what's the equation that computes "allowable following error" from [AXIS]FERROR, [AXIS]MIN_FERROR, and current speed?
you get FERROR at max speed
for speeds less than max, the limit is reduced proportionally
but not below MIN_FERROR
to MIN_FERROR at stopped?
current_ferror = max((current_vel/max_vel) * FERROR, MIN_FERROR)
not that the code says exactly what I wrote, but that's a formula to go with jmkasunich's words :)
(with which I concur)
see you guys later
EMC: 03seb 07v2_3_branch * 10emc2/configs/hm2-stepper/ (9 files): merge config fix from trunk: give stepgen accel headroom, and fix the ridiculous ferror settings
steves_logging is now known as steve_stallings
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/man/man9/thc.9: new manpage
hi, looks like I get to play inside, raining again so I cannot mow, darn 8-)
you must be very very sad
I will be when I finally get to it, growing like mad and there are 2 acres worth
you've got a riding mowing thing, don't you?
lets see, GPS, and g-code, I wonder...
I've thought about that lately
well, not GPS, but remote control at least
yea, 60" cut zero turn diesel
well shit, it should take about an hour then
big is good sometimes, but not at others, enough fussy stuff to slow me down
how do you shoehorn throttle + steering into emc's kinematics module? that's the big question in my opinion.
on a perfect day, it can be done in 2 hours
jepler, I bypass kinematics by connecting hal_input directly to the actuator outputs
then get a driving stick, and it's just like being there
only via wi-fi or something
real throttle stays constant, two levers control two hydraulic motors on rear wheels, interesting kinematics
oh, so it's not really MNC, it's just computer-mediated mowing
strangely, my wife is trying to convince me to buy a lawn tractor
so get HER one and let her play all she wants
Ingrid's just working on getting rid of the last bits of grass that are left
we've only got 1/2 acre, so the John Deere self-propelled mower ought to be enough
change of topic, has anyone tried a PCIexpress parallel port card, does the timing hold true
I have one to test, but have never gotten around to it
Jon Elson might have done some testing
who owns a machine new enough to have pci express and uses it with emc? :-/
my future customers?
ran into a customer with a machine that did not have conventional PCI only experss
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/man/man9/.cvsignore: add thc.9
steve_stallings, if you have the opportunity to get these people to spend $200 or so on a computer, it's possible to get a reasonably speedy dual-core atom, in a case, with 2GB memory and some few hundred GB disk space for that price
and it works well with EMC2/RTAI, has a parallel port, and has a PCI slot
and it's tiny (could be less expensive with a larger case, I think)
you have to be kidding, customers always want to use the thing they already have, no matter how much more it costs in dollars or greif
ok, so anyone who's asking for a recommendation, you can tell about this :)
I just got 15 such computers (as parts), with 32GB SSD instead of cheapo disks, and it was under $4k for the lot
so then you need a deal where they buy a breakout board, you throw in a computer for free
my name is not Jon
it'll be a savings in the long run, if your time is not free
no EPP for me, I'll just get all tangled up with the SmoothStepper
then why are you asking about parallel port timing, huh? :)
trying to avoid all the fun of debugging external devices for as long as I can
hey steve_stallings, could you make a list of the pinout and default axis setups for Mach? It might be nice to have an EMC2 setup that's plug and play for people trying it out
I'm not sure the default stepper setup matches at the moment
default pinouts for step and direction, and estop are the only meaningful things, all else
needs setup for user devices
ok, no charge pump by default?
I gather that's 50% or more of your support calls ...
"I bought this damned card and nothing works"
teaching people that they want a charge pump is only rewarded by having to teach them how to set it up
we ship cards with a wire jumper across estop input just so it works out of the box, charge pump feature is disabled by default
because of course that's the best way to use them ...
maybe your documentation isn't printed on brightly-enough colored paper
we even have a LED that says "outputs enabled" and people ignore it
you have to condense the instructions to one sentence so you can use microsoft word's biggest font and have it fit on that hot pink half-letter sheet
we do have problems with our documentation, it is too complex, gotta come up with more quick start kind of stuff
hmmm. methinks somebody doesn't know how to count: http://www.mini-box.com/PicoLCD-4X20-Sideshow
(I'm a little bit serious - a lot of stuff comes like this, and I bet people actually see the one sentence)
and yet they still have to put a piece of tape over the power inlet if there's a 110/220 switch ...
dang, I'll need to remember that, our upcoming PMDX-125 has just such a switch
you could use the trick of putting it in 220 mode at the factory - at least that should prevent damage
but if it doesn't work, then you're in trouble with tech support
yea, fun trouble, it will probably almost work but not enough headroom for linear regulators to work properly
modern 5 volt CMOS does a good job of faking it at 3 volts
hmm. you could use a wide input range switcher, and just have it run from 17V at 110 or 35 at 220 (for example)
oh - are you using TH or SMT?
SMT where possible, TH for connectors mostly, product is too low cost to use a discrete switching supply
your switch and dual-winding transformer, plus tech support headaches, ar probably more expensive than a switcher
[13:26:41] <steve_stallings> http://www.pmdx.com/PMDX-125
the parts cost for the switcher I'm using is $3.93 in small digikey quantities
that's total, including reverse voltage protection diode, input and output caps, controller IC, inductor, and another cap
isolating line voltage in to multiple isolated low voltage outputs for 3.93?
no, you still need a transformer anyway, since you want 1/10 the line voltage or thereabouts
how many rails do you have?
oh, OK, we use small buck regulators where cost effective, typically when powering a bit of logic on a device that has 50v for motor etx.
4 total, 5 and ~12 unreg on computer side, same again on machine side
ok, so you have two output taps anyway
what's the 5V current draw?
oh - with TH caps, the cost would probably be lower - the SMT electrolytics are relatively expensive
5V at < 200 mA typical
ok, you could use a linear regulator for that then
I did 8-)
this switcher is 500 mA I think. for a few cents more you can get 1A,or 2A
we have used both the National "simple switchers" for low Vin applications and the LM5xxx series for up to 80 volts in and 3.3 volts out
yep. this magic chip is the LM2674 (or 75 for 1A)
it's $2.36 in 25 qty. $1.79 at 100
at DK, so it might be less elsewhere
we have not uses the newer high fequency simple switchers yet
they seem simple :)
remarkably stable compared to older stuff
I've used these on several projects, and have never had a problem. voltages are stable (4.99 - 5.01 every time I've measured), low ripple when I've checked on a scope
and quite tolerant of capacitor and inductor selection
(I think on one design I used a 100 uH instead of 68 uH, and never noticed)
component tolerance and auto select of continious vs dis-continious is nice
haven't seen one get warm yet, though I don't run them near the current limit
they're good chips - they must be or I wouldn't be able to use them :)
still need to jump to LM5xxx series to get to 80 volt input tolerance
well, these aren't appropriate for drives
but for breakout boards, they have a wide enough range to allow for "auto-switching" 110/220 input
true, but then I have to upsize the transformer so a 220 volt capable one can pull the output current at 110 line
yeah, that's true
the DPDT switch has got to cost a buck or two though, especially if you get those nice ones that have the recessed actuator that says "110 220" on it
same problem for main input filter caps too
wow, connectors look like half the cost of that
well, they need to be 40V instead of 20 (or whatever)
hmmm. those are .150 spacing?
yes, connectors always drive the cost
what's the per-wire cost, out of curiosity?
the prices have finally gotten to a more or less reasonable number - I think I paid about $0.42/wire for pluggable headers
again in low quantity
we strive for 10 cents per wire, use same connector across lots of platforms, buy in 1K qty
yep. that helps
ah, that's a ton less than I figured
that price is for fixed terminal strip, not pluggable
all the terminal strips on that board are made up from the same basic 2 pin part
the pluggable ones used to be around $1/wire, but they're much more reasonable now (at least they were last week)
that is except for the high current on at the big relay
I like pluggable, but sometimes selling better than the competition just doesn't work, you have to conform to hold cost
we started out with pluggable on the PMDX-120, then changed over for later products
alex_joni: are you on line???
yeah, it's tough to get the right balance
SWPadnos: might you have a recommendation for an 'inexpensive' ssd ? 16g or so?
I looked on NewEgg last week, and the least expensive SSD they had (regardless of capacity) was $75 or so
there was a (supposedly much) faster one for about $15 more
I think they were 32GB, but the cheaper one could have been 16
oh, that assumes a hard disk (SATA or PATA) connection, if you want one of those little cards, they're a bit less
it's for my new Atom board. SATA connection.
ok. the ones I got were the Patriot Warp (though I cleaned them out of that one, you'll have to get something else :) )
got the board going. had to disable the built in lan in bios.
incidentally, plug the drive into SATA0, it wouldn't boot for me from SATA1
which file system?
ext2? or ???
um. ext3 or ext4 I guess
it's a stock Jaunty install at the moment
i have an interesting problem on a customers vertical lathe.
the rotational axis 'C' started getting flaky over a few weeks and became unuasbel.
i have noted that in jogging C positive, all is ok. jogging C in negative results in increasing ferror until ferror fault.
C does turn both directions.
checked most connections. swapped out motor and then drive. no change.
steppers, servos, analog, PWM ...
servos. m5i20 with 7i33
try swapping the motor connection to a different output
so change the HAL setup and the physical wiring to swap e.g. C and A connections
i swapped oiut the 7i33. no change.
have not changed out the 5i20 yet. but can.
check the limits on C
in the ini file - MIN_OUTPUT I think
(or whatever limits the PWM output)
which would lead to what??
well, if the axis moves in reverse, but doesn't move fast enough, then it's either mechanical (slip or something weird), or the DAC output can't go as far negative as it can go positive
or something else :)
but having swapped hardware around, I think you've eliminated a lot of that
i've restarted the machine numerous times, so it can't be wound up too far.
there is no 'last position file'
this has nothing to do with the min and max position settings
there is a separate setting (or there often used to be) for the min and max PID output for each axis
it's run for about 10 months now with no other significant problem.
the PID limits were called MIN_OUTPUT and MAX_OUTPUT (I think), whereas the travel limits are MIN(MAX)_LIMIT
(I think :) )
hmmm. I must be in emc1 thinking mode today. I don't see those in the m5i20 sample config
unfortunately the machine is 40 minutes away. and my 'day' job gets in the way. i can see the early, before the day job.
and i'm limited to about 1 hour or work.
anything you can think of on the 5i20 that could be flaky.
well, wait for someone else to give you advice. I think all mine hasn't been applicable for a few years
i do have the 4th axis available to switch C to.
oh right, it's XYC (or similar)
i'm going to convert there files to 2.3 and hostmot, but that will take time, and i want to get them back into production.
does this happen with all jog speeds?
or just the fastest
they've been really good and haven't messed with the files or anything. yes all jog speeds. handwheel and axis buttons.
I'd halscope the encoder counts and ferror
i'm still leaning to a connection problem.
also, how does the machine move? if you try to jog, does the machine move 99% as fast as it should and eventually fault, does it move 100% and fault more or less instantly?
bogus encoders sometimes count one way better than the other
yeah, I'm thinking there might be something funny with the feedback
if you see jumps in multiples of 4 counts or something odd like that, it's a smoking gun
when you swapped out the motor, did the encoder also get changed?
does it sound right when jogging?
(I was thinking of servos, where they'd probably be a single unit)
i've check connections, but really need to go through them again. yes, encoder is on motor.
ok, so changing the motor also changed the encoder
the motor is disconnected from the turntable.
testing that way may be detrimental since the PID is tuned for the extra mass
it may overshoot beyond FERROR
as cradek said, you need to halscope FERROR and feedback at minimum
i don't want thing going crazy. the turntable is big, 5' diameter.
yep, the motor is almost guaranteed to overshoot then :)
size 43 motor.
same as mine, but probably (hopefully) more powerful :)
another thing you can try just for kicks - swap the direction (negate scales where appropriate) and see if it then has problems going forward
i'll get the guys to re-install the motor, and hopefully get there in a day or 2.
you also mentioned that this happens while jogging - does it also happen for MDI or programmed moves?
yeah, several diagnostics to go through.
i believe it does. i was running late last time i was at the machine, so i did not notice everything.
just take a quick look and make sure the acc/vel limits are sane, and that they are in each AXIS_# section
thanks for the help. i'll be back, hopefully from the machine location. the day job is beckoning.
as i said, it function quite well for about 10 months.
ok, so something probably changed
check for a big piece of swarf in the gears
also remember what NOT to do - don't mess with the software, don't mess with the tuning
be sure you KNOW what the problem is before you fix it
measure / diagnose / correct
maybe we should make new NIST-style mugs with that
people always get that wrong - for instance in this case they'd mess with the tuning
or reinstall EMC, or reinstall the OS
well, that would be necessary once the motor gets removed from the big-ass table
at least they started with the mechanical parts in this case
yes it's true that motor won't run right, off the table
ok, time to make coffee. bbiab
anyone thought about G12 and G13 circular pocket for EMC?
can you get a link to some documentation or write a spec for it?
some info here http://www.practicalmachinist.com/vb/showthread.php?p=973754
* BJT-Work still looking
[15:01:18] <BJT-Work> http://www.cnczone.com/forums/archive/index.php/t-7071.html
* BJT-Work just e-mailed my Haas buddy about the spec
a wiki page about how it should work would be cool. it might be an easy project for someone at fest to tackle if we have a good spec.
ok, I'll work on that
seems like it would need to work with and without cutter comp turned on - that's the only stumbling block I can think of.
I wonder what other commonly-used cycles would be nice. the BOSS has a zillion of them - rectangular inside and outside pockets, facing, hole circles, on and on. (I never use them!)
argh. I may have to ditch the trip to Fest
my anilam bp has some nice ones for pockets
[15:42:20] <BJT-Work> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?G12-13
yuck, one week to go and you still don't know if you can go?
I just noticed that my flights all leave at 6:00 or 7:30 AM, and arrive back at 10:00-11:30PM
and the 3 trips I have scheduled are literally back to back - get home one night, leave the next morning
I might be able to reschedule the times of the Wichita trip to make it possible though
"If no cutter compensation is desired, a D00 must be specifi ed. If no D is specifi ed in the G12/G13 block, the last commanded D value will be used, even if it was previously canceled with a G40."
is it just me, or is this nutso?
it's not just you
that's right out of the manual :)
it makes sense to use the last value most of the time though, since you want to be able to set it once for a given tool and then do a lot of pockets with it
we can make it like we want :)
also seems weird that it uses radius
also it can do only one or two circles? it should be able to spiral or step out to cut a big hole.
my anilam uses diameter
q sets the stepover amount I think
oh ok, you are right, I kind of see how that would work
so the final depth is L*Z
it does approximately (K-I)/Q circles for each L
depth is specified as increment, count; radius is specified as start, increment, end
I think a crack monkey designed this
or somebody with very limited CPU and memory to work with
reload that page I added the Anilam spec
minimal calculation for the canned cycle, using the existing arc and cutter comp facilities
BJT-Work: that looks better but it uses some letters that we can't use
it's interesting that it doesn't let you specify direction. I bet it uses g41/g42 to decide
there are other things that I can pick from the conversational part
it would be neat if it would work with the other canned cycles (so you can G91-L repeat)
cradek: I added the Anilam description to the wiki page
the Anilam pocket cycles include Face, Rectangular Profile, Circular Profile, Rectangular, Frame, Hole, Irregular, Mold Rotation, Elbow Milling...
If your interested in any of them I can make a page for them
this gives us some good ideas. but with my usual hubris I'd probably invent a new scheme just for us. :-/
I wonder if a spiral is worse than concentric circles for some reason
there's no spiral code
a decent spiral can be made up of semicircles or quadrants
spiral is on another menu
I did a spiral on the counterbore py program :)
there is also a G5 Ellipse and the G6 Spiral
have you ever used G5 ellipse?
been a while
no kidding? what for?
to face of a part so the finish looked cool
face off a part
for another take on pockets, see HAAS for G12, G13, and G150 http://www.haascnc.com/customer_service/manual/vmc/96-8000.pdf
[16:48:11] <BJT-Work> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?G12-13
so, anybody want some of these: http://www.newegg.com/Product/Product.aspx?Item=N82E16811154093
perfect for a D945GCLF2 system ...
for free shipping with power supply, that's pretty good
yeah, I know
I'm still hoarding my full tower cases though...
I have 15 of them that I need to return, so I can probably offer an even better price :)
(the 15% restocking fee + a few bucks each for shipping takes like $10 away per)
"they're too big and ugly" for my customer
steve_stallings: I haven't got past the picture on page 1 of the pdf...
SWPadnos: that sounds like the hellish kind of customer
other times it's fun and lucrative stuff
cradek: slow connection?
steve_stallings: spiking neurons aren't _that_ slow :)
oh, page 1, not the first page in the PDF :)
steve_stallings: no, it's just funny
oh, page 1, not front cover, yea be careful with that pendant
oh, the yellow triangles go up through page 9 - wow
love the one with the lung warning!
"flying debris into unprotected eyes can cause loss of sight"
how about one for "reading this document can make your head hurt:?
aha, page 121
well, most PDF readers have a search function, you know...
cradek prefers to read 120 pages in 15 minutes
photographic memory ftw
yikes, HAAS's G12 has the same "uses the last D value even if cancelled" as BJT posted, must be copied from Yasnak
I don't see anything wrong with doing a circle without comp... I think lots of people don't touch the tool table ever.
circles are the easiest thing to do with comp - just make a smaller circle
it would be the height of coolness if someone could make inscape work with the new nurb/spline gcodes in trunk. Currently it splits up paths into little straight lines. http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl/emcinfo.pl?InkscapeHowto
in my copious free time :-/
so in trunk you can open a nurbs file and it gets transformed to g code?
no, there are g codes specifically for nurbs, I think g5.1/g5.2
the gcode specifies the control points for the curve
see butterfly.ngc in trunk
hey cool, I wondered where to find that document
well that's a very interesting shape I get
I'd hold off writing documentation for it
it's not clear to me that we'll be able to get it into shape
it's apparently not-quite-right in several ways