I wish there was an easy way to lock down two of the jaws on a four-jaw chuck
locate part 1 of 20, lock down two jaws, then when you insert parts 2 thru 20, just loosen and tighten the other two and it stays centered
gonna put some tape over the screws for two jaws, so I don't forget and turn the wrong ones, but they could still creep
btw, the reason I'm using a 4-jaw is that the work is square
Sounds like faceplate work if it is worth setting up
how would you clamp it? needs to clamp from the sides, like a chuck
because the part gets faced and drilled
slip between 2 rails on faceplate, up against stop, open end of 3 walls has screw to force against stop, sides may have side screws ? ( not a lathe guy )
that sounds like a 4 jaw with one (or two) screws locked down ;-)
well it was your idea ! ;)
if the screws were longer, you could just put a nut on the outside
Box type vice using screws ?
but they'd only be the right length for a particular jaw setting
there are square 5C collets...
I'm pretty committed to using the chuck - the job is big enough (20 pieces) that I don't want to mess around loading every piece, but not big enough to warrent spending lots of time on tooling
that's the real answer I bet
2" square rules out the collets
is there even a through hole on that spindle? I don't remember
I think there is - maybe 5/78 or something
yes, 1-1/8" diameter
oh, that's a lot bigger than the older models then
this isn't bar work, the parts are already bandsawed to length
I'm gonna be attempting poor man's "gang tooling"
I have a toolpost and a drill chuck mounted to the table
hmmm. the tailstock actually had something like an MT2 taper, right?
I have a sleeve that MT3 on the inside, and straight on the outside - its in a V-block
the v-block is on two sets of paired shallow steel wedges
the tailstock goes on the ways, doesn't it? (ie, unusable in the CNC configuration)
and the wedges are on parallels
SWPadnos: the tailstock is manually usable
you can have the program pull the table forward to clear the chuck, then towards the headstock
then pause, print a message "drill a hole X deep"
and let the user slide the tailstock forward and drill the hole
slide it back, remove the drill, and hit continue
programming the operator :)
yep - that's like I thought I remembered :)
but for these parts, the hole depth is important, so I'm going the gang tooling route, let the CNC drill it
and peck drill at that :)
fenn: Xandra, a vrml machine tool modeller, simulator, uses kinematic descriptions to relate parts (yes windows but the preview was cool ) "Xandra is a 3D visualization environment for machine tools. With Xandra you are able to load machine tool models in form of VRML Files, configure them and simulate kinematic/dynamic models with the loaded machine tool." at sourceforge
one of the demos was a hexapod, but didnt load
some source available
jmkasunich: make a video of some of your production run?
tomp: could you take a screenshot for me?
will try, i installed it at work this afternoon, now at home on linux
cradek: if it works
on my "big" lathe I often have to move the tailstock several times to drill a hole. it has a tiny travel. your way sounds much better.
fenn: i cheated, its from the src zip bundle, didnt run under wine
the travel on the shoptask tailstock is about 1.5 inches
that's to the point where the screw disengages the quill, which you don't want to do (lots of stress on the very tip of the thread)
really only about 1.4 usable
and when all the way in, it ejects the chuck or whatever - if the tooling has a tang or is just longish, the travel drops more
will you have to spot drill first for these parts?
I'm gonna attempt to make a divot with the facing tool
just thinking thru the program now
rough face (so I have a consistent starting length)
make the divot
thanks anyway tomp :)
ok, i sent another IM
Here's somthing different
[05:52:27] <SkinnYPupp> http://cgi.ebay.com/CHAIN-INDICATOR-HOLDER-MACHINIST-MILLWRIGHT_W0QQitemZ180225147362QQcmdZViewItem?hash=item180225147362&_trksid=p3286.c58.l1288
Good movie... http://imdb.com/title/tt0768212/
anyone know how PID parameters are set from the INI file? (halcmd setp is run internally?)
I'm thinking about PID-tuning via a pyVCP panel again...
is this still correct for sources.list: deb http://www.linuxcnc.org/emc2/
dapper emc2 ??
I get "cvs.linuxcnc.org: temporary failure in name resolution"
maybe I just have something wrong with the dns servers...
ya beat me to it alex_joni
ok, here's a user point of view question
we have a ini setting called RS274NGC_STARTUP_CODE
where would you expect it?
in what section?
[EMC], [RS274NGC], [DISPLAY], [EMCIO], ..
yeah, what section
let me look
without looking :P
I know where it *is* now..
I was just curious where you would expect it ..
from your choices [RS274NGC]
we also have [TASK], [TRAJ], [AXIS_*], [HAL], [EMCMOT]
I was going to look to see what section made the logical place not where it was now
I also think [RS274NGC] is the logical place
it even is like that in a couple of config files
I still say [RS274NGC]
but that doesn't work atm
currently it's coded for [EMC]
which I think is wrong, as *theoretically* one could use emc with a different interpreter
that's not logical
[11:39:14] <alex_joni> http://www.linuxcnc.org/docview/2.2/html//config_ini_config.html#sub:%5BEMC%5D-Section
with the latest from CVS, is "pre-2.3 CVS HEAD" what I should be seeing?
so what settings do you put in [RS274NGC] that would be used by the interperter?
only the var file
and the startup code so far
brb coffee is ready
alex_joni: I have an Ubuntu 6.06 under vmware, it probably runs a non-realtime kernel. is it correct if the update manager never suggests an update to 2.2.4? (since it probably detects I don't have real-time)
ahhh the first cup of joe in the morning before the sun comes up
awallin: no, it's not
awallin: what's the installed version you have?
dpkg -l emc2
awallin: then it's correct ;)
you probably have emc2 in your /etc/apt/sources.list
a manual install of 2.2.4 then?
if you want a 2.1 or 2.2 install you need to change source.list
replace emc2 with emc2.2
[11:46:18] <alex_joni> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?UpdatingTo2.2
hm, that's a cryptic way of saing "add four characters to your sources.list"
wow I learned two new linux commands this morning from alex_joni
awallin: most users aren't aware there is a file called /etc/apt/sources.list
BigJohnT: which one?
dpkg -l and dpkg -L
alex_joni: the package manager suggests 'sudo apt-get dist-upgrade'
awallin: is this breezy?
hmm.. then don't do that
only apt-get update & upgrade
* BigJohnT searches for floating torch mounts on google to see what's been done before
'the following packages have been kept back: emc2 emc2-dev'
awallin: apt-get install emc2
BigJohnT: I read that as floating point torch mount.. and it was like.. wtf?
my plan at the moment is to mount the torch on a floating mount and come down and "touch" the material and trip a micro switch that gives me the Z height of the material
hrm. "module rtai_hal does not exist in /proc/modules" even when selecting the sim/axis config
[12:01:44] <awallin> http://pastebin.org/24966
if anyone is interested
awallin: you can't run emc2 on a sim machine
not even with the sim/* configs
those are called sim, but they refer to a simulated machine, not simulated RT
if you want to run on a simulated RT machine you need the emc2.2-sim package
and emc2.2-sim in apt/sources.list
TRUNK works with --enable-simulator
couldn't find package emc2.2-sim
yup, that's the equivalent of emc2.2-sim
different package then the one you have installed :)
ok so how different are TRUNK and 2.2.4 right now?
hmm.. a bit
let me look
[12:07:27] <alex_joni> http://cvs.linuxcnc.org/cvs/emc2/debian/changelog?rev=1.33
the changelog only has a very small number of changes
there are also named O-words in TRUNK
and a LOT of radical changes planned coming up
atm you can still run TRUNK with 2.2.x configs, but that's about to change
ok. with emc2.2-sim added to sources.list 'sudo apt-get install emc2.2-sim' still fails with couln't find package
please tell us more about the radical changes!
sorry.. the package name is emc2-sim
the repo is called emc2.2-sim
[12:10:31] <alex_joni> http://www.linuxcnc.org/emc2/dists/dapper/emc2.2-sim/binary-i386/
awallin: basicly cleaning up joint/axis issues
[12:11:17] <alex_joni> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?JointsVsAxes
while your at it why don't you add HOME_VEL to the list
BigJohnT: what do you mean?
a setting for the velocity as you leave the home switch and proceed to the home position
currently it's faster than rapid it seems
it should NEVER move faster than rapid ;)
if it does, it's a bug, which we need to fix
it will jerk my machine farther than a rapid will
but I agree moving away from the HOME_OFFSET to HOME could be done at something other than rapid..
BigJohnT: it would be great if you can get a halscope snapshot proving that :)
home search and home latch are nice peaceful slow moves then BAM you head for home
I tried but couldn't get it to work like I expeceted
alex_joni: thanks, emc2-sim now works
awallin: please look at the docs around wiki and the docs at linuxcnc
I think it would take less effort to just add a HOME_VEL to the code
and let me know where you feel we can improve things
and I would if I could
BigJohnT: it's not that complicated :)
probably 20 mins of hacking
while you're at making changes, I still hope and wish the offset values could be made available to the intregrator through hal-pins for example (think pyVCP panel showing offsets!)
what file is it in?
BigJohnT: oh, all over the code
awallin: hmm.. which offsets?
the ones in interp?
yes, G92 G54 and so on
they aren't relevant to the currently executed motion
as interp reads ahead (quite much actually)
you probably want the *current* offsets
which don't exist atm
awallin: I'm not sure how to do that
well, I can't imagine it's *impossible* to show these as hal pins. both the current ones and the ones not in use but stored in interp would be interesting
awallin: as it is now: interp reads the file and calls CANON commands
at that level offsets are already applied
CANON queues up stuff, and sends them to task
then task dispatches them to motion and IO
so it's 2-3 levels down (through all of these offsets would have to be added)
awallin: otoh, one of the nice things to have is offsets in UVW
where UVW is attached to the end-effector of a puma for example
BigJohnT: still there?
from a user perspective it is quite strange that there are g-codes to set the various offsets, but no way of qurying what is stored in all the registers
I think some are available as #vars
I went for another cup of joe
* BigJohnT is looking at iniaxis.cc
looks like a few things would have to be added to that file...
looks like one line added to command.c
there is even a FIXME on line 1693 of control.c for home vel
alex_joni: still there?
* BigJohnT can grep now...
BigJohnT: I am now
alex_joni: me too now
alex_joni: still there?
(only for 10-15 minutes though)
ok, I found 3 files that need a couple of lines of code to add a HOME_VEL
iniaxis.cc, command.c, control.c
IF I knew how to do a diff I could try that
and a couple more :P
iniaxis works with EMC_*
command and control work with EMCMOT_*
(it's easy to confuse the two :)
there is another layer inbetween
nothing majour though
there is a fixme on line 1693 of control.c for the home vel
in iniaxis.cc around line 215 the homing parameter would have to be added
you must have another version than me
line numbers don't match :)
it's been a while...
yup.. homing is a different file atm
dang it I need to try and remember how to get the trunk
ok, found it .. give me a couple of minutes to see what I can come up with
I did a grep on home_search_vel and found the files real fast
BigJohnT: you were right.. no other level involved
I'm updating my cvs now
cvs up -dP
hmm, nope.. I was right in the first place
homing params can be set from the GUI too..
so that needs updating too
give me 5 more mins, I'll have a patch for you
BigJohnT: can you join #emc-devel ?
not that I know of
oh on irc?
Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emc/2008-03-23.txt
hi cnc people
jmkasunich: wheres that case cooling calculation page you did?
what are the consequences of using 192.168.1.x on a local lan?
tomp: windowses can have sometimes problems
tomp: I've used ssh here on my home system a lot and haven't had any problems - going between linux and mac os x machines
192.168.x.x are defined as a local lan
i was wondering about 192.168.1.x vs 192.168.0.x ( 192.168.0.x is what the spec suggests is used for privatenetworks along with 10.0-255.0.x and 172.16-31.0.x ranges , but to use 192.168.0x for class 'C' [home lans] )
right -192.168.x.x is a set of class C "private" networks
10.x.x.x is a class A private network
ok , i didnt know 1.x was as good as 0.x, thx
and the 172.whatever is for class B private networks
I have used both
it helps to use only one at a time :)
big thx, now i am comfortable :)
some consumer routers use 0 - some use 1. (linksys uses .1 netgear uses 01
I inadvertently changed from one to the other when I did a router replacement, and suddenly the printer stopped working :)
i use IPMASQ and a hub, so it's always my fault
well, the network printer thing is something to look out for. otherwise you can change as often as you like (as long as you're using DHCP)
tomp: a hub?!
not a router, just a hub
tomp: a HUB?! why not a switch?
hmmm. how much file moving do you do on your local network?
I guess if it's just one or two users, hub vs switch is pretty meaningless
a lot of files right now, trying to get things organized on one fileserver, just me and 4 desktops, one laptop
tomp: get a switch
will look, any suggestions for the cheap bastard in me?
unless you have any 10base-T devices - that would bring the entire network down to 10mbit (unless you have a "dual-speed" hub)
they're under $50 from every major manufacturer at this point
$10 from frys
10 < 50 :)
unless you get a gigabit switch,
how many of the client PCs are in use at a time?
ok, any others doing downloads or file copies while you're using the one?
SWPadnos: he said he has a fileserver
yes, streaming on the masq server while pushing files around
SWPadnos: If it's running smb, it'll be shitting all over the place =)
everyone talking to one machine means the difference between a hub and a switch is even more meaningless
tomp: streaming contant on the file server?
masq server == file server, or two machines moving data back/forth in addition to the PC you're using at the time?
hmmm, havent done the file server yet, was still organinzing. the stream is just BBC radio while i work. i imagine masq server == file server eventually
here's the deal: if you basically have only one pair of computers communicating at a time, then a switch isn't going to improve things much over a hub
if you have relatively large amounts of data, or frequent packets, going between more than one pair of computers, then a switch is a very good thing
ok, thx ( chewed on that one a bit :)
plus I remember there being some odd 100bT hub rules.. Like between a switch -> hub connections have to be less than 10ft or something like that.
oh- one other thing that may not have been clear: if you have more than one computer talking to the same computer (two clients to a server, for instance), then a switch won't help
no, the rule is that you can have at most a 100m cable from node to hub, 5m hub to hub, then 100m to the next node
That could be..
max 2 hubs + 205m cable total from any node to any other node within a collision domain
a switch separates collision domains
so you can have as many as you want
I don't think we have any hubs anymore.
so 2 computers cant ssh into the same computer?
I've still got one dual-speed (a 10-bT hub + 100bt hub + "switch" in one unit)
err - yes, they can
it's more a matter of how much data you want to move around
they can with a swx but not with a hub?
it's the amount of data
if you have 5 computers all trying to use 2mbits of bandwidth, then the total is 10mbits and you don't need a switch
if all those computers are connected to the same server, then a switch doesn't help anyway - all a switch does is connect any pair of ports on their own network (during a data transfer)
so if you have 2 PCs doing something, and another 2 PCs doing something else, it's like they're on separate hubs
but when you want to connect 2 clients to the same server, they still need to multiplex (the ethernet protocol takes care of that)
a switch makes your lan nice and tidy (basically)
so it's only when you are either using a very large percentage of available bandwidth (like 70-100% maybe), and you want multiple data paths, that you need to use a switch
there is another advantage of a switch: if a node goes belly up, it won't take down the whole network, which it can with a hub
tomp: Think of a hub as a dinner party where everyone is trying to have a conversation at the same time.
oh, and another one - switches can set up a full duplex path between nodes - hubs can't
it's like IRC ;)
a switch gives you a lot of private conversations
a hub is a channel like this
it makes no difference if you're trying to have a conversation with 6 people - there's still one of you, and you can only type into one window at a time
SWPadnos: wanna bet?
with 3 separate conversations, it's a net gain
unless you have some magic key event replicator, any keystroke you type will go to at most one window ;)
SWPadnos: dont make me release my evil bot on you ;)
where are the standard "comp" HAL things documented? it used to be 'blocks'??
man <whatever the comp name is>
like man and2
yeah... just strange that they have dissapeared from the manuals
if you want it, the pdf version of the manual pages is http://linuxcnc.org/docs/2.2/EMC2_Manual_Pages.pdf
ok. now I get 'operation not permitted' when trying loadrt freqgen
look in dmesg; the real error is there
for instance, if I specify step_type wrong, I get this in dmesg: [620012.465986] FREQGEN: ERROR: bad stepping type '33', axis 0
no channels configured...
you should specify step_type= with at least one value
step_type=type0 gives an error too
* alex_joni was faster :P
I was busy :P
yeah yeah.. excuses
hrm. the update function is called freqgen.make-pulses, but the pins have the zero in it: freqgen.0.step what's the logic there??
it's only one update for all freqgen instances
the .0 denotes the first instance
but you could have <nr> of them..
so freqgen.update-pulses, and freqgen.*.*
I vote for changing all parameters to pins in freqgen! (parameters are not accessible to pyVCP and thus evil.)
awallin: sounds ok to me..
patches would speed this up :P
trying to get the vfd to spin the motor just now... maybe coding later next week...
awallin: there's still time till 2.3 ;)
hm, not getting any output from freqgen. I'm setting freqgen.0.frequency (the pin) to something (like 1000), but freqgen.0.frequency (the parameter) stays zero
did you put 1 on enable?
ah, let me check.
btw.. man freqgen says *OBSOLETE* over here :)
it will probably be removed by 2.3
for enable, or for the whole freqgen?
stepgen does the same thing, freqgen is only a subset of stepgen
I want a pulse-tran for setting the speed of the VFD. what's a good way to do that? siggen followed by comp?
a step/dir pulse train?
just a square wave of varying frequency, 0-25 kHz is the default in the VFD but that can be scaled down a bit
there are separate pins for start/stop fw/rev
stepgen in velocity mode is like freqgen
in pwmgen the frequency is set by a parameter, so no go
looking at stepgen
you can run pwmgen in PDM mode
and use the value pin
stepgen will do it, but won't give you a 50%-ish duty cucle square-wave
stepgen in quadrature mode would
how is pwmgen set to PDM?
pwm-freq = 0
so in PDM mode, is value given in Hz?
I think you'll get essentially random speeds if you use PDM mode
it uses a high frequency pulse train, but the frequecy isn't controller, it's just a result of whether the output needs "more" or not
so it's stepgen I need then? does it have a velocity input?
hmm, right.. so stepgen type=0 ctrl_type=v is what you want
I think quadrature is the way to go, as jepler pointed out
so maybe step_type=1 ctrl_mode=v
(if 1 is quadrature)
hmm, right.. so stepgen type=2 ctrl_type=v is what you want
1 is up/down :)
velocity-cmd seems to be undocumented?
awallin: I got it from man stepgen
(did you . scripts/emc-environment ?)
otherwise you get the manpages from the installed emc2, not the run-in-place version
stepgen.N.velocity-cmd float in (velocity mode only)
Commanded velocity, in length units per second (see parameter position-scale).
[21:06:08] <alex_joni> http://www.linuxcnc.org/docview/html//man/man9/stepgen.9.html
I must have old manpages. anyway now stepgen seems to work, next I'll power on the vfd...
right. if nothing works, try plugging in the thing!
yay, it gets the frequency reference
yay, it runs forward!
strap it to the desk
yay, it runs backward
yeah, from 0 to max rpm makes it jump a bit. it's a 1.5kW motor, max rated 3000rpm
yeah, wouldn't want that motor running away
you can set accels for stepgen
how about spindle-speed-out, does that just jump abpruptly to whatever is set by an S word?
S0 -> 0
S1000 -> 1000
no ramps, no waiting either
the vfd has accel settings too
awallin: does it have an at-speed output?
yes, haven't hooked that up. I'm going to use a spindle encoder too
you can use !at-speed with feed-hold
it seems for a pulse train of frequency f=1/T i need a period of T/4 ?
with step_type=2 ?
sounds right (without pen & paper though)
with period=40 000 I get around 6 kHz (1/4*period would be 6.25khz)
what was the real-time test for testing how low period can reasonably be set?
[21:34:22] <JymmmEMC> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?TroubleShooting#RTAI_Latency_test
JymmmEMC: that's the rtai latency test, latency-test is the one provided with emc2
alex_joni: the GUI one seems inaccurate
it's done using emc2 components
so emc2 will behave very similar to the GUI one
not to the rtai latency test
I've gotten way different readings between the two
I get about 11 usec on latency-test, I think a tiny bit lower with rtai's test
I get on the magnatude of 4x difference
got the scaling working too, now I get the full speed range with the 0-6kHz pulse train. thanks for the help guys.
awallin: glad it works
* BigJohnT is still d/l files...
* alex_joni leaves too
good night all
good night alex
sometimes dividing is better than shifting :(
I like multiplying best
we can't have a mine shaft gap