does anyone have a suggestion for the following scenario....
...I want 5v-9v control going to a standard issue wiring box. Like a regular recepticle box that has a 110 outlet in it, plus relay.
...I've never been able to find hardware like that. except for X10, and X10 sucks.
you want it to actually fit in a regular outlet box, along with the outlet?
thats a 10lbs of crap in a 5lb bag problem
doesnt have to be a regular outlet box.
you want to put 120vac and 9vdc in the same box?
the point is that I have never found something that ties house wiring to DC-control.
they make 120V in and 120V out modules
dunno how many amps the outs are good for, probably not 15 like an outlet
*does a search on opto-22 block*
how much power do you want to switch?
yeah I need 20 amps.
that's a very tall order, and I bet you can't find it :)
even a standard light switch is only rated for 600W or so (5A)
[00:07:48] <jmkasunich> http://www.opto22.com/site/pr_cat_g.aspx?qs=1006&
SWPLinux: you can get a 20A rated wall switch at any home depot, etc
sure, for a heater or something
whoop. got a request from the wife unit to have dinner.
I don't see any high power switches on home depot online
15A, I guess - that's not bad
maybe I'm thinking of the timers and such for that 600W rating
there are opto-22 SSRs rated to 240V 25A (and maybe even higher)
03jmkasunich 07HEAD * 10emc2/src/hal/ (hal.h hal_lib.c halmodule.cc): Eliminated 16 bit hal pins and parameters. They were used only in a couple of places, and those have been extended to 32 bits. Eventually the 8 bit types will be going away too.
03jmkasunich 07HEAD * 10emc2/src/emc/usr_intf/halui.cc: Eliminated 16 bit hal pins and parameters. They were used only in a couple of places, and those have been extended to 32 bits. Eventually the 8 bit types will be going away too.
03jmkasunich 07HEAD * 10emc2/src/hal/components/ (6 files): Eliminated 16 bit hal pins and parameters. They were used only in a couple of places, and those have been extended to 32 bits. Eventually the 8 bit types will be going away too.
yeah - when you get the brick-shaped panel mount versions, they get pretty big
240V 45A, for $37
03jmkasunich 07HEAD * 10emc2/src/hal/utils/ (7 files): Eliminated 16 bit hal pins and parameters. They were used only in a couple of places, and those have been extended to 32 bits. Eventually the 8 bit types will be going away too.
120V 10A is $21
elson: are you still around?
120V 25A $23
this reminded me of something
argh - I have too many motion controllers in that bin
I can't close it
The high power SSRs require a heatsink though. I started to use one for my high temperature oven, and decided to just buy one with a heatsink. I think I got it from McMaster Carr.
anyone know the diff between reiserFS and ext3?
My understanding is that reiserFS has higher performance. ext3 does logging so can recover after a powerfail without doing an fsck. I don't think I've used reiser.
I've always used ext3. reiserfs is supposed to have certain performance advantages, but it is more complex than ext3.
I think reiser is faster if you have lots of small files.
anyone want to return back to the 9v input controlling 20A switch thread?
the thing is, I know that putting the components together is easy enough, I'm just wondering if someone's made an integrated unit that could drop into a "square box".
owhite: get a relay to do that for you.
signal -> diode -> transitors. Transistor controlls large relay switch.
Yup. McMaster has a 30amp 240volt SSR with a heatsink for $61. Save more than half that if you supply your own heatsink.
yeah yeah. I know. wired plenty of relays. but it just seems like industry probably has made something with a 20 amp relay in it, with nice terminal lugs and input for DC control, that's easily dropped into something you buy in the electrical aisle of the hardware store.
you could also use a much smaller relay or a transistor to switch the coil of an electromechanical relay
i've seen them for mounting in DIN rails
might be cheaper than a SSR, and won't need a heatsink
owhite: yeah... it's called a light switch... :) manual relay control.
what's the V and A ratings you need?
220v @ 15amps, and 110v @ 20 amps.
I've got a medium size relay here.. 12A 120VAC or 10A 28VDC.
owhite: VDC? VAC?
VDC in, VAC out
the dignal will be VDC.
what what's the LOAD on the really.
03jmkasunich 07HEAD * 10emc2/src/hal/components/abs.comp: new abs component for HAL, generates absolute value and a sign bit from an analog input
load on the relay: 20 Amps.
digikey part # PB905-ND $7.87USD.
you're missing the point. If I wanted a light dimmer, you could tell me to get some MOSFETs a potentiometer, and some resistors. Or I could buy a light dimmer.
i know i have seen solid state relays good for 20 amps
I know that there are plenty of components that work.
that are DC controlled
I want something that is four screw terminals and fits into conventional electrical box.
i have never seen what you want already with a standard AC receptacle
two terminals for AC, two terminals for DC. or something like that.
digikey # Z247-ND RELAY PWR DPDT 20A 120VAC PLUGIN
just take a standard box, put a solid state relay in with it, to control the A/C. Done.
I've done that.
I actually wouldn't mind some large relays... to controll on/off of my mill AC spindle motor.
me want something pre-made. that's what China is for.
owhite: I'll make it premade for you ... $250CDN.
You can have it for you in a week. includes shipping.
*writes a check*
nah... www.bidpay.com thank you.
*sends it to china*
*recieves 20 units*
or you can mail me an 'international postal money order'.
that was easy.
you're not informed correctly... they deal in bulk.
none of this cheese, under a shipping container full BS.
I see a marketing window for all the CNC enthusiasts out there.
stop complaining... as there's nothing out there that we know of... and I've made the offer of $250CDN... I'm sure someone will undercut me.
like I said, that's what china is for. :-)
That isn't a bad idea... box, two inputs, AC in, AC out, VDC in/out. relay, transistor, diode, a resistor and LED.
$75USD + shipping.
material costs would already be ~25...
I was sort of hoping for something around $20.
labour ain't free in North America, except Mexico.
I guess the reason they havent been commercialized is that the market is small. For a while I looked at lighting controls for large theaters, and didnt find much.
there are somethings that are pretty close for people that do stage lighting.
don't think those would handle the amps.
those are usually rampped up... lots of those types of lights don't like to be suddenly turned on...
thermal shock in some lamps
for example this hideous thing has plugs and relays all built in. it could be hacked. http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=300041529129
maybe there's something in "home automation" that would do what you want? http://www.smarthome.com/2207.html
(not this exact item, though)
yeah I hate the X10/home automation stuff. I have a friend that gave me a box of that stuff and it simply does not work.
*shrug* we'll make something, patent it, and direct the proceeds to emc2 development.
owhite, that's the mac total lost of $15Amps.
owhite it works good for me
'leave no child behind'... apparently it has failed, as proof by the line above.
I can even control things from dos/windows/linux too
iirc there is a 15 or 20a A/C control module too
John K, I'm trying to track through the way the ini file parameter PID_MAX_VEL works, and what it means in relation to PWM duty cycle.
Something in EMC2 seems to recalculate this to 1.25 (in my case) and resets the value back in the .ini file.
I overrode it in /configs/univpwm/univpwm.ini with a value of 20.0 and the pulse width got wider. I'm wondering if it even makes sense to tale this value from the .ini file, or just set it so I can get 99% pulse width.
what is iirc besides 'interactive illinois report card'?
I sure have the crappy connection lately
03jmkasunich 07HEAD * 10emc2/src/hal/components/abs.comp: fix license text
elson: you still here?
I was away for a bit, didn't see your question
are you sure that EMC2 overwrote a value that you put in the ini file for PID_MAX_VEL?
anyway, there are as many ways to deal with the scaling and limits as there are people I think
I strongly prefer the PID output to be in length units per second
IOW, if the PID output is "3", then the duty cycle should be whatever would make the motor run open loop at 3 inches/sec
the way to do that is to set the pwmgen scaling right
either test or calculate the approximate open loop speed you'd get at 100% duty cycle (or 99% if there are drive limitations on max-dc, you should set that first)
then set the pwmgen scaling to whatever that speed is
so if 100% gets you 2.5 ips, set the scale to 2.5 ips
that way, when the PWM output is say 1.0, you get 40% duty cycle, and about 40% of top speed, approx 1 inch/sec
then set PID_MAX_VEL to the same level, so it never asks for more than 100%
and set your machine limits (TRAJ and AXIS) to somewhat lower since you need some overhead for a closed loop system
if you got 2.5 ips with 100% duty cycle, I'd probalby use 2.0 ips = 120 ipm for my machine limits
no, cable now I think
cradek: Check the MTU
especially if you're using PPPoE or PPPoA
scratch the PPPoA if you're on cable =)
thanks but I don't have the slightest idea - it's not my connection
not at home?
working on sunday?
or messing with timeguy?
no, I'm at home, but timeguy isn't here
I think it's set up right though - it's almost always perfect
A-L-P-H-A is now known as a-l-p-h-a
Sorry, John, jumping to conclusions again. The REAL problem is that the PWM frequency was all wrong, and I didn't bother to read the scope settings. So, either the one-time setting for that 2-byte register isn't getting through, or it is getting overwritten later. The PWM is around 60 Hz! I should have noticed. So, of course, the on-time looked really small in comparison.
one time when loading the driver? or one time when you change the frequency parameter?
I really didn't think there was any one-time stuff
pwm freq in particular needs to be able to be recalced if you change the pwm frequency param
I think maybe the code puts it in the write cache for it to be written every time, but I'm guessing it is not being included in the range of regs to write. I'm reading through it now.
so it should be written every time (usually it will just re-write the values in the cache)
I didn't change the range to be written, just used the new API to set the same range as before
the driver does output the cache bitmap to the kernel log, so you can check it
Really, there's no purpose in rewriting the freq register every cycle. It should be OK to just write it once.
if neccessary, checkout another tree, with 2 day old code, and see if that is indeed using the same bitmap as the latest
cvs up -D'two days ago'
err '2 days ago'
yeah, but when you _do_ need to rewrite it, how are you gonna change the bitmap?
or are you talking about bypassing the cache when somebody changes the freq param?
Well, this may have been broken since the last change to the bitmap method. I tested with USC, but not with UPC, I think.
(that way lies madness, IMHO)
Why madness? Why shouldn't one-time settings just be done once?
writing to the hardware directly _and_ thru cache just seems like asking for trouble
its probably not that bad, but personally I'd avoid it
I'm also trying to keep this thing fairly streamlined. Last time I had the LA set up, the 4-axis cycle took 88 us, which is pretty good, but still much slower than the old EMC1 driver.
thats the tradeoff for a more flexible driver
what is your website? I want to look at the register map
I am already using the two ways to access in some of the other export routines, and the set encoder latch does it, because it writes to the same reg several times.
The web addr is http://jelinux.pico-systems.com/univpwm_regs.html
export is fine - that happens only at startup
but the PWM frequency can be changed on the fly
I always seem to forget the - between pico and systems, and go to some other companys page
I think it DOES get reset once when the univpwm_motion.hal file is processed. At this time, the board should be in E-stop.
OK, it is clear. I or'ed in the extra bit to write the PWM control register. So, it never writes the frew registers. If it doesn't bug you too much, I'd prefer to do this with selective writes when a change is needed.
ok, you are trying to avoid writing regs 1D and 1E all the time
Yeah, well picosystems and pico_systems were already taken - i tried!
Yes, changing PWM freq would be a rare event.
so how are you doing that? if you or the bits (2, for 1D and 1E) into the map, then they will be written all the time
oh, I see
you are oring in 1C
but not 1D and 1E
Well, now that I know why, I think I can fix it pretty easily. I did NOT think to add the extra bits for 1D and 1E.
are you writing to 1F?
seems like you need to, for the digital outs
so you only save one cycle by skipping 1D and 1E
Yes, clearly, 1F does need to be written. Maybe the overhead is damn small, then, if 1C and 1F need to be written every cycle!
addr, 1C, addr, 1F vs addr, 1C, 1D, 1E, 1F
OK, problem is moot, then. Your logic wins this time.
the cache could be made much more sophisticated for writing, and keep old and new copies, then only write the bytes that were changed
but I'm not volunteering to do that ;-)
But, my code in the spindle sync mechanism just hops in there and changes the one register directly, at those rare times when going into or out of spindle sync.
not much need for such logic, I don't think.
at least the bitmap lets you skip 18 thru 1B
you don't write any of the low block, do you? (00 thru 05)
05 is the one you "jump in" and change?
Strange thing, this machine probably needs a reboot, it is real short of free memory, and EMC runs like a DOG. It takes 2 minutes to bring up the editor. well, been up 42 days.
is it slow only when emc is running?
right 05 is the new register that resets an encoder counter on sensing the index.
No, incredibly slow with or without EMC, and lots of disk thrashing. My wife is watching something on the net, and this is the router, so I can't reboot it until she's done.
my uptime is only 15 days, but thats because I just installed dapper
Yes, never had this problem before. I hope it isn't a memory leak type problem in the RT code. With all the mucking about I've been doing, I wouldn't be surprised.
I've been up for very long times without running out of memory
Oh, we just had a power failure at work, I think my desktop Linux machine was up for 412 days!
a leak in the RT is very unlikely
But, I've been crashing things and other messes. Glad memory problems are not likely.
none of the RT parts of EMC do dynamic allocation, except for shared memory regions, and those are allocated once at startup
rtapi frees any shmem that you allocated when your module exits
no smoking guns in top sorted by memory size?
No, the first block 00-0F don't get written much. There is the encoder preset reg, and the timer, encoder setup regs.
No, memory usage in top seems modest. But, this machine only has 128 MB.
what distro are you running?
I'm not going to get worried unless this sluggishness recurs.
the pigs are usually GUI things, browsers, window managers, etc
This is from the BDI 4.30, 22.214.171.124-magma kernel
is BDI still supported?
normal ubuntu would be very unhappy with 128M I think, nothing to do with emc, just Gnome
a-l-p-h-a: yes and no
the code still compiles and probably runs on bdi
But, I'm also running apache, sshd, smtp & pop, and denyhosts, plus IP routing and NAT.
paul made some packages for the latest BDI, but he changed some things and won't reveal what he changed, so we're not so inclined to support people using those packages
all of those should be fairly "mild" in terms of memory usage, unless one of them is leaking
BDI can still be set up to compile the current EMC2. Yeah, I won't use what Paul has been doing to EMC because I don't know what is going on. He has a little war going on about some install issues. But, after installing the right optional packages, EMC2 will install just fine.
I think jmk was talking about paul's emc2 packages which were released without corresponding source
but I doubt many people are using them, so it's a very small matter in my mind
I wouldn't be going to the trouble except that this machine is kind of limited, and after going to all the trouble of getting all that server software running, I didn't want to change it.
yeah: "paul made some pacakges"
why are you using your server for emc developement?
nothing else handy?
Physical location. I already had the thing RIGHT next to my electronics bench. I have an even older PC there, and got tired of switching the kbd, mouse and video cable all the time.
next CNC workshop I'll make you such a deal on a 2-way KVM switch ;-)
you still have a bunch of those?
I have a mix of dumpster grade ones
but the one I was talking about was my old main one
I have another machine (the one that was at the Workshop this spring) that is connected to the minimill. Also, having a running EMC2 on the server means I don't have to boot a machine to look at the current code.
I'm using a 4-way now
cool - the Opterons did get cheaper
1210, $185 + $15 on ebay
The location is therefore right next to both analog and digital scopes and the logic analyzer. This is where I do production testing of boards before shipping them out.
Well, actually, is a backwards sense, this machine cost $14. I bought a computer for a customer (to preinstall EMC on) but it didn't work. The seller was a bum. I got a replacement from another outfit that I trusted, and it was damaged in transit. That seller sent me out a replacement mobo minus CPU. I swapped the damaged mo-bo into the DOA box, and just had to supply a CPU - $14 on eBay.
so the "damaged" mobo wasn't damaged after all?
My main destop machine runs Mandrake and can't use an RT kernel because I use VMware to run Windows (blech) CAD apps.
I'm running ubuntu dapper here, with a RT kernel, and have vmware with two virtual machines (so far)
haven't put doze in a VM yet, gonna do that with win95 for easycad one of these days
Yes, the CPU heat sink came off and smacked it up pretty good. The CMOS battery holder was in particles, and some other stuff may be damaged. But, I got it to run by picking parts off other boards. But, I wouldn't want to sell it to somebody after such repair.
Oh, well, that is now info, then. Last I checked (some time ago) VMware didn't handle a RT kernel.
I have found Win2K Pro to be an extremely reliable system under VMware/Linux.
So, am I too cheap?
was that meant to be rhetorical?
Yeah, I thought running my company's server on a $14 salvaged computer sounded pretty cheap!
elson: I say "hooray" to people who keep stuff from the landfill .. by contrast, I spend money too easily on new hardware.
I should probably eat something real... instead of cookies...
no real food since 7pm EDT yesterday
I buy off-lease Dell Optiplex machines from cacrc (corporate recycling council) in Baton Rouge. A real good outfit to deal with, most of these machines were in the $50 - 70 range, including shipping.
elson: then you should have some access to P4s.
YESTERDAY? That's not good for the body.
elson: I've been munching on cookies my sister baked for me.
it's not good for my system at all... just feel kinda blah to eat.
maybe I'll make a tuna sandwich.
Yes, the older ones are P3, the newer ones are P4.
use a P4... and go happy hardcore with EMC.
$70 shipped... that's not bad at all.
256/512 ram... with a 40gig HD, and you're laughing.
better then my 700mhz duron EMC machine.
Well, when the seller stands behind the units, and they are reliable, then yeah, I'm quite happy with them. Just like cars, doesn't make sense to buy new.
Have to test out the latest code changes, I just compiled.
elson: yes it does... if you want the original warrantee, style, new car smell.
and to loose 15% off the bat.
buying a used car, don't know who beat on it before, the history, etcetc.
could have been branded, and then unbranded. evil shit like that.
after the earlier discussion of dapper on 4GB drives, I created a vmware with a 4GB disk, installed dapper, emc2, and everything needed to build the CVS version of emc2. I still have 1.2GB free on the disk.
I'd have more like 1.5GB free if I removed all the stuff from apt's cache
jepler: it would be nice to note that on the wiki page (that nobody reads)
in fact I just added it there :)
elson: I'm browsing cacrc's website - nothing in there about resale
do they hide the info well, or do you have to know somebody?
he's sleeping with the hot secretary there...
notice no denials... I forgot to mention that the secretary is male.
a-l-p-h-a: don't be a .......
wow, so many words fit there
not ass... as that would be an "an" not an "a".
I get their stuff through eBay. They don't always have items on auction.
[03:15:22] <jmkasunich> http://www.hgrindustrialsurplus.com/search-products/product-detail.aspx?id=5589278&searchtable=2&sortExpression=&SortASC=&pageSize=50¤tPageIndex=0
is that a robotic hand?
Well, the changes seem to be successful! Now I get 100% duty cycle. (Just have to put back all the settings I changed.)
dunno what kind of hand is on the end, but its an arm
nicer one: http://www.hgrindustrialsurplus.com/search-products/product-detail.aspx?id=5585442&searchtable=2&sortExpression=&SortASC=&pageSize=50¤tPageIndex=0
imagine if you could build an arm with the dexterity of a human hand, and the genetleness of a womens touch....
wow, only $299 http://www.hgrindustrialsurplus.com/search-products/product-detail.aspx?id=5578974&searchtable=2&sortExpression=&SortASC=&pageSize=50¤tPageIndex=0
looks like a crane
Well, I'm going to bail out here and see if I can get this computer back to normal speed. Thanks much for all the help!
you're all going?
03jmelson 07HEAD * 10emc2/src/hal/drivers/hal_ppmc.c: slight adjustment of write cache limits for PWM controller
here's a hypothesis I pose. Are left sided road country drivers more coordinated than the right sided drivers... using their left hands to shift.
jymmmm is now known as Jymmmmmm
Jymmmmmm is now known as Jymmm
who are you harrassing now Jymmm.
bye. sleep, at and attempt at sleep
Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emc/2006-11-13.txt
ooh, java is apparently GPL'ed now
[12:03:42] <Lerneaen_Hydra> http://developers.slashdot.org/article.pl?sid=06/11/13/0724252
it's too bad java isnt useful :D
maybe it will become at least semi useful now ;)
Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emc/2006-11-13.txt
looks like jone may be getting more comfortable on irc - good.
that's a good thing
[13:26:18] <skunkworks> http://jelinux.pico-systems.com/
slow day today...
I'm doing a few hours extra
since I were sick a few days last week
alex_joni: my ir2111 h-bridge is still working ;)
alex isn't here today either?
alex is on a work trip
* alex_joni finally made it online
hey alex - hows work? :)
* alex_joni is dead tired
work is good
installed the bot today, did a testweld, all is good to go
nice little 1.5t turn-tilt table
nice (little?) :)
yeah.. the other one at this customer is 2.5 ;)
that's 1.5 workpiece, the table itself is ~4t
oh - I was going to ask, but didn't bother :)
I have a ~130 pound rotary table...
does that count?
oh, hi alex
when you're less tired, could you add a link to the lathe article/page at linuxcnc.org to my site that I finally scrounged up the effort to create?
lh: you can do that
oh - wait, not a wiki page, a joomla page - nevermind
* Lerneaen_Hydra is unsure of who has access there
anyway, the link would be http://lerneaenhydra.shacknet.nu
, for the time being
till/if I decide to buy a domain
Lerneaen_Hydra: sure can
* alex_joni checks if it's worhty :D
last one I promise http://www.cnczone.com/forums/showpost.php?p=217197&postcount=252
skunkworks: I don
't get it, what's so special about quadrature?
alex_joni: how do I find the page on linuxcnc.org if I loose the adress?
searching for lathe brought up tons of wiki hits
Lerneaen_Hydra: Documentation -> Technical Articles
so that's the only thing there as of yet?
Documentation -> Technical Article ;)
quadrature (normally used on servo systems) allow you to know which direction the shaft is rotating as well as how far.
yeah, so what's so amazing about that post?
it's also probably the most well-known method of tracking position incrementally
the amazing thing is that he didn't know about it before ...
and yes it is in most mice.
oh, I see
hmm.. wonder if he can come up with a better way to do it
somebody knows gantrykins.c
somebody does :)
there is some bugs on axis with ?
could be. what are you seeing?
instead of x,y,z, i have 0 1 2 ???
Didier: that is normal
you are using a system which is not a trivial kinematics
that means 0 doesn't necessarely correspond to X
not sure how AXIS does it, but in tkemc you usually home all joints then you can switch to "World View"
i see, i need to home all joints to see what happen
if I remember correctly, the homing process with gantrykins isn't 100% thought out
how does axis deal with <3 axes or rotational axes in the 3d plot?
thougth out = finished
thought out = not sure exactly how to do it :)
jeff may know, I believe he wrote gantrykins
but I think there are some issues with homing non-trivial machines in general
then i will see what happen and make code if necesary
(I don't have one, so I can't be sure)
sound sgood to me :)
i hope, i need to try it for a machine
SWPadnos: what about a 5-axis mill (3 translational, two rotating axes)
darned good question
you can't plot the toolpath well unless you know the geometry of the head / table ...
granted noone programs them by hand (ever), so the axis 3d plot is maybe not all too needed
the CAM app needs to know that however
[17:27:09] <Lerneaen_Hydra> http://www.rainnea.com/cnc_support.htm
<-- are a few different kinds of 5-axis mills
the bottom right one is actually rather easy to add to a 3-axis mill
no major overhauls wrt the spindle
just additons to the table
sure - I know the CNC_toolkit
I just wish I could convince Rab to redo it as a standalone app instead of a Rhino plug-in, which is probably what he'll do next
thank you, bye
SWPadnos: I suppose there is an issue with having the encoder resolution so high? .000208. will I loose some responsiveness as far as positioning? or do you think it can be tuned well enough?
if that makes sense
I think there's only a problem if the software can't update fast enough relative to the encoder count rate
I am thinking more along the lines of stopping - or getting pushed away from the commanded position.
it's a matter of the PID gains - Geckodrive recommends nothing higher than 1000 CPR, and they prefer lower count encoders, in the 100-500 CPR range (*4 for quadrature)
ah - so probably can get it to do what I want. Emc as the loop.
but you can't change the timing of their PID loop, only a couple of the gains (they have filtering that has soime time constant - that's the timing I'm talking about)
probably, but a PC may not be fast enough while generating the up/down outputs in software
time will tell :) Need to play with it some more.
think of it this way: if you output "up" for one BASE_PERIOD, and that causes the motor to move multiple counts forward, then it may not be possible to tune
(not that a base period will cause the motor to move at all, in all likelihood :) )
ok - but I am not seeing that yet. It seems to be very responsive to where is should be.
yeah - motors are really slow compared to computers :)
the other thing that can cause some instability would be the granularity of the PDM output
I just don't like how far I can move it - compared to where it should be. - But cradek said that should come with better tuning.
although what does he know ;)
I think cradek pointed out that you should be able to spin it freely withing the DEADBAND, but it should be a brick wall if you try to move it outside that range
(I can move it a good .005 with a wrench before I chicken out)
with it straining against you?
ok, that makes sense. motors aren't as powerful as you'd think
I can cause my 27 in-lb motors to fault a gecko
with a CNC handwheel on the shaft - 3" radius
if you calculate it out, I only need 27 in-lb / 3in = 9 lb of force to overpower the motor
* skunkworks needs a current meter in line to see what is going on.
good luck :)
you need a current probe and a scope, I'd bet
you can see if emc is doing its job by looking at the duty cycle of the PDM output though
I did that. pretty cool. I would move the servo and watch emc correct with the pwm signal.
just halscope or halmeter the PID velocity output
I suppose I should have tried my gecko experiment with a bulk power supp-ly (I think I had set the power supply current limit to 10A, the continuous limit for my motors)
are you running it yet?
still freeken cool :)
making parts for my electric bike
but now I'm too fscking tired
so I'll have to continue tomorrow...
I was wondering if anybody could help me deciding a I/O controller to control a cnc router I have
sure - what's the question?
the machine is 3-axis, I have 4 stepper driver modules and one frequency motor controller
specs are here: www.hi.is/~stefath/cnc
I think I need 16 I/O outputs but the cards on the linuxcnc site mostly have A/D controller cards and few digital I/Os
the mesa card has 48 I/Os, in addition to the quadrature and PWM connections
the least expensive solution would probably be a PCI dual parallel port card
[21:16:54] <Jymmm> http://koutech.com/proddetail.asp?linenumber=48
I have one, untested, but does support linux
you'll likely need some sort of "analog" output for the VFD - the a PWM on the mesa card should suffice, with a filter on the output
they have the magic Linux word on that page - it's probably supporte din the stock kernels
I think alex said he has this one too
I have the one from ByteRunner. it sisn't work with the USC, though I didn't try very hard
I haven't tried it as plain I/O
ok, is the parallel card fast enough? the table is 2.5 by 4 meters and the steppers are microstepped so I would think the I/O speed would need to be quite high
well, the mesa card has an FPGA on it, and sometime soon there should be step/direction output from it (maybe sooner if you mention it when you talk to them about buying the card)
the USC from Pico Systems is also good, it generates up to 4 axes of step output in hardware, so you won't have any problem with speed
03jepler 07HEAD * 10emc2/src/ (configure.in configure): having figured out which tclsh/wish has bwidget, invoke that one when running pickconfig -- fixes a problem on systems with more than one installation of tcl/tk
03jepler 07HEAD * 10emc2/scripts/emc.in: having figured out which tclsh/wish has bwidget, invoke that one when running pickconfig -- fixes a problem on systems with more than one installation of tcl/tk
the USC has 8 outputs and 16 inputs, in addition to the step outputs and encoder inputs
software step generation generally can't go higher than 20k pulses per second, so if you need more then dedicated step generating boards are needed.
(boards like USC go to 1MHz and beyond, IIRC)
depends on the version :)
(when writing the driver, jmk and I noticed that mine does 2.5 MHz, but his only went to 2.0 MHz - both of which are of course > 1 MHz)
ok, but onboard the cnc there are power modules that control the steppers so I think the USC is not appropriate
what kind of input do the power modules need?
there are inputs for direction, stepping, microstep and some error checking, you can see the specs on www.hi.is/~stefath/cnc
right - forgot to look there :)
USC's outputs are step/direction
it looks like step and direction are supported, which is exactly what the USC outputs
there are several products from Pico Systems - the USC or Universal Stepper Controller is the one you want
the USC works really well.. :)
how broad is the support for I/O cards, I just googled some and found these: http://www.measurementcomputing.com/cbicatalog/cbiproduct_new.asp?dept_id=142&pf_id=1702&mscssid=UM8DU6DN3WP49NU8F5N9L4TWPBGC97EA
there are other systems, including a complete servo drive kit, and a PWM controller nearly identical to the USC
those can be supported, but aren't at the moment
there's not much reason to do so, since they aren't isolated, and only provide I/O, for about the same cost as the USC or Mesa cards
(the mesa isn't isolated either, but the USC has relay sockets onboard for the outputs, and transient suppressors on the inputs)
regardless of the I/O board you use, if you don't use a device that generates step pulses, you'll be limited to the 20000 pulses per second or so that jepler mentioned
what kind of coupling is the steppers to the table? how many turns of the stepper move the table 1 inch?
I'm not quite sure, wait
127.300 pulses/mm or 3233420 pulses/inch
you'll want to reduce that, I think
I think the controller is telling this, it displays "127.300 pulse/mm"
the table is 13 feet by 8 feet so...
is that 127 point 3, or 127 thousand three hundred?
not sure, the controller is from japan and in europe we use comma "," instead of "." to indicate integer parts
that's why I asked :)
seems stupid to have .3 pulses ;)
but it's also stupid to have almost 128k pulses per mm, on a 4x2.5m table
though at the same time 1mm/127300
that's 0.00000078mm per pulse..
not really sane either
it's a really stupid controller, that's why we want to use linuxcnc instead
should probably be 127 point 3 pulses/mm
I would look at the setting of SK1 and SK2 on the motor drivers. Ilet us know what they're set to, but I think you'll be best off with 8x or 10x microstepping (maybe even only 4x)
well, 25.4/127300 = 0.0002, so that may be the steps per inch, not mm
well, close anyway
i'll look at SK1 and SK2, wait
0.00508mm per step that's snae
madmanLap, I'm going to have to run in a few minutes
SK1 is A and SK2 is A
ok - A is good, that's 10 microsteps, which is just about the limit of usefulness
you should figure out what step rate you need by figuring out (a) how many steps per inch (or mm) you want, and how fast you want the table to move
do you recommend a parallel port solution with a USC board or should I connect directly from the parallel port to the power controller?
that all depends on step rate
the power controller has optical isolation
the motors will have some top speed for the type of cutting you want to do - you'll have to find that out
then take the lower of how fast you want it to go or how fast it can go, multiply by steps / (inch or mm), and see how many steps per second you need
if you need <20000 steps/second, then a parallel port / other I/O solution should work fine
if you need > 30000 steps/second, then you'll have to use something like the USC or other step generator
between 20000 and 30000, you may be able to find a computer that will work, but if time costs anything, then it's probably not worth it, since the step generator is $250
if I take the step generator I would need two, right?
also, with a hardware step generator your computer will be responsive..
you have 4 axes?
only 3 axis
<- what swp said
ok. the USC can control up to 4 axes
I haven't seen the specs for the usc, where can I see it?
there's also a spindle DAC in the works, which could be used for the VFD control, and that plugs onto the USC board
[22:01:31] <SWPadnos> http://www.pico-systems.com/motion/
[22:01:34] <anonimasu> http://www.pico-systems.com/univstep.html
err - hold on
right - there
madmanLap: If you can do any programming, you can probably write a driver for that measurementcomputing.com board. but it will suffer from the limited software step generation rate just like a dumb parallel port.
ok - time for me to run. see you all later
madmanLap, good luck
c ya steve
another cheap 8255 board is http://futurlec.com/PCI8255.shtml
(72 I/O points)
how does EMC interface with a step-generator board?
as in, what format is the data in?
if it allows a steprate so much faster
it sets the rate of the step generators..
Lerneaen_Hydra: faster to send a word then it is to send pulses..
[22:06:42] <anonimasu> http://www.pico-systems.com/univstep_regs.html
at the bottom
anonimasu: oh, so some form of paralellisation
how do f.ex. circles or complex forms where the speed is constantly changing come out in the USC?
madmanLap: emc generates a "position command" once per ms. by taking the difference between the current position and thew new position command, you can find the velocity each axis must move in during the next 1ms
if you're using a stepper machine, then you simply command each axis to step the appropriate number of times to go that distance
spreading those steps evenly over the 1ms of time
with software step generation, the PC has to run 1000s of CPU instructions for each step. with hardware step generation like USC, a special-purpose circuit generates the steps so you can get much higher rates.
oh, so speed changes cannot be done more often than the TP period?
(then again, that makes sense)
I think I have what I need, I'll be looking at the USC
thanks for everything
anyone know if ext2/3 needs to be defraged?
it does not
if I remember it right
yeah you can defragment ext3/2
but, they should not frgment..
kind of odd that there are defragmentation util's out there though
[22:50:06] <jepler> http://cbbrowne.com/info/defrag.html
(looks like a rather old article)
I think I may do a fresh install of linux, and convert my ext2 to ext3...
my / is ext2
and there's a funny thing that happens when you try to convert / to ext3
so to avoid all that... just backup my ~ and restore it afterwards
know what I would really like to try and do... make a watch... but I've got none of the skills to do that.
and don't feel like splurging a few grand to learn how either... [I'm borked broke]
maybe I can sell a kidney. http://news.yahoo.com/s/ap/20061112/ap_on_he_me/pakistan_kidney_bazaar
better yet sell somone else's
anonimasu: ... can I come visit, I'll buy a beer or two.
* Jymmm invites anonimasu over for dinner
if you wake up in a pool of ice, with a cell phone... don't worry...
there will be a note attached to the cell phone.
not this 999 stuff in some other countries
well, without one kidney, I guess it dosent take much beer..
sweet... Cohen got punched in the face a few times. http://www.thesun.co.uk/article/0,,2-2006520669,00.html
what up people?
03jepler 07HEAD * 10emc2/src/rtapi/rtapi_bitops.h: when in SIM always obey ASM_BITOPS_H_USABLE
nothing. nothing happens when you're not here. :)
03jepler 07HEAD * 10emc2/src/Makefile:
do not install pdf documentation when --enable-build-documentation was not given
install html documentation when --enable-build-documentation is given
install man3 documentation
03jepler 07HEAD * 10emc2/docs/src/Submakefile:
do not install pdf documentation when --enable-build-documentation was not given
install html documentation when --enable-build-documentation is given
install man3 documentation