yay: I found a media that will at least reach the integrity check boo: it's an alpha of hardy, not final hardy
KimK_afk is now known as KimK
only 661 upgrades to install from hardy alpha6 :-P
darn, missed the A/D discussion
IMO, trying to initiate A/D reads in advance of the need for the data seems like a path to madness, at least for some percentage of our users
I have my own plan for A/D, since I have specific needs (I want to sample signals that may have significant high frequency noise and/or ripple on top of the desired lower frequency signal
16MHz (or so) SPI clock, connected to a serial A/D that continuously samples at clk/16 = 1MHz
FPGA generates a 32 bit register containing 10 bits of sample counter, and 22 bits of "sum of samples"
driver reads register in servo thread, takes difference between last sum and this sum, divides by difference between last count and this count
presto, average value over the servo period
and I just need analog anti-aliasing for the 1MHz or so sample rate - easy
I don't know if average dac-users want an instantaneous estimate or a 1ms average
clearly I know which you're after :-P
well, unless they have suitable anti-aliasing filters, or very clean signals, they want an average, whether they know it or not
sampling at 1KHz is incapable of handling any input over 500Hz - and if there is any, it gets aliased to lower frequencies
1KHz noise gets aliased to DC and becomes indistinguishable from the signal
my intuition, which maybe is just wrong, that just like you'd rather get the current feedback position of an axis instead of the average position of the axis over the last servo cycle, you'd rather get a current velocity feedback from a tach instead of a 1ms average velocity
but maybe that intuition is simply mistaken
the net effect of the average is to introduce a delay of 1/2 the servo period
if velocity is changing so fast that it makes a difference, then your servo period is too slow anyway
the control loop normally has an inherent delay of 1 period (minimum)
that assumes that measurements are made right at the beginning of the period, and that outputs have an immediate effect at the very beginning of the next period
if there is any noise in the system, you'll need filtering somewhere
you could do a 500Hz analog filter, but it's inherent delays are longer than the 1/2 period that comes from averaging
instead, I'd do a 50Khz analog filter, then let the averaging handle everything below that
it's starting to sink in
your application is something other than velocity feedback that I mentioned, though?
yes, current feedback in a drive
since the current contains PWM ripple, filtering is a big issue
jmk: you might consider min and max registers in your averaging SPI interface
that way you know the P-P noise/ripple.
You could also use the P-P for widening a halscope trace
to show the noise if int was used for general analog measurments
I still don't have an intuitive grasp of how many gates are in an fpga, so I try to keep the functionality there to a minimum
to me, a 12 bit magnitude comparator is a big thing
but you sure could fit min and max in a 32 bit register with room to spare, reading the register resets it for the next period
A single SPI interface is pretty small, you are probably talking ~100 gates for a fairly fancy one
before I got distracted I was looking at the analog devices website
I already had the A/D converters picked out
12 bit comparator is really an adder, using the built in carry chain, its small
12 bit, 1Msps, 6 pin SOT package, and I could have sworn they were under $2 each
Do you need isolation?
I'd like it
I'd be using hall devices for the current feedback, so I don't need high voltage isolation
but noise isolation would be good
I was thinking of the newish analog devices fast isolators
One issue with isolation is the isolation turnaround time for SPI
the A/Ds I was using may or may not have been SPI, but they definitely did not need turnaround
clock, strobe/chip-select, and a single data line IIRC
wish I could find them again
OH ok 1 channel so no data in...
ah, no wonder I couldn't find them on the AD website, they're TI
[03:44:00] <jmkasunich> http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=296-19308-1-ND
You still need to deal with 2 isolator delays (CLK up and data down)
Burr-Brown or TI division?
(skew is small and controlled but total delay is not)
a bit more $ than I remember, $4.28 in ones, $2.66 in hundreds
pcw: the new non-optical isolators have quite small delays
oh, maybe not as small as I thought
With Hall effects, Isolation may be unneeded
depends on power level - I'm looking at a kW or so, 300V DC bus
so there will be noise
Other solution (which I am thinking of adding to my SPI interface) is a receive data skew adjust
I'd rather have one power supply and ground for the drive interface, isolated from the PC
is it just me, or have component prices gone up?
We dont see anything going down...
I think they've gone up a little too
ADUM1400, 4 channel high speed digital isolator
digikey wants $6.17 at 1000 pcs, $9.88 for singles
for some reason I thought those things were around $4 in moderate quantity (100s)
Thats what we use on 7I65 (plus 1402 I think)
I think they were, but I could have been looking at the 10mbit ones
I think we paid around $5
The $5 was for the 90 mbit
$9.88 is crazy
in quantity they're less
I used one of those it the design of a cheap but accurate analog-in to analog-out isolator a year or two ago
OK being paged by SO
crap, how did it get to be almost 11pm
at least I got my xmas shopping done a day earlier than normal
I'm doing reverse christmas shopping - returning stuff :)
how long did that cordless saw last - about 6 cuts?
people already gave you stuff?
no, things I didn't use for the lathe haul
I'll probably return those cordless tools. they're supposed to last an hour on those batteries, but it looks like they don't
I'll wait for the lithiums to come down in price
cordless tools suck, but when there is no plug.....
I've resisted for a long time, but I figured that tying down the lathe would be a good time to try some
it was handy for sure, but 10 minutes or less of use and then a 1.5-2 hour recharge sucks
although to be honest, I'd make 8 cross-cuts in 2x4's with a sharp hand saw about as fast as that saw did
I brought one of those too, but didn't feel like using it toward the end of the day
yeah, I was certainly feeling less than energetic
I'm glad I didn't have to drive for the next 6 or so hours
I realized that the dinner we had was the only one I had on the trip
you didn't stop all evening?
otherwise it was beef jerky and donuts
not for food anyway
good thing you had breakfast ;-)
should have taken that frozen pizza
hmmm. maybe :)
put it in front of the heater vent, yummy
dried beef and sweets from my stash were fine
Congrats on your new lathe, SWP. I hope that eventually there will be a web post about your adventure with lots of trip photos, lathe photos (saw the one earlier), and specs
if only I had brought a camera ...
I took some photos while we were loading
Now I know you're pulling my leg
swp will have to take photos of the unload
I thought of it but forgot it
[04:03:40] <jmkasunich> http://jmkasunich.com/pics/swp-lathe-2765.jpg
[04:03:47] <jmkasunich> http://jmkasunich.com/pics/swp-lathe-control-2764.jpg
[04:03:53] <jmkasunich> http://jmkasunich.com/pics/swp-lathe-headstock-2763.jpg
[04:03:58] <jmkasunich> http://jmkasunich.com/pics/swp-lathe-otherside-2766.jpg
Nice, thanks! I've never seen a machine with those all-plexiglas (lexan?) "fighter-pilot-canopy" style covers, interesting. What's the approx part size and RPM? (What kind of a fast ball could that thing throw?)
I think the max turning diameter is about 6 inches
not sure about the spindle RPM either, but it's in the 3000-5000 RPM range I think
I think 3500 is the fastest speed of the factory setup
mine's set to go up to 2500 (vfd) and it seems like a good compromise
(I'd move the varispeed if I was doing a lot of very small work)
a 2x4 sawed to the right length sets the spindle gearing...
cradek: you are missing the ballscrew that adjusts that? or just don't want to mess with it?
the only part I'm missing is the brake
Well, it will make a nice project. Too bad it had to get started during a record ice storm, cold temps, etc. Here in the midwest, we are expecting a warmup after Christmas. Maybe coming your way afterwards?
the ballscrew and motor is there, but I don't seem to need it, I just set it to a nice middle setting and use the two gears
if you don't need full HP over the speed range, VFDs are great
do you know the ratio between hi and low?
my guess would be 3.something to 1
setp mux2.0.in0 .0038
setp mux2.0.in1 .0154
SWPadnos: dunno if you've looked in there
KimK, we usually get your weather a few days later, and right now the forecasts look pretty good for the weekend (around 40 every day except Thursday and Friday)
I haven't removed the tarp
three phase AC motor with two pulleys
jmkasunich: so looks like 4:1
driving two pulleys on the countershaft
one pair is small driving large, the other is large driving small
two electric clutches on the countershaft, connect on pulley or the other to the shaft
brake on the countershaft, spring applied, electric remove
countershaft drives one varispeed sheave
then there is a floating shaft, with two varispeed sheaves (I think), and a final varispeed on the spindle itself
a motor/ballscrew moves the floating shaft up and down, and the varispeeds adjust themselves to suit
nice arrangement IMO
I think the motor turning that ballscrew is a stepper but I'm not sure
it has crazy limit/home switches on the ballscrew
needs low end torque, not speed
yeah, and needs to be cheaper than an axis
it is a low-precision ballscrew
KimK is now known as KimK_brb
KimK_brb is now known as KimK
EMC: 03seb 07TRUNK * 10emc2/src/hal/utils/halrmt.c: Same fix as for emcrsh earlier - check the return value of read()
KimK is now known as KimK_afk
KimK_afk is now known as KimK
KimK is now known as KimK_afk
EMC: 03bigjohnt 07TRUNK * 10emc2/debian/extras-Ubuntu-5.10/usr/share/applications/emc2-halmanual.desktop: add hal manual to the menu
EMC: 03bigjohnt 07TRUNK * 10emc2/debian/extras-Ubuntu-8.04/usr/share/applications/emc2-halmanual.desktop: add hal manual to the menu
EMC: 03bigjohnt 07TRUNK * 10emc2/debian/extras-Ubuntu-7.10/usr/share/applications/emc2-halmanual.desktop: add hal manual to the menu
EMC: 03bigjohnt 07TRUNK * 10emc2/debian/extras-sim-Ubuntu-8.04/usr/share/applications/emc2-halmanual.desktop: add hal manual to the menu
EMC: 03bigjohnt 07TRUNK * 10emc2/debian/extras-sim-Ubuntu-6.06/usr/share/applications/emc2-halmanual.desktop: add hal manual to the menu
EMC: 03bigjohnt 07TRUNK * 10emc2/debian/extras-Ubuntu-6.06/usr/share/applications/emc2-halmanual.desktop: add hal manual to the menu
EMC: 03bigjohnt 07TRUNK * 10emc2/debian/extras-sim-Ubuntu-5.10/usr/share/applications/emc2-halmanual.desktop: add hal manual to the menu
cradek_ is now known as cradek
EMC: 03bigjohnt 07v2_2_branch * 10emc2/debian/extras-Ubuntu-5.10/usr/share/applications/emc2-halmanual.desktop: add hal manual to the menu
EMC: 03bigjohnt 07v2_2_branch * 10emc2/debian/extras-Ubuntu-6.06/usr/share/applications/emc2-halmanual.desktop: add hal manual to the menu
EMC: 03bigjohnt 07v2_2_branch * 10emc2/debian/extras-sim-Ubuntu-5.10/usr/share/applications/emc2-halmanual.desktop: add hal manual to the menu
EMC: 03bigjohnt 07v2_2_branch * 10emc2/debian/extras-Ubuntu-8.04/usr/share/applications/emc2-halmanual.desktop: add hal manual to the menu
EMC: 03bigjohnt 07v2_2_branch * 10emc2/debian/extras-Ubuntu-7.10/usr/share/applications/emc2-halmanual.desktop: add hal manual to the menu
EMC: 03bigjohnt 07v2_2_branch * 10emc2/debian/extras-sim-Ubuntu-6.06/usr/share/applications/emc2-halmanual.desktop: add hal manual to the menu
EMC: 03bigjohnt 07v2_2_branch * 10emc2/debian/extras-sim-Ubuntu-8.04/usr/share/applications/emc2-halmanual.desktop: add hal manual to the menu
BigJohnT: seems you managed it
yes I think I did :)
you only need to add it to the files file
[14:05:44] <alex_joni> http://cvs.linuxcnc.org/cvs/emc2/debian/extras-Ubuntu-8.04/emc2.files?rev=1.1
one for each arch where you added it
does that have anything to do with the order that they are put in the menu?
no, I think they get put alfabetically
those files only specify what files get included in the deb
if you have files which are not stated as such, there's an error during building of the debs
hmmm 2.2.x/debian/extras-sim-Ubuntu-8.04 does not have the emc2.files
well, we haven't built packages for 8.04 sim yet
so that might be the explanation ;)
hmmm trunk sim 8.04 has one
EMC: 03bigjohnt 07TRUNK * 10emc2/debian/extras-Ubuntu-6.06/emc2.files: add hal manual to the menu
EMC: 03bigjohnt 07TRUNK * 10emc2/debian/extras-Ubuntu-7.10/emc2.files: add hal manual to the menu
EMC: 03bigjohnt 07TRUNK * 10emc2/debian/extras-Ubuntu-5.10/emc2.files: add hal manual to the menu
EMC: 03bigjohnt 07TRUNK * 10emc2/debian/extras-sim-Ubuntu-6.06/emc2.files: add hal manual to the menu
EMC: 03bigjohnt 07TRUNK * 10emc2/debian/extras-sim-Ubuntu-8.04/emc2.files: add hal manual to the menu
EMC: 03bigjohnt 07TRUNK * 10emc2/debian/extras-Ubuntu-8.04/emc2.files: add hal manual to the menu
EMC: 03bigjohnt 07TRUNK * 10emc2/debian/extras-sim-Ubuntu-5.10/emc2.files: add hal manual to the menu
EMC: 03bigjohnt 07v2_2_branch * 10emc2/debian/extras-Ubuntu-8.04/emc2.files: add hal manual to the menu
EMC: 03bigjohnt 07v2_2_branch * 10emc2/debian/extras-Ubuntu-7.10/emc2.files: add hal manual to the menu
EMC: 03bigjohnt 07v2_2_branch * 10emc2/debian/extras-Ubuntu-5.10/emc2.files: add hal manual to the menu
EMC: 03bigjohnt 07v2_2_branch * 10emc2/debian/extras-Ubuntu-6.06/emc2.files: add hal manual to the menu
EMC: 03bigjohnt 07v2_2_branch * 10emc2/debian/extras-sim-Ubuntu-5.10/emc2.files: add hal manual to the menu
EMC: 03bigjohnt 07v2_2_branch * 10emc2/debian/extras-sim-Ubuntu-6.06/emc2.files: add hal manual to the menu
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/config/stepper.lyx: shorten chapter title
EMC: 03bigjohnt 07v2_2_branch * 10emc2/docs/src/config/stepper.lyx: shorten chapter title
* BigJohnT heads out for a while
oops, John T is right. I didn't notice that was on -devel.
building emc2-sim also builds the hostmot2 packages
cradek: I have a file "engraving-psfonts.tar.gz" that I think I got from you a while back
most of the fonts in it are .PFB, not .ttf - can TTT use those?
those are postscript fonts and they work fine
are any of them single line fonts?
yes I think so
[16:41:44] <cradek> http://timeguy.com/cradek-files/emc/linefonts.png
"Firefox is already running but not responding. To open a new window, you must first close the existing Firefox process, or restart your system"
(which I've seen before)
jmkasunich@shoptask:~/emc2/nc_files$ ps -A | grep fire
that is new
that shell is on another box
you haven't lived if you haven't accidentally shut down the wrong machine
any reason all the font names in that tarball start with ^ ?
I'm sure that's how I found them
odd - when I double click on that font, I get a bunch of oddly shaped dots, instead of characters
ick, they are not single line. the "l" in helv goes from Y0 to 2953 back to 0
bet the / does the same then
I need to engrave "5/8"
yep they are all closed paths (retracing the path I guess)
making a replacement die for a hole punch that my dad made 20+ years ago
it'll still work fine I bet
I bet that is because TT defines a glyph as a path that can be filled
yes I agree
postscript is the same way I'm sure
these are font formats for raster devices, not vector
I still want to add the hershey true-single-line fonts to TTT
one of these days....
you need hershey
I dunno whether it makes more sense to add it to TTT, or just make a new program hershey-gcode
there probably won't be much common code at all
well you could reuse TTT's great UI ...
the command line you mean? that is a great UI
well, the saw just got done cutting my blank - should stop messing with the labeling and start working on the part (first things first)
don't want to use stamps? that would be simplest
yeah, but that isn't CNC
the originals weren't labeled at all, the label is just showing off
I've been tempted to make a punch holder that goes in the spindle of the mill somehow, to let me punch with consistent spacing and in a straight line
I made a holder a while back, just a block of steel with a groove in it
I draw a line on the part and line the block up with that
that would be good enough to get them upright
hm, maybe simple is best
I sometimes put hash marks along the line for spacing
but that fails because it really wants proportional spacing if you are doing actual words, not just a few chars
for 3 characters you could very easily edit the gcode if you want single line
yeah proportional is good.
I'll see how well/fast making the actual part goes
if I'm gonna label it, I need to do it before heat-treat
[17:00:55] <cradek> http://timeguy.com/cradek-files/emc/5-8-singleline-biarc.ngc
seb_kuzminsky: hi, what happened to the other buildslave?
alex_joni: paul_c's buildslaves? we had a hissy slap-fight about whether or not to enable CONFIG_PCI_LEGACY in the kernel config, he refused, so i removed his slaves from the build farm...
* seb_kuzminsky rolls eyes
bbl, gotta go work on the Newtonmas presents
can't believe I still do that
heh, you'll learn eventually
one too many alt-a's ?
* alex_joni didn't figure that out .. yet
it's nice, but I still mess up sometimes - it's not the program's fault.
anyways.. back to part 2 of the christmas dinner ;)
odd - this time I didn't load an infinite loop or anything, but axis seems to have locked up
locked up how?
no keyboard or mouse response
which pointer is it showing? what did you do last?
I can mouse to another window and it will take focus, when I go back to axis, it doesn't redraw the part that was coverd by the other window
normal looking pointer
I've seen it get stuck with the "busy" pointer, after opening a program
it happens more often on a very slow machine or X connection
about 15 mins ago, I started the spindle with F9, (after running multiple programs and doing all kinds of other stuff), then I hit esc to stop it after filing some burrs
then I wandered around looking for an allen wrench
eventually removed the work, put something else in the chuck, went to jog, nothing
oh so you didn't just load a program or anything like that. it was just sitting after an abort
that's a new one on me
what's top say? is AXIS busy?
anything in dmesg?
mouse pointer was on the touch-off button, that tooltip is displayed (and gets repainted when I return to the axis window)
top has milltask and axis at the top, but neither is hogging CPU
task 10%, axis 1%
any python traceback on stderr...?
no - I opened emc from a shell, nothing there but normal stuff
EMC has been running for a couple days
nothing weird in dmesg either?
I've been doing ctrl-K to clear the backplot on occasion
[39452.284511] RTAPI: ERROR: Unexpected realtime delay on task 1
[39452.284519] This Message will only display once per session.
[39452.284523] Run the Latency Test and resolve before continuing.<6>input: AT Translated Set 2 keyboard as /class/input/input3
the unexpected delay message is old - it happens every once in a while (once per run), but I haven't chased it
the weird thing is the "input: AT......." stuff at the end
yeah that is strange
your keyboard cable isn't somewhere you can trip on or yank is it?
what's the easiest way to put another message in dmesg? I want to see the current timestamp, and if it properly handles the next line
is there a missing \n at the end of that message, or is it cut off by a limited buffersize somewhere?
I think the message is missing a /n, I scrolled back and found another instance with a different thing tacked on the end
keyboard cable is tie-wrapped and routed from the board to the PC, it didn't get whacked or anything
and when I'm in another shell, it seems to work fine
does the fact that the touch-off button tooltip gets redrawn when I switch back to the axis window mean anything?
the portion of the preview window that was covered by my shell doesn't redraw at all
the non-preview parts that were covered wind up solid gray
except for that tooltop
I just realised that the tooltip is being drawn even when my shell has focus - it draws right on top of the shell
2118 PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
7 jmkasuni 15 0 6404 1800 1332 S 10.6 0.7 133:40.68 milltask
21188 jmkasuni 15 0 55376 32m 5444 S 1.0 13.1 2364:24 axis
the time consumed by axis is incrementing (slowly)
any thoughts or info requests, or should I just kill it and get on with life?
EMC: 03jmkasunich 07TRUNK * 10emc2/src/rtapi/rtai_rtapi.c: add missing newline
jmkasunich: I'm clueless, sorry
EMC: 03jmkasunich 07v2_2_branch * 10emc2/src/rtapi/rtai_rtapi.c: add missing newline
if it's not too late, killing axis with SIGINT might produce a traceback on stdout that I'd take a look at it
other than that I have no ideas either
I just tried a plain kill, and it didn't die
so, its not to late
jmkasunich@shoptask:~$ ps -A | grep axis
21188 pts/0 1-15:24:34 axis
jmkasunich@shoptask:~$ kill -s 2 21188
jmkasunich@shoptask:~$ ps -A | grep axis
21188 pts/0 1-15:24:34 axis
am I doing something wrong?
not that I se
does axis intentionally catch some signals, so it can clean up?
kill 9 got it
there are tracebacks on the shell, but they are all reporting "Keyboard Interrupt", I suspect they are axis refusing my earlier kill attempts
yeah, that's the traceback I wanted to see
just a sec
Exception in Tkinter callback
Traceback (most recent call last):
File "/usr/lib/python2.4/lib-tk/Tkinter.py", line 1345, in __call__
File "/usr/bin/axis", line 2533, in task_stop
/usr/bin/emc: line 607: 21188 Killed $EMC2_BIN_DIR/$EMCDISPLAY -ini $INIFILE $EMCDISPLAYARGS $EXTRA_ARGS
taking that at face value, it was trying to send an 'abort' but was stuck
or at least it was trying to send an abort right when the signal arrived..
the section from Execption to KeyboardInterrupt was repeated 4 times, once for each kill attempt
I'm half tempted to reboot before I do anything else, just on general principles
EMC: 03cradek 07TRUNK * 10emc2/docs/src/gcode/main.lyx: degression is not a typo
check 2.2 branch too?
jepler: I did; it's not wrong there
thanks for catching that cradek