good night all
cradek: what's the next project?
hello, did somebody try to run steppers through a USB->Parallell adapter cable?
probably not going to work
cool stuff: http://www.theregister.co.uk/2008/12/05/fridge_sized_raygun/
you always have the cool toys Alex
hmm.. didn't spot this yet: http://www.python.org/download/releases/3.0/
wow that changes everything
[16:04:12] <alex_joni> http://pastebin.ca/1278824
I am trying to get to the logitech thread on gmane.. http://news.gmane.org/find-root.php?message_id=%3cE1L8xaD%2d0006Vc%2d30%40h25xhf1.ch3.sourceforge.com%3e
did this show up in the logitech thread? http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?A_New_Approach_For_Using_Joypads_With_EMC2
I seem to be having issues with the list. (my issues)
skunkworks: oddly that thread is missing from gmane
not sure if that page was in the thread, but the data there seems like it
I sent an email to the list - from my web interface - has not shown up though
I think this has a good bunch of info also http://www.linuxcnc.org/irc/irc.freenode.net:6667/emc/2008-08-22.txt
Roguish_ is now known as Roguish
[19:00:02] <alex_joni> http://dsplabs.utt.ro/~juve/blog/index.cgi/photography/01228674766
awww you didn't cnc it?
(looks nice though)
too small cnc
and manual adds more reward value :D
skunkworks: both your emails made it to the list
[19:51:29] <dmess> http://www.pastebin.ca/1278998
does this look workable???
has any one used sercos servo protocol at all with emc yet??
rs232 serial and usb serial are both not going to work with emc2 today. http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Emc2HardwareDesign#USB
for position control, anyway -- for setting spindle speed and not in css mode, it's a different story
jepler: I wonder if the USB protocol has an instruction to not poll the (all/some of the) bus. In essence, making (at least) one usb address synchronous
I still feel ethernet is the way to go though =)
JymmmEMC: in any case, what it needs is a dedicated programmer with the right expertise
sercos might be doable
especially SERCOS-III (which is using ethernet links)
jepler: If you know, what is sent out of the paraport?
logic zero and logic 1
on several pins
[20:22:31] <alex_joni> http://ethernet.industrial-networking.com/articles/articledisplay.asp?id=942
as a stream, or byte?
byte for step/dir itnerface
as pin states. it's not data, it's logic signals
as used for CNC anyway
but the parport is also used in EPP mode for connecting a pluto or a mesa 7i<mumble>
yes JymmmEMC, it's like having 12 or 13 fast switches under program control
light switch types, not networking :)
so basicaly a square wace
the nice thing about parport and PCI is that you write a value to some memory location or I/O port, and the outputs change
you don't have to make up a packet of information, checksum it, tell the controller to send it somewhere, wait for a response, and make sure it's a valid response
SWPadnos: does 10Mbs have even timing variance?
controller being the USB / ethernet / firewire / SATA controller
SWPadnos: so it's just a matter of latency at that point?
not very much - at the bare metal level, ethernet is a lot like a very fast serial port
there are several issues
well, would 100Mbps help?
it's faster, but just as complex
SWPadnos: but is it better?
it's a little like an elevator instead of an escalator
the parallel port is an escalator - any time you want to send some data upstairs, just dump it on the bottom step
SWPadnos: I'm just trying to eleiminate won't won't work
it can work, it just doesn't right now
it's more complex than the systems we have now, and it may repreetn a significant change in how EMC2 splits up the work of motion control
SWPadnos: For testing a single stepper motor, the 10MBps ehternet chips are cheap and abundant
But if 10Mbps may not cut it, jump straight to 100/1000
I think everyone agrees that an ethernet-interfaced control thingie would be nice, it's just not simple, and there are other options that already exist and are probably less expensive (like PCI cards)
SWPadnos: well, ethernet itself is cheap and everywhere, Paraport are rare, and PCI are expensive.
how inexpensive do you think the ethernet-connected controller will be?
a Mesa PCI card is $200, plus a bit for the I/O interfaces - maybe $350 total
ethernet in --> step/dir -or- cw/ccw out
uh-huh. like I said, how cheap do you expect that to be?
SWPadnos: Well, I'll tell you if you answer my first question
I don't care if it's 1.5Mbit/sec arcnet
the data rate is relative inconsequential
SWPadnos: come on, I'm being serious. I really dont know the timing requirements and if 10Mbps ethernet with 40% overhead will cut it, or just jump to 100mbps
JymmmEMC: you can't send 100MBps or 1000MBps by direct access
ignore the data rate, and assume that any ethernet chip is enough
alex_joni: Yes, I know.
that kind of throughput works only with DMA (no processor involved)
we don't need that
dmess has any one used sercos servo protocol at all with emc yet??
not that I'm aware of anyway
we need something that sends some velocity prescriptions, reads back some status, etc
and all should be around 0.2 msec
I have Indramat servos whitch intakes sercos
An I have Indramat PPC contoller
[20:35:02] <JymmmEMC> http://www.embeddedethernet.com/
that usis sercos
and are you goint to use EMC and make a driver??
I ment motion controller
that has RTOS in it too
JymmmEMC: there are countless solutions out there
all that's needed is for people to stop talking about it and actually doing it ;)
but it seems there's few people who actually want to do that
I think the way for this to have any chance of success is for someone to do it who has something to gain of it
that's the most likely scenario :)
either they want to connect some hardware they already have, or they want to sell some hardware that works with emc2
I see a number of possible solution scenarios
a close friend is starting a gantry and im tying to sway him away from mach3
1. there are small embedded boards out there (running linux+rtai on ARM) that could run some part of emc2 (maybe even the motion controller)
This kind of http://www.boschrexroth.nl/country_units/america/united_states/en/products/brc/ProductDocumentation/CurrentProducts/Controls/PPC-R/PPC-R_PR03.pdf
model is PPC-R01.2
2. there are small embedded boards that can run eth and generate step/dir signals
3. use special sercos/ethercat/whatever hardware
RTnet has been proven workable with emc2 in the past
sercos is overkill for a milling machine IMHO
er.. ethercat i mean
[20:42:15] <JymmmEMC> http://www.beyondlogic.org/etherip/ip.htm
looking at it another way, there are several control layers at which you can split the workload
1) actual motor drive
2) position/velocity/torque commands to motor drive
oops, forgot 1a and 1b) physical input/output
2a) encoder input or other "higher level" inputs
2 doesn't need to be to a motor drive
3) multi-axis position planning
but yeah, I agree with the general structure
JymmmEMC: realtime is too overloaded in modern PC literature
SWPadnos: in ref to http://www.uclinux.org/ucsimm/
4) path planning and other higher level control aspects
(like multi-axis gearing)
5) all but GUI
JymmmEMC, I know what realtime is, and that there are 1000 implementations of it
and some even use ethernet or USB or firewire or (name your interface)
and 1000 different things it means
make it all irrelevant, get it working on an ARM board ;)
I'm not talking about offloading from EMC, just the physical connections from PC to devices
I've even used ecos and ELKS
replacement for paraport basically
for example a realtime control system doing fly-by-wire in an airplane is totally different than a linux patched with realtime extensions so that your video plugins work faster
JymmmEMC, you have to understand what kinds of data there are, and where they need to be processed, to be able to discuss this at any level beyond "it would be great" ...
JymmmEMC: depending on the comm solution chosen, you need to offload some from emc
specifically what kinds of data and processing are needed in EMC2, that is
if you want to transfer each transition (like on the parport) you won't find a protocol/bus that works
you basicly need packets each 20 usec or so
and you need some feedback they arrived properly, etc
alex_joni: paraport is in Kbps, ehternet is in Mbps, I don't see why emc has to offload anything in that respect
data rate has nothing to do with this
it doesn't matter if it sends 100000TBps
if it sens one packet (10000TB large) every 0.5 seconds, it's useless
and does it?
you can probably make a driver that uses ethernet (with raw access to the eth chip) that sends each transition for steps
the idea is that you will have lots of packages with a LOT of overhead
<- that's the full size image of that scope plot. note the timebase
and could it send out for three stepper motos as example?
250 uS per division, so they're seeing >500 uS jitter
one eth frame is 64 to 1500 bytes long
so you send 2 bytes (needed) and the rest is not used
not really optimal
on the other end you still need some smart device, which gets the package, returns an Ack/whatever, and puts the data on output pins
alex_joni: But, you're talking networking comms, not wher ethe end point will be a motor or ADC
JymmmEMC: yeah, but comms is kinda related to endpoint too
my point is that if you still have to have the smart endpoint, why not send velocity requests at the servo rate
and then your comms will surely be fine
alex_joni: why does it have to be "smart" ?
to talk ethernet it needs to be smart
SWPadnos: btw that img is still loading =)
alex_joni: does it?
because there's no dumb ethernet chip that will automatically convert ethernet packets to step/dir or PWM
yeah, their site is slow - try refreshing :)
SWPadnos: Are you sure?
none that I've seen
about no dumb chips
you could use a microcontroller to do it, but I'd count that as a "smart" endpoint
there are chips that provide some eth/serial eth/parport conversion
SWPadnos: Ok, what about the uC thingy with RTOS in it, couldn't that recreate teh step/dir pulses?
there are chips or modules that have 10/100 on one side and a serial port on the other
but almost all I've seen are using micros and they are damn slow :)
heh - yeah, that's the other thing :)
JymmmEMC: it could
you have a 25 Mz MCU with a 10/100 port on it, so it's dog slow at interpreting commands and sending results
but an fpga could do that probably even better
oh hey - that BeyondLogic page shows the XPort - I have one of those
and a DigiConnect EM or ME - don't remember which
that would do what for you??
not much, that's why they're back in their boxes at the moment
they are very useful if you have a serial device that you want to connect to ethernet/the internet
* alex_joni uses ser2net for that
how fast would it be... i have older fanuc machines with only rs232
it runs ok up to 38400
about Fanuc speed ^2
these can do 921.6 kbaud
well.. I use it together with some windows freeware which maps it as a local port
hmm interesting the tidbits you pick up in this channel
there are drivers with the Lantronix and Digi units
which do that - you get a virtual comm port
same way a USB port works, kind of
well.. I use it for my PBX at work, so speed is not that important
Fine forget ethernet.... VESA to Step/dir FTW!!!
what about running something like RTSP on ethernet?
which is how I learned that if you write a Windows program that needs to open a port higher than COM9, you have to use different system calls
JymmmEMC: imo the only way to go is ethernet
and have it fault if it sees a drop or packet outside of the time window
realtime serial protocol?
[21:00:11] <alex_joni> http://en.wikipedia.org/wiki/Real_Time_Streaming_Protocol
it is the protocol used for VoIP
another crappy use of "realtime"
granted a lost packet in voice is harmless whereas a lost packet controlling a moving object can be very bad
alex_joni: Me too, but if ya'll are gonna bitch at me on the timing, VESA VGA to STEP/DIR. Can't go faster than the speed of light... IT'S THE LAW! =)
and it's a hard problem too
do you send velocities (and have the device keep going in the absence of updates)?
alex_joni: -fx ?
do you send positions (and have the device stop in the absence of updates)?
or use the LCD controller on a microcontroller to drive a stepper instead of a LCD
I'd just use the PWMs
I think the comms need to be at servo level
sends velocities/get feedback
alex_joni: I'm too poor to own fiber
if no feedback -> ferror
JymmmEMC: you wanted speed of light
and it's not that expensive
it's OK, electrons go almost as fast as light anyway
alex_joni: Have you proced GBIC's lately?
it's not important over short distances (like anywhere on earth)
SWPadnos: they don't :)
0.69 is the factor quoted for most ethernet cable :)
or right around there
SWPadnos: I go down to 60%
I tested it a ways back
Because of collisions between electrons and atoms, the drift velocity of electrons in a conductor is on the order of millimetres per second.
you tested the propagation speed in ethernet cable?
However, the speed at which a current at one point in the material causes a current in other parts of the material, the velocity of propagation, is typically about 75% of light speed.
SWPadnos: random data (no hw I/O) across xover cable gave apx 60% thruput
that's not the same thing :)
SWPadnos: I don't care, if I can only get 60%, I cna only get 60%, not 69% dad burnit!
you're talking throughput, I'm talking latency ;)
were you using Windows or Mac machines in that test? (vs. both Linux)
SWPadnos: Keep it up, and no kimchi for you!
SWPadnos: Actually MS-DOS
with packet driver
3c509 and 3c905TX cards
oh, so 100mbit
yeah, both had apx 60%
ethernet has a heavy overhead I found out =)
gbic's are cheap..
the overhead is partly dependent on the packet size, and is a fixed amoung
depends what you are sending and at what level
SWPadnos: If you care to setup two boxes running MS-DOS, I'll find the utility
you need 18 bytes for the framing info - that's it
regardless of the packet size
if you send raw eth packets, then you might get better performance
alex_joni: but is that something you would want/need to do for this?
the stack, chipset, and CPUs make a big difference
JymmmEMC: I think that makes thigns simpler
especially on the endpoint part
alex_joni: "Just becasue you can, doens't always mean you should" =)
Hmmmm looks interesting... http://www.gridconnect.com/gc-ex-184b.html?source=goog&kw=ethernet+chip&gclid=CMn-yo22r5cCFRIcawodRRlijA
184 pin BGA, 1.8V core, 2.5V I/O, 12mm x 12mm
other than that it looks good
maybe a bit too smart for what we need
and note that it's ~$20 in low to medium quantity
this is why it comes down to what you expect to pay per axis
SWPadnos: There ya go... $20/axis
and what you expect the device to do
uh huh. if you can make that chip work unconnected to anything but bare ethernet wires and a motor drive, more power to you :)
it does have 32 IO's if I'm reading it right
so it should be ok for more than one axis/joint
SWPadnos: Well, if you have a five axis machine and all the I/O and oyu can do it across any PC with EMC installed and ehtenrt card, why not
SWPadnos: even your craptop could run it! <ducks>
yep, and the free PC board, power supply, industrial I/O interfacing, ethernet connector, and terminal connections will make it a really cheap product!
SWPadnos: sell those as add-on's
SWPadnos: not much different than having to have a 25 wire db25 to db25 cable, connectors, etc
except that the DB connectors are available at Radio Shack (who knows for how long though)
I can get db25-db25 for 3$
SWPadnos: look at that $50 breakout box for the mesa you have
SWPadnos: so not really all that much, and RJ45 much cheaper
JymmmEMC: I think that's an excellent point
SWPadnos did pay 50$ for it and lots more for a mesa
but I can get a Mesa PCI card, 2 I/O interfaces (with terminal blocks), and a PWM servo interface (with terminal blocks) for $400 total
SWPadnos: GREAT! AWESOME! toss it in your craptop
laptop's don't work ok for RT anyways..
SWPadnos: What I'm saying is you could run it on anything.
using a laptop for a milling machine control is stupid and unnecesary
anything that has an extra ethernet port
my laptop, and most that I've seen, have at most one ethernet connection
there are very few laptops with 2 eth's
if I want the laptop on the internet, then I can't use an ethernet motion doodad
alex_joni: ether switch/hub
mine has 2.. if I use the docking station
you need exclusive access to the ethernet chip for the RT comms
unless I buy an ExpressCard adapter
alex_joni: right, a hub
SWPadnos: or that
JymmmEMC, get a clue, please :)
JymmmEMC: I mean the ethernet chip on board
that means no other access for the same chip
YOU MUST HAVE AN INDEPENDENT ETHERNET CONTROLLER THAT LINUX DOES NOT INTERFERE WITH FOR RT ETHERNET!
is that clear?
SWPadnos: you can still use the wifi for regular networking
then let me say it again ;)
YOU MUST HAVE AN INDEPENDENT ETHERNET CONTROLLER THAT LINUX DOES NOT INTERFERE WITH FOR RT ETHERNET!
or use ethernet over firewire
most laptops have firewire (for some odd reason)
and it must be so independent, that it shares no registers with any other ethernet controllers (which could eliminate some dual-eth chipsets)
sure, wifi would work
SWPadnos: See, all you guys try to use emc box for everything, irc, etc, I'm talking a dedicated box.
a dedicated laptop, you mean
with no access to updates
JymmmEMC: then build a dedicated box and be done with it
no need for ethernet
SWPadnos: you dont run EMC oon a laptop silly.
of course not, who thought of that? :)
SWPadnos: Um, you did --> (01:24:36 PM) SWPadnos: a dedicated laptop, you mean
[16:22:09]<SWPadnos>using a laptop for a milling machine control is stupid and unnecesary
JymmmEMC: do you actually have a point for this conversation?
damn. now why didn't I think to ask that?
alex_joni: You mean besides gettign your panties all in a bunch over it?
besides wasting time
Well, I obviously don't know the timings, but all I hear from you guys is "no". and I still dont see "why not".
you need to read our responses more closely then :)
nobody around here said no for ethernet
we did say no to usb (with timing examples)
what we've both said is that ethernet could work, but there isn't any existing implementation for EMC2
SWPadnos: I read them, doens't mena I understnad them.
or why there is so much opposition
first you fretted over cost, ok so I found $20
and that there isn't much impetus for any of the volunteer developers to make one, since there's a lot of work involved for little gain (when we can spend $400 or less and have a better implementation that already works)
you found a $20 IC, not a $20 smart axis controller
and I can find you a $9 ARM chip with ethernet anyway - that's not an issue
there isn't opposition to someone developing such a device and driver set for EMC2
it's just not something that anyone has bothered to do on a volunteer basis
SWPadnos: Ok, you found MESA, jepler foudn Pluto, before then there was Jon's. these are all propritary Iface. If Peter got hit by a car, where's your mesa then?
I'm probably one of the most likely to volunteer, but it's a big project and I don't care enough about it at the moment to spend my time developing it
I still have the VHDL and C code to make it run
SWPadnos: And if/when it craps out, like everything eventually does, then what?
then again, if I start making an ethernet controller, what happens to my customers if I get hit by a bus?
MESA is a bit bigger than Peter
afterall it's a company..
if nowadays bill gates gets run over by a car I doubt anything will happen to microsoft :)
ok, fine, the bldg blows up, whaterever. All I'm trying to say is they are all propritary ifaces
alex_joni: scroll back ?
mesa is not an interface
JymmmEMC, an ethernet controller would also be a proprietary interface
PCI is the interface for the 5i20 for example
I tink were splitting hairs here.
no, I'm trying to understand your point
yes - bus accidents are bad for customers, there's no doubt about it
iface between EMC and drivers
ok .. so "logical" interface
there is nothing proprietary in any of the devices you mentioned except for PPMC
or pc and motor drivers
there is for mesa too
SWPadnos: if mesa went away, you couldn't build your own boards I think
all of the logic source code (verilog and VHDL) is available for Mesa/Pluto, and the software is available for all of them
I could, but you couldn't ;)
you know what I mean..
oh sorry, thought you were JymmmEMC :)
maybe I could too.. after a couple of attemps and a bunch of broken mobo's
Take it to the point where ppl can build their own drivers
JymmmEMC: I doubt there are 2 people on the emc list that can solder a bga
you mean having open source schematics and board layouts>
I cant remember those two driver chips they use
I don't know anyone who can do it by hand
and soem others
l297/298 are really crappy :)
alex_joni: Sure, but ppl CNA build them
that doesn't mean some people don't like to build them
JymmmEMC, you need to read about the layers in an EMC2 system, but unfortunately I don't think there's one I can point you to at the moment
JymmmEMC: right.. but going from building something like that to an embedded system with eth..
there's a HUGE step
why not have an ethernet thingy too?
this one is a start: http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Emc2HardwareDesign
JymmmEMC: it's almost like saying build your own simple motherboard
most ethernet chips are BGA or QFN packages, which are impossible or difficult to solder
alex_joni: Well, ppl in the 70's did that wirewraped their own mobo.
any that are TQFP (basically all the others) are also ethernet only, no microcontroller
you can get crappy PIC's or whatnot in DIP packages with some eth access
I tink even cradek did that with his altair
but those are really crappy performance wise
JymmmEMC, stop changing the subject
SWPadnos: I didn't, was just responding to alex_joni
I have a system I built for tracking employers
basicly time accounting
it uses an ATMega + ethernet chip + a linux server with mysql and php
well no shit, but there's obviously a huge difference between wire-wrapping a 40-pin CPU and soldering a 0.25mm-pitch QFN package
that has 128 pins
SWPadnos: Aren't those forms available in a plugin factor?
nothing too fancy, maybe something that people could actually do
plugin - no
but performance wise it's crappy.. I usually have ping times of 1..3msec, but sometimes it can be as high as 700msec
as Alex said, there are some PIC/AVR chips that may have eth and are DIP (though I don't know which ones)
there are eth modules that are DIP20 or so
but those aren't 100MBps capable
SWPadnos: Well, it's not easy, but you CAN solder a flatpack
ok, but if that company goes belly up, you'd hav eto make your own (with the components they use, which are probably QFN or TQFP)
I can solder TQFP in my sleep, and QFN if I've had enough coffee
JymmmEMC: sure, I did a couple TQFP by hand
but not BGA on any day
speaking of which, time to make some tea. bbiab
it's not that hard once you learn it.. but it's not everyones dreamjob
I think hardest to solder whas a PLCC socket
[21:43:28] <alex_joni> http://www.global-b2b-network.com/direct/dbimage/50221753/SMT_Type_PLCC_Socket.jpg
<- that kind
has PWM http://www.atmel.com/dyn/products/Product_card.asp?part_id=4117&ListAllAttributes=1
Terminal Pitch (mm) 0.50
JymmmEMC, think about what you'd want this hypothetical controller to do, and think about what you'd be willing to pay for it to do that
and consider that a very high performance PCI system can be had for ~$100/axis, including all necessary industrial I/O filtering/protection and termination blocks
an ethernet system might have some advantages thoug
I agree - it's a question of what it needs to do and how little it has to cost
if you have tons of IO all over the machine then an ethernet connection might make more sense
DA10 Change before SDCK Rising Edge
SWPadnos: Price $12/ digikey
JymmmEMC, that's a little like looking at a $12 hammer and asking why a house costs more than $12 (and why isn't it built yet?)
SWPadnos: Not really. See, all I hear is opposition in making it reality. Everyone likes it in concept. Well, $12 for the heart + extra for the misc components., so say $50 built. But I *think* you are looking at it from a revenue POV, not an alternative to ParaPort.
no, I'm being practical
please search the emc-users list and read the conversations on this topic there
give me the short and sweet
the fact is that reliable terminal connections cost far more than the CPU
and interface electronics (like static protection and short-circuit protection) cost more
and a paraport has these?
Jon Elson has said on several occasions that the cost of terminal strips for his boards cost him over $50 per board
SWPadnos: you're looking at it from an industrial over-engineer POV
I'm trying to tell you what obstacles are in your way
I am not trying to dissuade you from pursuing this idea
for you, being able to plug into any old PC is a significant advantage, so ethernet is a good interface to choose
SWPadnos: I know those removable term strips are expensive as hell. though they shouldn't be
well they are, unless you buy a million of them
but there are other term strips far cheaper too
look at the cost of a PC
what percentage of that is the CPU?
hell, look at a single LED for that matter
we used to have customers ask us if we could take out some features (from our remote controls) and make them cheaper
well, we could have eliminated the $1.20 CPU, and made them do nothing
at an overall savings of $5-6 to the end user
(out of $150)
the point there is that the CPU is not the major cost component in a device like this
it's the I/O
and the PC board
Ok, i estimated at $50, lets 4x it... $200 for an ethernet controller that has step/dir and/or cw/ccw plus I/O. Is that more in line?
sure, that's closer
remember also that people building their own have to buy parts at singles cost, not qty. 1000 cost
SWPadnos: I think $200 is ONE-off pricing
yes, it could be one-off pricing or close to the cost of a finished product, if the manufacturer can exploit wuantity pricing
etch/cnc your own PCB
well if you already have a CNC ... ;)
or use a service for $30/each
or even $50/ea PCB
ok, now we're getting somewhere :)
ok, RJ45 PCB mount, what $0.25/ea at an surplus store
not for ethernet
they're $4 or so in medium quantities
at least - you need magnetics in there
[22:07:27] <JymmmEMC> http://www.excesssolutions.com/cgi-bin/item/ES4526
that is not an appropriate connector, unless you add in external magnetics, like a Pulse transformer
SWPadnos: Ok $25 for magnetics
up to $100
hold on. we're not going to design the hardware on IRC today
neither one of us knows enough of the specifics of what's needed to do that
SWPadnos: You've been saying it isn't practical. So I'm going overboard here on pricing.
we got a PCB, chip, magnetics, and RJ45 connector for $100
and at some point, if you're just going to be using TTL I/O on one side and ethernet on the other, you may as well use an ethernut or other ready-made board
I thought those didn't have the timing?
so there's an upper limit to the hardware cost
well, that gets you to stage 2: writing the software for the ethernet-connected device
y'all sure can talk
JymmmEMC, they may not, but to be honest, I don't know if an AVR32 or ARM does either (though I suspect they're both fine)
fenn: can't we though =)
SWPadnos: lets assume the AVR is, I notice 2nS in the timing sections
that has nothing to do with CPU speed
but VLC sees my wireless IP webcam... couldnt that sort of code work??
12MHz / 10 = 1.2MHz even
still beats paraport I believe
its wireless G
JymmmEMC, what are you thinking of when you write those numbers down?
I mean, what do you think they mean?
SWPadnos: Ok, here's the deal.... Yes, I've harped on you over the years about this crap. But I don't expect you or anyone else to do anything about it.
I've thought about the problem a lot more than you might think
SWPadnos: I just get tired of hearing that it cna't be done for what the fuck ever reason, or that it's unrealistic.
you need to listen to the reasons
it's not impossible, but it is impractical for a few reasons
SWPadnos: What I HAVE heard is that it's a bitch and " I dun wanna go to bed".
anything can be done with the proper $$$$ incentives
or non-$$$ incentives
SWPadnos: whining that it's hard, basically.
actually, you're whining that it's easy
SWPadnos: I'm ranting that it's realistic.
any of us could have just said "ok JymmmEMC, go make it and submit patches for EMC"
SWPadnos: yeah, and ya'll know that I'm clueless there too.
of course ;)
if we had 1 million ppl in this channel it would be done in days
Well, there have been some might smart and ehtusiastic ppl here over the years too that have come and gone.
I think part of the problem here is that there are obstacles to implementing this, it's not a simple problem, and nobody who has the time and/or capability of doing it is motivated to do so
that says nothing about whether it's possible, just that it's not imperative
sounds like our root cause?? so EMC is dead
as Alex said, the most likely scenario for making one of these is if someone who stands to make some money does it first
dmess, what the hell are you talking about?
just throwing it out there... no we wont support that.. its too hard to interface.... i thought we wanted it to hook up to ANYthing..
SWPadnos: Ok, forget EMC... How hard would it be to prototype to control a single stepper motor via ethernet? timing/speed be damned.
well, you need to get some hardware, then write the software that runs that hardware, then write the software that tells it what to do (from the PC)
dmess: remember - the developers are volunteers
dmess, while we'd like EMC2 to support as much as possible, it doesn't mean we're going to go out and design interfaces for non-existent hardware, or that we'll go out and volunteer to develop new hardware when there's plenty of god stuff that already exists
SWPadnos: are there any protocols/API's that already exist that could help?
if there were a piece of hardware that existed, and "all we had to do" was to write a driver for it, then there would be more of an impetus to do it
i do.. so remember that.. and i appreciate ALL they have done.. ive been a follower for some time... and now have moving drives ;)
JymmmEMC, I don't know what's there, that we can access and use at no cost
(of course there's UDP, I'm thinking of higher level "motion-oriented" protocol stuff
SWPadnos: Ok, at waht OSI layer would you think would be needed?
all 7! ;)
SWPadnos: Come on... you're trying to offload from EMC using all that
let me talk to a few freinds in lower places... how fast do we WAnt/need/can handle as a min
I can make a serially-connected device run a stepper (or other step-dir controlled) drive
SWPadnos: There is more than TCP/IP protocols
in fact I have one already designed and built
that would be pretty easy to run over ethernet by changing the microcontroller
it can do PWM too, it acutally uses a hardware PWM to generate steps
not really, it's just an AVR board :)
how do we close the loop?
remember, JymmmEMC said "timing/speed be damned" :)
this is actually an MPG-like thing, but I stuck a serial port on it and it's reprogrammable
so whatever AVR software you want could be used :)
so she runs like jack the bear but we dont know where we are
SWPadnos: SLIP protocol?
[22:30:38] <SWPadnos> http://www.cncgear.com/MPG/
like I said, it's a pulse generator that has a manual rate control knob, but it's reprogrammable ;)
SWPadnos: Hey PLIP!!!
JymmmEMC, UDP is fine, and probably what we'd want
i have a FREE MPG on its way for my lathe and havent given a thought to how to hook it up.. go figure
I don't know if RTNet has any drivers for chipset-integrated ethernet controllers
no PLIP either (I know what it is)
cool. send me one next time OK
SWPadnos: it's actually the inverse of what we want =)
yeah, but it's reprogrammable :)
i could.. he has 4
it has 4 transistor outputs, so it can do 2x step/dir
not CE approved
non CE MPG - who cares?
thats what i said... i'll ask him for 2 tomorrow
used.. off of some of our OKK's
OKK by me :)
some scratches included
ahhh, the smell of machining oils
hmmm. I bet my tea water got hot then cooled off again
or has melted the pot
you pay shipping...
I can do that. hey, at that price I'd buy 2 :)
i cant be greedy... been to that same trough for my free vfd
should i leave the big plug on it or can i chop that and put on ebay for 100 bucks
oh, I don't need connectors
just screw terminals on it
(assuming that's what it has)
these are all mil-std connectors worth a mint
yeah, I know the type
1 plug.. the other end will be $$
I paid more for connectors than I did for motors
(and I have Baldor servos with 1000l BEI encoders)
sucks to be us
im gonna cut it off and ebay it
this channel has some of the smartest minds on the planet ... (excluding myself) as i have see MANY issues solved in 1 overnight session's... worth of hack
looks like the crew... EH