#linuxcnc-devel | Logs for 2014-05-15

[09:42:09] <KGB-linuxcnc> 03Sebastian Kuzminsky 05master 13395a1 06linuxcnc Merge remote-tracking branch 'origin/2.6' * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=13395a1
[11:33:14] <cradek> hmm, rtapi_vsnprintf can return -1
[11:33:28] <cradek> it's not at all clear to me that it normally acts like vsnprintf
[11:55:06] <cradek> oh maybe that's normal for vsnprintf
[12:39:11] <pcw_home> Ick still homing to index thump issues not sure if is motion/halfile/hm2 driver/firmware
[12:46:37] <pcw_home> cradek: you must use home to index, have you ever seen any thump on index issues?
[13:15:11] <cradek> pcw_home: nah, not since I fixed it
[13:18:06] <cradek> I hope something hasn't rotted. My mill hasn't been updated forever. I have no idea what's running on it.
[13:19:43] <cradek> pcw_home: is the setup having trouble using any of the deriv inputs?
[13:21:52] <pcw_home> Not sure where the problem is :-(
[13:27:04] <pcw_home> I can understand Dabits dual feedback setup being tricky, (axis.index-enableN must be fed to both encoders)
[13:27:06] <pcw_home> but Todd has a fairly standard config
[13:27:06] <cradek> pcw_home: comments say that if using previous_target there will be a one-cycle ferror proportional to velocity
[13:27:24] <cradek> comments in pid.c
[13:28:00] <pcw_home> in Todds case the velocity is .05 IPS
[13:28:16] <pcw_home> (latch velocity)
[13:28:24] <cradek> yeah the comment says "but velocity is usually very small during index search"
[13:28:40] <cradek> does he have any FF0?
[13:29:04] <cradek> one of us needs to reproduce this :-/
[13:30:18] <cradek> pcw_home: it would be useful if he would turn on debug, and scope the resulting errorI, errorD, commandD, commandDD pins
[13:31:13] <pcw_home> Yeah I I wonder what hardware I used to test the error-prev stuff
[13:31:58] <pcw_home> Dabit suspected an error in motion, but what motion does is really simple
[13:32:56] <pcw_home> just sets commanded to FB when index-enable is low (when in proper homing state)
[13:34:31] <pcw_home> I will try and set this up today
[13:34:33] <pcw_home> hmmm is there a max/min component?
[13:34:45] <cradek> there's the various limit components
[13:35:03] <cradek> it'd be best if you could set up hardware
[13:35:14] <cradek> failing that I recommend he turn on pid debug and make us some plots
[13:37:07] <cradek> previous-target reintroduced a thump on 1/12/13, but then I fixed it on 1/25
[13:37:10] <pcw_home> I can probably avoid restarting linuxcnc by resetting the encoder (so theres a jump every time it homes, not just the first time)
[13:37:32] <cradek> well you only need to plot it once...
[13:37:53] <pcw_home> yeah but apparently it doesnt always happen
[13:38:05] <cradek> that feels unlikely to me
[13:38:28] <pcw_home> maybe count step size dependent
[13:38:30] <cradek> some homings have a larger position jump than others...
[13:39:26] <cradek> I don't see anything that is likely to have broken it (I studied the changes from v2.3.0 forward)
[13:39:32] <cradek> changes to pid.c, I mean
[13:40:32] <cradek> hm2 stepping the position and dropping index-enable on different cycles would sure break everything
[13:40:39] <cradek> so it could sure be elsewhere
[13:41:24] <pcw_home> Yeah I worry that is a long time bug that people have just lived with (since it only happens on the first home after power-up)
[13:42:36] <cradek> well I'll try to help you if you find it's broken
[13:42:41] <pcw_home> I looked briefly at encoder.c and its seems correct
[13:42:56] <cradek> I am absolutely sure my mill no longer thumps. it was extremely violent before I fixed it.
[13:43:17] <cradek> there is no way you could miss it
[13:43:41] <cradek> they're not using software encoders are they?
[13:43:54] <pcw_home> No both hm2
[13:47:16] <pcw_home> strange. well I will put some hardware together
[15:51:12] <kwallace> I keep losing my autosave.halscope. Is there a way to load a halscope saved config when using "loadusr halscope"? Unless I'm missing something, it doesn't look like halscope takes any arguments.
[15:51:26] <cradek> kwallace: it's on the menu in halscope
[15:59:08] <kwallace> My problem is that I use "loadusr halscope" in a hal file to invoke halscope. Another issue is that halscope looks for autosave.conf and loads it, which is fine except my application doesn't close halscope, it does an "unloadusr all" in my .hal file. This clobbers the autosave.conf and the next run doesn't run halscope with the saved config.
[16:03:57] <kwallace> It would be nice to be able invoke halscope with the desired config file name. Another option might be to copy my conf file over autosave.conf before I start my application, but I'd like to avoid that.
[16:04:24] <cradek> perhaps you could look into fixing halscope to still do the autosave when you kill it that way
[16:04:48] <cradek> but this sort of feels like a "doctor doctor it hurts when I do this" kind of problem
[16:09:47] <kwallace> But I would not normally touch halscope. I just type in a couple values into a ui.py display, halscope shows if the signals are correct, then I close the ui.py and it takes halscope with it when it leaves. The config is okay if I close with the X box in halscope before closing the ui, but that's not intuitive.
[16:17:08] <cradek> halscope already takes some arguments - not entirely clear what - it would be awesome if you would add the ability to load a config with the command line
[16:20:23] <kwallace> cradek: Well, I suppose I could look at the halscope source and see if anything comes to mind.
[17:02:24] <kwallace> cradek: Just in case you or anyone else might be interested: http://www.wallacecompany.com/pulse/
[17:03:11] <kwallace> I am playing with a transistor ignition and need a pulse source to trigger it.
[17:25:54] <CaptHindsight> kwallace: I'm doing similar with piezo injectors
[17:27:01] <CaptHindsight> they are a low cost source of piezo stacks for making fast high pressure valves
[17:32:00] <CaptHindsight> what periods are these? loadrt threads name1=fast period1=50000 fp1=0 name2=slow period2=1000000
[17:32:19] <CaptHindsight> what is the reference clock?
[17:32:49] <kwallace> CaptHindsite: I need to learn more about piezo injectors. I'd like to fit an injector to my garden tractor but I'd like to not need a high pressure pump to feed the injector.
[17:33:48] <kwallace> I believe the reference is the base thread.
[17:34:26] <CaptHindsight> so is 50000 x the base thread for the fast period?
[17:35:25] <CaptHindsight> kwallace: there are some low pressure input injectors ~20-40psi
[17:35:54] <CaptHindsight> they have a hydraulic amplifier inside
[17:37:36] <CaptHindsight> http://www.youtube.com/watch?v=ftchx1TDNJo
[17:38:23] <CaptHindsight> http://www.youtube.com/watch?v=67GSNUR7eqM
[17:38:39] <kwallace> It has been a while since I studied thread timing. I use the edge component for the pulse length and it uses nanosecond units. I use a UI entry of 1, then multiply by 1,000,000 to get the input value for 1 millisecond.
[17:39:46] <CaptHindsight> ok, so it is in nanoseconds
[17:42:03] <kwallace> CaptHindsight: Here is a screenshot: http://www.wallacecompany.com/pulse/Screenshot.png
[17:44:33] <CaptHindsight> kwallace: do you want to inject into the port or directly into the cylinder?
[17:46:12] <CaptHindsight> all the port type injectors are low pressure
[17:46:16] <kwallace> The plan is to put the injector into a throttle body on the existing manifold feeding two cylinders.
[17:46:33] <CaptHindsight> http://www.youtube.com/watch?v=H6p0sVdMeeQ
[17:47:32] <CaptHindsight> ok so look for older PFI injectors (port) vs direct
[17:50:39] <CaptHindsight> http://www.ebay.com/itm/8-piece-24-pound-Chevy-Corvette-fuel-injectors-/171321539749 for example
[17:50:56] <kwallace> I have some injectors from a Mazda I had many years ago, but I assumed the fuel pump was high pressure. For small engines the fuel system really needs to be dirt cheap and simple.
[17:51:27] <CaptHindsight> yes 20-40 psi, you'll just need to find ones with a low GPH
[17:53:51] <kwallace> Another option is to drip fuel in with a peristaltic pump, but I'm thinking the pump tube might not last very long.
[17:54:30] <CaptHindsight> http://www.rceng.com/technical.aspx
[17:56:55] <CaptHindsight> it's roughly .50 lbs. of fuel per hour for each horsepower for normally aspirated engines per cylinder
[17:57:59] <kwallace> I have a Honda Hurricane that I made some throttle bodies for years ago. I think I have some injectors from RC Eng. sitting on the shelf somewhere. Another unfinished project.
[17:59:02] <CaptHindsight> I'll try out your pulse generator soon
[17:59:56] <CaptHindsight> up until now I've been using tandem systems, Linuxcnc for motion and another to control printheads, lasers, injectors etc
[18:01:07] <kwallace> My next step is to hook up a BIP373 from http://www.diyautotune.com/catalog/bosch-bip373-coil-driver-mod-kit-p-230.html
[18:02:25] <CaptHindsight> http://maya-auto.com/wp-content/uploads/bip373_datasheet.pdf
[18:02:50] <kwallace> All I can find are circuits that hook them up directly without any support or protection.
[18:03:40] <CaptHindsight> they are a fancy darlington with built in protection
[18:04:14] <CaptHindsight> they have built in thermal protection and ESD
[18:04:27] <kwallace> Seems too good to be true. But wait ... there's more.
[18:05:48] <kwallace> I'm used to adding a condensor and zeners for over voltage protection.
[18:08:23] <kwallace> Like this Velleman kit : http://nexstar4.free.fr/images/allumage.png
[18:10:23] <CaptHindsight> from the limited bosch data sheet it looks like everything is in that IC
[18:12:04] <kwallace> I guess I'll find out soon, maybe tomorrow.
[18:59:20] <dgarr> kwallace: for options try: halscope -h --> Usage: halscope [-h] [-i infile] [-o outfile] [num_samples]
[18:59:41] <dgarr> so you can say loadusr halscope -i yourconfig.halscope
[19:00:00] <dgarr> commit d625b66353b
[19:08:49] <kwallace2> dgarr: Thank you, I'll look into it. ... Oh, did you just add that feature? Double thank you.
[19:09:13] <dgarr> added in 2008
[19:09:57] <kwallace2> Well, thanks any way.
[20:45:45] <kwallace> dgarr: I made the halscope -i change and it seems to work.
[20:54:42] <pcw_home> I've noticed that halscope forgets it config file rather frequently, Is this a known bug?
[21:43:36] <kwallace> pcw_home: I think my problem was that halscope doesn't autosave unless one closes halscope with a X or quit. Just killing it seems to clobber the autosave file. A wild guess may be that scope opens the autosave file and leaves it open until scope closes. If something funky happens, it clobbers the open file.
[21:44:50] <kwallace> Using the '-i my.halscope' option works.
[21:53:42] <kwallace> CaptHindsight: The BIP373 works just as advertised - nearly naked, only wearing 330 Ohm resistor on the base pin. I hooked the resistor to a parport pin and it work fairly well, but may need a buffer chip or pull up resistor since parports usually don't source very well. The next step is to wire up a gear tooth sensor and disk on an engine.