well, 1) wouldn't it be neat if the rtapi code could actaully be severed off from emc and provided as a "platform neutral" rtapi solution? 2) do you really want the rtapi directory to depend on libnml? 3) wouldn't it be better to tear the timer, shmem, and mutex stuff out of libnml and into a convenience library that is then used by libnml and rtapi?
i don't want to duplicate code, but i'd rather do that than create a needless dependency
there already seems to be a lot of duplicated code for these functions as it is, just looking at the rtapi directory. now you're telling me that libnml has its own little api too.
1) rtapi is already independant of emc2, it is a long way from providing a neutral RT solution. 2) I am more concerned with re-using code rather than duplicating it time and time again. 3) libnml is already a library.
libnml predates rtapi.
Oh, and there are already "standard" methods/calls for implementing much of what rtapi does.
so why isn't emc using such?
but JMK wanted to provide policy and method for anyone wanting to use it.
regarding 3) and 3a), are you saying you don't want to touch libnml?
I'm currently doing some work on libnml.
okay, so you don't want *me* to touch it, that for sure ;)
is there any reason that libnml can not be rewritten (someday) to use rtapi?
There is a hell of a good reason that libnml will NOT use rtapi
because a) libnml looks to be in C++, right? b) rtapi is in C, c) rtapi's scope is much more focused on threading/IPC/etc
libnml could provide a wrapper around C calls, if needed... but i suppose you will explain why that isn't going to happen
rtapi provides both policy and method - I find tha offensive from a library
well, what if i said i'd rewrite it to give you "pluggable policy personalities"?
the services that libnml require are provided by libc.
* zwelch will need you to provide a little more detail about what "policy" vs "method" details you want separated
so will the sim implementation of rtapi, right?
library calls provide "methods" to perform a task.
(pthreads is libpthread, but it's now part of glibc...)
Policy is all about restrictions of those methods.
what restrictions do you feel are imposed currently?
most of what i've seen seems to be a bridge abstraction for implementing an API that gets implemented by several different underlying APIs
it goes further than just rtapi, so let's not rake it over again.
i guess what i'm offering to do is provide an even lower level access to the rtapi functions
something that replaces the equivalent code in libnml in C, and can be used by the higher level policy functions of rtapi
rtapi could have been implemented with a few simple macros and header files.
Good evening folks!
well... That would get rid of the policy.
I still fail to see what rtapi could possibly provide that libc doesn't already do.
as far as libnml is concerned.
Does anyone know where information can be found that explains what type of linear encoders and interface hardware work with Emc2?
Theres just so many available i don't know what to ask for!
You're pretty much limited to digital quadrature encoders (or scales)
rules out the 11uA and 1V p-to-p scales.
Hi paul, no analog eh!
Not unless you have cards going spare.
Paul, not sure what you mean?
as for interface hardware, would suggest looking in the hal/drivers tree and see what is in there.
Ok I take a look, will that determine what encoders i could buy?
Paul, what about incremental encoders with TTl output?
another way of saying quadrature encoder
digital output covers differential & TTL
Most of the USDigital stuff is suitable.
ok, this won't be an easy thing to do by the sounds of this?
for the current status of the hal drivers, bug JMK.
does jmk ever come on irc?
Hi Paul, Us Digitals 2 versions of linear encoders are only 1" to 2". Not much use!
Paul I'll look there!
It's only when you start looking at their <4um scales that the prices start looking good.
Paul they have two versions SHG and MHG, which one would you reccomend?
MHG are very small if you are limited on space.
Paul, are they basically the same and I can make my decision based on price?
paul_c: i also wanted to ask you about the bdi-4.27 branch; is that your work tree?
one sec - got to update a package on the server.
* zwelch likes to know where the other active LOD are, so he knows who to bug before making potentially conflicting changes
bdi-4.27 is a tag for the bdi-4.2.x releases.
Paul sorry for all the questions but this will be a pretty major expense for me and I want to know what I buy will work!
You want to get some quotes for the Newall stuff - It isn't cheap.
What app is it for ?
I want to use it on my mill
do you need 1um resolution ?
what is that?
I'd be happy with 0.0005
I'm a Canadian that doesn't do metric
a 20um scale in quadrature will give you a resolution of ~0.00025"
Thats plenty fine enough!
There's an outfit over in California doing cheap glass scales...
Paul, Newwell has two signal converters SCC-100 and SCC-200, will they interface with Emc2?
What do I have to look for then?
[01:06:08] <paul_c> http://www.usdigital.com/products/pci4e/
(no emc2 driver) http://www.vitalsystem.com/web/motion/motion.php
(maybe a driver soon)
Ok Thanks!! you've been a great help!
Bye for now! have a great evening.
* Jymmm pokes paul_c with a 400KV cattle prod
* paul_c looks to see who has ops.
whats happen paul_c?
got someone from the RTAI team bugging me for a contribution...
financial or development?
tell em when it can handle 100KHz+, you'll do it
not even going to get 10KHz out of fusion.
huh? 90IPM is about 27KHz
38KHz @ 1um res.
I'm not doing 1um =)
and 2.28m/min is pretty slow.
damned begging letters..
move it to the spam folder for scanning later.
Ok, gots me my foil tape!
10yards should be enough
* zwelch just got sim_rtapi.c to compile under the new API, but now faces the same task for sim_utapi.c
When I was at the electronics swapmeet, I found a plastic bender heating element. You mount it to a plank, but the instructions call for two layers of foil tape as a thermal shield
this stuff is only 2" wide, so I'll have to double up on the width.
zwelch: I would expect rt & usr sim to be pretty much the same code.
paul_c: the amount of code duplication for task/mutex/shmem handling is rather staggering
so don't duplicate then.
nope. in fact, i've already written some code that can be re-used by all of the rtapi implementations to eliminate a bunch of common code
that should be interesting....
well, i don't plan to do anything outside of sim* for now
the code can be re-used, but i'll coordinate first
right now, i've got past the ulapi.c problems by a phenomenal amount of code use; the file now reads... '#include "rtapi.c"' :)
erm... add sim- to that where appropriate ;)
unfortunately, i've now realized that my Makefile hacks for building the sim need to be added to the rtapi subdirs too (e.g. examples)
development... it's like a big text adventure game. fight past one bug only to be faced by the next one :)
zwelch: I haven't been paying much attention to this conversation, but I am interested in being able to run emc2 in simulator mode, to make sure AXIS works there. Will you post to emc-developers when there's something I can use?
"works there" means "works in emc2"
i'll make a post as soon as i have something that works (in any reasonable capacity) for me
#include "foo.c" is an ugly, evil hack for avoiding linking (in my opinion) and is something that should be avoided for all but exceptional cases.
i.e. it must at least compile :)
paul_c: it's a hack ;)
i don't plan to commit that change
i know what i have to do there, so i'm just looking ahead to see what other problems were lurking
eventally, the thing will actually finish building :/
well, putting source in a header file is common technique when you need inlining
as not many linkers can inline code (and gnu ld isn't one of them)
Suggestion: commit little & often - there may be others that will spot bugs that you might have missed.
i know this is a naive question, but if you need higher resolution, why not offload the realtime stepping control to a microcontroller?
i think the short answer is, it would have to do everything the main CPU is doing
of course, the real question is: if you need better resolution than what a general purpose processor can provide, why aren't you looking at existing (e.g. DSP-based) solutions? certainly, it would be cheaper to buy one of those than designing a RT "EMC offload engine" ;)
it has somehting to do with being convinced that I, or we, can do better than these existing proprietary solutoins
the winner usually has the most resources, since translating ideas into working code takes time. closed source shops making money off their products can afford to pay programmers full-time to improve them (so they sell more, grow, etc.)
i don't doubt open source is the better model, but i think "we win" only when we are able to produce more, faster, and for less.
* icee shrugs
I don't always know open source is good for the industry
producing free/open solutions does have a tendency to lower market prices and commoditize things
that said, i believe the community needs to find funding to "win"... i.e. it has to compete in the marketplace, in order to acquire the resources required to become and stay better
why? It's providing the tools for machine tool vendors to backwards integrate and stop licensing control components
once it reaches a critical mass it'll be cheaper for them to integrate in what they want, rather than license.
i agree, but my point is that "critical mass" is not fixed; it is largely dictated by existing market leaders
that's not true
critical mass is dictated by the use cases of the end users
only to the extent that the market leaders can create and foster new use cases is that true
Anyways, I think the whole idea of 'winning' is superfluous
i can scratch my own itch with emc; i don't care if emc takes over the industry
in the context of the market, "winning" simply means "becoming a market leader"
to the extent that it stays well supported and maintained it's in my interest
the best way to keep an open source project vital is by having a large userbase
a large userbase? more like a large number of developers using it.
and in a vertical market, that means taking marketshare from existing players, which means competing with them for their users
well, once there is a large user base, then there is demand for developers... who can supply their services at the going market rate
perhaps, perhaps not.
* zwelch shrugs
there's friction in the market that prevents end users of solutions like this from directly employing developers to solve problems
e.g. they don't have the expertise to spec and manage the project and to hire well.
then that becomes the developer's responsibility
I think the model you're looking at isn't how open source has been adopted in vertical markets
What happens is a die hard cadre of a few users builds a solution to solve a hobbyist problem, and the project grows attracting only technical users; then, someone who is licensing IP realizes they can instead just use the open source solution; they have experience developing and integrating so they can manage a project and integrate it into their product.. and that's how you get critical mass.
e.g. what uclinux has done in the low ends of the rtos space, or the high end server market with linux, etc.
and to an extent what we see happening with emc and sherline and a few other low end tool vendors
or what i see starting to happen with mythtv in my exposure to semiconductors
so you don't want to actually do anything, just sit around and wait for this magical "critical mass" to just happen, eh?
you don't think there's any room for "grabbing the bull by the horns", so to speak?
what difference does it make to me whether EMC reaches that critical mass?
it's not worth me spending my development time to fix problems that i don't have
i guess my question was along the lines of: i already have to buy substantial hardware to get this going (pc, steppers, stepper drivers) .. why not another $100 for a little micro and logic board to offload the hard real time stuff
mrallen: because rtlinux is already pretty good for the step rates you get on a small installation?
i have done a bit of micro development (wrote my own rt scheduler for Atmel AVR) in assembly
the parallel port under rtlinux is good to over a hundred kilohertz
i could do an open source AVR driver + schematics that could put all the hard real time stuff on a 3x2 board
on the AVR, that gives you less than 16 cycles
to choose what you're going to do and stick it out GPIO to match the parallel port.
er. my bad. math is hard ;)
200 cycles on 20mhz versions
and you could maybe use the timer counters but that would get tricky
if you wanted to do this, why not use something like tms320f2810?
because i don't know it :)
well, that's what the commercial solutions use, and for good reason
unless you're going to do the math on accelerating axes, etc, in the realtime controller, it's not a big win.. and AVR isn't studly enough for that at reasonable step rates
and really, maybe AT91 would be.. but use the right tool for the job.
i wouldn't imagine offloading everything
mrallen: what real-world performance advantage would you hope to get?
but set up a com channel feeding pre-digested bits to the micro ... it would do make sure it executed at hte
mrallen: the problem is that the "bits" take into account feedback from the system (right?)
icee: really don't know. i'm not familiar enough with the low levels of emc
rtlinux is pretty good, and the parallel port is acceptable.
you can't just generate a bit stream and deliver it in real-time; you may have to be re-calculating things based on feedback from the system - in real-time
zwelch: as i said, i have a naive view of this. i picked up on the comments above from paul that it couldn't move fast enough due to RT limitations
right, and to use EMC "offloading", you'd have to port EMC to the actual target embedded hardware and run rtlinux on it :)
thanks for the clarifications, though.
zwelch: that would suck
then you just run emc over ssh over usb ;)
i.e. run emc on a gumstix
then the gumstix is the RT computer
and a USB peripheral
but you'd have to build a board to drive the hardware that interfaces with it
exactly what limitations was paul speaking of? because i didn't see it
* zwelch didn't either
he didn't exactly answer my question
the only thing i remember was... "gotta tune stuff" or something
vague, yet unhelpful :)
[03:16:38] <icee> http://pico-systems.com/univstep.html
if you really want a microconrtoller board.
looks like they went with a FPGA to implement it
* Jacky^ ywnssssss
jcolley is now known as siado
is it possible to use dual loop feedback using linear encoders on slide position and rotary encoders on the servo shaft with the Vital card?
icee: incidentally, I did a quad step+dir to phase drive (is this the right term? 4 outputs, 1 or 2 on at a time) on 16MHz AVR. It was fast enough to 48000 steps per second per axis, but I'm not sure how much faster than that it would have gone.
er, I probably mean 9600 steps per second per axis
this is atmega32, triggering interrupts on each DIR pulse (you can configure the timer inputs to act like external interrupts)
atmega16 would have worked too
but I just don't think there's time to do anything much more sophisticated than that
jepler: what KHz would that be?
[21:49:33] <ValarQ> morning folks =)
06:49 < Jymmm> [21:49:33] <ValarQ> morning folks =)
Allthese huricanes are bugging me...
were do you live?
Calif, just the principal. FEMA says NY, New Orleans, then San Francisco, which I'm 40 minutes from.
and that was in 2004
It's been almost 100 years, so a big earthquake to brewing.
FEAM == Federal Emergency Management Adminitration
The federal agency that does disaster mgnt stuff
I have some disaster suppplies, but as much as I would feel comfortable with.
so i actually just managed to get emc2 to build without RT bits
however, i almost immediately discovered they don't work :)
* zwelch tosses those bits in the round filing bucket and begins to stalk his elusive prey: "buggus rutimeous"
zwelch: PUT DOWN THE REMOTE AND STEP AWAY FROM THE TV!
paul_c: got a sec?
hm, I have a question about jog wheel
Ask, and I'll see if I can answer it.
would it be possible to modify the USC driver so that I could use one of the counters as input for a jog wheel
I don't see why not.
and, how well does the bdi version of emc2 with the usc driver work?
Wouldn't limit it to just te USC... Make the same change on the other drivers, and you'd have jog wheels on all of them.
maybe I should have a look at it
being without jog wheel is a hassl
I don't have any of Jon's hardware... All I can say is the driver works.
just look at the recent emc-user postings
to see if it works?
I am not too happy with emc1..
so it's probably lots better
confirms USC works for one user at least.
the fact that it uses the max accel for rapid moves on all axis:es..
my stepper on z goes whirr when I move it..
there is no USC driver for emc2
didnt you say there were before?
in the bdi version?
JMK started on the framework of one at the codeFest, but that was as far as it got.
the BDI-4 drivers are EMC1.
well, the usc works great..
emc does not.. :/
you think emc2 is going to be any better ?
paul_c: remember that I've been making lots of parts with emc2..
emc1 would work good if it werent for the fact that it dosent limit the accel of my Z for some reason
weren't you having "issues" with the tkemc buttons in emc2 ?
in emc1 also..
if I remember it right..
hi anonimasu :)
how's things going?
here, rain ..
looking for inexpensive servo driver
i found an interesting chips
no, not gecko
what's wrong with geckos?
and the controller needed for encoders etc..
you dont need any controller with geckos
you feed them step & dir just as you would with stepper drivers
where the encoders are connected ?
into the gecko..
to the driver ?
they have the pid loop internally
total control to the driver, minimal to emc so ..
well price/performancewise you wont beat them..
i'm looking at pic-servo
theres an interesting chip
if you feel like building something build a pulse generator
for using with geckos :)
but an rs232 controller is needed
btw, why are you looking at servos when you cant afford cheap servodrives?
or are you going to ignore inertia and just use cheap dc motors?
for 2 reasons
first, i want to start to play around servomotor
you know good encoders are $150..
or well ok ones..
without spend a lot of money
second, my cousin is interesting in build a machine for his plasma cutter
anonimasu: i found some nice encoder for just $ 19
how nice is nice`?
mmhh cant remember the resolution
but it could be enough for me
searhcink the link ..
[13:35:47] <Jacky^> http://www.usdigital.com/products/e4p/
# 120 to 300 cycles per revolution (CPR)
# 480 to 1200 pulses per revolution (PPR)
2 channel quadrature TTL squarewave outputs
i cant start buyng encoders of $ 150 each :\
also, ill use surplus motors ..
i hope done it with about E. 350-400
but i need to check for shipping cost, tax, etc ..
before to buy
g101 is selling now
how much ?
and 135$ for the breakout board
20 A 80 V ?
you still need drives.
a g340 costs 148$
if understood ok, they are just amp ?
what other is needed ?
but it depends if you want to use software generated pulses..
or hardware generated like from a usc
i would like to use emc, of course
paul_c: sayd servos module has step, dir and encoder quadrature control
when you use geckos you arent using the servo module, for emc but stepgen..
maybe also PID could be controlled by emc
but the geckos takes care of the pid loop for each servo..
it working stepper-like whit step-dir
the driver do all the rest
so.. the jecko can be connected to the parport
like normal stepper drives..
i want to calculate anything now
about thefinal cost
if you need more speed you can get a usc..
if the difference isnt too much
or build a pulse generator
148*2 for the x&y axis..
id be inclined for gecko
you wont need a servo for the Z..
there are a cheaper ver g320 but you wont have the pulse multiplier there..
and you defenetively want it :)
ive seen something
and motors + encoders
ok.. i must go now
i'm going with a friend to help him in the first dsl installation ..
aefully quiet in here this morning
got a replacement part yet?
arent you anxious to start to cut parts?
"It's very difficult to soar with the Eagles when you're working with turkeys"
I've been looking at some flashlights the last couple weeks... big variety out there
Mostly LED based, trying to avoid the specialized batteries
I found a single AA one with 10 hour battery life, but the reviews say the beam shape is not very desirable.
$20 USD http://www.flashlightreviews.com/reviews/inova_x1.htm
This one is on sale for $40 USD, but uses litium batteries and only a 1-2hr life. http://www.flashlightreviews.com/reviews/inova_xo3.htm
paul_c: on 4.29, AXIS works for us, the problems we attributed to the FP bug are gone. http://axis.unpy.net/index.cgi/01127491762
Jymmm: did you ever find out if updating emc fixed your problems with axis crashing
bdi 4.29 seems to fix the problems we were seeing with axis back in 4.25
* alex_joni recalls reading smthg like that in the logs
glad to hear it's ok now
catch you later
jepler: Yes it did fix the segfault. I was also able to lick up the system too =)
you mean, it locks up with 4.29, or it did lock up with an older release?
Well, I was never able to use axis at all previously. After I updated the kernel and what not and got axis up and going. I was mucking around with axis and locked up the system. Finally was able to exit axis, but couldn't even kill -9 the background emc tasks that were running.
jepler I did have a question.... Does axis "falback" on the gl stuff if it can't keep up on older video HW ?
what the advanage of a pulse multiplier on a servodrive ?
i'm looking at gecko G340 ..
it allows you to scale down high res encoders
so you get more speed..
i'm a little confused about the encoders..
it seem geckodrive has internal control fo encoders right ?
you are not involving emc in the pid loop at all
going away to a friend
so, emc will not check the quadrature encoder ..
Jacky^: I forget, what is the travel on your machine?
actually is about 30 cm
but i'm asking for drivers to use on a new machine
not this i've now..
what abut encoders connection to emc ?
if drive like gecko are used it seem no coonecion needed
but how can emc know if an error occour ?
there is an error pin on the gecko
you can use that as estop in to emc
emc, in this case cant do nothing
but best would be to have the encoder conected to emc too
* alex_joni tried that
error pin ?, i've seen
but if you want a proper solution
this pin should be connect to parport ?
use an USC
like motenc ?
iirc the USC has the possibility to take care of encoders properly, even if the gecko errors
the one JonE sells
never seen ..
got a link ?
[20:38:23] <alex_joni> http://pico-systems.com/univstep.html
with this board emc can take the control on quadrature encoders i suppose
one small problem ;)
no emc2 driver yet
alex_joni: how's things?
going home tomorrow
anonimasu: how about you?
not too great
havent been to work today
or.. sick of it ;)
woke up and couldnt get out of bed.
too bad :(
btw, I'm coming to sweden ;)
yeah you told me about that earlier
will be the 8th country in 2 months :))
alex_joni Just as long as you stay in .eu!
I usually am outside ;)
so I don't actually care
alex_joni: No, stay in Europe that is!
Jymmm: where else should I go?
nothing interesting there
alex_joni: if not USC used, G320 can work fine the same on emc ?
alex_joni Can you feel the love?
Jymmm: melting my heart
Hi Alex, hi all
what's new Martin?
hm, still working on the breakout boards for the mesa card
the io board ready, and waits for the components
i hope to finish teh DAC and encoder board as well this weekend
5i20 pci anythin io card
are you programming it yourself?
first i need the hardware
they're 40 minutes from me.
8 hours i think
what pulerate can you get out of it and how many axis?
pulse rate (KHz)
that's PR ;)
i want to controll servos, so i'm away from this puls rate question
we are two gys here in germany working on that suff
we are making some breakoutboards, to read encoders, and to give out a +-10V signal for the servos, and for the io stuff
personal or production use?
i need it for a machine, that one is for production
ok, when you get 99.999999% ready to release, send over 10,000 cases and I'll beta test them for you =)
or i send you the files, then you can make it by ureself
no, not will never do...
I have to maintain my lazy status =)
maybe we will sell some, but that is not our focus
though, you could get some boards made and sell tham as a kit.
screw that, seel them. I'm sure more than one person out would like an alternative ready to go.
what is a DRO ?
you could sell just the boards, and let them buy/stuff the components.
Digital Read Out
Imperator_ shipping would be easy too for a flat PCB
alex_joni: this USC is veri interesting, i like hoe it work :D
the problem is that if we sell them we must support the cusomors, so maybe it is better to sell them only to some well known people
Imperator_, good afternoon
Jacky^: ask anonimasu, he has one iirc
Imperator_, I am very intersted in a bare pcb if you are having some made
Imperator_ no it doesn't.... just do like everyone else does... yahoo groups for (free) support.
Imperator_: See, your first customer.
Jymmm: if we sell only pcbs to everyone we will have only trouble, i think
Imperator_ gimme my 10% finders fee bitch!
Imperator_ Not if you document it well.
it's just a breakout board (passive), isn't it?
we will set the stuff under GPL if it is finished
then np. If someone is using an anything card, they should have the technical skill to solder header connectors.
maybe we wil sell some, we will see
Imperator_, to clock the signals into the dacs, I assume you are planning an emc 2 fpga mod?
how do you mean that Lawrence ?
how are you planning to write the drive values to the dacs?
that are 16bit words
its a quad DAC DAC7744
yes... but are you using the fpga to send parallel or serial data to the dac?
ther is a other one with a serial interface
DAC7743 i think
ok... no problem... I assumed you were using a high speed serial interface with "16 bit uarts" in the fpga
I'll take another look at the interface on the data sheet
parallel is much more simple
I was thinking one could reuse the pwm and dir lines and do an fpga config so that a single 16 bit wirte to the fpga would cause the data to be sent to the appropriate dac registers
Imperator_, can you please send me an updated eagle project? (ve7it at shaw dot ca)
pc_op` is now known as pc_op
maybe we go one step ahead and let the FPGA do the PID stuff
the fpga does have one config that includes a processor option that does do complete servo in th fpga
I would be very happy to let emc do the servo (easier config changes) and just have hardware registers that one read a 32bit position and wrote a 16 bit drive word
ah, i got the message that the IO board is working now
are you well?
50 % well ..
the rest is baaaad
another driver burned
stopped again ..
i'm right thinking it
what is this gecko thingie you folks always talk about?
im thinking the G340 ..
geckodrive.com maker of cheap and good stepper drives
I know ..
Jacky^: servos? just use emc and do it properly
i also seen the USC by picosystem
robin_sz: cheap and good?
i like it
ValarQ: geckos? yeah cheap and very good
i dont know how accurate will be the encoders control by gecko
Jacky^: yeah, if you wanna drive steppers with emc, you'll need that
no, not stepper
i'm thinking servos
im tired with stepper
the only test i will do, if ive time.., will be to addencoders to stepper
as we talked some mouth ago ..
but i want to use DC motors
i wonder what happening ,, if a big error on quadrature occour on gecko
if undestood well, it will go in fault
if you end up that far away your machine is _wron_
I've never had a gecko trip when I didnt accel it way too much
Jacky^: for plasma?
this.. would replicate the same problem ive with the steppers now ..
robin_sz: yes, plasma cutter
Jacky^: no, the pid loop would catch up..
if you get any deviation..
Jacky^: if you get larger deviation your motors are too small..
robin_sz: to cut flanges
or there's something in the way..
Jacky^: how big is the machine you want to build?
1m x 1.5m?
1.5 mt should be enough
1x1 is steppers in my opnion
to remain in a low cost machine
robin_sz: I know ..
what are you going to do for torch height control?
but my cousin absloute dont want to use steppers..
why is that?
robin_sz: i dont know what he exactly want do ..
the oly thing he sayd he dont like steppers
Jacky^: larger steppers is the solution to the missed step problem.
well, you want my advice?
leave him to get on with it. ask him to give you a phone call when its finished
anonimasu: i need very high speed for woodwirking
too many pass at high speed
Jacky^: machine rigitity.
Jacky^: define high speed
robin_sz: he build amachine 3 years ago ..
to mill steel plate
with a router?
5 cm thickness
no motor.. 380 V
he used DC motor with encoder
and some primitive driver
ah look, we did this conversation last week ... a small 1mx1m plasma is perfect for steppers. no missed steps
he was only swicthing from 12 V to 5 V the motors
are you aware that 12v motors are toy motors..
I know ..
using servos will give you many more problems on a plasma
suitable for rc cars..
probably new nema 34 are ok
unless they are 12v 20a
and good drives
but he dont like it steppers..
well, let him build it then
i need to start building a new machine
not the toy ive now ..
Jacky^: what machine is built to mill 5cm plate?
is it a router?
anonimasu: i know he used a big AC motor
does he know *anyhting* about building plasma with servo? ...
Jacky^: was it a router?
anonimasu: I suppose so ..
Jacky^: you know how servo works, with an encoder?
robin_sz: i've an half idea..
Jacky^: you know how a plasma torch is ignited with about 2Kv of RF?
robin_sz: nope.. i' dont know anything about plasma cutter ..
he only sayd it work with high voltage and air ..
he sayd have seen these cutter working
well, to ignite the arc, they use about 2kv of RF ... it has two uses ... one is to ignite the arc, the second is to destroy any electronics it can find
they generate LOTS OF RF interference..
they are a bit better, but the cheap/older plasmas kills stuff even faster
you mean this can disturb the regular work of drivers ?
if you can afford to build it two or 3 machines to find out how to get it right (like I did) then expect to kill your servo card at least twice
oh shit yes
so ... feel free "not to like steppers" .. but having built and sold more than 20 tables ... that would be my choice
good to know ..
steppers could be better for the interference ?
Jacky^: yeah, because the drivers can be near the PC in a shielded cabinet, no need to take sensitve encoders up to the table
of course, with the right shielding it is no problem
but .. expect to kill a couple of encoders or cards getting the problems sorted
robin_sz: thanks :)
for my case, instead, should be ok
at least.. if i will get high speed
Jacky^: have you looked at servo prices?
Jacky^: how fast do you want to go?
anonimasu: how much is possible ..
woodworking required high speed
this is my issue :(
the limit is how much $ you are willing to throw at the problem..
and i want a machine dont lost steps ..
you shouldnt be losing steps unless your steppers are too small..
well, my router managed 600ipm
i assumed no limit to start
my budget is not infinite
i'm an obbist
i can afford 1-2 k euro
for electronic only ..
so, forget it then. if you cannot afford 10 eur for 4 micro witches, you will not affrod the machine
i can only say, i can choose
betwwen gecko G340, encoders etc ..
or a good steppers system
uh huh ..
3 axis machine
the issue also comes down to what servos you have..
i dont have any servos at the moment
nema23 servos are a bit more then 400eur each..
for a router, wither is a good choice I guess .. but at 1m x 1m steppers might out perform servos for same price
you can use dc motors with encoders..
my cousin sayd he can found a lot in north italy at low cost
but the motors will be adeguate to the drivers..
servos are nice. VERY nice ..
but expensive ...
robin_sz: what you mean as servos ?
i mean DC motor with encoders
well, depends on what motor
servos are dc motors with encoders that has low inertia rotors..
or well motors.. that's optimized..
and bi-directional brushes
or better still, AC servos :)
but the G340 can drive it ?
i think not .. just DC right ?
dont worry about AC motors..
AC is not in your budget
one drive retails at 1/3 of the budget..
the cheap ones.
1500 an axis for drive and motor (ac)
but ... whatever you use, use limit switches. cheap microswitch is fine
this make non sense for my need ..
2 eur each
the switches ?
being able to reference the machine is important
for me its quite clean
i want to start a job and finish it without lost steps, and at high speed
but i work only on wood
for my cousin is very different..
i will say him about what we talked
your cousin needs less speed..
and also less resolution
and a small 1mx1m machine is steppers really
gear for 1 turn of motor = 10mm movement and it is correct
robin_sz: for the plasma cutter ?
use about 3 to 5 Nm steppers
for router use 5mm per turn
he already has some leadscrew ..
but i cant remember which diameter
in the old machine
its not ok leadscrew ?
you cannot use leadscrew with plasma
for plasma ?
you seen the dust it produces?
very VERY abbrasive dust?
never seen in action ..
it gets EVERYWHERE
mount with rack facing down
robin_sz: you can do preload with a cheap rack, isnt that right?
this is plasma ... its going to be smelly and nasty .. and covered in crap anyway
plasma wanders about anyway
* Jacky^ googling for some samples of toothed belt
look for AT series ..
either AT5 or AT5L if you can fford a bit more
suggest 25mm or 35mm wide
2mm wil be fine on 1mx1m machine
big machine use 50mm, AT10
but ive not understood as they are mounted on a machine :(
i'm seraching for some picture
if you are very cheap use the T5 series belts, but there is backlash of about 0.5mm
yeah, it couldnt a prob ..