Advanced Search

People with low education = easy to control
People with high education = not easy to control
Quote from a historian i am pretty sure was/is British : If you want to delete a nation/culture/people, destroy their monuments
Tesla, the real one not this elon prick, : Only a fool makes the same mistake twice
Einstein : universe and human stupidity have no boundaries
Someone on the internet : nothing is foolproof to a talented fool
Me personally : How the hell did we end up in a world run by really, really dumb humans???
Again me personally : There is no humanity left in humans, they just look like humans.
Thomas Jefferson : anyone capable of deciding about his people on the spot- should be shot in public << referring to presidents and government, and the fact those things need time to think through
...

rodw wrote:

Couldn't help myself and posted at Phoronix...

Could not find it there, or using different nic?

Moved to "ethercat" section.

Usually, cutting to slow or lack of air pressure.

Blender is a monstrosity and a magnificent piece of software, so yes, it is always worth the time and effort to learn it, not only for CNC.

Sorry took a while but I did hook up a scope to a highdef plasma with HF start. Measurements taken directly between electrode and work from the 50:1 output.
I could not capture kerf crossing that well as the change in voltage was so slow and my portable scope is what it is... Closest thing to kerf crossing was running out of plate and that I could capture so I did that.
How I see from these scope captures is pretty much as I though. You can see from both 50A and 100A pierce that during pilot the voltage hits a "ceiling" that is the OCV, then drops when it transfers to plate. Then when you run out of the plate it looks to me that the voltage hits a "ceiling" that is the max DC output of the plasma, 50A and 100A look very similar.
I dont have a current probe, now at least, so I can't capture the current ramp up but in the highdef stuff I would assume this is measure in some hundreds of ms instead of ns again I cant really prove this as I dont have a current probe.

EDIT: Some typos

I started a build for new CNC plasma table so here's my starting point.

CPU Ryzen 5 6C/12T 3.2Ghz 32G RAM, 550G SSD, Dual Ethernet ports
Ubuntu 24.04.02 - I know it's not "debian" but is what all other machines here run. Compatibility and support are the key deciding factors
PREEMPT-RT installed, Fairly minimal system with snap completely removed.
Stepper Drives (4ea) Stepper Online DM542T - have had excellent results with these in the past
Stepper Motors (4ea) Stepper Online 23HS30-2804S - again have worked well in past
Custom built frame from 2x2 16ga steel tubing. All Linear Rails SBR20. All Ball Screws SFU1605
Motion Controller Mesa Electronics 7i95T & THCAD2 - ordered and expected arrival in next few days
LinuxCNC version 2.10 being built from source.

I can successfully build the rip version without error and it appears to work correctly however as I mentioned I do not have the 7i95T in hand quite yet and it's not listed as an option in the opening config choices.

To get the full package I'm working to compile to get the .deb version and having some make problems.

Problem 1
I get lots of errors during the build of man pages. Some work but many report this error and often multiple times.

Making manpage sampler.9
<unknown>:1: SyntaxWarning: invalid escape sequence '\S'

I suspect this has to do with the fact that Ubuntu 24.04 uses python3.12. I haven't dug into to it but have had many problems in many projects with python version compatibility and this is known issue from 3.11 to 3.12. Any suggestions on how to fix this are welcome and sooner or later it'll have to be addressed anyway.


Problem 2
This is the end summary of the build

Compiling module_helper/module_helper.c
<command-line>: error: "_FORTIFY_SOURCE" redefined [-Werror]
<command-line>: note: this is the location of the previous definition
Compiling localized message catalog ../share/locale/ar/LC_MESSAGES/linuxcnc.mo
Compiling localized message catalog ../share/locale/bg/LC_MESSAGES/linuxcnc.mo
cc1: all warnings being treated as errors
make[2]: *** [Makefile:278: objects/module_helper/module_helper.o] Error 1
make[2]: *** Waiting for unfinished jobs....
make[2]: Leaving directory '/home/plasma/Documents/linuxcnc-source-dir/src'
dh_auto_build: error: cd src && make -j12 build-software returned exit code 2
make[1]: *** [debian/rules:55: override_dh_auto_build-arch] Error 2
make[1]: Leaving directory '/home/plasma/Documents/linuxcnc-source-dir'
make: *** [debian/rules:43: build] Error 2
dpkg-buildpackage: error: debian/rules build subprocess returned exit status 2

The _FORTIFY_SOURCE error shows up many times as well and probably isn't the only problem.

Again I know it's not debian but Ubuntu is derived from debian and the rip version compiles and works so I've very optimistic once I work out a few kinks this will too.

cc (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0
Copyright (C) 2023 Free Software Foundation, Inc.

Thanks for any and all suggestions. Mike

My thoughts regarding my decision to create my own Break Out Board were influenced by my background. I've very recently retired from a 40 year career in hardware and software development. The largest chunk was in integrated circuit design. Where I don't have any experience is in milling and CNC, and I've been finding the learning curve on that to be more than a bit frustrating.

Going more or less chronologically:

I had decided that I wanted to be able to both control the speed of the spindle and read the "actual" spindle rpm. I wanted these lines to be electrically isolated from the mill's electronics. To isolate the spindle rpm signal I wanted to use an opto-isolator. So now I needed a board of some type whether purchased or made by me. Initially I bought a SparkFun Opto-isolator Breakout for this purpose. Spindle RPM control on the PM-728VT is a 0 - 5v analog input. So I searched for a PWM input to analog out board to purchase, because that's circuitry I'm not comfortable designing. I could not find a  board that accepted 5v input, nor could I find one that was 0-5v output. That's not saying no such thing exists, but I couldn't find one. So I went with the board listed above that's 3.3v digital and can output 0-10v. Five volt switching levels are nice in a noisy environment, but from a purely digital design standpoint they are archaic. Which might be why I had a hard time finding a PWM DAC with a 5v input. So if I ran the 7C81 in 5v tolerant mode I'd need to level shift that input. Not a big deal. Initially I was planning to simply put a 138 ohm resistor in series and call it a day. But I started to wonder if between it being a 5v tolerant output, and not a true 5v driver, and the series resistor, would the PWM waveform be sharp or distorted. So I thought the better solution for this particular signal was a true 3.3v signal. I went with the KBSI-240D Signal Isolator board between the PWM DAC and the PM-728vt circuit board because someone on another forum with a PM-718vt used one and was very positive about it.

I wanted a BOB to convert the 26 pin "ribbon cable" connectors on the 7C81 to some type of more robust cabling from the controller box to the mill. I didn't find any boards with three 26 pin "ribbon cable" connectors on one end. What I was seeing were a lot of boards that converted a single connector of some number of lines to terminal blocks, or to 25 pin connectors. I did not want to run a 25 strand cable to the mill and split it at the mill to the individual servos. Also, I personally don't like handling a large number of signal lines via terminal blocks. I also started to think about buffering the 7C81 from the servo and mill electronics. On the off chance that those electronics went south, I didn't like the thought of it possibly toasting the FPGA. Especially since the FPGA is connected directly to the RPi that I spent a lot of time setting up with Linux and LinuxCNC.

So at this point I wanted to make my own BOB that was space efficient at converting the three 26 pin "ribbon connectors" to lower pincount connectors where I could run low wire count cables directly to  each individual servo, and a couple others for other purposes such as limit switches etc. Since I was going to the effort of making a BOB, it seemed only logical to include the opto-isolator rather than having an additional board for that purpose. Figured I might as well put a voltage divider on my BOB that would make the 0-10v out of the PWM DAC 0-5v to feed into the KBSI-240D. Since the BOB now needed power for the opto-isolator anyway, and I felt why not buffer the signals from/to the 7C81 and improve robustness, So I decided to put a couple signal driver ICs on the BOB. Transceiver ICs that level shift are common. Since the  PWM DAC input was 3.3v, and the mill and servos are 5v signals, running the 7C81 at 3.3v and using level shifting transceiver ICs was a logical choice.. Two 74ALVC164245 bus transceivers handled the number of signals I needed and have good specs. At the time designing the BOB was a fun diversion. Amazing what one can do with the FOSS available today. I'm very impressed with KiCAD. My BOB is 2.5" x 3.5" which is pretty compact for the functionality it contains, and I'm rather proud of it.

Best, Gene

One thing is to find the edges of the plate, another is using scrap plate that can have holes from already cut parts. If you just want to find edges and always have square plates the you can use capacitive, laser range sensors or almost anything that senses metal.
In my mind the time saved is quite minimal and only works for squares or rectangles. Well of course you could run a line on all edges but would it be worth it, not really sure.

The camera stuff would be way nicer, example stuff like this Trumpf cut and drop . Libellula Optia the video could show case it better but basically you take a picture of the sheet and then you can nest to it.

Another possibility would be to utilize the spindle at speed section of Linuxcnc. Manual sections 5.6.2 and 5.6.2.1.
if you want to use classic ladder you may also want to explore the linuxcnc manual section 8
 

Not sure but it seems simpler to have a standard X/Y plasma table
with a roll-fed extension. The roll feed could be implemented in hal,
perhaps using the limit component to generate the motion profile
with the required Y distance and a PID loop using the encoder as
feedback.

I think i found the AXIS.GUI source code in qtaxis.ui to edit it with Qt 5 Designer . Is this correct?

Thanks a lot.

Sorry tommylight. I thought you added cammera.zip, but it was cmorley. So sorry.

I've been trying to figure out this post processor with F360 as well. Thanks Hakan for your help, I think I also found that there should be an "if centerpunch" statement in the onRapid function around line 407.
if (centerPunch) {
writeBlock(String("G0"), x, y);
}
else {
writeBlock(gMotionModal.format(0), x, y);
}
After adding this, for the one part I was testing it seemed to issue all the G0 commands prior to the M3 commands. However, when I tested a part with more holes, on the last one it failed to issue the M3 start command and was not going through small hole section at all in the post processor.

I found that the difference evaluated by this line 502: "if (Vector.diff(circleBuffer.start, circleBuffer.end).length < toPreciseUnit(0.01, MM)) { " was 0.22 something which was greater than the allowed tolerance, but I am unclear as to why. All the other holes of same size were showing 0 for this difference. If I raise the threshold to 0.25, the centerpunch command gets executed, but I'm not sure what the ramifications are on doing that. If I don't change that, the else gets evaluated on line 577, causing the centerpunch to get skipped: writeBlock(gMotionModal.format(circleData.clockwise ? 2 : 3), xOutput.format(circleData.end.x), yOutput.format(circleData.end.y), iOutput.format(circleData.center.x - circleData.start.x, 0), jOutput.format(circleData.center.y - circleData.start.y, 0));

N375 M5 $-1
N380 G0 X12.92307 Y10.51936
N385 G0 X37.03129 Y123.7121
(36.1442, 124.511, 0)
(37.0313, 123.712, 0)
(37.0313, 123.712, 0)
(0)
(0)
(TRUE)
(TRUE)
(CENTER PUNCH THE HOLE ONLY)
N390 G0 X36.1442 Y124.511
N395 M3 $2 S1
N400 G91
N405 G1 X0.00001
N410 G90
N415 M5 $-1
N420 G0 X37.03129 Y123.7121
N425 G0 X82.04218 Y124.49036
(83.2358, 124.511, 0) //Circle buffer center
(82.0422, 124.49, 0) // Circle buffer start
(82.042, 124.511, 0) // Circle buffer end
(0.020638258210129726) // Difference
(0) // Plane
(TRUE) // Small hole section
(TRUE) // Centerpunch
N430 G3 X82.042 Y124.511 I1.19363 J0.02064
N435 G1 X82.04218 Y124.49036
N440 M5 $-1
(END CENTERPUNCH)

Im sorry i got a little of target here.
My real questions are: should the shutdown.hal run at start up? 
Is this expected behavior?
How fast are loaded user components removed after a shutdown?
Is there a way to pause for a few seconds after the shutdown.hal is called before user components are removed?

All that being said.  
Maha thanks for the link.  i had missed a python library for at exit.  this will almost fix my issue.  i had looked at that a while ago to answer other question i had, for got what he was working on.