I could get fancy, use a 5KHz thread for the diffs and the scope
ok, changed the max velocities to 12 ips
20 milliseconds is large compared to the couple milliseconds of uncertainty
max and default accel are 20 ips^2
I guess 720 IPM is fast enough for testing ;)
ha... I already had vel and acc differentiators in the sim hal file
just took 10 seconds to make a 1" move at F30
should only take 2 sec
(and did only take 2 sec before)
did you change any thread rates?
gonna change the ini back to 1.2 ips max and see what happens
try a G0
to late, already shut down
damn, mini isn't
yeah - fullscreen
at least the controls resize
ok, thats gotta be a bug
with limits set to 1.2 ips, a 0.5 ips takes 2 seconds to move 1 inch, as expected
with limits set to 12 ips, a 0.5 ips move takes 10 seconds to move 1inch
no other changes?
changed both traj and axis limits
what's the step rate (I assume you're using freqmod)
should try changing one at a time
sorry - stepgen?
no steps at all, position command out of the motion controller is just looped back to feedback
emc2's version of sim mode
oh yeah - right
anybody with a working emc2 checkout can do the same thing
(if only you could compile that on a non-rt system)
sudo scripts/emc.run configs/sim.ini
those are floats in motion, right?
i need to learn how to use this scope thing
it sounds useful
I'll bet there's an overflow somewhere (though I don't immediately see how)
the sim ini even starts the scope for you
I'd try a binary search for the breakover point
1.2 works, 12 doesn't - try 6.6
gonna check and see if jogs work right
too bad the pico hardware isn't supported by emc2
also G0 - just for kicks
it is - kind of
kind of? :P
I'm working on the driver, when I'm not chasing bugs
yeah -so shhhh :P
cool, i'll need one ;)
jog slider set to 30 ipm (0.5 ips) and thats what I get (with 1.2 max)
G0 X1 took a little less than one sec, matches 1.2 ips max
gonna change to 12 again
[00:18:01] <Jymmm> http://puffinplc.control.com/
ok, jog at 30 ipm is correct
how do the accel profiles look with G0?
G0 X1 took about 500mS
vel peaked at 4
accel went +20, then -20
with some ramping
so you get similar rounding?
about 20mS to go from 0 to 20ips^2
about 30mS to go from +20 ips^2 to -20 ips^2
in three distinct 10mS chunks
traj runs at a 10ms period, right?
thats gotta be interpolation
ok - so there's some cubic thing going on
(or whatever it is now)
anyway, that was G0, gonna try G1 F30
also try longer moves with G0
this move took 2 seconds, as it should
I think everything will be correct (less "rounding" errors) with this accel
I'm curious where the high accel breaks
I'm running 20 accel, 12 maxV
same thing that broke earlier
ok - I thought you were at 1.2 again
did 12, saw the prob, did 1.2, no prob, back at 12 again
hm - that's not encouraging (intermittent problems)
maybe right at startup problems?
dam thats annoying
mini blasting onto the whole screen
you can set the geometry pretty easily, though I've forgotten how
it's especially annoying with a 1920x1200 screen ;)
G1 F240 works right
X5, so that it can actually reach max vel before it needs to slow down
I'm starting to doubt my sanity
I only saw the one bad move, and maybe I typed the F word wrong
go doesn't fix it
G0 is just as slow as G1?
at 180 commanded...about 300 ms each for accel and decell .... so better than 12 sec at full velocity
dave-e: can you try a different .ini that uses steppers (and disconnect anything from the parport), and try that?
I suppose so...maybe later...
it would be good to experiment with accels / vels that would be dangerous in real life :)
well I left accel at 20 but set max velo to unrealistic levels
dave: what GUI are you using?
the system would drive x to about 266 real ips
ipm I hopw
launch that table into orbit
oh , there go those units again
it is about 600 Kg...anyway
alas, orbit is km/s
doesn't matter if you are in the way
several km/s, that is; it would be a very big number in even ips
if your standing there, 266 ips would seem like orbit
(as you go to clean out your pants)
jmka: or try and pry yourself out from wherever the machine/table ended up ;P
ah....gotta go...see ya later
ok, I got one funny (slower than expected) move the first time I fired up sim.ini (which was with 12 for max vel)
i'm not sure it's good for me to hang out on this channel
the previous test with 1.2 was using emc.ini
hearing stuff like this just makes me fundamentally nervous
i mean, slower-than-expected is better than faster-than-expected
that's what E-stop is for ;)
doesn't matter what speed, don't trust SW, only hardwired estop switches
and test them too...
jmka: sometimes I'd rather bury my head inthe sand ;)
estop doesn't work, you might
separately from the rest of you
yah, I know.
we actually had the mazak estop fail - there was gummed up oil/coolant in the contactor, remove coil power and it would drop out 2 or 3 seconds later
it still makes me nervous.. i would like the software to work, and for the estop and limit switches to be an unnecessary redundancy
i've only seen emc do something really bad once
the machine hung, but the axis kept turning at the same rate forever
i e-stopped it in time.. (my mini mill -doesn't- haveh ard wired limit switches)
was this emc2, and recent?
yah, emc2 from a few weeks ago
I may be the guilty party
I put in a fix for one problem that probably caused another
any real machine i build where there's a danger to more than property will have a robust estop system thuogh
though worst cases (full speed into a limit switch) may result in an unpleasant stop for the machine
well damn... I can't replicate yet another bug
maybe wasn't even a bug
wanna talk about half-wave rectifiers?
Harmonic Frequency Magnitude Phase Norm. Mag Norm. Phase
-------- --------- --------- ----- --------- -----------
0 0 -2.7591 0 0 0
1 60 167.761 -0.069571 1 0
2 120 5.01699 93.6125 0.0299056 93.6821
3 180 1.84187 15.7644 0.0109791 15.834
4 240 3.71185 -82.647 0.0221258 -82.577
there's a fourier analysis of the voltage inside the chassis on a simplified version of my circuit
-2.75V of dc bias on the line
what are you assuming for line impedance?
that's with .2 ohms + 50 microhenries of reactance on the line
and full load on the bus?
* jmkasunich gets out the copulator
I am going to characterize some transformers today; I can't find good specifications o nthem
have no idea of DCR so I don't know what kind of current/heating that will induce
.018 ohms Xl
that seems way too low if you're saying that's the DCR i should expect
50uH at 60 Hz = 0.018 ohms
so a total impedance of .22 ohms or so
we're talking about the AC line here, the only transformer is the utility one on the pole, right?
.2 seems hi
well, I'm going to have some transformers for DC supplies in the chassis
so they're who i should worry about screwing up with the dc bias, right?
for low voltage DC supplies
15V and 24V
stop, slow down,
I don't know wtf you are talking about
the MAIN 170V dc bus, does it have a transformer or not
though SWP says I'm suicidal for not having one ;)
ok, the other supplies are irrelevant
all the drives I work on don't have transformers, I'm still alive
jmka: well, they are.. if 2V of dc bias on their input will make them explode
but you're a pro ;)
that's the reason to care about the DC bias right? what it does to inductive loads
I see, you're worried about main bus Idc * Rline = too much DC volts for your 20-50VA aux supply transformer
or are there other reasons to care about the DC bias
I'm worried about what the DC does to the 25KVA transformer on the pole outside your window
hm.. the pole shouldn't see much of that DC bias
because it's on the other sides of the line reactances
what line reactances?
well, the .1 ohms of impedance from my breaker panel to the drive
the fraction of an ohm to the pole transformer, etc
nothing between you and that xfmr but wire
bet its less than that
that 2V is in the chassis; it's divided by the resistance of the wire
and most of the source impedance is inside that transformer
where do you get 0.2 ohms from
I made it up as a worst case number
ok, lets back up and work this problem from the beginning
14 ga wire is 2.6 ohms / 1000 feet, 12 ga is 1.7 ohms/1000 feet
I assume you have US standard, 120/120 split phase power?
L1, L2, and N
120 from L1 or L2 to N, 240 from L1 to L2
how many amps is your feed?
so if i have 100 feet from the breaker panel to the drive at 12 ga, that's .17 ohms
hmmm, the 12ga might be significant
I'm used to the wire being just a little fatter
but lets keep going with these calcs my way, we'll get to the wire eventually
i'm not sure of the NEC wire size for a 200A service
but i can look it up if you want
200A * 240V = 48KVA
indoors, AWG6 is about 60A, AWG4 is about 100A
yah, i think service wiring needs to be larger though
it should be half that - it's 200A for 120, so 100A per phase (I think)
here, i'll see if i can find it easily
but the utility gets away with murder, they use tiny wires, on the expectation that you'll never load that panel to 200A
and if you do, the wires are overhead in free air and get cooled that way
SWP - a 200A panel has a 200A, 2 pole main breaker
ok - interesting
(maybe I can have a full machine shop here :) )
anyway, transformer impedance is usually about 5%
that means they drop 5% of rated volts at rated amps
so on each phase, 5% of 120 = 6V
6V drop at 200A = 30 milliohms total impedance (inductive and reactive)
and nonreactive, yah
inductive and resistive, I meant
when I do sims, I assume 70-80% of that is inductive, 20-30% resistive
if the feed from the xfmr to your panel is long, that might increase the resistive part
I don't think the pole transformer will see any DC bias though
since I'm recitfying both phases
any bias after inrush, that is
I think I agree with that
in my simulations, there's no DC bias when i fourier analyze the two hot wires with respect to each other
you will be pushing DC thru the neutral conductor
but that should be OK
gotta make sure the source is balanced
that's true. to the extent that the source is unbalanced, some of that DC bias will continue to exist
if one side is 120V and the other is 121, you'll have net DC, cause one side will provide more power
the bigger the bus cap, the worse it will be
yah. 3300uF * 4 plus a bunch of 100uF in this case
ultimage worst case, the bus cap is so big that after the 121 side charges it to 121*sqrt(2), it doesn't drop below 120*sqrt(2) and the other side delivers nothing
not bloody likely...
cap the size of a large desk to do that probably
yah. i don't think that's much of a risk, there's significant decay in bus voltage
i have about 20-30 degrees of conduction at nominal load on both phases
thats good... anyway, getting back to source impedance
30mohms at the panel
how much did you say AWG12 was, 1.7mohms/ft?
yah, let me run the simulation with .030 ohms input impedance, figure out the 'worst case' the pole could see
ok, its gonna be hard to wire it with less than 20 feet of 12 AWG, so the branch ckt is gonna contribute more Z than the feed and transformer
and mostly resistive
those mohm/ft ratings are for single conductors, right? so you shuold have to multiply by 2 to account for both legs?
i can actually get the wiring pretty short, but i view having a little resistance and reactance there not as a bad thing
some filtering for the rest of the house
I'm not used to that... most of my stuff, the drive load is much closer to the overall transformer size, so the wire Z is much lower compared to the xfmr
I was assuming 20 ft round trip, 10ft one way
panel in the shop
if the shop is across the basement, round trip could easily reach 50 ft or more
so, hooked up direct to the pole, nominally:
Harmonic Frequency Magnitude Phase Norm. Mag Norm. Phase
-------- --------- --------- ----- --------- -----------
0 0 -0.34013 0 0 0
1 60 169.537 0.43743 1 0
2 120 0.702606 97.8887 0.00414426 97.4512
the main panel for the house is actually pretty much opposite on the wall where i'm going to put my big tool
it's on the outside, the tool is on the inside
ok, so you'll have a long feed
no, i mean, i could have a 5 foot feed each way
in the ridiculous case
so could be very short
what is your load anyway? 1 HP, fractional, multiple?
1kW servo motors * 4
are you running just axis servos from this, spindle motor is separate?
though i'm setting up the design of the circuit so it can be used as a VFD/inverter too
1kW / 170V = 5.88A per servo
at the DC bus
X4 = almost 25A
not gonna use #12
use #10 at least - because of the high crest factor, low conduction angle, the RMS will be higher
yah, but I am willing to limit my output
i don't picture turning all 4 motors at max torque simultaneously
note that the neutral needs to handle the DC bus current
the two hots handle half
i've taken 2 axes full power-- 12A-- as my worst case loading
true about all for at max, and also servos almost never maintain full load, it's just pulsts
seems fair, but don't forget the spindle, that does run continuous
anyway, should be under 20A
yah, but it doesn't provide DC bias unless i also use one of my drives for it
and may be on its own circuit
easy way to deal with the aux transformer issue: use a 240V primary, from L1 to L2, don't use N
many small xfmrs have dual secondaries so you can connect them for 120 or 240
yah, i'm looking for good isolation transformers, if i find one cheap enough i'll go for it
dual primaries I mean
240V primary, 60V secondary would be ideal, because then i'd do full wave rectification
and there's many isolation transformers i could set up for that
what are you making with your aux supply, 120VAC for computer and such, or some kind of DC supply?
getting me confused again, what is the 60V for?
jmka: 60V RMS full-wave rectified would give me 170V p-t-p
you want 170V DC?
it works pretty nicely with my 100VDC motors
p-t-p is a useless number
i know, i'll get 10-25V less
120VAC full wave rectified gives 170VDC
vpp doesn't factor into it at all
120V AC + 4 diodes = 170VDC
240V AC center tapped + 2 diodes = 170VDC
i always get confused about this.
the pole transformer is 240V AC center tapped, with the tap grounded
yah, i know that
there's 2 aux supplies
I think I know where there is a drawing of various rectifier ckts on the web, gonna look for it
ok, tell me about the aux loads, how many V DC and how many amps
1 24V, that runs the brake, that is stepped down to 15V for gate drive, and to 5V for the axis microcontrollers
about 4A @ 24V
90% for the brake I bet
the brakes take 2.4A; the other things probably don't really take 1.6A
but i was very conservative in adding them up
and then there's a few hundred mA isolated 5V supply
to speak to the computer at ground-referenced, rather than in-chassis-neutral referenced, levels
it'll probably just be a simple linear supply
ok, total is about 100 watts or so
yah, the first set is neutral referenced, though, and the second set is floating (but probably grounded to safety ground via the parallel port)
you mean the first set is referenced to DC bus negative?
not my first choice, but OK
don't like direct connections between control power and main power
there would be a need for a lot of optoisolators otherwise
you are building H-bridges, right?
24, plus expensive high-common-mode amplifiers or analog optos with feedback or pulse transformer setups to get current feedbacks back
i'm using IRF integrated IGBT modules
H bridges, 3 phase bridges?
ac servo motors or something?
yah, sanyo denki 3 phase ac servo motors
the ones surpluscenter had for $40 apiece
so you have 4 bridges, 6 switches in each bridge
each bridge has 3 low side and 3 high side switches
how are you gonna drive the high side ones?
the module takes care of that magic for me
i just need to provide it with bootstrap caps
Hello - is there an "if" operator in the gcode interpeter? I have been looking for an hour and have come up with nothing. I mean if #8>10 then n100 or something like that.
it has 5V logic level inputs
skunkworks: not in standard g-code
it also enforces dead time, etc.. it has an IRF IGBT driver integrated
along with shunt resistor for measuring current, etc
I'm not sure if the latest lerman additions to the interpreter support that, it's brand new stuff
ok, so you have 6 control signals referenced to negative DC bus
yah, 6 per axis
one shunt or three?
one shunt back, and one thermistor of feedback
where is the shunt?
between the dc neg bus and the low side igbts
so you can't measure individual phase currents, just total low-side current
so i can control overall current in a control loop..
though i depend on the encoders being accurate to commutate properly
what kind of motor control are you doing? trapezoid or something
I don't see how you can control sinusoidal currents in three phases with one current sensor
well, near sinusodial; 256 sine table entries
my point is you cant control three currents with one sensor
jmka: I run a PID control loop on current; the output from that I multiply by the sine table entries
thanks jmkasnich - is there one implimented in emc?
to get the PWM duty cycles for each output
there's some additional details (clipping of the PID to prevent current from running away too badly if the commutation is messed up)
skunkworks: I just told you, not in standard G-code (which is what emc does)
the angles i get from the encoder
there is a brand new bleeding edge g-code interp for emc that might do what you want, but I've never used it
icee: I can't visualise what you are talkking about
at any given time, current will be going out one phase and coming back in the other two, or vice versa
a single sensor looking at the bus current can't tell you what is going on at the motor
here's what i'm doing. I take the current motor position as measured by the encoder
I add up to 60 degrees depending on the current speed to figure out where i should try and place the magnetic field
under some condigions, you could have zero bus current while at the same time have enough motor curent to pop the IGBTs
yah, only under severe deceleration, though, right?
I clip torque changes for that exact reasons
you are talking about "where I should try and place the magnetic field"
that is good, that is vector control
but a magnetic field vector has two degrees of freedom, angle and magnitude
it is simply impossible to control a two degree of freedom system with one feedback
then i take the sine of that angle, plus 120, plus 240, to figure out the 'relative strengths' each phase should be driven with-- the relative voltages
it may be possible to actually measure indvidual phase currents using that one sensor, if you make the measurements at the right time during the PWM cycle
the angle is open loop
the magnitude is closed loop (the feedback)
doesn't work that way
I understand I'm controlling bus current rather than total phase current
this is frustrataing, it's the kind of stuff where I need to draw sinewaves, and vector diagrams, and point a the pictures
DC bus current times DC bus voltage = output power.... agreed?
and that total phase current can differ from bus current based on speed and acceleration/deceleration
over the long term, that's true
long term = averaged over one complete PWM cycle
assume for the moment constant output speed
assuming reasonable definitions of 'output power'
so we can ignore accel/decel
assume constant output torque too
output power from motor = speed times torque
input power to drive = DC volts * DC amps
in a properly balanced three phase system, power flow is constant throughout the 60Hz (or whatever) power cycle
i'm measuring speed and torque
well, speed and dc amps, sorry
position and amps.. doesnt that count as two feedbacks?
we understand power out the shaft - spd * torq
we understand power into the drive - vdc * idc
torque being an unknown under my system, since i'm not measuring phase currents
but the flow of power from drive to motor consists of 3 voltages and 3 currents
never mind, I'm babbling
basically, you have two choices
open loop, where you produce three phase voltage, and the motor draws whatever current it needs
and closed loop, where you actively regulate the individual phase currents
isn't there a middle option?
jmkasunich - thanks again - I will do some more looking.
produce three phase voltages, scaling ALL THREE based on observed power consumption?
can't figure out how to put into words
I understand that there's a degree of uncertainty i pick up in that on each phase's current
what you describe might work, IF you can control the phase relationship between motor position and output voltage vector
you really want to control the output CURRENT vector
based on the acceleration/deceleration of the motor in the short term, imbalances between IGBTs, coil reactances, etc
you are talking about measureing the magnitude only of the current vector
I don't think you can do that with a sensor on the DC bus
so you're saying if you're doing voltage control of 3 phases, a DC bus current feedback is of no value?
I understand you have no idea how much power is getting pumped from ground to ground by the motor inductance
it can be used for overcurrent protection and such, but not for closed loop current vector control
are your encoders absolute, or do you have hall sensors for commutation?
incremental encoders with index
i'm planning to ignore the halls
then you don't have angle either
incremental encoder doesn't tell you the angle at power up
a startup sequence will gather that
force it to turn until you get an index?
you'll just run open loop until you get the index I assume
btw how does the micro tell the computer when it's on an index?
f: the index is just used to adjust the encoder count
fenn: in what context: general, emc, other? emc doesn't have micros telling computers anything
jmka: the mid-end commercial drives do something similar, though they use the halls for crude commutation first
jmkasunich: as i understand it the micro tells the computer via parallel port a 32 bit encoder count and 8 bits of "stuff"
then when they first pass an index pulse (or in some cases, a hall transition, though that seems ugly) they can more precisely control the angle
er, 24 bit not 32
icee: I believe that you can't do closed loop sine vector unless you directly or indirectly measure at least two phase currents
it may be possible to use the single shunt to do that if you make the measurements at the right point in the PWM cycle, but it is certainly non-trivial
j: I know I can't tightly control the phase currents, but I believe I can control total current better than without bus current feedback at all
lets walk thru the PWM cycle
at some point in the cycle, all three phases are high
i'm going to filter the output from the shunt
to only pass things below the PWM freq
at that point, current can circulate in the motor, thru one IGBT, in one lead, out another, thru a freewheel diode, and back out again
your sensor reads zeros
yah, i know.
later, phase A is low, B and C are high
sensor reads A
later, A and B (or maybe A and C) are low, the third is high, sensor reads A+B (which equals -C, since all three sum to zero)
still later, all three are low, sensor reads zero again
let's say I turn my drive on, run it awhile, stop the motor, turn the drive off
the integral of the shunt voltage should be proportional to the total power consumed in the motors, yes?
at the end of that time?
total power, not torque!
yah, but power is torque * speed, and i'm measuring speed
you could be drawing 0.1A at 170V from the bus, and delivering 17A at 1V to the motor
well, not quite that extreme, but yes
I exagerated only a little
assume motor stopped
maybe a factor of 3
all three phases high, current circulates freely
torque control will be relatively poor at low speed
decaying slowly as resistance dissipates energy
then B and C go low
because of greater uncertainties in speed and greater measurement error of bus power
current begins to ramp up, going from + bus, to A, out B and C, to -Bus thru sensor
a short time later A goes low, now all three are low again, and current circulates
yah. the ramp rate of coil current is something like 3A per PWM cycle at 15KHz
skunkworks: the lerman additions to the interpreter DO support IF constructs. See: http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl/emcinfo.pl?AdditionalGCodeFunctionality
the period when A is hi and B,C are low can be very short, but still keep a lot of current going
skunkworks: straight from the expert
yah, the decay rate is very slow
Jacky^ is now known as Jacky^afk
Expert (from the latin) ex meaning former, spurt meaning gushing
so if you actually want to control the current, you need to measure during the period when A is hi and B,C are low
not average over a full PWM cycle
if you average, the actual current will be a LOT higher than you think it is
basically, the ramp rates are about 45000 A/s, and the decay rate is something like 500 A/s
so 90:1 ratio
i understand the power that flows between the body diodes :(
and that it is unmeasured
this is all at zero speed of course, when there is no counter emf to make things even more complicated
you know this is probably a futile conversation
very nice - thanks again. I love reading this channel - very informative. I assume you guys are talking about the 3 phase servo card? icee and lawrenceg (among others) are working on.
there is far more to this than can ever be conveyed over IRC
you either already know what you are doing, in which case you don't need me confusing things
i understand that we are 'modulating' the coil, using high voltage to cause current to begin flowing, and then it continues to flow through the body diodes after i turn off the phases
or you don't, in which case I won't be much help
* icee nods
you can always just measure it and see if you were right
a few thousand times a second?
that's what scopes are for right?
if you are running open loop, sure
closed loop implies feedback
I have a DSO
remember that 45000A/sec ramp rate
i can data log to test my models
the ramp rate doesn't matter. i can measure the current with halls or resistors on each phase and filter
and check against my model
halls on each phase... now your talkin!
he's talking about testing phase of development
well, for tuneup and developing this it's an option, and one i'll likely employ
halls on each phase (two anyway, the third can be calculated) is the right way to do this IMHO
"three offloads the cpu"
fenn: calculating the third isn't hard
the third can be calculated with an opamp and three resistors
it's the bigger of the two if they're opposite in sign
or the sum of the two if they're the same
is the total current
it is minus the sum of the two, regardless of sign
the total current is what i said
ok, I thought we were talking about the third phase
and yah, i know the currents total to 0
for control you care only about Ia, Ib, and Ic
Idc isn't very usefull
anyways, it's just a few operations to come up with total magnitude, and a few more to come up with angle
that's not Idc, that's Ia+Ib+Ic
Ia + Ib + Ic = 0
so if you measure Ia and Ib, you can calc Ic = -(Ia + Ib)
that's all I was saying
Idc is what you get from your shunt
yah, sorry, it's not Idc either, though
congverting Idc into usefull readings is non-trivial at best
the bigger of the two if opposite in sign, the sum of the two if they're the same though
is the magnitude of the current vector
in the motor
draw three sine waves
then do your formula at several points along the cycle
the magnitude is a constant, your expression isn't
anyway, we've digressed a lot
yah, you're right.
I just want to say one thing
well, actually 2
first - unless you want to toast either your computer or yourself, at least make your first unit with either full isolation between control and power (best) or with isolation between power circuits and the line
second: build and test your first unit at 50 volts or less
the latter is planned
and in the current circuit, the computer is fully isolated
the computer talks to a microcontroler over the parallel port, which talks to the other micros over an optoisolated SPI
ok, that is encouraging
so you have a mivro at the negative bus, but totally isolated from the control signals
yah. it saves me from optoisolating 24 signals, as i can opto 3 spi + 3 selects instead
i know otherwise drive returns currents would try and make their way through the parallel port into the computer's safety ground
even if it doesn't fry stuff, the noise will just drive you nuts
yah.. i think with the couple volts of voltage drop across neutral, at very low impedance.. it'd fry stuff ;)
you most likely won't be able to scope your isolated micro for the same reason - noise as soon as you connect the ground clip
I have access to a differential probe
the ground clip on a normal probe sounds like a good way to blow up my scope
* fenn has to catch a plane in the morning.. g'night
anyways, i know i don't entirely know what i'm doing, but i'm proceeding cautiously, and will through the software development etc.
worst case, it will be a voltage controller
But I think the information from the current shunt is of nonzero value and I hope to use it to improve its performance above voltage controllers
it is of non-zero value, I'll grant you that
but it will be non-trivial to extract that value
hmm, one thought crosses my mind
there are three output phases, and each can be high or low
that is 8 states
or open, during the deadtime
for each state, the shunt reads either zero, or the value of one phase, or the inverse of the value of one phase
inverse of hte value of one phase.. (thinking)
yah, that's true.
the ADC isn't fast enough to do anything like what you're describing though
but not necessarily in phase, right?
you digital folks, jeez....
1 to 8 analog demux chip
select inputs driven with the gate sigs
single input driven with the shunt signal
can't do that - the A/D is a successive approximation type, and takes (16 or 60) us per conversion
caps on the outputs as sample-holds
no flash samples
SWP: there's a sample and hold
in the dsPIC
* icee checks
when the gate pattern is such that the shunt output = Ia, connect the signal to the Ia cap
when the gates are such that the shunt gives you -Ic, connect it to the -Ic cap
could be in the dsPIC, I think there isn't in the regular PICs and AVRs
when a cap isn't connected, it holds the previous value
presto, 6 reconstructed current signals
swp: the AVRs have htem
true - it would be pretty close (though the uC pins aren't exactly opens
page 391 of the data sheet, there's S/H's
no fscking micro
analog mux chip, from analog devices or something
oh - OK. I thought the voltage had to be stable during the conversion
heh - you analog guys
not communicating again
I'm talking about building three (maybe 6) sample and holds
one per phase
I understand :)
the output of that will be stable during the conversion...
it's a question of how fast they drain through the micro
the pin isn't multi-megohm impedance
they're being refreshed at the PWM rate
the impedance is pretty high
pic the right cap, you're good
not gonna be +/- 2%, but better than nothing
charge injection from the analog switches will probably be more of a problem than droop
that and charge time when duty cycle is very low
I've got to get those datasheets onto this computer
and filtering switching noise off the shunt signal
you analog guys :)
all the world is analog
or just building a model based on motor speed, commanded duty cycle, and DC bus current
you just have to look fast enough, or small enough
if you look close enough, all the world may be digital ;)
i mean, we don't know whether at some level time and space are ultimately quantized
true, but very, very few people need to look that close (thank god, that shit makes my head spin)
quatum piss sicks
lol @ jmkasunich
part #emc superconducting circuits == superfluidic aether's the way to go
one naive thing i could do is keep the 'lowest' phase at any time nailed to DCBUS-, and then bus current + total positive duty cycle for the other 2 phases will be very close to Imotor
bus current / total positive duty cycle that is
yah, that comes very close to working without even any cleanup with speed
(there's still uncertainties in accuracy of phase angle, switching times, blah blah blah)
maybe i should just build a massive spice model of this
try using forward slashes in front of those commands (/) :)
sorry didn't know these were seen by others...
they aren't if you put a slash in front :)
03paul_c * 10emc2-auto/wiki/ (18 files in 14 dirs): "Auto update wiki from a cron job. Wed Nov 23 05:30:01 GMT 2005 "
SWPadnos_ is now known as SWP_Away
Jacky^afk is now known as Jacky^
anyone here read the International Herald Tribune ?
please see pag. 7 !
just getting ready to go to the shop
rewiring the cnc
just a little...for the new spindle
replacing the spindle power cable with 3 conductor shielded
and a second cable for the cooling fan
oh well, did you have interferance issue ?
upping the limit/home switches from 5v to 24
good hacking :)
Hello - last night someone gave me a link to additional g-code funtionality - http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl/emcinfo.pl?AdditionalGCodeFunctionality
- Now I am not a linux person. at all. I was looking at source forge and trying to muddle though. I am lost. first is this mod available and how would I go about installing it. Be Gentle ;)
It's just that the Colombo is bigger, and needs bigger wires
im reading the news..
[12:33:48] <Jacky^> http://www.beppegrillo.it/english.php
to get an article published we need to go aoutside of italy :/
that what is happening here :(
hi skunkworks. I'm not a linux guy either...but many on here are. I would guess you have to download the new interpreter code and recompile.
Is the economy bad over there jacky?
les_w: only debits
but its absurd
Things seem ok here. Ok for me at least. This is the best year ever for me.
Hi les_w - that is what I figure also. Just lost. It was funning as I was reading the knowledge base artical and the very first fix is a problem with g92 after an abort - I have only been playing with it for a few days and have already run into this issue. Zeroed z axis and x and y had random posisons
I think that is because I started a business.
its ok to start a business, if you have money
skunkworks, I did not have a problem with the previous behavior of g92. I do like the addition of subroutines though.
isnt ok to start business just with debits
here is happening
firt\st two years...no money.
Currently skunkworks I am more concerned with the quality of motion in emc. I have a new spindle and cannot fully use it's capabilities.
So I am kind of on the sidelines as far as the other stuff.
ma non dire scemit�
les_w - what kind of issues?
However I am a commercial user so my priorities are very different.
skunkworks, poor trajectory planning at high speeds.
On a Bridgeport or minimill or something it's fine.
at 600 IPM it's not.
Are we talking x/y/z at higher feed rates
les_w: here all are doing the capitalist with debits, no money around ..
just using public money of citizens
who want to do the capitalist should use his money
We use foreign notes to finance oil crusades ;(
what are you using - servos to go? Motenc? what servos and drivers? Just nosey
skunkworks, what machine are you using? servo or stepper?
full servo STG for me
copley 412 drivers
I am using stepper on a larger gantry - max is about 100ipm. We have a large K&T mill we would like to retrofit with servos. This is more of a learning experiance on the gantry just to see how emc works
mine is a gantry
let me get you a link
[12:47:27] <les_w> http://www.lmwatts.com/cnc.html
that page is a little dated
skunkworks, I am planning to get a larger vmc as well. Might use emc.
I was just there a few days ago. - funny. our gantry is about that size 2'X6'X4'
The gantry is light metal capable, but I need a third machine for metalworking
none yet - this is our first full 3 axis machine - the k&t was 2.5 axis. only picture is of a ball in a cage (not finished) just playing around - http://www.lmwatts.com/cnc.html
looking at a hurco with blown/old control
oops wrong past http://www.electronicsam.com/images/woodcube.jpg
are you a commercial operation?
no - not yet - just a very big hobby. Nice having the machines for my own projects. we do side projects but not serious. maybe some day.
[12:56:28] <skunkworks> http://www.electronicsam.com/images/turbo/turbo.JPG
boring out an exaust housing to fit a larger turbine wheel for may Twin turbo stealth. that is on the K&T millwakeematic. still with the 60s era descreet componant controller - germanium transisters. scary
fuuny site. Low prices per pixel.
north of lacrosse in holmen.
i'm from chicago. Flew little planes a lot near you on the way to oskosh and appleton
twin turbo stealth huh? I am car shopping right now.
I have to drive to chicago a lot.
I like the stealth - same as the mitsubishi 3000gt. Lots of power potential 300 from factory on a little 3 liter
well, I am getting old. Time to sell the BMW and get an old man car. 53.
like anything though it has issues - being awd it has a weak transfer case/output shaft. people are working on that though.
Yeah I am looking at AWD audis.
my coworker is a bmw fan. he has an 86 m5
I had an audi before...very comfortable.
drives to chicago in the 325 are not so comfortable
It doe have great handling though
tiring on a long trip.
right - my coworker used to race his at the big tracks in california. big in the bmw club
Last time, i rented a (gulp) cadilliac to go up there....and I (double gulp) liked it!
but I am not ready for quite that much an old man car to buy.
I sent my dad the link to your site - he really liked your lead screw pitch correctio
couple hundred year old technology
simple and sweet
emc can map single screws but it cannot match pairs
important on a gantry.
well, off to the shop. Nice talking skunkworks. BBL
can you annodize mirrored aluminum w/o mucking up the mirroring?
Jymmm is now known as Red70sShow
Red70sShow is now known as Jymmm
* alex_joni drops a needle
* Jymmm turns on the electric magnet
* alex_joni drops another needle
* Jymmm aims the magnet in alex_joni direction and hits the REVERSE button
* alex_joni drops
Jymmm: tried notpron?
never heard of it
[16:58:47] <alex_joni> http://deathball.net/notpron/
something to keep you busy
please guys, post your comments here: http://www.beppegrillo.it/eng/2005/11/stand_up_clean_up_the_parliame_1.html#comments
cradek: "cvs update" and tell me if axis still works for you
jepler_ is now known as jepler
cradek: I added a progress bar while loading a file, and changed to a different tabset implementation, so beat on those two things particularly.
the whole area left of the preview is now overly wide
there is extra space to the right of the slider and to the left of the labels on the F3 tab
the progress bar is nifty
I'm not sure why the tabset is wider overall.
did you take out the watch cursor?
on a huge file, there is a long pause before the progress bar starts
on a huge file, there is a long pause with no feedback before the progress bar starts
the loading goes in two phases: first, the text widget is fully populated. second, the g-code is interpreted. The progress bar covers the second step only
yes, I think the "watch" cursor got lost along the way, I'll put it back
I think it hung on Reload
did you get an error on the terminal?
any reason it would be slower? I loaded a 100k line file, then hacked off half of it and hit reload
so I would expect it to be faster than the initial load
is it spinning or idle?
#0 0x40155b39 in TkBTreeNumLines () from /usr/lib/libtk8.3.so
#1 0x401532f0 in TkBTreeInsertChars () from /usr/lib/libtk8.3.so
that probably means it's busy messing with the text widget
#2 0x4015357d in TkBTreeDeleteChars () from /usr/lib/libtk8.3.so
#3 0x40151042 in Tk_TextCmd () from /usr/lib/libtk8.3.so
#4 0x4014f8fd in Tk_TextCmd () from /usr/lib/libtk8.3.so
#5 0x4019978b in TclInvokeStringCommand () from /usr/lib/libtcl8.3.so
#6 0x401d5517 in TclExpandTokenArray () from /usr/lib/libtcl8.3.so
#7 0x401d5638 in Tcl_EvalObjv () from /usr/lib/libtcl8.3.so
is it trying to empty the widget??
I'm going to kill it and try again
reload fails this way even on a 500k file
weird. I can hit "reload" on 3D_Chips 'till the cows come home.
jepler "The cow goes 'moooooooooooooooooooooooooooooooooooooo'" =)
It must be alex_joni's bedtime
finished my massage ;)
just got back from the lumber store...
got some ironwood w/ purrty grain in this piece.
yeah.. almost the same thig
hey les...you awake?
sshhhh.. he's sleeping
well let's rattle the cage!
les_w WAKE UP YOU LAZY BUM!
dave-e hows that?
<Jymmm> les_w WAKE UP YOU LAZY BUM!
it apparently wasn't LOUD enough
Jymmm: almost didn't hear that
alex_joni: Just call me a a church mouse.
how's it up there?
cold and snowing.
huh.. frozen lake?
Yes. Almost all of it.
Seems like that happens within a couple of days of our thanksgiving holiday.
oh.. right, almost forgot about thanksgiving
Why should you remember? It's a pretty small part of the world that celebrates.
heh.. guess so :)
but we got some people up (down?) there we do business with..
Our big holiday now is hunting.
but I guess you didn't hear about titan
That shuts everything down.
They make spraying equipments
Titantool Inc. iirc
okay tell me more
[21:01:16] <alex_joni> http://www.titantool.com/
they were bought by Wagner a few years ago
I know that name.
and Wagner is the company we work for (for spraying equipment)
getting the titan site now.
not very dial-up friendly I see
Got the front page. What are they doing to your company?
we got spraying nozzles from them
as do all the spraying equipment manufacturers ;)
adjustable tips .. very nice
just got les on landline
hey dave.. you were using a STG iirc?
he has timed and doesn't have a problem....
He going to test for you.
his machine is apart right now but he will confirm in a week or so
yes alex I run with a stg
any one know emc 2 install? I followd the install tips here http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl/emcinfo.pl?BDI-4_Compile_EMC2
but when I go to run emc - it start up and then stops. The error I do see is no dispay name and no $display envorment variable amoung other errors
* alex_joni was wondering... ;)
skunkworks: you're running as root
not the user that owns the X
try running as user, and place a sudo in front of the command
I thought I did it both ways - right now I am logged under my x user
the other possibility is that you have a faulty CVS checkout, try an 'cvs up' in the emc2 dir
I actually installed it under root and my x window user
skunkworks: try installing it for the user the X server belongs to
alex_joni: I see a great need coming to update the Hal_Introduction paper.
the only thing I did as the x user wah the cvs -d:pserver........ line and on - should I start over and do it from the begining as the x user?
We need units in there for iocontrol and classicladder
and updates to halcmd
skunkworks: just login as the user, and sudo scripts/emc.run should work
the only thing is that the user won't be able to edit files, or do changes
try using : chown user:group -R * inside the emc2 dir
replace user:group to fit your needs
rayh: waiting for the devel to stabilize a bit, before doing that
or else, it would be done twice or more often
skunkworks: you're pretty near to rayh :P
at least that's what frappr shows..
well near is an understatement (over here, in europe there would be 2 countries on the way... *g*)
I think it's about 5 hours drive.
at our permitted speeds.
how much is that? (permitted speeds?)
55 mph for most side roads
65 or 70 mph on freeways.
darn slow by international or eu standards.
70 x 1.609 = 112 kmh
not THAT slow
in eu mostly 120 and 130 on highways
except germany ;) there's no speed limit there
That 112 is the fastest in the states.
I will say again - I am not a linux user. ;) Where is ray?
it was 136 kph where I grew up.
Da UP 'eh?
You're near LaCrosse?
thats you - I am bad with names - I talked to you yesterday - or the day before
right - near lacrosse in holmen
That would be a drive ;)
right next door
skunkworks: maybe you can afford rayh's hourly fees
Says 6 hours.
ok.. guys, take care
how many beers per hour?
Catch you later Alex.
gotta go, early start tomorrow (err.. today lateron)
skunkworks: usually 30 miles / gallon :D
skunkworks: Well I'm slowing down a bit in my old age!
I think I am going to reinstall the bdi 4.3 - I may have screwed it up when I tried to patch the first time. I typed in the cvs install wrong and kept right on going. I am close though
Now I can't start http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl/emcinfo.pl?BDI-4_Compile_EMC2
from the begining as the x user can I? I would assume if I am updating the kernel I need to be logged in as root
No you can "sudo" root at any time if the installer did sudo setup correctly.
That might have screwed me up then. I did everything in root then reinstalled the cvs -d:pserver..... and on in the x windows user account - if that makes sense. Did I mention I am not a linux person ;)
alex had mentioned trying chown user:group -R
How do I find out what group I am
What user name did you give when you installed?
Then the install makes a group names sam as well as a user sam.
when I type in chown sam -R is says too few arguments
If you are still root, but you want the emc2 install to be owned by sam.
Though I reinstalled emc as the user sam
Okay. Then you are all set.
nope - emc2 starts to load - then unloads itself
What directory are you in when you try to start it.
In a terminal, you can type pwd <enter> and it will tell you where it is.
^ no making fun
hold on - I am going to try a few things and get back - thanks for the pointers
hey Jymmm whatchadoinwiththembackwardsslashes?
k -- going for dinner.
rayh is now known as rayh_away
rayh_away: C: Dos, C: Dos run, Run DOS run!
skunkworks: any luck getting emc2 to run?
not yet - I just got home and as we speak almost have the bdi installed. I will then do the emc2 thing as the x user from begining to end. I will report back. thanks
ok. I just went through getting emc2 and compiling it yesterday - did it all as the x user, and it does come up and run.
I am sure I screwed something up. ;)
2nd times the charm
I started on the wrong wiki page the first time and ./configure errored out
think it assumed I knew a lot more about compiling emc than I did...
stupid question - I get to the command console by ctrl-alt-f1 (f1 to f6)?
Yes - and not so stupid - I had to try it out to see.
but I dont' think you need to do that.
you're installing the BDI-4.30, right?
and it's coming up in the KDE uh, windowing environment, I guess?
If so, there's a terminal icon down on the taskbar, next to the calculator. That will open up a command window where you can type in the commands.
it's nice to have the wiki page open next to the terminal window.
so I was using the cntl-alt-f1 as the console to install and try to run emc2
well I installed it again and got the same results - display issue
what do you mean by display issue?
well I guess you have to run it through the terminal in the x windows. started right up.
I guess I am learning
all X (graphical) programs are like that
they connect to the X server (windowing system) named DISPLAY
so if you have networked machines, you can run the program on one machine and display it on another
now I know. when I would try to run it through the cntrl-alt-f1 it would come up with something like - no display global varible
that is cool
thanks ot jtr for mentioning the terminal window in x windows
I would be still banging my head against the wall
I'm new to IRC - can you fall off the channel without knowing it?
I think I left myself wide open there...
jtr: usually an irc client will tell you when it has lost contact
or miss messages? I've been monitoring, and haven't seen any text between skunkworks reply "yes" to me, and his "so I was using..."
just a couple of users leaving and rejoining.
there was no text between those (20 mins)
just some folks coming and going
Ok, good - just looked like it was starting up in the middle of a conversation - not that I don't do that. ;)