skunkworks: let's come in here... what do you use for sacrificial drilling?
my top was made of plywood. I was just asking you what you where going to use :)
maybe polycabonate or plexiglass?
so far I've used the perfboard with all the holes in it
I made the holes in this table so they (should) line up
so *maybe* that'll still work
worth a shot.
double sided boards will be very easy for you now. you won't have to think ;)
what do you have for a vacuum pump?
how many small vertical holes did you drill
perfboard has 100 per square inch
I haven't drilled them yet, but the internal channels are .400 apart
should not matter - some will line up. - I think my holes where 1 inch on center 1/8 diameter. worked quite well.
wow that's big holes
is big or small better?
well, I finally have X+Y servo mounts
for the wrong size motor, but that's easy enough to fix
[00:17:06] <skunkworks> http://www.electronicsam.com/images/KandT/vactop.JPG
[00:17:24] <skunkworks> http://www.electronicsam.com/images/KandT/vacback.JPG
or, I could procrastinate by looking for AC servos that are closer to NEMA34 than my NEMA42 motors are :)
cradek: it is all about square inches. (I never calculated it;))
aren't you supposed to be working on an analog PCB
yes - I actually received a few today!
skunkworks: wow that's a much simpler design than mine :-)
oop, gotta run, bbl
I did get the LEDs to blink on the analog card, but I haven't been brave enough to hook up $700 worth of parts to the +/- 15V supply yet :)
well, 50 mA can kill it, if it's going into the wrong pins ...
silly dog... he's whining for his evening walk
doesn't know it's raining outside
3 bidders.. life is good.
"let" (ie, kick) him out, and see how long it takes for him to whine to get back in
it doesn't work that way
no fence around the yard
and no trench to let him play int
and slightly insane dog will run away/into traffic, etc if he sees squirrel, other dog, cat, skunk, etc
skunk - eww
he chased a skunk once
by the time I caught up with him, the skunk was chasing him
unfortunately I don't think he learned a thing from that experience
so jmkasunich, do you think I've destroyed the tantalum caps that were reversed? :)
I did have the current limit set to a couple hunddred mA, and there are 3 caps in parallel
(no smoke or bad smeels, which is good)
how high did the voltage get before it current limited?
so probably a watt or two
(dissipated in the caps)
well, it may be more than 3 caps, or I have other problems
damned VSS/VDD labels
who reversed them?
you or the assy shop
crap either way, for the first 3 boards
istr asking if you could rename the nets
shit. I just missed the digikey cutoff too
yeah, I know
I never use the default names, I use "+5V", "-15V", etc
yep. I think I used vss/vdd because there are some chips that have hidden pins that connect to them
I must have overlooked things too - I reviewd the schematic
eh - no biggie. I just have to find them all (and hope it's not all of them :)
hmmm. I wonder if one of the tantalum cap models has the wrong polarity for the pin numbers
Vdd is +15, and Vss is -15, right?
(had to check my cheat sheet)
C11 on the input conditioning sheet is wrong - I just changed those 3
yeah, I just spotted that one
C33 seems to be wrong
(and 35, 37, 39)
hmmm - bummer
oh man, just about every cap on that page is backwards
somehow I always get 90% of 50/50 chances wrong
C19 and 31 are OK ;-/
how'd you manage to get C44 wrong?
and C42 ...
probably cut/paste errors, actually
I can understand having one that should be upside down rightside up, but that one you had to flip over
hmm - should C42 have a supply connected to it?
or is that pin like the reference pins, powered from inside the chip
DVCC is connected on my sheet - that may have changed after I sent you those
it's a little late now ( I should have spoken up earlier ), but one trick I use - I NEVER put a polarized cap upside down on the schematic
I probably did it once, then copied it a few (hundred) times
I draw a negative rail at the bottom, a positive rail at the top, and ground in the middle
at least that means the ~1W was spread around 12 or more caps :)
9 on the A/D sheet
yep, +3 on the 3 copies of input conditioning. haven't checked the DAC yet
although 42 and 44 aren't on the 15V rail, they're 5V or 3.3V
the dac sheet looks OK
(but should be double or triple checked anyway)
looks OK to me as well
actually I think all of the ones on the ADC sheet are on 5V or less
ok, so only 12 caps to reverse - no problem on the first 3 boards, and the assy house will get the right XYRS data for the remaining 7
AVcc, and REF
hmmm. true, except for 19 and 31
they're not backwards
right - yay!
so then the 3 caps may have taken most of the power from the reversal
I noticed that one was a little warm, but only a little
do you have spares?
I was feeling around for awrmth and didn't notice it, I only noticed because the side of my finger brushed against one of the C11s
no, unfortunately they're all in Colorado
that's where your quick-turn board house is?
I do have 3 boards though, so I can pull from the spare if necessary
just outside Denver
if you really have a spare, I suggest that you red-tag the board you are working on and declare it the spare
reverse the caps on one of the others
I'll try reversing on this one, and see if ot works. I have the boards numbered so I can tell them apart
if it doesn't work, I'll reverse all the caps on a different board before applying power
that works - my point is to not kill one board fixing another
if you do have toasted chips on this board you're gonna need them
and if you have a deadline, better to not spend time trying to troubleshoot weird stuff
if it doesn't get better right away, pause, take a deep breath, and go to the next one
heh - yep. all I need right now (for the next few days anyway) is a single board that I can test the FPGA code with
I did manage to use the simple I/O config to get the LEDs on the board to turn on and off, so I know that at least part of the board isn't screwed up :)
well, time to move caps around. bbiab
and the address and bits were even the ones I expected for the LEDs
man. it's funny how everything happens at once. I'm working on this project, the dental appointment I've waited 6 months for comes up, my company gets audited, and today I had to figure out and pay estimated tax. what a pain.
well, I think I found the real problem, and it's a real problem
the OPA2131 has - you guessed it - reversed VSS and VDD
but it isn't rotationally symmetric like the quad op-amp
what kind of package is it?
SO-something, or some tiny bastard
thats easy then
nig - 0.05 spacing ;)
lift pins, ECO wire
lift the leads off the pads, and run wire-wrap wire
oh wait - I got that one right - just didn't see the dot on the PCB
I swapped all the caps that seemed to need it but it didn't fix the problem. I'm ringing out a bare board to see if the -15 line connects anywhere suspicious
SWPadnos_ is now known as SWPadnos
ok - the good news is that I see a difference between the powered-up-and-not-working board and one of the other boards
the bad news is that the difference is by the DACs (I measure ~35 ohms across the VCC filter cap, vs. open circuit on the other board)
the worse news is that if one or both DACs are blown, those are the most expensive chips on the board, at ~$54 each
when you say "across the VCC filter", that could mean anywhere on the board almost
any chip or cap connected to VCC
there aren't all that many
but that's true - it could be any of them
I just happened to notice it next to the most expensive chips on the board, so murphy's law kicked in ...
I'm running a loop composed of: g0z-.001
but halscope of zpos-cmd looks quite odd
I expected something between a sawtooth and a sine with .001 p-p
[17:08:40] <jepler> http://emergent.unpy.net/index.cgi-files/sandbox/commanding-1mil-moves
but I get something irregular
anda bout .0015 p-p
if I program G61 I get <.0005 p-p which is a surprise too
ack. stupid browser doesn't know what to do with your .png because it isn't named *.png
wow that's a strange behavior
two browsers actually, Mozilla and Forefox (on Windows)
err - Firefox
heh - IE opened it just fine :)
of course - that's because it's insecure ;)
I like living on the edge
what's the step resolution?
I'm used to graphical programs automatically tacking on extensions, so I didn't type one
.png added to URL
SWPadnos: SCALE 8000
but looking at -cmd, SCALE doesn't enter into it
true, and -fb is tracking pretty well
[17:44:08] <cradek> http://timeguy.com/cradek-files/emc/mils.png
I get something a lot more believable with G61
with G64 I don't get any motion, it just sits halfway between
hmmm. how about ferror settings?
as a test, loop -cmd to -fb and set ferror to 1e-10 or something
also, is DEADBAND used in motion, or is it just for PID?
as an unfortunate aside, I notice pause/step/resume don't work right
step just resumes
config files: http://git.unpy.net/view?p=zenbot.git;a=tree
does step work for anyone?
I guess my tree is a little old
no, seems broken here too
did you try to fix step-from-beginning?
I'm digging in the history now
yes, I thought I had ...
[17:55:14] <cradek> http://cvs.linuxcnc.org/cgi-bin/cvsweb.cgi/emc2/src/emc/task/emctaskmain.cc.diff?r1=1.91;r2=1.92
"step while pause should resume & step" ?
err - almost :)
hmmm. is there even a concept of "run-to" in the interp?
I don't think so
hmmm. so single-step is an execution mode?
step is implemented in motion, by pausing when the motion ID changes to the next number
ah - handled in checkPreconditions
so, step actually stops slightly past the "corner"
though I'm reading comments, not code
alex touched it last!
gads I have no idea how this all works
maybe I'll go out to the garage and drill some holes or something
maybe I'll step away from the computer for a bit
* alex_joni is back
cradek: what kind of step is broken? last I touched that part it worked both ways
it sounded like step from pause resumed, rather than single-stepping
I'll look at it in a second
"step" didn't seem to have any effect when running, and was just like "run" when paused
in tkemc step works before a program is run
in axis that is greyed out
in 2.1, not in TRUNK
that was referring to what part? your comment or mine?
in TRUNK, "Run" and "Step" are both available when a program is loaded and machine is on
and in fact in that case ("Run" never pressed) step seems to work
and step after pause resumes in TRUNK?
* alex_joni is still compiling
hm but something in stat is right because when I start with "step", the pause button should press in on the toolbar but it doesn't...
I can confirm that both axis and tkemc behave the same, step after pause actually resumes
jepler: can you remind me how to use 2-3 commands of gdb?
n/m figured it out
stepping is really a big POS
can someone check if I got it right this time>
yes I'll test it now
yes it looks like step works right
there's still a problem with the pause button in axis
I'm looking into it now
axis expects interp_state to be paused for the "pause" button on the toolbar to press in
but if you start with step, interp_state is "waiting"
I don't think you can resume once you started with step
I do it by hitting pause twice
EMC_TASK_PLAN_RESUME is what is sent by the second (effective) click, and if I send that manually it resumes from a start-with-step
so .. any way to get "pause" into the stat buffer when a step is done?
don't sweat it
can you try something?
I shut down my devel box
put 'emcStatus->task.interpState = EMC_TASK_INTERP_PAUSED;' at line 763 in emctaskmain.cc
that does something bad
the second "step" acts like stop for some reason
forget it then .. I'll try to look at it when I have something to test
the other thing what I would have tried would be to put:
interpResumeState = EMC_TASK_INTERP_PAUSED;
but it probably does something bad too
good night all