic tools to figure out what that error is 20:49:04 halscope is probably the best tool at the moment, because at the moment, nobody knows exactly what's going on 20:49:18 and without knowing what's happening, it's hard to fix it 20:49:58 wasnt i supposed to add something to the ini file to point to the encoder scale? 20:50:07 jlmjvm, not necessary 20:50:20 but you should make sure the scale is correct before going on 20:50:38 it seems by your account - it is working correctly - 20 turns of the leadscrew is 1 inch. 20:50:47 in the display 20:50:59 ie, manually turn the motor shaft/screw to move the table 1 inch, and verify that the display shows 1 inch of movement 20:51:21 its doing that 20:51:29 does it go the right way? 20:51:32 yes 20:51:47 ok, and if you don't use the encoder feedback, but use the stepgen feedback, can you jog? 20:51:57 and when you do that, does it jog in the correct direction? 20:52:19 yes,was moving it yesterday 20:52:20 you can experiment a little by setting ferror very large - like 1 inch 20:53:08 are you sure software counting can keep up with your encoder rate? 20:53:10 will try,but was i supposed to add something to the ini file? 20:53:13 just to be sure, you are exiting from EMC and restarting every time you change anything in the hal or ini files, right? 20:53:25 yes 20:53:33 ok - just had to check 20:53:38 no prob 20:53:43 jlmjvm: no, you did not need to modify the .ini file if you put the encoder scale explicitly in the .hal file 20:53:53 you have the option of using an ini file setting for the scale (or anything else), but it's not required to use one 20:53:53 k 20:54:10 just wanted to make sure 20:54:23 lemmee open up the ferror setting 20:55:54 open up both of them. 21:01:38 its working 21:01:56 some big dummy had the pins backwards 21:02:16 heh 21:02:32 i swear when the power was off the motors it was reading right 21:02:47 it has to be right, and also it has to match the motor 21:03:00 it's more important for it to match the motor 21:03:11 (as far as following errors go) 21:03:14 swapped the a and b pin # and its working 21:03:30 ok. you could also have negated the scale as skunkworks suggested 21:03:37 jlmjvm: negating the encoder scale I think would have done the same thing.. 21:03:38 21:04:01 * skunkworks thinks now jlmjvm should now play with adaptive feed controll.... 21:04:28 lets rock n roll,lol 21:04:34 * SWPadnos thinkd skunkworks needs to take his meds 21:04:41 err - thinks 21:04:43 i really appreciate the help from everyone 21:05:06 any time. just remember to put it on the wiki if you discover anything major about steppers+encoders 21:05:15 jlmjvm: you really need to post this on cnczone. This is always brought up. 21:05:16 im the only guy on the block with encoders on his steppers 21:05:25 really 21:05:42 So - now if you stall - do you get a following error? 21:05:49 yall can have a copy of my hal and ini if you want 21:06:00 no thanks. I'd rather have you put it on the wiki 21:06:07 im sure it will 21:06:09 hint hint 21:06:29 lemmee give it a big accel and stall it 21:06:32 brb 21:06:48 me,on the wiki 21:07:04 just set the axis vel limit to 1000IPM and try G1X10F500 21:07:15 but set the ferror back down first 21:07:49 he should be able to measure the following error - and as it increases slow the feed rate down using adaptive feedrate.. I think it would be a cool experiment.. 21:12:55 oh yeah,jacked the acceleratio up from 6 to 60 and it gives an immediate joint following error,doesnt even move 21:13:41 thats gotta be working 21:14:42 plus its showing up when i home out the axis,-.0002,.0001 etc 21:15:21 i think this is gonna really help 21:15:25 cool 21:15:51 im ready to have a keg party now,this was magor to me 21:16:18 what base period are you running? 21:16:22 10000 21:17:17 its reading at 60 ipm no problem 21:17:41 yah - you should be able to do 150ipm no problem 21:17:41 should be able to get 100 ipm like i did in mach yesterday 21:17:58 150 would be awesome 21:18:08 what is your input scale. 21:18:11 ? 21:18:22 remember - you can output at least twice as many steps as you can read 21:18:23 the cool thing is i had these encoders all along 21:18:29 10000 21:18:35 the encoder counting is going to be your real limmit 21:19:01 you probably won't be able to count past ~90 IPM reliably 21:19:03 SWPadnos: explain? you mean doing doublefreQ? 21:19:09 yes,they were higher count than i wanted,but they were on the motors i bought 21:19:13 yes, doublefreq is part of it 21:19:36 but you need some margin for reading quadrature that isn't needed for step generation 21:19:42 and they were 250.00 encoders,wish they had an index 21:19:48 AH - but he should be able to count at 50khz shouldn't he? with base period set to 10000? 21:20:02 at most 21:20:09 atleast 21:20:16 oh - I think I had an extra factor of two in my head 21:20:18 so that is 146ipm 21:20:21 100 is prolly all i will go 21:20:21 lerneaen_hydra has joined #emc 21:20:52 no, at most. if latency is 0 and the encoder has perfect 90 degree phases, then 50 KHz is doable 21:21:34 and at 60Hz? 21:21:34 if latency isn't zero, then the least time the encoder needs to be at one state is 10000 + latency to guarantee that the software sees it 21:21:52 oh, nm didn't see the K 21:21:53 if the encoder isn't perfect, then that restriction is on the shortest phase 21:23:06 so if it's got a "short pulse" in one quadrature state - say only 75 electrical degrees instead of 90, then that short state needs to be BASE_PERIOD + latency_max long to be noticed 21:23:55 in that example, 75 degrees is ~1/5 of a full cycle, so a cycle of 4 counts = 5 *(base_period+lat_max) 21:24:03 huh 21:24:08 heh 21:24:20 so it works great, until you start looking at the fact that nothing is perfect 21:24:33 heh 21:24:48 alright lets get started on that adaptive feedrate,lol 21:24:52 heh 21:25:00 just kidding 21:25:03 opk - time for me to get back to paying work! 21:25:30 thanks for the help 21:25:35 I could have swarn though that I was counting encoders close to base period... but I don't know for sure.. 21:25:45 sworn 21:26:04 but that is an argument for another time.. 21:26:05 you could have been 21:26:10 thats how we say it down south 21:26:20 ahh coulda swahn 21:26:33 lol 21:27:16 SWPadnos: this is all I have http://www.electronicsam.com/images/...astermouse.png 21:28:53 jlmjvm: atleast get a wiki up and I will post it on cnczone 21:29:01 btw,couldnt this same scenario be used with a gecko g320 setup by also connecting the a and b encoder wires to a 2nd breakout board also 21:30:15 how do you get a wiki up,never done 1,and no smart remarks,lol 21:30:34 never posted before 21:30:54 still in intense learning mode 21:34:38 back in a bit,gotta eat supper 21:39:24 jlmjvm: glad it works 21:39:48 thank alex 21:39:54 thanks 21:40:04 jlmjvm: the wiki is pretty straightforward 21:40:21 gonna check it out 21:40:33 first thing read & try to follow this page: http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?BasicSteps 21:40:52 i think this is an awesome safety feature 21:40:57 will do 21:41:28 once set up right, you can have unstallable steppers :P 21:41:48 basicly if the stepper starts to lag behind, you adjust the feedrate to go slower 21:41:53 just like marris is doing 21:41:58 yup 21:42:03 sweet 21:42:21 if you prove it working 21:42:32 that would be awesome 21:42:45 we all think it should, but so far no one bothered enough to try it out *blush* 21:43:05 i will try it 21:43:48 gotta get my 2nd paraport installed tomorrow 21:44:38 then i can wire it all correctly,was outta inputs,had to disable y limit switch 21:45:37 2 ports,many inputs 21:46:18 I would personally try to get my hands on some hardware counter for the encoders 21:46:42 I'd trust one a bit more than software counting 21:47:25 may look in to that later 21:48:17 just glad its working 21:49:56 alex joni:shouldnt i be able to do this with gecko servos also,run a and b to gecko and second parport?

Next his mission it to see if when measuring following error within emc - have emc start slowing down as the folloing error gets larger. 'adaptive feedrate'

Cool stuff - great work!!.

sam

[epineh]

That is so cool !

I can't wait to start playing with EMC, I was going to ask which IRC server you guys use but I decided not to be lazy and found out myself

Will be interesting to follow this.

Cheers.

Russell.

[samco]

forgot to mention it is a cnc bridgeort,older boss mill.

He put aftermarket steppers on it - I will try to get the model.

edit - He just got a second printer port working - so he is sending step/dir out of one and reading encoders in the other.

sam

[skullworks]

I have to quit sleeping thru all the really interesting subjects...

[cut more]

Wow,
I have been wondering if this could be done. Does anyone know what is being used for drives?
Has a wiki been posted on this? I looked & had no luck.
Thanks,
Cutmore

[samco]

IIRC they are gecko vampires running directly from the printer port.

He is planning on settting up a wiki page for it.

sam

[cut more]

Any new info on this?

Thanks,
Cutmore

[sansbury]

What about using a linear encoder to measure the actual motion of the axis, as opposed to just the turns of the motor?

[nd55]

[Warning: Noob in the area]

> What about using a linear encoder to measure the actual motion of the axis, as opposed to just the turns of the motor?

This is OK for the initial acceleration and traverse, but as the machine gets towards the target area, backlash limits the ability to accurately locate the machine location and the controller ends up 'hunting' back and forwards across the desired location.

Effectively, destroying the motors in short order.

Some controllers use a 'dual closed-loop', one the motor shaft encoder and two the linear glass scales on the machine, but that is a more complicated/expensive solution which would better be solved using higher quality components in the first place.

on a side note, Yaskawa 1540 have the capabilty to use dual-closed loop feedback. Anybody tried it?

Nick.