i do Atmel AVR's
I think jepler and cradek also program avr's
SWPadnos seems to be the expert
am I wrong?
well, if I'm theexpert, we're in trouble :)
I probably do have the most background. but I suspect jepler has surpassed my practical knowledge by now ;)
i was just wondering about the programming software that is available for Xilinx stuff, like the avr studio from atmel....
Xilinx ISE WebPack is available free for Windows and Linux
ahhh, the magic word "free" i will dl it now....linux too kewl....
Altera has free Windows tools, but you have to pay for the Linux version (unfortunately)
mmmm meatloaf is done.
mmmm - mac+cheese is gone :)
gee, i'm gonna order a chicken ceaser salad from next door....
tho mac + cheese sounds good too.....
hmmm - I don't think my neighbors serve chicken caesar salads
my neighbor happens to be a restaurant.....
pricy but worth it....
I am 5 miles from any restaurant. but a few deliver :)
I guess I'm at the midrange then - I have a bunch of restaurants within a mile, and there are a bunch of others that deliver :)
my potato buds are a little soupy. thats ok though.
boy its hard to tell when things are done when your color blind
another 10 minutes to be safe :)
what on earth is a potato bud, and do people eat them?
it is some sort of dried potato flakes that you mix with boiling water,milk and butter. sort of makes a mashed potato like substance
its funny - as a kid we alway had real mashed potatos. when I moved out and the girlfriend made instant potatos - I thought it was the best thing ever :)
i c potatoes that can survive a nuclear blast.....
right - just not meal worms :)
mel worms - that sounds like something a Start Trek alien race would say
Please Zagreeb, we have almost no meal worms left
mdynac: other than the edm - any other machines planned for conversion?
yes credak, I will buy your meal worms for 100 quoinants
or gold pressed platinum
yes, there is one in mind......some type of HUGE cam making machine.......one linear z axic and one rotational axis...
while the cam rotates the z axis moves up and down cutting a cam follower slot in the cam....
are we talking automotive cams?
no we are talking huge cams in machinery....
quite unlike an auto cam
or is it gold pressed latinum? I could never understand quark
or spell it even
go to www.camsinc.net he is my customer......
wow. cool. didn't know there was such a big need.
check out the follower groove in the flash display there, that is what i mean....
what was that made for?
desperatly needed in the manufacture of the Kotex!!!!
and let's not forget the tampon.....
playtex is his biggest customer.....
the machine grinnds the follower groove......
a very niche business, he is the cam king tho.....
he bought a machine that has an old 8088 pc controller, some university in england wrote the o/s for the machine, and he needs to upgrade....
ok - meatloaf looks done now - at least it seems to be the same shade all the way through
the o/s is a one time only thingy...
yah - I get all excided thinking about how powerful emc2 is
and i don't know if it even works, lest find someone who knows it...
it's the greatest thing since sliced bread....
or potato buds even....
[00:35:58] <skunkworks> http://www.electronicsam.com/images/KandT/cncworkshop/cncworkshopemc.jpg
we do have a bridgeport at the shop that is just begging to have a cnc control on it.....and a clausing lathe....
skunkworks, i'm glad i wasn't in that line up!!!!!
i somehow managed to dodge all those flashbulbs that week...
who is eletronic sam??? and how's the color business?
[00:44:58] <skunkworks> http://www.electronicsam.com/images/KandT/cncworkshop/memazak.JPG
I am on the right - jepler is on the left
the funny thing is the site comes up in a google search - which I find funny.
jeff i remeber, you are kinda vague.....
yah - I was sort of an emc latch key kid
swpadnos i have a ISE Webpack question?
ok - I may be able to answer
you live in cheeseland?
(I don't use the ISE directly, mostly)
well, sort of - Vermont :)
I do - WI
they have not been doing very well :)
that was for skunkworks...sorry i am not irc savvy
whatever - I just have to remember the phrase when we visit the in-laws ;)
i am running slackware 11 here and the webpack is for red hat , any suggestions?
can slackware use RPMs?
or is it a .tar.gz?
not sure, that is new territory for me, but what better way to learn....
actually a .tgz
ok, then I'd just try doing the install however they tell you to - it may work without too much effort
i think i saw a rpm2tgz somewhere in my travels...
if it's a tgz, then you need the opposite ;)
(which may not exist)
well i will just go ahead and grab the rpm and figure it out as i go along....
no slackware uses .tgz packages
oh - I thought you were saying that the file is a tgz
wow the webpack is one gig!!!!
the programming tools are about 360 megs, is that good enough to start with?
I don't know. you do need the chip support files, and probably the documentation would be a good thing
k i'll fire up the dl then....
apparently it is one big shell script....
yes - it has the install data embedded within the script
wow excellent xfer rate is should have it in 20 mins
she's settling down to about 660KB/sec
like I said - that's quick :)
i used to use redhat , back in the day....
me too, back when Walnut Creek CD-ROM was the cool place to get CDs
I think I still have some of the Linux disc sets from ~1994 or so
kewl, that reminds me of my old Telix days.....
my fav bbs surfer....
back before the net was the web
back in 1987 we had one gig on a bbs out of jax, fla.
was that Channel1?
picked up a benchtop Delta drill press off craiglist for $25 last night
a whole bedroom of hard drives....
at least you didn't have to swap 3000 floppies around ;)
we started out that way...
yep - RBBS and TBBS were great back in the day
the old FIDOnet etc....
mr holmgren, yor seahawks did well last night....
if compiling a simple FPGA project (127 lines) can take 2 minutes it must be real frustrating for "real" projects
45 minutes + for a LabView FPGA project
on an Opteron 244 system (single threaded app, so cores don't matter)
is there some way to work with smaller parts of the project, instead of rebuilding it all?
for some parts, yes, but the final place+route is the big memory hog, I think
to think I get crabby at my day job because the final link of the software takes something over 30 seconds ..
you need a faster computer (cluster)
feel free to tell your boss I said that
I tell cradek that all the time
I can't convince him a modern 4-core opteron system will beat this dual xeon by that much
hmmm - what's the time for an emc2 make on that unit?
after make clean, of course
dunno -- I doubt it has the build requirements
even for sim
I think I did some tests on the Linux kernel - something like 2 minutes or thereabouts, I think
using the .config from Ubuntu
with modules? or bzImage only?
that seems really fast
dunno - I can check
ok - it's building now - I'll let you know in a few minutes
ok - full kernel (2.6.15) + modules build was 8m11.927s real, 27m39.884s user, 4m8.104s system
that was the output from `time make -j8 bzImage modules`
after make clean, with a mostly cold cache (mostly because I had started a make but had forgotten the -j8, so I stopped it and did make clean again)
cradek would have killed for that when he was working on the RT kernel packages
it's almost getting to the point where Gentoo is viable
yeah I bought a new machine because my PIII was not going to work for building the kernel a lot of times.
a fascinating bug report from Ed Nisley on -users
credek: I would think others would be having similar problems.
yep, I would too
I read through it - what about a rtai problem?
if he doesn't see it in halscope, it's quite likely his hardware
I'm really anxious to see his answer
I like how so many people say - 'I never had problems with emc1'
cradek: oh, did you e-mail him too?
jepler: yes cc:
skunkworks: I sure did (and I fixed a lot of them!)
I was busy writing something very much like what you wrote
I also mentioned how to downgrade the emc2 package, which may be premature at this point
yeah I was holding off on that, but it's fine
I really, really, really doubt we broke stepgen in 2.0.4.
I really, really do too
I did touch the motion controller, however
if the motion controller was reversing on stepgen, he'd get a following error immediately because of accel constraint violation
so it's stepgen or hardware/OS
he'd also see ramps back down, not a smooth step ramp with reversing direction
like I said, I'm anxious to see the answer!
cradek: i widened the quadrature counters in the FPGA but I haven't updated the hal module yet
is it getting full now?
lets take bets - on board video :)
I'm now trying to figure out if the pins on the "JTAG connector" can be used or not
skunkworks: my bet is it's something stranger - and switching to a different machine will fix it
jepler: are some pins on the fpga not hooked to anything?
if he had a big problem with realtime, 2.0.4 would report it and he surely would have mentioned it, his bug report was so precise in every other way
skunkworks: do the math -- there are 66 I/Os available on the acex1k 100-pin package, but only ~30 on the two headers and ~17 on the parport...
it appears that the pins on the so-called "JTAG connector" are just hooked to I/Os, so this device can be used as a JTAG interface, not so it can be JTAGged
how many more is that?
there are 22-ohm series resistors. will that affect their ability to work as I/O?
cradek: enough to get a full 8 din + 8 dout
don't think so
simple matter of soldering anyway...
you could tack 0 ohms over them, or in their place
yep very true
this is a lot of control for $60 isn't it
$60 plus hours and hours of jepler's work anyway
and now that the digital inputs don't fit in a 32-bit register with one quadrature output, I can put the quadA, quadB and quadZ signals in the new 32-bit register so that someone with fewer than 4 axes or no index pulse can use them as inputs
cradek: I would have wasted my time anyway
jmk-away is now known as jmkasunich
skunkworks: I think cradek's assigned me to work on a high speed logic scope next
we gotta get jepler a m5i20
that fpga as 200,000 gates ;-)
:) - you have a lot projects
uh oh -- I just hit 95% full
sure, I'll take a donated m5i20
guess it's done then
wow - that is a cool little fpga. even cripled it will do a lot. Nice find jepler
what did you do to get 95% full?
If I take out the vanity register it's back down to 88%
skunkworks: adding little stuff mostly
yeah, a register that always read 0x4a656666
(that is, an EPP register)
jmkasunich: FF2 works, and it really helps a lot on my lathe
jepler: how many gates does that cost you? :-)
cradek: good to hear it
cradek: about 32 apparently
er, 32 LEs
yep (I was reading back over the last couple of days)
jmkasunich: I need to make a real ower supply now - the .5kw transformer isn't cutting it :)
cradek: bah that's only .1G!
jepler: it seems pretty snappy to me! I don't hear the accel
that's accel to full speed in 1/40th sec
do you notice the difference from the previous tune?
yeah it's much smoother
strangely adding much I now makes it less, not more stable - I don't know if that's normal
it works pretty well even at I=0
you get very low steady state error?
yes seems like
usually I=0 means that the output can go to zero and the motor stop while you still have some error
yeah it doesn't do that, it sticks within the deadband even if I try to turn it
I do have some I now
P=45 I=8 on both axes
jepler: so you have come up with 18 extra i/o - above the encoders and pwm pins?
skunkworks: 16, I think
on the JTAG connector there are 7 I/O, 2 GND and 1 VCC
I think I gotta buy me one of those widgets
mine is being shipped :)
I guess that does make 18 -- there must be two I still haven't used
jmkasunich: for me it was worth it for a chance to learn some verilog
jmkasunich: but beware the gratis design software from altera is windows only, and I'm guessing not win9x..
2000 and XP, their website says
I should just save my $ and get the m5i20
that has the capacity to do what I _really_ want
what do you want?
4 axis stepgen, 4 axis encoder for X,Y,Z and a future rotary table stepper, plus
1 encoder, 3 analog inputs, 3 pwm, and a few digital I/O for an AC servo drive for the lathe spindle
you want steppers and encoders?
plus misc I/O for jogwheels, etc
cradek: steps out, and looped back to close the loop
thats what jon e does on the USC card
use step/dir to quadrature logic, then encoder counters
if I someday decide to go servo, just change the stepgen to pwm and use real encoders
oh, almost forgot - need another encoder channel for the spindle
yeah you'll need a slightly bigger device
I'm beat .. the pc software changes to go with all these fpga changes will have to wait
I'll also want to be able to transfer all that data to/from the FPGA at several KHz
PCI is much faster than EPP for the data transfer
I bet this will do 5kHz
seems elson's page claims 10kHz is possible
for 4 channels maybe
if you want more B/W just go USB :-P
now I'm really leavin
* jmkasunich starts fighting with BDI-4.50 again
jmkasunich: are you back home? I thought you were out til thursday
only a two day trip
what do you think Ed Nisley's problem is?
not enough info
if its is momentarily reversing at the very end of a move, it might be a tiny overshoot
if its happening at some other time, I haven't a clue
you and jeff gave him good advice, I'm waiting for the next round of info
However, -sometimes- the motors turn (slowly!) the wrong
way; start correctly, jerk backwards, then run backwards;
run correctly for a short time and then stall; and so
oh, that part....
me too :-)
looking at it with halscope will determine if its some software bug or a hardware issue
I suspect (hope?) halscope isn't going to show it
yeah there's not much code past what halscope sees is there
ok - bed time. night
SWPadnos_ is now known as SWPadnos
do anyone know how currents are specified in 3 phase AC servos?
specs say its RMS current but is it different for 3 phase than 1 phase?
for 1 phase it is 1/sqrt(2) times peak current
[17:28:18] <jepler> http://www.phys.unsw.edu.au/~jw/power.html
"An AC generator may have more than one coil. If it has three coils, mounted at relative angles of 120°, then it will produce three sinusoidal emfs with relative phases of 120°, as shown in the upper figure at right. The power delivered to a resistive load by each of these is proportional to V2. The sum of the three V2 terms is a constant. We saw above that the average of V2 is half the peak value, so this constant is 1.5 times the peak amplitude for any
so ignoring that the AC servo is an inductive load, my understanding is that the RMS current is the same all the time
but the question is what manufacturers mean by RMS. is it per phase peak *1.414 or sum of all phases
tele the manufacturer. :) The sales people and tech people are usually helpful. They're job is to know the products they sell.
and support them.
usually the support people don't know the answers for the questions I ask :)
the peak is still sqrt(2) * RMS for sinusoidal voltages
power calculations are different though, and power is 1.5x the value for single phase
look at it this way - with 2 wires (one phase), you get X power for a given voltage and load. if you add 1 more wire, that's 1.5x the number of current paths, so the current is 1.5x as high
on a separate note: http://www.youtube.com/watch?v=MqDAf_lg9Xs
I should get one of those
jepler: have you bought the flashy board yet?
skunkworks: yes, I bought a flashy
1 or 2 channel?
skunkworks: I have a rudimentary program to sample from it under linux http://emergent.unpy.net/files/sandbox/scope.png
1-channel, pluto only handles 1 channel with his scope firmware apparently
SWPadnos: I wonder how that came about. (thought proccess wise)
but it is way cool
I can jugle 3 objects. I read a book on jugling when I was in high school. It was funny. They said 3 objects is very easy. 4 is just a bit harder. 5 objects you need to practice almost every day to keep good at it.
yeah - I've never been able to get too far with 5
I can juggle 3 random objects pretty well though (I did a banana, a grapefruit, and a miniature marshmallow once)
believe it or not, juggling was an option for a gym class in high school :)
funny - gym was not one of my favorite classes :)
mine either, but with such subjects as "juggling" and "orienteering", it ended up being OK :)
jepler, here's a nice little compile server: http://www.tyan.com/products/html/typhoon_b2881.html
SWPadnos: had to look up "orienterring" Never heard of it. Looks like it would be fun - and probably earn you a merit badge in boy scouts ;)
yep -there was one
also useful in the military
The big thing now is gps treasur hunting.
"geocaching" - that's pretty fun too :)
thats the name.
a few people from my astronomy club do that.
I got my wife a GPS for Christmas last year or the year before - we've managed to find a few of them
astronomy club.. How dorkey is that :)
but it is fun
we managed to get a look at Haley's comet from the observatory at a nearby college - it was very cool
Cool. the club members have a bunch of nice scopes.. one just got a 12" reflector.
we bought a telescope (a 5" or so Meade) for Haley's comet, but the higher magnification eyepieces never seemed to work, so we returned it
telescope positioning and tracking is one of those other hobby applications for motor control
cradek: you *can* have two separate supply voltages on the ACEX chip (VCC_IO and VCC_CORE) but I am pretty sure they're the same (3.3v) on the pluto-p board
jepler: how was lunch?
skunkworks: pretty good
should have my board at the end of the week.
the other i/o boards should be here by then also.
skunkworks: there's a breakfast-all-day diner we go to sometimes. They have a large enough vegetarian clientele that there is a veggie burger and veggie bacon available. that's where we went today
veggie bacon? is that salty seitan?
cool. wife only lets me eat veggie bacon.
have you ever had a walnut burger?
SWPadnos: I'm not sure what this kind is made from. my favorite kind is made from tempeh, which is a soy product.
skunkworks: no, but it sounds good
[19:05:45] <skunkworks> http://www.walnutburger.com/
oh yeah - tempeh is good
this place is just the right mix of small town and big town diner
big enough for veggie burgers, small enough for the regulars to be seated with a grunt and a point at the big table in the corner
skunkworks: did you order a "flashy"?
No.. Not yet.
skunkworks: Do you have any kind of oscilloscope?
skunkworks: my impression is that flashy is not all that useful, compared to a real analog scope
we have a b&k and a tecktronixs. both 2 channel. but a nice digital storage scope would be nice.
I thint the b&k is 30mhz and the tronix is 100mhz. both very used and beat
flashy is pretty basic -- 8-bit sample resolution (0-255), no AC-coupled option, only 512 samples (at least on pluto-p with the supplied firmware)
the pluto board doesn't have enough memory for a reasonable record length, so the effective capture rate is limited by parallel port bandwidth (which we know can be flaky)
cradek: trempeauleau hotel?
skunkworks: no, this diner in lincoln
jepler: why are you thinking of making a hal modual for emc out if it?
skunkworks: on further consideration that's not such a great idea, because then you're limited to BASE_PERIOD as the sample rate.
inless you can bulk transfer several samples in one BASE_PERIOD
but you're still limited to the I/O rate, unless you enable DMA
ie, ~1 uS per byte
I guess It would be nice to use hal scope to see outside world signals... Say the output of the fpga's pwm
but like you say - the flashy seems pretty limited.
even if you can get more than one sample per BASE_PERIOD, halscope won't be able to graph it
hmmm - true
I did find that on my PC I can do 32-bit EPP transfers, and it is somewhat faster than 4 8-bit transfers
are we talking transfering programs to the fpga or actually reading and writing registers from emc2 - or both?
I was wondering about that
actually, I think the USC code can be sped up by writing a word with both address and data
but I'm not sure
I don't know about the USC code, but in mine I only write the address at the beginning of each read or write cycle. the address register in the fpga auto-increments at each data read/write operation.
also, the parport code should be better with 16-bit or 32-bit writes - I think the CPU only waits when a second port write occurs before the first has completed
same with the USC
but I'm not sure how much of EPP he implemented
IIRC there's a "read more" and a "write more" function, which doesn't do an address cycle
ok, so on the PC end, if you set some control bits some way, then the bus cycle is data only, otherwise it may be address then data?
* SWPadnos hasn't studied the EPP specs much
the port at base+3 is the address register, base+4 is the data register
do you outb(0, base+3); byte1=inb(base+4); byte2=inb(base+4)
+4? I thought the parport only used 4 addresses (I guess it must be 8)
SPP has 3 registers, EPP adds at least 2 more
see how much (little) I know? ;)
I'm always pleased when I know more than someone else
Quick question on G2/G3 parsing, If I provide a a G2/G3 command which translates to the arc having the same start and end point with any given radius would that result in a 360° circle or something else ?
a full circle
(of course you can't use R format)
cradek: great, thanks, just wanted to make sure before I implement, hmm we might wan't to add that to our g-code handbook
cradek: so only I/J for a full ring ?
yes that's the only way to describe one circle
I wonder if round-off might cause problems for 360° (or so) circles
beware that if the points are "just a little bit apart" then you can get a very tiny (nearly 0 degrees) arc instead
radius and one point on the circle doesn't quite cut it
cradek: if you know the rotational direction and if only x y is given then .....
cradek: different if you are dealing with x y z
hmmm - how does the interpreter decide which circle to mill when given the radius and 2 points?
Rugludallur: start/endpoint and radius gives you infinitely many circles
there should be a mirror image problem
SWPadnos: using the spec :-)
oh, well that explains it ;)
actually there are 4 arcs
right, left, larger, smaller
G2/G3 should select one pair
and positive or negative radius selects the other
right and +-R makes the other choice
I knew the NIST folks had to know more math than me
"A positive radius indicates that the arc turns through 180 degrees or less, while a negative radius indicates a turn of 180 degrees to 359.999 degrees. "
a 359.999 degree arc in R format would be a really bad idea...
that's why I wrote this -- perhaps I should specify 360 degrees too
Prefer ``Center-format'' arcs
Center-format arcs (which use I- J- K- instead of R-) behave more consistently than R-format arcs, particularly for included angles near 180 degrees.
to say "near 180 or 360 degrees"
So IJK is the way to go, just drop R
R is easier to write by hand sometimes - I don't see any reason for generated gcode to use it
great, thanks for excellent help (as always)
we even all agree for once
Hmm I wonder if I should put the radius in a comment for human readability ?
03jepler 07HEAD * 10emc2/docs/src/gcode/main.lyx: 360 degrees is also bad for R-format arcs
I agree also :) our old machine only does I/J arcs anyways.
I think IJK is actually more standard
yeah, what he said
the new fangled R is for slackers.
you know, it's amazing how a British accent gets people going
well, here in the US at least - I'm sure people in the UK don't care
yeah - I called my car insurance company to see if I'm covered in the UK, and I said "right then" in an english accent - the girl went nuts
"do that again"
"are you from there"
"that's so cool!"
when she gets out of high school she'll get over it
I said "cheerio" at the end of the conversation - she almost fell out of her chair laughing, then thanked me for doing that
I've just started to do automatic builds of the PDF and HTML documentation from the v2_1_branch: http://linuxcnc.org/docs/2.1/
hmmm - she's probably out of high school already - unless she has no classes 7th and 8th periods :)
the first upload is nearly done
8th period? you had 8 periods?
we only had 7 in high school around here
ah. I think we had longer days but fewer of them
or the opposite - I don't know (my 20th reunion is this year, so I have a right to forget these things)
jepler: this should be prominent on the main Documentation page
cradek: you mean inside joomla? Yes, it probably should, but I've forgotten my joomla password again
03jepler 07v2_1_branch * 10emc2/docs/src/gcode/main.lyx: merge rev 1.14
is there andy advantage having velocity and position loops in cascade compared to just position loop?
I see many servo systems have position and veloctity loops used simultaneously
Are you seeing "nested" loops?
just like on page 2 here: http://www.danahermotion.com/support/troubleshooting/PDF_Resources/2000-07%20Cascaded%20pos-vel%20loops.pdf
* rayh mumble mumble about file names with spaces.
* SWPadnos mumble mumble stupid Mozilla / Windows / Alamo website
hi again folks
rayh: I'm wondering if cascaded loops has some benefit over just position loop alone
Oh yea. The diagram on the top of the second page shows the nesting of velocity inside position pretty well, approx.
In fact it shows "current loop" the third common loop as well.
but the other possibility seems to be position PID cascaded directly to current loop
which is better?
I've seen systems using EMC that work both ways.
Tuning is a bit more critical if you are command current directly from position.
JonE does it all the time with his drive boards.
In fact there is no velocity feedback to his drives.
I couldn't find any information about differences
Most stand alone servo drives can do either.
seems like I have to try both
My plan is only a posistion loop.
Read the first paragraph of the link for a description of velocity command.
cradeks lathe is set up the same way.
an analog voltage command is compared to an analog voltage feedback.
You've got to have a tach or in dahner's case a d/dt from encoder pulses going by.
but has anyone tried position-velocity-acceleration-torque cascade? :)
torque is current
but separate acceleration loop, too
I don't see how accel is different from velocity repeated over time.
difference of acceleration vs velocity is similar to velocity vs position
* rayh shuts his eyes trying to imagine what a variable accel profile might look inside a velocity loop.
and how the size of a change in velocity command from one loop to the next would change the accel ramp.
Are you seeing something like an S accel ramp with the ability to change the steepness of the S if the velocity delta is large?
no, I'm just wondering if dividing control into smaller parts has some benefit
We have found that position to current works as long as the interval between command updates is small.
I currently have 20kHz current loop and 2kHz position PID
JonE and I both like to point out that our old AllenBradley 7300 controls update about 100 times a second.
In a control like the AB the velocity command and velocity feedback from a tach were very important.
the default setup of emc2 runs the position PID at 1kHz, but with a quick change you can try 2kHz (or any other value) and see if it is an improvement.
I'm running position loop in drive for minimized loop delay
approx, Those numbers sound plenty fast.
"position loop in drive"??
I'm planning to make PID 4kHz since I still see stairy response in current loop :)
yes. drive receives position commads
Oh that. You will always see a stairy response in a digital system.
yes but these stairs are caused by very fast current loop
resolvers and tach and analog velocity signals always did produce a better surface finish on parts.
approx, are you using emc to send position commands?
Serial HAL module?
maybe stepping output or some serial output
I'm developing my own servo drive
approx says "some serial output" and has my undivided attention?
currently I can control drive via SPI
velocity is always a problem with stepping types of output.
cause that's position alone.
Even gecko's pll does a poor job of spreading motor steps between command steps.
I guess I would get smooth enough operation with 2000 steps/rev
encoders are 8000 CPR
here you can see this stairy behavior: http://dal.ton.tut.fi/pub3/torquestairs.png
Sawtooth velocities are one reason that many tool makers have gone to very high encoder counts.
green is torque command from 2kHz position loop and blue is acutal achieved torque
unless you want to support legacy (step-dir only) software I don't understand why one would use step-to-servo in a new design
because it is de facto stanrad
Nah. It's a cheapo way to expand a servo drive's potential market.
rayh: because of legacy software, or some other reason?
but user may want to use it instead of spending days of programming support for some serial protocol
Not legacy, it's the other way round.
Step and direction is easy to create and ship to a drive.
There are a lot of step and dirction control software out there mach and turbocnc and such. But very little softare that is cheap that will run a servo loop. (like emc2 can :))
real position and velocity commands are much harder.
machinist wants just position so I guess fast step/dir is sufficient
quadrature is better than step/dir too, but that doesn't mean anyone uses it...
Not the machinists I know. Surface finish is everything.
but if position is accurate, then surface should be fine
On a critical part any one of then will use an older mazak v10 than a new VTC.
any time you change velocity you change cutter characteristics.
sawtooth velocity and you wind up with steps in the finish.
So they sawtooth very quickly to make the steps small.
But then you come to the limit of a PC's ability to generate step pulses.
but doesn't PC just limit top speed of positioning with steps?
You can see the issues in the lines on your trace that you linked above.
Each time a new step command is issued you get a flattening of actual torque.
that happens always if current loop bandwith is higher than controller update frequency
no matter how position information gets transferred
Imagine for a moment that you want an axis to travel at 0.001 ipm.
How do you step command such a thing.
with step pulses that are not much smaller than that.
it doesn't make any difference if position is transferred by pulses or by serial data
Right, unless with serial data you also issue a velocity.
hmm.. haven't seen such controller that uses that kind of mixed control mode
but sounds interesting
A lot of drives seem to compute velocity for the next command based on the velocity required for the last command loop.
But this always glitches on the first and last command of a series.
Interesting discussion, approx. Gotta run to town. bbl.