#linuxcnc Logs
May 17 2020
#linuxcnc Calendar
09:58 AM Tom_L: so i see
10:05 AM * JT-Cave checks the radar before going back to the splitter
10:05 AM Tom_L: dunno why it wasn't writing to the file
10:26 AM HexaCube: hmmm, the 'Planet CNC' controller looks pretty interesting too
10:32 AM HexaCube: will something like this still work with linux cnc? https://www.aliexpress.com/item/33013843649.html Or is that Mach 3 only?
10:32 AM JT-Cave: you can just use the parallel port or a parallel port and a BOB
10:33 AM JT-Cave: https://www.cnc4pc.com/c10-bi-directional-parallel-port-interface-card.html
10:37 AM dirty_d: so with arch linux arm running preempt_rt 5.4.40 kernel on rpi 4 I'm getting 28us latency so far
10:37 AM dirty_d: All I did was apply the rt patch and disable kernel debugging options, not sure if there are more options that improve latency further
10:38 AM dirty_d: I had to modify the linuxcnc source, it seems older kernels reported "PREEMPT RT" when you run uname, this kernel reports "PREEMPT_RT"
10:38 AM dirty_d: so it was incorrectly detecting a non-realtime kernel was running
10:44 AM Tom_L: HexaCube, likely not since it probably needs it's onw ethernet driver
10:47 AM Tom_L: besides, it's just not as sexy as a mesa card
10:50 AM JT-Shop: lol
10:52 AM Tom_L: are the chicks used to you yet?
10:59 AM dirty_d: Tom_L, so whats the deal with DPLL, what is it?
11:00 AM Tom_L: you'd have to ask pcw for specifics but it makes things run better. it's an entry in the .vhd file for your bit file
11:00 AM Tom_L: if you're using a mesa card you probably have it and don't know it
11:03 AM dirty_d: Yea I have a 7i96. I wonder if linuxcnc is actually using it though.
11:03 AM dirty_d: http://linuxcnc.org/docs/html/man/man9/hostmot2.9.html
11:03 AM dirty_d: it looks like it needs to be set up to do so
11:04 AM dirty_d: there's a dpl section
11:04 AM pcw_home: on systems with significant jitter, the DPLL retimes the sampling times of critical data (stepgen position, encoder positions etc) to reduce the jitters influence on motion
11:05 AM dirty_d: pcw_home, I've only set my stuff up using pnconf, so you think my setup is alreay taking advantage of DPLL?
11:05 AM pcw_home: typically just 2-3 statements in the hal file
11:06 AM Tom_L: might be useful on the rpi
11:06 AM pcw_home: newer versions of pncconf may set it up, not sure
11:06 AM pcw_home: yes its useful on the RPI
11:08 AM pcw_home: for stepgens:
11:08 AM pcw_home: # latch 50 usec before nominal read time:
11:08 AM pcw_home: setp hm2_[HOSTMOT2](BOARD).0.dpll.01.timer-us -50
11:08 AM pcw_home: setp hm2_[HOSTMOT2](BOARD).0.stepgen.timer-number 1
11:08 AM dirty_d: pcw_home, there is no dpll stuff in my hal file
11:08 AM skunkworks: The steppers sound perfect with DPPl inabled
11:08 AM skunkworks: enabled
11:08 AM dirty_d: pcw_home, that's all I need to add?
11:09 AM pcw_home: for stepgens, yes
11:10 AM Tom_L: i see skunkworks is using -200 but he's running servos?
11:10 AM pcw_home: you may have to increase the time depending on how much baseline wander you have
11:10 AM Tom_L: and no stepgen.timer-number
11:10 AM Tom_L: since he's running servos
11:10 AM skunkworks: I am running -200
11:11 AM skunkworks: heh
11:11 AM Tom_L: you _are_ running servos aren't you?
11:11 AM skunkworks: steppers for the green machinee
11:11 AM Tom_L: ooo
11:11 AM dirty_d: pcw_home, so like: setp hm2_7i96.0.dpll.01.timer-us -100
11:12 AM dirty_d: the last number should be larger than the highest ever latency right?
11:12 AM pcw_home: you can get an idea of the baseline wander by plotting dpll.phase-error-us
11:12 AM dirty_d: or typical latency?
11:12 AM pcw_home: not latency
11:13 AM dirty_d: oh, servo period?
11:13 AM pcw_home: no, baseline wander
11:13 AM pcw_home: basically the worst average shift in latency
11:13 AM dirty_d: ahh
11:16 AM skunkworks: but I was trying to figure out the issues with the time shifting. So - I could probably re-visit the number
11:16 AM pcw_home: for example you runa YT video and the average latency shifts 100 usec, the DPLL adjusts to accommodate now you close you browser and suddenly the average latency is 10 usec
11:16 AM dirty_d: ok
11:17 AM pcw_home: if you did not sample 90 usec before the normal time, the sample time could be off a full servo period
11:17 AM Tom_L: how do you know if the dpll number is too high / low?
11:17 AM pcw_home: too high just reduces you control bandwidth a bit
11:19 AM pcw_home: too low will occasionally have large timing errors (halscope dpll.phase-error-us to see the baseline wander)
11:20 AM pcw_home: on stepgen/velocity mode servo systems you can compensate for the added delay with FF2
11:21 AM pcw_home: (FF2 = seconds of delay)
11:22 AM pcw_home: (between position sample and velocity update)
11:23 AM dirty_d: I need to get around to tuning my system. I only have open loop steppers, but Sometimes while jogging an axis would oscillate a lot for a few seconds. I fixed that by setting PID_D to 5, but now the motors make weird noise sometimes when they're not moving
11:23 AM dirty_d: I wonder if this DPLL setting will help that
11:24 AM skunkworks: for a pid setup - the P should be your servo period.
11:24 AM skunkworks: for steppers running mesa
11:25 AM dirty_d: Ahh, yea my P is 1000, and my servo-period is 1ms
11:25 AM skunkworks: FF1 should be 1 and ff2 should be tweeked for best following error
11:25 AM dirty_d: everything else was 0, except FF1 which was 1
11:25 AM skunkworks: (very small number)
11:26 AM dirty_d: I'm familiar with PID control, but I've never used FF terms and I have no idea what they do
11:26 AM skunkworks: you also want a max error - MAX_ERROR =0.0005 (for english - I don't know what mm should be)
11:26 AM skunkworks: pcw_home: would know
11:27 AM dirty_d: skunkworks, I don't really have following error, but if I jog erratically with the MPG wheel I'd get crazy overshoot and oscillation
11:27 AM dirty_d: like large oscillation, inches
11:27 AM pcw_home: thats a pncconf bug
11:28 AM pcw_home: Metric system?
11:28 AM skunkworks: if you are getting crazy overshoot - you have following error ;)
11:28 AM dirty_d: pcw_home, which thing is a bug?
11:29 AM dirty_d: skunkworks, I'm still trying to wrap my head around how there can be following error in a system with no feedback
11:29 AM pcw_home: pncconf sets the PID maxerror to 0.0005 which is not appropriate for metric systems
11:29 AM dirty_d: ahh
11:29 AM dirty_d: im using inches
11:29 AM pcw_home: either set it to 0 or say .01
11:30 AM dirty_d: skunkworks, so you think my problem is just that FF2 is 0?
11:30 AM pcw_home: well if using inches then you should not have overshoot
11:30 AM pcw_home: no FF2 is fine at 0
11:30 AM skunkworks: dirty_d: interally - linuxcnc treats everyting like a servo (with feedback)
11:30 AM dirty_d: when I say I'm using inches I have my units set to INCES in the ini file, if that's what you mean
11:31 AM dirty_d: INCHES
11:31 AM dirty_d: actually, LINEAR_UNITS = inch
11:31 AM pcw_home: If you have large overshoot its probably a wrong velocity or acceleration setting
11:31 AM skunkworks: pcw_home: I think his P was really low
11:31 AM dirty_d: pcw_home, I have very low acceleration because I'm stuck with only one Y axis ballscrew and motor for now
11:32 AM dirty_d: skunkworks, my P is 1000
11:32 AM Tom_L: my p is yellow
11:32 AM dirty_d: haha
11:32 AM pcw_home: is the stepgen accel at least 25% greater then the axis/joint accel limit?
11:33 AM dirty_d: yes
11:33 AM pcw_home: is the stepgen maxvel set to 0?
11:33 AM dirty_d: no, its 25% larger than MAX_VELOCITY
11:34 AM dirty_d: pnconf set these
11:34 AM pcw_home: OK thats sounds right, usually if you have overshoot its because you have insufficient stepgen headroom
11:34 AM dirty_d: FERROR = 5, MIN_FERROR = 0.05
11:35 AM skunkworks: I think you are seeing the 'large overshoot' as maybe low acelleration?
11:35 AM dirty_d: skunkworks, I thought maybe at first, but no I checked with the hal scope
11:35 AM pcw_home: FF1 = 1?
11:35 AM dirty_d: yup FF1 = 1
11:36 AM pcw_home: strange, did you plot the following error?
11:36 AM dirty_d: yea I did
11:36 AM dirty_d: if it helps, setting D to 5 or 10 completely eliminated the overshoot and oscillation
11:37 AM pcw_home: no D should ever be needed
11:37 AM dirty_d: But the motors wine and make weird noises
11:37 AM dirty_d: ok
11:37 AM dirty_d: pcw_home, could it be an ethernet latency issue or something?
11:37 AM pcw_home: No
11:38 AM pcw_home: is you servo period 1 ms?
11:38 AM pcw_home: your
11:38 AM dirty_d: yup 1ms
11:41 AM dirty_d: this is my ini file https://pastebin.com/Ct2KJwXy
11:44 AM dirty_d: and hal https://pastebin.com/QPhXKewJ
11:47 AM pcw_home: might try changing this to 0: setp pid.y.maxerror .0005
11:48 AM pcw_home: 0.0005 is normally ok for inch machines unless there is a big timebase error (host 1 ms is not really 1 ms)
11:56 AM dirty_d: change 0.0005 to 0?
11:56 AM pcw_home: yes
11:56 AM dirty_d: ok, does 0 mean ignore the error?
11:57 AM pcw_home: no, pid max_error is a bound on the error (difference between command/feedback)
11:58 AM pcw_home: that is, a bound on the error used in the PID calculations
11:59 AM dirty_d: ahhh ok
12:02 PM pcw_home: I'm not sure whats going on. Normally the ferror is less than 100 u inches or so without any fussing
12:03 PM dirty_d: this only happens while jogging
12:03 PM dirty_d: and I kind of have to try to make it happen
12:04 PM dirty_d: I think I have my MPG to jog in step velocity mode
12:04 PM dirty_d: and its only when I have the step size set to 0.01"
12:06 PM dirty_d: not sure if that's what it's called, my MPG is driving axis.x.jog-counts
12:06 PM dirty_d: and axis.x.jog-vel-mode is set to 1
12:06 PM pcw_home: normally you would have much smaller ferror settings (say .0005 and .002) and you would never expect to violate these
12:07 PM Thorhian: Where can I get the mesaflash tool for RPi (ARM) pcw_home?
12:08 PM Tom_L: git clone github.com/Linuxcnc/mesaflash
12:08 PM Thorhian: Thanks tom
12:09 PM dirty_d: I'll be back to test this stuff on my actual linuxcnc machine
12:39 PM Thorhian: I managed to get a connection to the 7c81 with mesaflash :)
12:39 PM Tom_L: nice
12:44 PM dirty_d: I took a video of the oscillation, waiting for it to upload
12:46 PM dirty_d: https://photos.app.goo.gl/9t6wz374mVmhay7k9
12:50 PM pcw_home: probably need to plot commanded and feedback position (and maybe commanded velocity) in halscope to debug
12:52 PM pcw_home: most common cause of this kind of behavior is lack of stepgen headroom or pid max_error set too low
12:52 PM pcw_home: linuxcnc violating accel or velocity constraints would also do this
12:53 PM pcw_home: as would a sizeable timebase error in the host (enough so the 0.0005 inch max_error is exceeded)
12:54 PM dirty_d: pos-fb oscillates around pos-cmd
12:54 PM dirty_d: i set the max error to 0
12:55 PM dirty_d: like you said in the hal file, and set the FERROR and MIN_ERROR high, so I can see the oscillations before linuxcnc errors out
12:59 PM dirty_d: pcw_home, I added that DPLL stuff to my hal file and I can't get it to oscillate anymore
01:00 PM pcw_home: setting maxerror to 0 did not stop the oscillations?
01:00 PM dirty_d: setp pid.x.maxerror 0 that? no it didn't
01:01 PM dirty_d: it didn't have any effect as far as I can tell
01:01 PM pcw_home: Did you set all to 0
01:01 PM pcw_home: or is that just x you were testing
01:04 PM dirty_d: only x
01:06 PM dirty_d: that dpll phase-error-us looks wacky, I'm not sure what it's supposed to look like
01:06 PM dirty_d: it looks like noise
01:08 PM dirty_d: its spiking up to 80
01:14 PM dirty_d: I spoke too soon, with the dpll stuff it's much better, but I can still get it to oscillate
01:15 PM dirty_d: I had to use a 0.01" jog step scale before to get it to oscillate, but now it won't oscillate with the step size, but it does with a 0.001" step size
01:15 PM dirty_d: so weird
01:16 PM pcw_home: try setting the stepgen maxvel and maxaccel to double their current values
01:22 PM dirty_d: ok doubling them with the DLPP stuff commented out fixed the problem 100%.
01:22 PM dirty_d: before I doubled them I tried setting that last DPLL number to -200 and it made the problem much worse. setting it to -50 fixed it 100%
01:22 PM pcw_home: what linuxcnc version?
01:22 PM dirty_d: 2.8
01:23 PM dirty_d: I'm still not sure how to choose that dpll number
01:23 PM dirty_d: do I want it to be the average value of the phase error?
01:24 PM pcw_home: watch the dpll.error-us (scaled appropriately)
01:25 PM pcw_home: the peaks are latency spikes (read happen that much later than the baseline)
01:26 PM pcw_home: but you should also see the baseline wander if you say launch a browser or move the mouse around the screen
01:27 PM pcw_home: the pre read time (the time setting) must be larger than the largest possible base line shift
01:27 PM dirty_d: yea it does move around when I move the mouse
01:28 PM pcw_home: What I would do is see the larges baseline shift you can generate and double that value
01:29 PM pcw_home: what host CPU do you have?
01:29 PM dirty_d: by baseline do you mean the average value of the noise?
01:29 PM dirty_d: it's centered around zero
01:29 PM dirty_d: when I move the mouse around the average doesn't move much pas 10 from zero
01:29 PM dirty_d: I'm using a rpi 4
01:30 PM pcw_home: thats the base line shift
01:30 PM pcw_home: Yeah (dont try videos...)
01:30 PM dirty_d: yea I only have linuxcnc open when I'm using it
01:30 PM dirty_d: why do you think setting the dpll value to -200 made the problem worse?
01:31 PM pcw_home: I think you can get away with videos at 1 2 ms servo thread
01:31 PM pcw_home: I dont think it would make much of a difference
01:32 PM dirty_d: -50 is perfect, -100 is pretty good, -200 was the same as when I wasnt using dpll
01:33 PM pcw_home: That doesn't make any kind of sense to me unless something was on the edge of not working
01:34 PM dirty_d: with the stepgen max accel and vel set to 200% setting dpll to -200 works fine
01:34 PM dirty_d: I wonder why I need it set to 200% instead of 125%
01:35 PM pcw_home: sounds to me like a constraint violation
01:37 PM dirty_d: why would that happen though?
01:40 PM pcw_home: No sure, maybe a timebase issue (on a related note are you running ntp? timesyncd has issues)
01:41 PM dirty_d: yea I switched from timesyncd to ntp
01:43 PM pcw_home: might help to plot the commanded velocity and acceleration
01:43 PM pcw_home: (when jogging)
01:44 PM dirty_d: I'll try that
01:47 PM dirty_d: joint.0.acc-cmd appear to always be 0
01:52 PM dirty_d: join.0.vel-cmd looks like I'd expect
01:58 PM pcw_home: halscope has a ddt function, if you put the cursor on the velocity trace you should be able to read the accel
02:00 PM dirty_d: the velocity goes between +/- MAX_ACCELERATION
02:00 PM dirty_d: i mean ddt of vel-cmd
02:01 PM Thorhian: So I have the bit files for the 7C81 downloaded, and I have a single 7i76 installed. I found a 7c81_7i76x2d.bin and pin file, but I'm wondering what the x2d represents? Not sure which file to flash the board with.
02:02 PM Tom_L: Thorhian, it's pcw_home's secret code
02:02 PM Thorhian: Lol Tom.
02:02 PM pcw_home: its for 2 7I76s but should be fine
02:03 PM pcw_home: you flash with the .bit file (the .pin file is a text file of the pinout from mesaflash)
02:03 PM Thorhian: Ah, sounds good then. THere wasn't a file marked with jsut the 7i76x1s.
02:06 PM dirty_d: im gonna boot up my other arch linux arm image and see if I have the same oscillation problem there
02:16 PM skunkworks: His thread orders might be wrong
02:17 PM skunkworks: That would also cause oscolations.
02:17 PM pcw_home: It looks right (pncconf created config)
02:17 PM skunkworks: ok
02:18 PM pcw_home: I've seen some other weird customer errors in RPI4
02:18 PM pcw_home: on RPI4
02:19 PM rmu|w: -EWEIRD_CUSTOMER
02:19 PM skunkworks: huh. I have not seen that... yet
02:19 PM pcw_home: one was sserial related (that I could not duplicate)
02:20 PM rmu|w: maybe the problem is with the "steppers with feedback"
02:22 PM rmu|w: if the stepper controller uses the feedback to adjust for lost steps the pid and the stepper controller would work against each other
02:22 PM pcw_home: Pretty sure this is pure open loop
02:24 PM rmu|w: oh sorry i misread an earlier statement
02:24 PM skunkworks: I am also using smart serial
02:24 PM skunkworks: But sometimes I am lucky
02:27 PM pcw_home: yeah it was a very weird error (7I87 not detected if PWM was present on the host FPGA card)
02:28 PM pcw_home: X86 worked for customer, RPI4 didn't
02:28 PM pcw_home: RPI4 worked for me
02:29 PM rmu|w: connecting 7i90 via sserial to 7c80 also works
02:30 PM rmu|w: (rpi4, but should not matter)
02:30 PM skunkworks: it shouldn't... ;)
02:30 PM rmu|w: also my franken-stmbl is recognized
02:31 PM skunkworks: (that is what I am using smart serial to communicate with.
02:32 PM skunkworks: And using my own line driver strait from the 7i92 i/o
02:32 PM skunkworks: So I would call the whole machine franken-green
02:33 PM skunkworks: I stopped playing with the servo'ed emco when I got the spindle hooked up. 1 issue was I needed a decent spindle encodeer and issue 2 was when the spindle was on I would get random estops.
02:33 PM skunkworks: I wonder why.. I had no grounds hooked up - anywhere.
02:34 PM Tom_L: those weren't in the kit
02:35 PM skunkworks: heh
02:35 PM skunkworks: I had 3 power plugs hooked into it..
02:35 PM Tom_L: i ordered a hdmi->vga so i can test on the mill next week
02:37 PM skunkworks: So - starting from the power side.. lets do it mostly right... http://electronicsam.com/images/greenmachine/IMG_20200517_142557.jpg
02:38 PM Tom_L: where's you get the fancy choke?
02:39 PM skunkworks: comptuer power supply...
02:39 PM skunkworks: we have bad ones stacked up like cord wood at work.
02:39 PM skunkworks: This should handle the power.
02:47 PM skunkworks: i did get funny noise issuse with a crap wired keyboard and mouse... on the rpi4 and the green machine. (random key presses)
02:54 PM Thorhian: I just flashed the firmware :D
02:55 PM Tom_L: did you verify it?
02:56 PM Thorhian: Is there an exact procdure for verifying? I reloaded the board and it was reporting back via SPI just fine.
02:56 PM Tom_L: you can --readhmid to see the pinout changes
02:57 PM Tom_L: the write automatically verifies itself
02:57 PM Thorhian: Yeah the pinout changes looked correct.
02:57 PM Tom_L: ur an expert now.
02:57 PM Thorhian: XD
03:03 PM Thorhian: Now I just need to learn how to configure LinuxCNC and start testing the motors out.
03:26 PM Deejay: gn8
03:38 PM Thorhian: ls
03:47 PM Thorhian_: Drawing power form the Pi for the Mesa cards seems to work so far. We will see what happens when I get some motors running off of it.
03:48 PM Thorhian_: If need be I can power the 7i76 sperately and just let the pi power the 7c81.
03:56 PM Tom_L: i put both my mesa cards on a small 5v smps
04:04 PM skunkworks: I think the mesa and pi are run off the same power supply in the green machine
04:05 PM Tom_L: 3A ?
04:05 PM Tom_L: one of those little chinese smps would do i bet
04:05 PM skunkworks: 4
04:05 PM skunkworks: 4a
04:07 PM Tom_L: so are you committed to the Rpi4 for that cnc now?
04:07 PM skunkworks: probably.. for now.
04:07 PM skunkworks: it seems to do just fine
04:07 PM skunkworks: for what I need
04:07 PM Tom_L: i'm not sure what i'd do with all the extra space
04:09 PM skunkworks: plus it allows me to 'test' the pi
04:30 PM Thorhian_: I'm kind of laughing at some of this hal file syntax. Reminds of lisp.
04:34 PM Thorhian_: So in the hal file there is the loadrt statement for hm2. I replaced hm2_5i25 with hm2_rpspi. Should it be hm2_7c81 and should I replace any mention of hm2_5i25?
04:44 PM Thorhian_: ls
04:57 PM Tom_L: hm2_rspi, i don't see a comp called hm2_7c81
04:57 PM Tom_L: http://linuxcnc.org/docs/2.8/html/
04:58 PM Thorhian_: I ended up choosing hm2_7c81. The only error at startup now is that it's saying that my watchdog timeout_ns variable is not found.
04:59 PM Thorhian_: sorry, hm2_rpspi lol
04:59 PM Tom_L: mmm i thought they did away with that
05:00 PM Thorhian_: "parameter or pin 'hm2_rpspi.0.watchdog.timeout_ns' not found"
05:00 PM Tom_L: # # # setp hm2_[HOSTMOT2](BOARD).0.watchdog.timeout_ns 10000000
05:00 PM Thorhian_: Oh wait, it's complaining that it exists I think.
05:00 PM Tom_L: comment that out then
05:01 PM Thorhian_: Done
05:12 PM Thorhian_: Awesome, Axis finally came up. Thanks Tom!
05:12 PM Tom_L: np
05:15 PM pcw_home: the watchdog should be there and set to a reasonable value
05:20 PM Tom_L: pcw_home, i thought they did away with that in the hal file
05:21 PM pcw_home: they did away with the pet_watchdog function not the timeout setting
05:22 PM andypugh: https://youtu.be/TX0MDlRF7gY
05:24 PM pcw_home: OK _that_ is really weird
05:24 PM XXCoder: non round screw
05:25 PM pcw_home: XXcoder is making it sound weirder
05:25 PM andypugh: Relieving for a tap.
05:27 PM pcw_home: Thats a tap?
05:27 PM andypugh: It will be, when I add flutes and a taper and harden it.
05:32 PM Tom_L: Thorhian_, did you get that about the watchdog?
05:34 PM XXCoder: interesting
05:34 PM XXCoder: rather large tap tool indeed
06:09 PM skunkworks: andypugh: awesome!
06:14 PM skunkworks: andypugh: that finish looks great!
06:15 PM andypugh: It looks better than it really is, I suspect.
06:16 PM skunkworks: You have an resolver for the spindle? what kind of effective resolution is that?
06:16 PM andypugh: And it’s anyone’s guess what it will be like after heat-treatment,
06:17 PM skunkworks: andypugh: how many flutes is that?
06:17 PM andypugh: Resolution is equivalent to about 24 bits, but the actual linearity isn’t as good as an encoder, typically.
06:17 PM andypugh: 4 flutes.
06:17 PM skunkworks: (dad was wondering)
06:18 PM andypugh: I don’t like how hard it is to measure 3 flutes.
06:18 PM skunkworks: heh
06:20 PM unterhaus: andypugh what thread is that tap for?
06:21 PM andypugh: Lathe cross-slide. Non-standard LH Acme
06:21 PM andypugh: (9/16 8tpi Acme. I challenge you to find a tap at any price….)
06:21 PM unterhaus: I recently looked for an acme tap for Norris style wood planes, outrageous
06:22 PM unterhaus: could resurrect Mr. Norris for less
06:23 PM andypugh: The chap who I am making this tap for was quoted £500 for one.
06:23 PM andypugh: For the more common sizes you can get them for £35 or so,
06:25 PM Thorhian_: Sorry pcw_home, I was at dinner. I commented out the line I quoted since LinuxCNC didn't recognize it. How would I set it in the case of hm2_rpspi?
06:28 PM Tom_L: Thorhian_, you need the watchdog timeout just not the pet_watchdog function
06:28 PM Tom_L: that line i posted should be sufficient
06:29 PM Tom_L: comment out addf hm2_[HOSTMOT2](BOARD).0.pet_watchdog servo-thread
06:29 PM Tom_L: if it's there
06:31 PM Thorhian_: I never had a pet_watchdog line. It was this: hm2_rpspi.0.watchdog.timeout_ns
06:32 PM Thorhian_: setp hm2_rpspi.0.watchdog.timeout_ns 5000000
06:32 PM Thorhian_: That is the exact line.
06:35 PM Tom_L: you need that i'm told
06:36 PM Thorhian_: Linux CNC refused to run with that line.
06:37 PM andypugh: Does the pin exist?
06:43 PM Thorhian_: I'm not sure, but mesaflash reports that the WatchDog module does exist.
06:44 PM andypugh: comment out the line that causes trouble, start linuxcnc, then get a pin list with “halcmd show pin” and a parameter list with “halcmd show param”
06:44 PM Tom_L: mm i must not have the dpll in my bitfile on the mill
06:46 PM Tom_L: i also had that watchdog time commented out in my hal so i uncommented it and has no ill effect
06:46 PM Tom_L: nice video andypugh
06:49 PM Thorhian_: andypugh: How do you run hal commands in linuxcnc?
06:49 PM Thorhian_: I'm using the axis interface.
06:50 PM andypugh: Open a terminal.
06:51 PM andypugh: You can either type “halcmd” for each command, or join the running HAL session with “halcmd -kf”
06:51 PM andypugh: Then any commmand that could be in a HAL file is valis.
06:51 PM Thorhian_: Ah, didn't realize that halcmd was a program in the PATH. Sorry.
06:51 PM andypugh: So you can load and unload copmponents, make nets, set pin values etc.
06:55 PM Thorhian_: I found the parameter. It's listed using hm2_7c81 *facepalm*
06:56 PM andypugh: See, I would never have guessed that.
06:56 PM Tom_L: heh
06:56 PM Thorhian_: Is that sarcasm lol?
06:56 PM Tom_L: nnnaaaaa
06:56 PM Thorhian_: XD
06:56 PM Thorhian_: Sorry you have to deal with a noob on here.
06:57 PM andypugh: No, I am saying that I am glad I suggested halcmd rather than trying to guess the problem.
06:57 PM Thorhian_: Well thanks Andy. All of the other hal commands that failed might need the same treatment. I better get searching lol
07:18 PM Thorhian_: Okay, I fixed all of the lines I commented out and linuxCNC is now starting up :)
08:12 PM veegee: If I have one cylinder of pure argon, and a separate cylinder of pure helium, what kind of regulator thing do I buy to be able to connect them together and create a mixed gas output with a custom ratio?
08:13 PM veegee: I plan to experiment a lot while I'm learning TIG welding. Welding copper might require pure helium
08:14 PM veegee: Also make videos with a squeaky high pitch voice
08:26 PM roycroft: veegee: there are gas mixing devices made to do that
08:27 PM roycroft: it would be a lot cheaper to just buy an extra cylinder with the mix you want, in addition to your pure gas cylinders
08:28 PM roycroft: you're talking $2500+ to get a gas blender
08:30 PM dirty_d: veegee, you could just combine the outputs from two separate flow regulators on each tank, and do math to figure out what to set each one
08:31 PM veegee: dirty_d I like that
08:31 PM veegee: Just wanted to make sure I wasn't missing anything
08:31 PM veegee: Have to make sure to prevent backflow I think
08:31 PM veegee: well actually that shouldn't matter
08:32 PM SpeedEvil: roycroft: or 3d print
08:32 PM veegee: Renting this 3,200 sq ft industrial unit for hobby purposes is giving me anxiety. Need to figure out something I can do to offset the costs
08:32 PM veegee: maybe I'll come up with something
08:35 PM XXCoder: maybe some simple production on side of your hobby
08:36 PM jdh: what would weld that was copper?
08:37 PM dirty_d: veegee, yea I don't think you need to worry about that
08:37 PM veegee: jdh bus bars for high current power supplies
08:38 PM dirty_d: if the pressure from the other affected the other one, you'd just see the ball float lower
08:38 PM dirty_d: right?
08:38 PM veegee: dirty_d it won't matter during actual use since the pressure inside would be 0 anyway
08:38 PM veegee: the flow meters would both be accurate
08:39 PM veegee: the flow is the important part, so as long as I adjust it so both the flow gauges read what I want, that's all that matters
08:40 PM roycroft: except it won't work - the mixer has to measure the actual flow rate of each gas, which can vary due to pressure/temprature changes
08:40 PM dirty_d: even if the pressure isn't 0 I think they'd still be accurate
08:40 PM dirty_d: pressure will reduce flow, which you will see
08:40 PM roycroft: that's why the mixers are expensive
08:40 PM dirty_d: the flow you see is the flow you get
08:41 PM veegee: roycroft the flow meter on each cylinder is already doing that
08:41 PM dirty_d: roycroft, he's using flow regulators, not pressure regulators
08:41 PM veegee: it's the same thing as just having two separate cylinders just venting into the room
08:41 PM veegee: if you adjust both the flow meters, they'll both be accurate
08:41 PM roycroft: what about cross-mixing?
08:42 PM veegee: what do you mean?
08:42 PM dirty_d: that's the idea?
08:42 PM veegee: The whole point is to make an adjustable mixed ratio gas
08:42 PM roycroft: the gas at the higher flow rate pushing back on the one with the lower flow rate
08:42 PM veegee: roycroft that won't happen
08:43 PM dirty_d: well, it will, but it doesn't matter
08:43 PM veegee: it will happen when the valve at the welder is turned off
08:43 PM veegee: but when in use, the pressure inside the outlet hose is close to 0
08:43 PM veegee: it may affect the flow of the other gas ever so slightly, but then I just adjust it to what I want
08:44 PM veegee: venturi effect and all that
08:44 PM dirty_d: exactly
08:44 PM dirty_d: that's a pretty cool idea though
08:44 PM veegee: I can create a little mixing chamber to reduce that effect and potentially get better mixing
08:45 PM veegee: it can sit between the welding machine and the output hoses
08:45 PM jdh: takes a lotta amps to tig copper
08:45 PM veegee: yeah that's why you need pure helium
08:45 PM veegee: My machine can do 325A no problem
08:45 PM jdh: how thick are the bars?
08:46 PM veegee: I don't have a specific project in mind yet, I'm just saying that it will happen at some point
08:48 PM jdh: could be disappointing!
08:48 PM veegee: I've watched some videos, apparently it's fine as long as you're running pure helium
08:49 PM veegee: Can also preheat it to help. Either way, I'm not worried
08:49 PM XXCoder: isnt helium rather expensive now?
08:50 PM veegee: if you can do aluminium, you can do copper. C110 alloy copper thermal conductivity is not /that/ much higher than aluminum
08:50 PM veegee: XXCoder about 3x the cost of argon, but going to be using it on average at 10% the rate of argon
08:50 PM veegee: so I'm not worried about that
08:50 PM XXCoder: indeed
08:51 PM veegee: The damn industrial unit costs $3,000/month (but sharing with a friend). That's the real source of my anxiety
08:51 PM XXCoder: helium is renewable in sense that you can just wait for decay to produce more
08:51 PM veegee: well the gasses don't actually go anywhere
08:51 PM XXCoder: but up and gone :)
08:51 PM veegee: they just go back into the atmosphere and you just do fractional distillation again
08:51 PM XXCoder: ya know
08:51 PM jdh: fractional distillation of helium?
08:52 PM XXCoder: you can liquidify air heh but helium is not really recoverab;e
08:52 PM jdh: from air?
08:52 PM veegee: jdh at least from natural gas
08:53 PM veegee: Oh right I confused myself, was thinking about argon fractional distillation
08:53 PM unterhaus: someone was saying we were going to run out of helium. My guess is most of it in natural gas is wasted
08:53 PM XXCoder: would nitrogen work as netural gas?
08:54 PM veegee: but they do fractional distillation of helium from natural gas right?
08:54 PM XXCoder: neural finally got spelling right
08:54 PM unterhaus: I'm confused about nitrogen, can't you get it to explode?
08:56 PM XXCoder: "
08:56 PM XXCoder: Pure Gases
08:56 PM XXCoder: Acetylene
08:56 PM XXCoder: Air
08:56 PM XXCoder: Argon
08:56 PM XXCoder: Carbon Dioxide
08:56 PM XXCoder: Helium
08:56 PM XXCoder: Hydrogen
08:57 PM XXCoder: ooops had to exit before it spammed more lol
08:57 PM XXCoder: "Nitrogen
08:57 PM XXCoder: GENERAL FACTS:
08:57 PM XXCoder: Nitrogen (N2) is a diatomic gas which comprises 78% of the earth's atmosphere. Colorless, odorless, tasteless, and nontoxic, nitrogen exists as a nonflammable gas at atmospheric temperatures and pressures. Nitrogen is utilized as a purge gas with stainless steel tube welding. Nitrogen can enhance plasma cutting, food processing, heat-treating. Small additions to argon based shielding gases can be used
08:57 PM XXCoder: for welding stainless steel by the Gas Metal Arc Welding (GMAW or MIG) process. "
08:57 PM XXCoder: looks like it works as addition but not by itself
08:57 PM jdh: nitrogen embrittlement
08:58 PM unterhaus: okay, solved my confusion, bombs use ammonium nitrate, which produces oxygen when burned
08:59 PM roycroft: nitrogen is fairly reactive - it's a poor substitute for a noble gas
08:59 PM XXCoder: too bad really, if it was usable, then you can make your own distillery to get that
09:00 PM XXCoder: then you will neevr run out of it
09:01 PM SpeedEvil: err
09:01 PM SpeedEvil: Argon is 1% of the air
09:01 PM SpeedEvil: Also, vacuum can work
09:01 PM XXCoder: i wasnt sure if could get stirling air liquidifer cold enough for that
09:01 PM SpeedEvil: though has its own issues
09:02 PM Tom_L: weld in space
09:05 PM roycroft: that even poses a lot of problems, tom_l
09:05 PM XXCoder: man too bad we dont have portal tech yet
09:05 PM veegee: First thing I have to do is build a loading dock ramp: https://secureservercdn.net/198.71.233.47/rpo.dde.myftpupload.com/wp-content/uploads/2019/12/MOBILE-YARD-RAMP-STOCK-SERIES-YRM2084635-1024x768.jpeg
09:05 PM XXCoder: so we'd have infinite free vacuum
09:06 PM veegee: The garage door to the unit is at truck level unfortunately
09:06 PM veegee: These damn ramps cost a minimum of $5,000
09:06 PM XXCoder: we would make 2nd law our bitch with that. cn make infinite power
09:06 PM veegee: But I don't need their weight capacity nor the length.
09:08 PM skunkworks: I had 2 drives fail in succession on a raid 6 array...
09:09 PM veegee: skunkworks ZFS/BTRFS or just mdadm?
09:09 PM veegee: or worse, hardware?
09:09 PM skunkworks: mdadm
09:09 PM veegee: either way, RAID 6 can tolerate 2 failures
09:09 PM veegee: yeah that's fine
09:10 PM skunkworks: first drive is 90% rebuilt..
09:10 PM veegee: just hope something doesn't fail during rebuild
09:10 PM skunkworks: took almost 24 hours..
09:10 PM veegee: I only use RAID10
09:10 PM veegee: super fast rebuilds but damn that capacity loss
09:10 PM veegee: 24 x 8TB drives
09:10 PM veegee: 96TB lost to redundancy :(
09:11 PM skunkworks: heh - this is 8 6TB drives
09:11 PM veegee: I do regular tape backups though and use ZFS for checksumming and bitrot protection
09:11 PM skunkworks: we are starting to run zfs at work. It seems nice.
09:12 PM veegee: also use T10 DIF; my drives are the newer SAS 3 drives and support it
09:12 PM veegee: Got a new dual socket EPYC server the other day. PCIe 4 lanes for days
09:13 PM veegee: fully saturates the infiniband cards
09:13 PM skunkworks: we just got a gigabyte EPYC server too - dual socket. it is nice
09:13 PM veegee: Yeah that's the one I have
09:13 PM veegee: R282 Z92
09:13 PM skunkworks: running proxmox currently.
09:13 PM skunkworks: Is that the 24 bay one?
09:14 PM veegee: The 24 bay NVMe one yeah
09:14 PM veegee: all bays populated with 4TB 3 DWPD NVMe drives
09:15 PM veegee: Anyway need to figure out how to design this truck loading bay ramp. Time to learn some mechanical engineering stuff
09:15 PM skunkworks: good luck
09:15 PM veegee: I'm guessing rectangle tube will be good enough
09:16 PM veegee: If there were any decent threadripper motherboards, that gives you way more compute per dollar than EPYC
09:16 PM veegee: sacrifice some PCIe lanes and half the memory bandwidth, but still really solid
09:19 PM skunkworks: it is amazing what is out there now..
09:19 PM veegee: It's amazing Intel still has nothing to show for it lol
09:19 PM skunkworks: heh
09:19 PM veegee: fuck Intel. I'm so glad AMD came back
09:20 PM XXCoder: its why competitors is good
09:20 PM XXCoder: nothing makes you run faster than wolves nipping your ankles
09:21 PM veegee: or a coughing chinaman
09:21 PM veegee: I KID I KID
09:21 PM * XXCoder sighs
09:21 PM veegee: Just highlighting the racism I see around me
09:21 PM skunkworks: [===================>.] recovery = 99.2% (5815813708/5860390912) finish=11.4min speed=64894K/sec
09:21 PM veegee: It's subtle but it's there even in Toronto
09:23 PM veegee: skunkworks check the actual disk throughput using `iostat`
09:23 PM veegee: Rebuilding is largely a sequential read/write, see if the disks are being used to their full potential
09:24 PM veegee: `iostat -x 1`
09:24 PM Tom_L: Thorhian, you around?
09:25 PM skunkworks: veegee: this is pretty crap hardware - offsite backup..
09:25 PM skunkworks: it is a external box with 2 Esata connections to it - crap card.
09:25 PM veegee: wow that's terrible
09:25 PM veegee: Not even SAS HBAs
09:25 PM skunkworks: old
09:26 PM veegee: Consider tape backup
09:26 PM veegee: Far more reliable in every way. LTO-6 is affordable now on ebay
09:27 PM skunkworks: we have a tape drive - was plannning on getting that set up. no one used it.
09:27 PM veegee: You need a tape library
09:27 PM skunkworks: it is.
09:27 PM veegee: That's the fun way to do it. Little robot inside autoloading the drives
09:27 PM skunkworks: what was the model. Let me think
09:27 PM veegee: I have an MSL4048
09:28 PM veegee: What brand?
09:28 PM veegee: Dell? HP? Quantum?
09:29 PM skunkworks: dell
09:29 PM skunkworks: it is old - but the last time I looked a few months ago - new tapes - it would have enough capacity. IIRC
09:30 PM veegee: powervault TL2000
09:30 PM veegee: TL4000
09:30 PM skunkworks: looking
09:30 PM veegee: Yeah those libraries last forever. The drives are the expensive bit
09:33 PM skunkworks: uh
09:33 PM skunkworks: PowerVault 124T
09:33 PM veegee: Ah yeah the smaller one, decent unit
09:34 PM skunkworks: Dual Magazine
09:35 PM Tom_Rpi: mmm getting an error loading the spi module
09:36 PM Tom_Rpi: hm2_rpspi: dlopen: /usr/lib/modules/hm2_rpspi.so: undefined symbol: hm2_register
09:39 PM Tom_Rpi: oh i may have figured it out
09:40 PM * skunkworks is waiting...
09:40 PM Tom_Rpi: yay
09:41 PM Tom_Rpi: i forgot you have to load the hostmot2 driver ahead of the hm2_rpspi one
09:42 PM Tom_Rpi: so i've tested mesaflash on epp, eth, spi and linuxcnc on eth and spi on the rpi4
09:43 PM Tom_Rpi: and added the dpll stuff to my mill config (it was old)
09:43 PM Tom_Rpi: the 2.9 axis screen is screwed up though
09:44 PM Tom_Rpi: maybe
09:44 PM * Tom_Rpi goes to the shop to check the old axis
09:46 PM skunkworks: Tom_Rpi: Cool!
09:46 PM skunkworks: after both drives dropped out - the volume went read-only. Need to figure that out after then 2nd drive rebuilds
09:46 PM skunkworks: seems I have had that before.
09:47 PM skunkworks: yah probably.. [Sun May 17 21:29:03 2020] EXT4-fs (md126): error count since last fsck: 3
09:48 PM skunkworks: a fsck is in it's future.
09:48 PM Tom_Rpi: yeah on 2.9.0~pre0 beside the X Y Z coordinates upper left there are 3 arrows there now <- beside each axis
09:49 PM Tom_Rpi: that's a new 'feature' i didn't have before
10:06 PM XXCoder: damn he has airblade protected camera. im so jelious
10:07 PM dirty_d: I think there was a video where he made it
10:07 PM dirty_d: maybe
10:12 PM Tom_itx: skunkworks, that cpu utility i mentioned to you... i now reaized you can display the core temp and cpu useage on the top bar
10:19 PM skunkworks: cool - i have not played with it yet
10:20 PM Tom_itx: just right click on the bar -> add remove pannel items
10:21 PM Tom_itx: it's nice seeing the temp at a glance
10:22 PM XXCoder: that guy didnt measure depth before permently inserting rods. geez
10:22 PM XXCoder: hes far better than me, but that mistake jeez
10:24 PM XXCoder: dirty_d: thanks for link, subbed
10:33 PM dirty_d: no problem
10:34 PM Tom_itx is now known as Tom_L