probin: that is great news
I scared him away
probin: welcome, glad to hear your fpga card is going well
jmkasunich_ is now known as jmkasunich
SWPadnos_ is now known as SWPadnos
alex_jon1 is now known as alex_joni
unfortunately, I think I spotted a bug on pluto_step last night
when running at the maximum frequency command, sometimes the step signal HIGH time is 2x as long just after a direction change
I fear that this may be 2 step pulses issued without an intervening LOW time
is it always 2x, regardless of the HIGH time setting?
I didn't think to test that
what setting have you been testing with?
smallest possible (1.6uS?) step HIGH and LOW time
I don't remember what the scope settings said, but if I understood right it was displaying captures of single events, not envelopes or averages
ok, that's still multiple cycles of the internal FPGA clock though, right?
the FPGA clock is 40MHz (25ns), but the step generators use a "divided by 64" (1.6uS) signal for everything
oh. in that case it may be a single extra clock cycle glitch
rather than 2 stape
nothing in the step generator should be having transitions except at the 1.6uS rate
the next things I'll test are longer step HIGH times, and using the "signal present for more than" trigger mode to see if there are any 3.2uS glitches except at direction changes
set the HIGH time to 3.2 uS, and trigger on anything longer than 4.8 uS (+ a little)
if you get no triggers, then you know it's a single cycle problem
then, go to a 3.2+ uS trigger, to see if it's gone with higher HIGH settings
(it could be that the minumum step size is 2 for some reason, not that you always get N+1)
I'm not 100% sure I follow all that
and I have to work all day before I get to play with it more anyway
heh. me too :)
I think it's most likely that it's 2 step pulses without an intervening LOW time because the LOW time is ensured only by a limit on the frequency commands given by the HAL driver
ok. that may be the key
like the HAL software step generator, the accumulator is supposed to pause counting during a direction change
does it correctly wait the DIRHOLD (or DIRSPACE)?
whichever one is the "time after reversal before generating a step"
dirhold I think
another corner cut: there is a single DIRTIME that is enforced before and after a direction change
you need a 7i43 ;)
I suppose I do
the single DIRTIME isn't a big deal, since direction changes only happen at slow/no speed
even with a 20 uS hold time, it's only doubled to 40 us
yeah, for applications with any acceleration it's not a big burden
(assuming it's not buggy anyway)
rayh: hi ray.
jepler: ever looked at comedi?
alex_joni: no, I only know it's some acronym or name associated with rtai
I just took a look there yesterday (again)
they have a kernel and userspace interface to their drivers, which I thought could be usefull to look at
jepler: they support a wide range of data aquisition boards
do you think there might be a use for this in the context of emc, or for some other thing I've mentioned?
or maybe for something you have in mind ? ;-)
scope scope scope
badger badger badger
when shopping for a new scope, I dismissed any PC-attached data acquisition cards because I assumed the software would be windows only .. maybe I was wrong.
on the other hand, the two GUI apps linked from the main comedi site are ugly ugly ugly
standalone scopes have much much (much!) better triggering options, and the UI is almost universally better
I also imagine '80s HP test equipment has to be better quality than '00s non-name USB capture equpment
I'm sure of it
if you can get '60s HP test equipment, it's even better
I don't know what year the old tek 465b was
the old signal generators had true frequency output, not something that "averages out" correct
HI jepler SWPadnos
these days, with PLLs, you only get something that's statistically correct
was away in a meeting
I think I may have found a good embedded PC for EMC
I'll be getting 2 today, so I'll have test info later on
jepler: I was thinking of a HAL wrapper around the comedi drivers
Oh let me know how the boards test out and what they cost.
alex_joni: go for it! that'll also force us to tackle the issue where HAL driver "A" requires non-HAL module "B" to be loaded, which lurks for users of rtai_net and rtai_serial type devices too
alex_joni: do you have some comedia-supported hardware now?
rayh, will do. I think the board itself is $450 in singles, plus storage, CPU, and RAM - so around $600 complete
but it's from an industrial supplier, so hopefully the parts won't change much from day to day
jepler: unfortunately no
did you ever get that other one to boot?
That jetway itx is supposed to be in production for 4+ years.
Sent em back
uh-huh. did the second one ever boot? :)
did you look at the hardware the guy who was in here yesterday uses?
also ITX iirc
yep - looks like ITX + a custom FPGA card
[13:58:34] <alex_joni> http://atelierrobin.net/mini-itx.jpg
mini ITX, from the image name
[13:58:45] <alex_joni> http://atelierrobin.net/PCIO100.jpg
"Runs EMC 2.x standalone"
hmmm. I don't see the output card for that - I'm assuming it plugs into the card-edge connector on top
almost the same mini itx I've tested here.
yeah I also assume that's where the outputs are
this is the one I've got coming: http://www.acrosser.com/Product/Embedded%20PC/Embedded%20Systems/mbedded_system_celeronM_ar-es0892.html
there are others that don't have the PCIe and PCI slots
Ah I saw that. Looked really good.
that case has a 120W power suplpy built in, and the qty. 1 cost is $557 (+CPU/RAM/HD)
I need a 5i22 in my system, so I went with that one. Also, if the integrated graphics screw things up, I can stick in a cheap Nvidia PCIe video card (if I need video at all)
Sure. I've not had any rt issues with what they call the unichrome onboard.
excellent - I hope to have the same lack of problems
it's bad when a 2000A power supply goes tits-up due to my controller :)
"I have developed a real time protocol for controlling machines via
the USB port on a Windows Xp based computer. "
alex_joni: sure you have
jepler: not me.. I just found that :P
SWPadnos: is this the same project you were talking about back at fest, where you were hoping to have 16 or 32 PIDs running at a few kHz?
There is a guy in the netherlands that does it with parts of emc.
"It has almost, +-20ms, real time on screen graphical animation
going on when the tool path is being executed on the machine. "
yes, but it's morphed quite a bit
SWPadnos: I thought for awhile it was shelved, or at least the HAL part of it
jepler: now we have a politically correct definition of real-time (+/- 20 ms)
jepler, yep, that's accurate
alex_joni, that's more or less within one frame (and well under a frame time for cinema)
I bet you could have a tolerable low-speed machining system with 20ms position commands
USB can be great for any system that doesn't need feedback fro control
cradek, I don't see how your changes would have affected parameter 5220
he didn't say they did
it made a bunch more required fields in the var file
that's just the error that pops up
the interp incorrectly adds them I guess, I didn't dig into it too much
ok, so it's a bogus message
well I think the interp generates a new var file with all "required" fields set to 0
... which is invalid since #5220 can't be 0
please fix it better if you care
(I just put proper var files in all the configs)
^ one obvious trivial fix
hmmm. do you suppose that may be happening even if 5220 is 1 (writing the file due to missing UVW params, but still writing 0 to 5220)
you now know everything I know, sorry
there was a random bug that somehow set it to 0.. It happened to me once but I could not reproduce it..
I have not seen that for years now
I think it was tkemc actually (it reads and writes to the var file itself, so does the interpreter, neither of them has any locking scheme)
but that's only speculation, I never tried to find/fix it
one of these days I'll make it all work through NML
on a separate note, I've done some testing on that embedded PC, and it looks - unspectacular
what's that supposed to mean?
latencies are usually pretty good, but max_latency is ~12000-13000
not very good?
that's ok I guess
what CPU freq?
it's not too great for a 1.8 GHz processor, I think
might get the same from a 800 C3
it's an Intel board as well
no SMI issues or such I hope
maddash had reported good luck with a very similar chipset, but no luck here
I don't think so
I actually couldn't get any video from the onboard graphics (at all - no boot), luckily I had gotten an Nvidia card just in case
there was an SMI-like latency anomaly once. something like 500-600 ms latency at one point
and this was without X running
I had just done a networking restart though, so that may have something to do with it
(hard to tell when you have to switch terminals to see, and there's no backscroll)
no, text mode (I think)
it's an Intel 965, so I'm not sure it's supported well by Dapper
(I think it's a 965)
I had a SBC on which I had to enable FB and disable scrollback
in order to get proper things displayed
some of the new intel chips have incompatible PIDE (including cdrom)
actually, I was using the 7300 at the time, since I never got any display from the onboard video
I used my laptop drive with EMC2+dapper on it :)
so I didn't even need to boot from CD
but this was worse - no BIOS POST screen at all from onboard video
when did ide change to pide.. Do many poeple got confused?
I'm also kind of bummed because I've found the perfect A/D chip - 16 bits, 6 channels, simultaneous sampling, 250 KHz
not available until the end of the month :(
$25.80 in low qty
cheap, for what it is
* alex_joni agrees
and what your going to make on the project ;)
Analog Devices AD7656, in case you want to know
I make more if parts are available
I guess I should go check the core duo unit. bbl
can you isolate one of the 'processors' on a core duo?
cradek: which was the usable SMP kernel? 2.6.17 or 2.6.20?
I guess it would be .20
they both work but .17 had a keyboard issue (for me) and .20 has no bootsplash
I might as well download half the files in /experimental instead of all of them :)
hmmm. presumably linux-source contains everything in linux-headers?
but you still want to install linux-headers
if you want to compile & distribute stuff
it's better to compile against linux-headers
I'll likely want to compile
makes it for less stuff to distribute
12 bit, 2+2 channels, 500 ksamplings
that's nice, but 12 bits kills it for this project
that's a burr-brown (TI) part, right?
just in case you're not reading #emc, Stuart Stevenson sent me some pics of his Dah-Lih conversion http://www.anderswallin.net/2007/08/dah-lih-emc2-conversion/
err - I have 2.6.20 running on this machine (woohoo), but I can't run the latency test
[21:07:57] <SWPLinux> http://pastebin.ca/643510
when I ./run the latency test, all the /dev/rtf? files go away. do I need to set up some udev thing?
argh. and I can't run EMC either
here's a compile error I have gotten a couple of times:Syntax checking python script pumagui
Traceback (most recent call last):
File "<string>", line 1, in ?
File "emc/usr_intf/axis/scripts/axis.py", line 3145
SyntaxError: invalid syntax
well, I guess it's a syntax check error, but still
what version of emc is this axis.py from ?
the closest line to that in my copy ^^^
those are lines 3144 and 3145
hmmm. maybe it didn't update correctly - CVS says "M"
"from" is a python reserved word
"M" means you modified it -- trying to add a minimum spindle override number?
couild have been
to use the -from of a tk widget, write "from_=min_spindle_override" (see the added _ in from_)
U just deleted and updated. we'll know what happens in a few seconds
it did work once today, that's the interesting thing
huh, I see that MIN_SPINDLE_OVERRIDE was listed in the 2.1.0 release notes...
at least, I'm pretty sure it did
I wonder how much this laptop hard drive is slowing things down
whoah. it may be time for a clean checkout
I get endless "waiting for s.axes != 0 -- Had to poll again..."
unfortunately, that message may obscure the actual emc startup failure
what was the CVS incantation to reset to TRUNK/HEAD and overwrite anything?
you might want to remove the print or increase the wait time from 100 times per second..
-AC or something
cvs -H up
I have no -H available
-A to go back to TRUNK from another tag; -C to get rid of modified files (but often in my experience that fails)
I was playing "teach a man to fish", but I guess it didn't work :-P
for me, -C often fails to get rid of the old file, and prints some errors
then I give up and get a fresh checkout instead
sorry - food deprived and overheated :)
hmmm. anybody notice that isolcpus is a map, not a quantity?
actually it's a list, not a map
I guess I should reboot and try it :)