#emc | Logs for 2009-09-15

[00:00:49] <MarkusBec_away> MarkusBec_away is now known as MarkusBec
[00:39:21] <Allen_NZ> hi elperepat ... what is your setup ?
[00:40:09] <elperepat> Home-made XYZ milling with stepping motors. EMC2 live-cd updated to 2.3.3
[00:40:10] <Allen_NZ> sounds like you are working on the same problem as me
[00:40:24] <Allen_NZ> steppers with linear encoders ??
[00:40:32] <elperepat> no encoders
[00:40:36] <Allen_NZ> i.e. scales ??
[00:40:37] <elperepat> open-loop
[00:40:39] <Allen_NZ> Oh ok
[00:40:59] <Allen_NZ> Set it in ini with BACKLASH= on each axis
[00:41:28] <Allen_NZ> works pretty well
[00:41:46] <elperepat> ok. Thank you. I'll try that. I hesitated to edit the ini because of stepconf, but I don't think I'll need to use it again anyway.
[00:42:22] <Allen_NZ> can make it more accurate if you position the table in the middle of the backlash at start-up
[00:43:19] <Allen_NZ> I am trying to get backlash correction better with steppers and scales
[00:45:50] <elperepat> ok. My setup is not very critical: I'll use the milling to etch PCB. But I remembered that I improved the results by setting backlash compensation last time I tried emc2 (a years ago). I'm re-starting the milling and used a new computer.
[00:46:45] <Allen_NZ> If the control of the table is good after using the wizard you shouldn't need it again
[00:47:11] <elperepat> seems ok. I'll make a copy just to be sure.
[00:48:47] <Allen_NZ> check the headroom on the STEPGEN MAX VEL AND ACCEL though otherwise backlash correction won't do anything
[00:49:32] <Allen_NZ> STEPGEN_MAXACCEL = 2 x MAX_ACCELERATION to allow backlash
[00:51:35] <Allen_NZ> and make sure that your steppers can keep up
[00:52:21] <elperepat> I'll check all these and verify the results
[00:52:22] <elperepat> Thanks for all the ino
[00:52:23] <elperepat> info
[00:52:23] <Allen_NZ> during the backlash correction
[00:52:28] <elperepat> ok
[01:07:18] <elperepat> Thanks Allen_NZ, its working properly! It had 8mil on axis Y and now it's almost 0. For a home-made milling, it's alright for me... Time to etch some PCB now ;-)
[01:11:31] <MarkusBec> MarkusBec is now known as MarkusBec_away
[01:12:02] <Allen_NZ> sounds good :)
[01:12:12] <Allen_NZ> Go for it
[01:12:18] <elperepat> Thanks for your help!
[01:13:10] <Allen_NZ> No problem just happen to be delving into the same area so answers at front of brain :)
[01:13:27] <tomp> tomp is now known as tom3p
[01:20:44] <Valen> w00t The mill made its first cut last night
[01:26:59] <tom3p> if a tooltip is cutting on a surface that's revolving on an A axis (around X), the spindle on Z is changing height too. both motions are constant while cutting. is that path a cycloid arc?
[01:26:59] <tom3p>  where can i find the length given dA, start end radii & Z begin & end?
[01:27:26] <tom3p> i mean , anyone know the formula?
[01:30:38] <tom3p> whats that math site ? wolfram? isnt that german for carbide?
[01:30:57] <tom3p> or tungsten? 'wolfram kupfer'
[01:32:44] <tom3p> http://mathworld.wolfram.com/Cycloid.html doesnt look like the Z rising A turning path
[01:42:12] <tom3p> doesnt look like an ellipse because dZ isnt uniform for dA
[01:46:15] <toastydeath> hm
[01:46:27] <tom3p> yeah, the z doesnt >have< to go up then down, it can just keep rising thru all the A motion, so it might be a spiral given more than 360 degrees
[01:46:36] <toastydeath> yeah, it's a helix of some sort
[01:46:53] <toastydeath> where R varies along X
[01:46:56] <toastydeath> weird!
[01:47:00] <toastydeath> i don't know the answer, but it's interesting.
[01:47:12] <toastydeath> if i knew more calculus it should be simple to figure out
[01:47:38] <tom3p> ok, so maybe its a kind of spiral ( in the ZY plane ) and i should think R not Z
[01:48:04] <toastydeath> yeah, it is totally a spiral, like a G2/G3 spiral but more fancy?
[01:48:10] <toastydeath> or a g32 move on a lathe
[01:48:21] <toastydeath> would be a better analogue
[01:51:32] <tom3p> the new Angle is oldA + dA, the new R is oldR + dR
[01:52:23] <toastydeath> this really depends on how you want to map it, i guess - i could see this whole thing as dependant on X, or on theta
[01:54:44] <skunkworks> Patrick Swayze died..
[01:54:58] <Valen> what of?
[01:55:04] <skunkworks> patriatic cancer
[01:55:17] <skunkworks> pancriatic..
[01:55:27] <Valen> yeah I heard he had that
[01:55:29] <Valen> sucky
[01:57:39] <toastydeath> tom3p: i don't see exactly what you're trying to solve
[01:58:08] <toastydeath> like what is the end product
[02:00:08] <toastydeath> all I see is f(x) = <a,z>
[02:00:15] <toastydeath> and that will depend on the shape you want
[02:01:50] <tom3p> toastydeath: i want to find the length of the cut. this will help determine the correct joint velocities.
[02:01:59] <toastydeath> oh damn
[02:02:20] <tom3p> its surface feet , to determine sf/min
[02:04:27] <tom3p> http://mathworld.wolfram.com/ArchimedeanSpiral.html has the arc length, and it does seem to be a form of archimedean spiral
[02:06:07] <tom3p> hah! that sounded good till it turned out to be a 'hypergeomtric function' ( isnt that a Dr Who catch phrase? )
[02:07:32] <toastydeath> like, I don't know how you're going to calculate this
[02:07:55] <toastydeath> i can see the calc solution (i can't do it) but i have no idea how you'd map it to points
[02:12:36] <tom3p> the arc length is simpler here http://mathworld.wolfram.com/ArchimedesSpiral.html
[02:13:39] <tom3p> (back after laundry)
[02:18:09] <toastydeath> yeah, it's simple there because the curve is defined
[02:18:19] <toastydeath> and you can just take the derivative of the path
[02:19:09] <toastydeath> but for any arbitrary function f(x) = <a,z> you're going to have a completely arbitrary df/dx
[02:20:13] <toastydeath> with the spiral, f(x) is clearly defined
[02:20:34] <Valen> first cut on the mill
[02:20:37] <Valen> http://www.vapourforge.com/mill/VIDEO0010.mp4
[02:20:38] <Valen> w00t
[02:20:45] <Valen> (little on the large side)
[02:21:01] <Valen> ferrror is ~0.02mm on a G0 at this stage
[02:21:14] <Valen> thats with no feed forward though
[02:21:36] <Valen> on that cut the ferror was ~0.008mm or so
[02:22:28] <Valen> blew the fuse on a 7i40 by trying to push 2 axies at 15A so i'm going to look at modding it with one 15A fuse per side
[02:28:16] <skunkworks> Valen: congratulation!
[02:28:34] <Valen> there doesn't seem to be any issues with backlash or the like
[02:28:46] <Valen> but we can only run one axis atm lol
[02:29:15] <Valen> the linear scales seem to be working ok, I was worried that there wouldn't be enough counts for it to close the loop well
[02:29:48] <Valen> theres a bunch more pics and vids at http://www.vapourforge.com/mill/
[02:30:23] <Valen> http://www2.vapourforge.com/mill/VIDEO0011.mp4 is it doing a set of 50mm seeks
[02:30:50] <Valen> those are 36v motors running at 12v because the contollers cant handle the power
[02:46:11] <skunkworks> very neat
[02:46:20] <skunkworks> you need a youtube account :)
[02:50:12] <skunkworks> pcw_home: <Valen> blew the fuse on a 7i40 by trying to push 2 axies at 15A so i'm going to look at modding it with one 15A fuse per side
[02:53:44] <pcw_home> Thats likely at 12V primary power. at 36V the primary current would be ~1/3 for the same motions
[02:55:45] <pcw_home> Like a switching regulator, as the primary input voltage drops, the primary current increases
[02:55:47] <pcw_home> (assuming a constant motor power loss)
[03:02:11] <Valen> 2 problems
[03:02:23] <Valen> 1) the current limit is on the motor power side
[03:03:08] <Valen> and 2) the 48v power supply I got was sposed to have 10% adjustment, to bring it down to ~40v only has 1v adjustment so will come down to 47 volts which is still too high
[03:03:22] <Valen> but you raise an interesting point wrt fuse though
[03:03:49] <Valen> I'm looking at putting a stack of 10V zeners onto the power supply to drop the voltage to being in range
[03:15:03] <pcw_home> Much less likely to blow the fuse at 40V.
[03:15:04] <pcw_home> If its a switching supply, maybe you can tweak the feedback divider to get it into range
[03:17:37] <pcw_home> It does sound like you need a much higher current HBridge.
[03:17:39] <pcw_home> Ill bet those motors want ~50A of peak current
[03:34:50] <Valen> yeah probably
[03:35:02] <Valen> drivers in that range don't seem to be off the shelf items though
[03:35:07] <Valen> so I may need to make some
[03:37:29] <Valen> unless you know of any?
[03:37:45] <cradek> original amps are gone?
[03:37:54] <Valen> there are no origional amps
[03:37:58] <cradek> ah
[03:38:09] <Valen> this is a conversion of a manual mill
[03:38:17] <cradek> gotcha
[03:45:37] <Valen> know of any 50A drivers?
[03:49:05] <pcw_home> We've considered some heftier amps, but they will likely be serial. not PWM
[03:49:37] <Valen> would that work with EMC?
[03:49:52] <Valen> other thing then is they wont run the encoders
[03:50:03] <Valen> well not with mesa controller card anyway
[03:50:38] <pcw_home> Eventually I would hope they would be supported (and yes by a FPGA card)
[03:50:57] <Valen> its basically re-creating the mesa cards
[03:52:31] <Valen> you thinking of making them PCW_home?
[03:52:59] <pcw_home> The amps would be like our 8I20, Just a dumb current amp but
[03:53:00] <pcw_home> serial interface instead of PWM, so FPGA config would have encoder inputs
[03:53:02] <pcw_home> and serial ports (we use 5 Mbit/sec ATM)
[03:53:14] <Valen> oh you work for mesa?
[03:53:24] <pcw_home> Yep
[03:53:29] <Valen> that makes more sense lol
[03:55:00] <Valen> are the 2 large smt resistors on the board the current limit sense resistors on the 7i40?
[03:55:32] <pcw_home> Should be 4 (2 each side)
[03:55:43] <Valen> yeah, 2 per bridge i meant lol
[03:56:10] <Valen> I was thinking of taking the switching offboard and halving that sense resistor to up the current handling
[03:56:18] <pcw_home> Yep .005 Ohm IICRC
[03:56:40] <Valen> using 1405's (i have a boatload of them for other speed controller stuff)
[03:57:03] <pcw_home> The PCB is also a limit
[03:57:06] <Valen> 0.01 I think they are
[03:57:12] <pcw_home> (and connectors)
[03:57:24] <Valen> yeah I was going to run bussbars so there wouldn't be any actual current flow through the PCB
[03:58:11] <Valen> other than the fets are there any components on the low voltage version of the card that limit the voltage? other also than the overvoltage protection?
[03:59:18] <pcw_home> Well then you have to wonder if the deadzone time is enough...
[03:59:19] <pcw_home> I expect the bypass capacitors would be a limit, Not sure if they are 50 or 63 V
[03:59:57] <pcw_home> MOSFETS are 55V
[04:01:16] <Valen> I was thinking of upgrading those to 100v models
[04:01:24] <Valen> some new nice fets floating about ;->
[04:02:04] <Valen> actually looking at the board now I dont see any caps on the motor power side
[04:02:14] <Valen> oh yeah there they are
[04:02:17] <Valen> some teensy ones
[04:03:56] <Valen> yeah looks like a high voltage mod of the low voltage part is a bad idea
[09:23:57] <piasdom> g'mornin all
[09:24:04] <pjm> good morning
[09:32:00] <cnc_valen> might have a line on why we were tripping current limits so easily
[09:32:14] <cnc_valen> I think these are low inductance motors
[09:33:55] <cnc_valen> at 24v it tripped out really fast
[09:34:10] <cnc_valen> upped the pwm to 80khz and it went much better
[09:34:20] <cnc_valen> hacking up an inductor notw
[09:39:55] <cnc_valen> well that worked untill the inductor started cooking
[09:45:08] <archivist> thicker wire :)
[09:47:45] <cnc_valen> working on it ;=>
[09:48:51] <cnc_valen> see the videos archivist/
[09:48:53] <cnc_valen> ?
[09:51:45] <archivist> no
[09:56:46] <cnc_valen> www.vapourforge.com/mill
[09:57:16] <cnc_valen> 10 is its first cut
[09:57:25] <cnc_valen> 11 is some seeks
[10:37:07] <cnc_> cnc_ is now known as cnc_valen
[10:51:05] <MarkusBec_away> MarkusBec_away is now known as MarkusBec
[11:53:46] <cnc_> cnc_ is now known as cnc_valen
[13:57:35] <Valen> yeah so looks like the 7i20's lack the powah we need
[14:00:57] <skunkworks> what kind of power do you need?
[14:01:23] <Valen> 15A is having trouble getting over sticky spots in the ballscrews
[14:01:58] <skunkworks> did I see you direct driving the screws?
[14:02:03] <Valen> yeah
[14:02:21] <Valen> because we are using linear scales we must fight backlash at every stage
[14:02:38] <skunkworks> belt drive will have no real backlash..
[14:02:39] <Valen> otherwise it'll hunt
[14:02:57] <Valen> I'd rather get beefier drivers
[14:03:01] <cradek> is it scales only or do you have double feedback?
[14:03:07] <Valen> scales only
[14:03:17] <Valen> it seems to be working fairly well really though
[14:03:37] <cradek> I'm sold on double feedback after seeing how great it worked on the G&L at MPM
[14:03:45] <Valen> ferror is ~0.01mm or so on average on a G0
[14:03:53] <Valen> double would be better no doubt
[14:04:08] <Valen> URL for that thingie cradek?
[14:04:14] <cradek> oh I didn't look at the pictures before! I thought you were talking about a big machine.
[14:04:31] <Valen> its bigger than most of the stuff people are running emc on ;->
[14:05:12] <Valen> we made circles tonight, Tried to run the axis G-Code but it ran out of power on half of the moves to the right (i left the ferror limits really really big)
[14:05:56] <cradek> hm, it moves left easier than right?
[14:06:07] <Valen> yeah
[14:06:21] <cradek> that's odd - is it fairly level?
[14:06:22] <archivist> fix mechanics
[14:06:24] <Valen> it might be some small brush timing issue or something
[14:06:35] <Valen> archivist, the problem is there isn't enough margin in the machine
[14:07:01] <archivist> does sound like you are on the limits
[14:07:02] <Valen> its hitting the current limit when moving to the right fairly slowly and then only in places
[14:07:05] <cradek> yeah, sounds like being underpowered is a big part of the problem, but there might be other little things too
[14:07:36] <Valen> If we can get it to move reliably then it can probably still do usefull work,
[14:07:47] <Valen> I'll see about winding down the backlash springs
[14:08:12] <cradek> is it pretty worn? dovetail machines wear unevenly
[14:08:24] <Valen> near new would be my description
[14:08:39] <cradek> ah, good
[14:09:01] <Valen> it jams all over the place
[14:09:11] <Valen> but there are clear "pockets"
[14:09:12] <cradek> no lube system I imagine - do you use real way oil?
[14:09:37] <Valen> the ballscrews etc have an oiling system but the ways are manual
[14:09:52] <Valen> and I think its using motor oil
[14:10:09] <Valen> the ballscrews have a hydraulic oil reccomended by the oil people
[14:10:26] <cradek> does it have oil nipples/zerks or do you just have to smear it on the ways?
[14:10:33] <Valen> nipples
[14:10:45] <Valen> I have never heard the term zerk before lol
[14:11:00] <Valen> little ball bearing on a spring doovey you stick the oil can into and push
[14:11:15] <cradek> http://www.mowpart.com/product_info.php?products_id=3979
[14:11:32] <cradek> like this? there's a special gun you use that injects the oil
[14:11:35] <EbiDK> EbiDK is now known as EbiDK|AWAY
[14:11:52] <Valen> yeah one of those but they are embedded
[14:11:54] <cradek> (NOT a grease gun like for a car, but same idea)
[14:12:11] <Valen> we made one up with a standard oil can and a fancy bit on the end
[14:12:17] <Valen> the pump type oil can
[14:12:44] <Valen> turning the ballscrew by hand the force isn't excessive, and is way down on the acme thread version
[14:13:25] <cradek> oil gun: http://www.mcmaster.com/#1879k48/
[14:13:42] <Valen> mcmaster doesn't like us forign types anymore
[14:14:07] <Valen> yeah those are the sort of "gun" we have
[14:14:33] <Valen> ours just looks like the tilt oiler but it has a button pump in it
[14:14:39] <Valen> works pretty well really
[14:14:40] <cradek> sorry if I'm telling you stuff you already know...
[14:14:43] <cradek> ok, cool
[14:14:58] <cradek> sometimes people don't know about their lube system or what oil to use
[14:15:13] <Valen> its a generic motor oil on the ways
[14:15:29] <Valen> what type of oil is best for them (other than "way oil" ;-P)
[14:15:40] <archivist> way oil!
[14:15:53] <Valen> dont make me get out my dd commands archivist
[14:15:55] <Valen> ;-P
[14:15:55] <cradek> I'd use something like vactra 2
[14:16:06] <archivist> motor oil does not stick the same
[14:17:48] <Valen> mega$ or do you need a PHD to find it or anything?
[14:18:15] <cradek> over here it's easy to find it one gallon at a time
[14:18:32] <Valen> probably not going to fix the problem though
[14:18:33] <cradek> looks like about USD20/gal
[14:19:15] <Valen> we have a pair of ballscrews with some belville washers providing spring force to counter the backlash, I'll see about winding down the spring force
[14:19:16] <cradek> no, you'll still need more power, but the right oil is important for the long term
[14:19:33] <Valen> need to pull it apart to do it though I think :-<
[14:20:10] <Valen> I wonder how that hydraulic oil would go
[14:20:19] <archivist> too thin
[14:20:39] <Valen> it is reccomended for bearing lubrication and ballscrews etc
[14:20:45] <archivist> only farmers use it everywhere
[14:20:47] <cradek> most two-nut systems don't use belvilles, they use a solid spacer between the two nuts that is very carefully ground to the right thickness to give a fixed preload
[14:21:10] <Valen> they probably have really good ballscrews as well ;->
[14:21:18] <cradek> heh
[14:21:53] <cradek> mills with one-shot lube (all big ones) use the same way oil for spindle bearings/rails/ways/ballscrews
[14:21:55] <Valen> I thought of doing it in a similar way but if there is any error in the ballscrew it'll jam up or float
[14:22:15] <cradek> yeah I bet it has to be pretty darn perfect to work that way
[14:22:50] <Valen> but yeah on our other axis the errors are in the 3rd decimal place of milimeters
[14:22:59] <Valen> thats with about 10 minutes of "tuning"
[14:23:17] <Valen> P = 5 I = 3 D = 0
[14:23:43] <Valen> so I think our washers are doing a good enough job ;->
[14:24:25] <Valen> atm when it stops it over or undershoots by ~0.004 but then it pushes it to 0.000
[14:26:08] <Valen> so know of any beefy servo drivers?
[14:26:18] <Valen> 30A would probably do
[14:26:23] <Valen> 48V would be nice
[14:27:04] <archivist> beef up skunkworks design
[14:27:13] <Valen> url?
[14:27:16] <cradek> pico amps?
[14:27:27] <skunkworks> they are only 20 iirc (pico)
[14:27:44] <archivist> thats the beef up bit :)
[14:28:21] <Valen> I'll probably just roll my own
[14:28:38] <Valen> might as well shoot for 100A at 80V for it, I have all the bits laying around the place
[14:30:19] <skunkworks> will those servos handle that?
[14:30:31] <Valen> for a period of time
[14:30:39] <Valen> ;->
[14:30:53] <Valen> they are actually 500W scooter motors
[14:35:32] <skunkworks> so... at 36v that is 15a.... :)
[14:35:45] <Valen> the nameplate says 18A
[14:36:31] <Valen> we are also planning to water cool the motors if they get too hot but running them at ~15 seems to only have warmed them up some
[14:37:20] <cradek> your motor mounts look nice
[14:37:43] <Valen> dad made them, He was using the same mill but with the handles on and the scales through EMC as a DRO
[14:37:51] <Valen> using scales on both mill and lathe
[14:38:02] <Valen> it all feels nice and solid though
[14:38:22] <Valen> pushing along an axis as hard as both of us can gets it to move ~.01mm
[14:40:45] <Valen> ROFL!, just got a price from onlinemetals to ship a hunk of AL worth $275 to me, the shipping cost is $785 for express of $586 for cheap
[14:42:24] <cradek> woo
[14:51:49] <Valen> It'll be good practise for making our brushless motor drives I spose
[15:11:06] <Valen> any idea what the FPGA on the 7i40 does? looks like it runs the bridge and probably has something to do with current limiting but combinational logic can do that pretty easily?
[15:11:15] <Valen> spose it might generate the deadtime
[15:14:03] <Valen1> Valen1 is now known as Valen
[15:44:26] <Dave911> >>>ROFL!, just got a price from onlinemetals to ship a hunk of AL worth $275 to me, the shipping cost is $785 for express of $586 for cheap
[15:44:28] <Dave911> Wouldn't you think that they would spot such silliness before sending out a quote. I bought some AL tooling plate off Ebay for a good price once. It was a cut out, but plenty big for what I needed. I use it as the bed on my engraver. Nice and flat. There is a AL company in MI that has good prices. http://www.schupan.com They have no problem shipping. There are probably others out...
[15:44:29] <Dave911> ...there as well.
[15:45:09] <archivist> he is in .au
[15:53:19] <Valen> it is a 15x20x2" chunk too
[18:11:08] <piasdom> skunkworks; did you hear from MarkPictor?
[18:15:57] <ZeroFlex_afk> hallihallo
[18:16:16] <ZeroFlex_afk> ZeroFlex_afk is now known as ZeroFlex
[18:52:18] <ZeroFlex1> what programs do you people use for creating a drawing for a machine ?
[18:52:28] <ZeroFlex1> to be converted later to g-code ?
[18:55:01] <pjm> i use 'cambam' for windows
[18:56:10] <archivist> * archivist uses inside rear of skull
[18:56:25] <pjm> LOL
[18:56:38] <pjm> ah yes i use the same too, but its not really a program
[18:56:52] <archivist> it creates the programs :)
[18:57:35] <archivist> damned stepper driver died in the new toy
[18:58:08] <pjm> oh typical
[18:58:09] <pjm> that is a pain
[18:58:24] <archivist> Im waiting for chips
[19:00:02] <ZeroFlex1> :)
[19:00:34] <pjm> i've started with my 7i43 board, mounted it and the IO cards on a ally panel ready for some testing activities
[19:00:57] <ZeroFlex1> hmmm 7i43 ?
[19:01:05] <ZeroFlex1> is it a mesa board ?
[19:01:09] <pjm> yah
[19:01:17] <archivist> I need to start earning money for toys like that
[19:01:19] <ZeroFlex1> better then de 5 series ?
[19:01:38] <pjm> probably not better, just alternative so them
[19:05:42] <archivist> just contemplating a workalike http://www.youtube.com/watch?v=5dvLbb-HNsE
[19:06:14] <pjm> yeah my 1st step with the mesa is to make a pyvcp for all the I/O pins based on that one for the 5i20
[19:06:28] <pjm> once that is done, i can start moving from parport to mesa
[19:10:24] <Guest688> hello all
[19:10:35] <Guest688> i have a big concern about emc 2.3.0
[19:11:20] <Guest688> i am using it for 4 months now and i still "loose" sometimes steps
[19:11:27] <Guest688> i actualy gain to many steps
[19:11:34] <Guest688> i don;t know why
[19:11:58] <archivist> dont run steppers too close to spec
[19:12:10] <cradek> "lose"
[19:12:13] <archivist> dont power down during use
[19:12:44] <cradek> your problem is either with your hardware or with your emc configuration - what troubleshooting have you done?
[19:13:35] <Guest688> hu
[19:13:54] <Guest688> i use servos with general numeric drivers
[19:14:05] <Guest688> the drivers are step/dir drivers
[19:14:18] <Guest688> so far i done about anything possible
[19:15:12] <micges> Guest688: you're using software stepgens via parport?
[19:15:40] <Guest688> the pc i am using a compaq PIII at 833 mhz with 512 mb of ram gives a good latency test result of about 8900
[19:15:49] <archivist> have you set the pulse width correctly for the dirives
[19:15:54] <Guest688> i put the latency at 15.000 just to be shure
[19:16:04] <Guest688> yes the pulse width is set correctlu
[19:16:08] <cradek> so you have encoders going to the drives and not emc?
[19:16:22] <Guest688> the drivers are build by my fathers company and work with other systems perfectly
[19:16:23] <cradek> how do you know you are losing/gaining steps? what is the actual behavior you see?
[19:16:31] <cradek> so you have encoders going to the drives and not emc?
[19:16:39] <Guest688> yes the encoders 1000 steps/rot go to the driver
[19:16:54] <Guest688> but the fallowing error is set to 0
[19:17:03] <cradek> how do you know you are losing/gaining steps? what is the actual behavior you see?
[19:17:21] <Guest688> ok, so i have a big file, 50 k lines of gcode
[19:17:26] <Guest688> i run it
[19:17:48] <Guest688> the z axes for example must go down to z-0.12 mm and up again to z1 mm
[19:18:07] <Guest688> for aprox. 4k times in this program
[19:18:27] <Guest688> at the end of the programm the axis goes us about 1 mm
[19:18:52] <cradek> do you mean when the program ends, the z axis is 1mm higher than when it started?
[19:18:58] <Guest688> yes
[19:18:59] <Guest688> about
[19:19:05] <Guest688> sometimes more sometimes less
[19:19:15] <Guest688> the same thing happens sometimes also on x and y
[19:19:20] <Guest688> but not allways
[19:19:34] <Guest688> i didn;t figure it out yet when it does happen
[19:19:40] <cradek> if your timing settings are wrong, the drives may sometimes go one step in the wrong direction because they have not yet noticed the direction signal has changed
[19:19:54] <cradek> another possibility is the drives are not correctly and reliably reading the encoder signals
[19:20:13] <archivist> electrical noise on the step lines as well
[19:20:20] <cradek> that's also possible
[19:20:46] <Guest688> this possibility i would exlude because the drives are used usualy in very bad industrial conditions and allways work perfectly
[19:21:04] <cradek> it would be a bad idea to exclude any possibility without testing it
[19:21:12] <Guest688> the noise on the dir/step lines would not be great, it is a 1.5 m printer cable
[19:21:14] <cradek> bad grounding of your encoder wires could be a fault
[19:21:26] <archivist> if ok without the spindle running and fails with shows noise
[19:21:57] <cradek> archivist: pwm servo drives can make just as much noise too, if your wiring/grounding is wrong
[19:22:09] <archivist> I use a printer cable and HAVE suffered noise
[19:22:22] <Guest688> the servo drives are pid drives
[19:22:24] <archivist> yup
[19:22:28] <pjm> yeah me too, worst was from the VFD, it got into everything
[19:22:44] <Guest688> what is the vfd?
[19:22:57] <pjm> its the variable freq drive for the spindle motor
[19:23:06] <archivist> inverter driving the spindle motor
[19:23:10] <Guest688> thanks, it does not use any
[19:23:18] <Guest688> it has a simple isel ufm1050 motor
[19:23:41] <archivist> a brushed motor?
[19:24:02] <Guest688> yes
[19:24:10] <archivist> if so can easily produce electrical noise
[19:24:47] <archivist> grounding and screening needs checking
[19:25:15] <Guest688> here you are right but why does this thing sometimes happen only on z and some other times on x and y too?
[19:25:58] <cradek> because that's how noise works - it's unpredictable
[19:26:30] <cradek> I also think you should consider it might be stepping in the wrong direction due to incorrect timings
[19:26:35] <Guest688> as for the screening and grounding i think it is good, it is an isel machine with aluminium body, all axes are gronded individually
[19:27:39] <Guest688> do you think it could be the mainboard or the pc,
[19:27:40] <Guest688> ??
[19:27:43] <cradek> I do not understand why you are discounting every suggestion. Clearly it is working incorrectly. These are very good suggestions from several people about things you should check in order to troubleshoot.
[19:28:08] <Guest688> yes but i checked the grounding and the screening
[19:28:15] <cradek> how did you check it?
[19:28:32] <archivist> we are suggesting electrical noise be cause of old age and experience
[19:28:48] <Guest688> i run the same pc and same software on two different machines
[19:28:58] <Guest688> both servos
[19:29:00] <Guest688> same drives, other cables
[19:29:05] <Guest688> one not grounded at all
[19:29:22] <Guest688> and the other one grounded on all axes and every cable screened
[19:29:28] <Guest688> but the same results
[19:30:17] <cradek> ok that is not definitive but might point more to a problem with step/dir timings
[19:30:25] <Guest688> ok
[19:31:12] <cradek> what is the name of the drives? can we see the specs online?
[19:31:18] <Guest688> we calculated the timings, since we have acces to the source code of the drivers internal firmware and put double values in
[19:31:23] <Guest688> but no better results
[19:31:51] <Guest688> the drivers are made by my fathers company and used for over 12 years in industrial applications
[19:32:16] <cradek> what is the name of the drives? can we see the specs online?
[19:32:17] <Guest688> especialy plasma cutting where the noise can be so bad a pc nearby would crash and restart
[19:32:23] <Guest688> the drivers work just fine
[19:32:40] <Guest688> the company is general numeric
[19:33:02] <Guest688> i don;t think you know them
[19:33:09] <Guest688> it is generalnumeric.ro
[19:33:49] <cradek> are the step/dir lines read by a microcontroller in this drive?
[19:33:57] <Guest688> they use a own avr based cnc for the motion of the cnc
[19:34:02] <Guest688> yes
[19:34:05] <Guest688> it is an avr
[19:34:42] <cradek> that's a notorious way to have difficult timing requirements...
[19:35:38] <cradek> I think it's very possible you are still getting steps in the wrong direction.
[19:35:41] <Guest688> yes, i see what you point at
[19:36:07] <cradek> is the quadrature from the encoders also read by the avr?
[19:36:32] <Guest688> but i cannot explain it to myself why it works perfectly when controlled by the own cnc controller also with avr-s
[19:36:42] <Guest688> yes the encorder is read by the same avr
[19:36:49] <cradek> I can - the step timings are different
[19:37:18] <Guest688> yes, but we calculated the timings and doubled them
[19:37:32] <archivist> double again
[19:37:34] <cradek> bad or too-slow quadrature decoding is also notorious in microcontroller based drives
[19:37:54] <cradek> it's very easy to write incorrect or quadrature decoding
[19:38:19] <cradek> you could even verify the timings with a scope
[19:38:58] <cradek> does the drive step on high-going step signal or low-going?
[19:39:10] <cradek> if you have that set wrong, the timings will be (very) wrong
[19:39:15] <Guest688> it steps on low going
[19:39:33] <cradek> do you have the step line inverted in emc's configuration?
[19:39:47] <Guest688> no
[19:39:56] <Guest688> only dir lines on some axes
[19:40:05] <cradek> that is surely one problem
[19:40:18] <Guest688> why??
[19:40:34] <cradek> emc ensures the timings are correct based on STEP. your drives step on /STEP
[19:40:41] <cradek> so, the timings are not guaranteed to be right
[19:41:32] <Guest688> ok, so we have to reprogramm the drives
[19:41:39] <Guest688> that will be not a big problem
[19:41:40] <cradek> or configure EMC correctly
[20:10:27] <Guest688> :)
[20:10:36] <Guest688> i just talked to my father
[20:10:45] <Guest688> about this, and to the programmers
[20:10:59] <Guest688> the timing is more than enaught
[20:11:26] <Guest688> 5000 ns is way to much, they process a step in about 1000 ns
[20:12:10] <Guest688> we will dig furture into this, and change the pc for the start because probably the printer port does not get enaught amps for the optocouplers
[20:12:19] <Guest688> thanks to everybody
[20:12:41] <archivist> but is you process the wrong direction because you are too near the edge
[20:12:46] <archivist> is if
[20:13:39] <archivist> not just width its also the coincidence of the dir change and step
[20:15:31] <Guest688> what do you mean by coincidence?
[20:15:49] <Guest688> of dir and step?
[20:15:59] <Guest688> they come at the exactly same time?
[20:17:31] <archivist> dir to step delay and needs should be documented
[20:18:29] <Guest688> yes the timing in also under 1000 ns because at each step we analyse the dir
[20:18:40] <Guest688> and the whole procedure takes about 1000 ns
[20:18:51] <archivist> on the rising or falling edge
[20:19:18] <Guest688> on the falling edge
[20:20:26] <archivist> I remind you of <cradek> emc ensures the timings are correct based on STEP. your drives step on /STEP
[20:20:50] <Guest688> yes but,
[20:21:24] <archivist> invert the step signal see if it cures it
[20:21:36] <Guest688> i will try it
[20:22:00] <Guest688> thanks fot the help
[20:22:21] <Guest688> i will try it in a couple of hours and inform you tomorrow obout the results
[21:51:17] <jthornton> http://www.linuxcnc.org/component/option,com_kunena/Itemid,20/func,view/id,835/catid,9/
[21:54:36] <archivist> I would not go there...laptop., is he trying to multiboot to xp or ubuntu
[22:00:56] <jthornton> that was kinda my thoughts...
[22:17:09] <jthornton> well my network is working now on the gateway so it is update city time :)