Anyone Actually Using External Offsets Branch for THC?

Hello

Is anyone actually using the External Offsets branch for torch height control on an operational table? Please tell me about what you had to to to actually get cutting. I have been lurking frequently but not working on my LinuxCNC setup for quite some time. I think that I have one of the earlier release of this branch installed. The really embarrassing thing is I don't even remember how to download the source code and compile it get a newer version. I want to try and get my table changed over to running under LinuxCNC now that I have all of the new Z axis mechanics and torch holder complete.

What I have and what I think works.

LinuxCNC 2.8. Pre with external offsets branch of some unknown vintage installed and apparently working with Z axis on bench.
Gmoccapy as GUI. My setup is in imperial units so I don't think I have EO scaling problem to deal with.

I can home table and jog all axis reliably. Hardware limits work on all axis. I can run simple Gcode test files to exercise X and Y axis reliably. I have working touch off that works with both ohmic sensing and float switch. I still need to add more error detection to touch off code. I already have more than most but know from hacking my present external THC that more is better if you don't want torch crashes. I can read arc voltage reliably when cutting (LinuxCNC running in parallel with my present controller). I can read arcok (arc transfer) switch from plasma power supply and start the arc (when connected). I actually figured out how to browse my Windoze network today to get files from my NAS and other Windoze boxes.

What I know I need before operation.

Get torch crash switch ( I can read it now) to implement feedhold ( actually I need to get feedhold working from GUI as well) and shut off torch and retract Z axis to safe height.

Get a working PP for SheetCAM that is setup for my touch off without all the extra stuff that most of the LinuxCNC ones have.

Tune PID for THC, I have no idea where to start as I am not using Axis.

Test cut a million times to try and find errors in my code.

What else do I need? I plasma cut almost every day so basic CNC cutting operations are on auto pilot most of the time.

John

John, I'm not aware of anybody that has got it going, but I am sure it can be achieved as I cut a 400mm long part which buckled 25mm while cutting at one stage.

Do this to update your code
1) if building from source:
a) $ git fetch
b) $ git reset --hard origin/dgarr/external_offsets
c) $ cd the_src_dir; make && sudo make setuid

This will break your old config as there has been a lot of changes. Dewey has implemented a number of changes from the feedback I've given him both tho the sample gcode and to fix a couple of bugs he found. He's also kinda locked in all of the complex stuff in a library hal file that you don't need to touch and its not populated to your config folder.

I would strongly recommend that you cut and paste your setup in its simplest form into a new Axis version of your setup based on the eoffset_pid sim as per the detailed instructions. I eventually did that. It will only take a couple of hours. Do your tuning with the tools provided then fold it into your Gmocappy setup.

I've probably got a usable Sheetcam PP for the current eoffset_pid setup.

I've also got a nice voltage sampling component that makes sure the set point starts off on the actual voltage plot as well as a velocity hold component.

I've been a bit like you, just no time plus its been very hot here so the garage is fairly uninviting.

The biggest thing you could do to help Dewey is to provide some halscope plots showing the torch voltage rise time so he can fine tune his software model of the torch. This would allow a much more accurate software tuning process which when you read about PID tuning is how most non-CNC PID tuning is done. (eg in a software model). He has quite specific requirements.

In the sim the torch is the only simulated component as the rest of the machine movement, velocities and acceleration is based on the actual machine hardware. I think this would be much easier for you as you have a known working table as a reference whereas I had to start from scratch with a plasma machine of dubious quality.

I'll dig a bit of stuff up for you while you update your config to the latest version and configure a sim that uses your actual velocity and acceleration parameters in the in i file.

Rod

Thank you for the reply. I was afraid that I was going to be in for a major rework of the code. I am sure all the tricks and procedures that I followed from others work on this forum will come back as I get going. Do you know where Dewey has the requirements for the HAL Scope files defined? I was actually planning on making some HAL Scope recordings in the next few days using my present external THC.

I may have a go at getting an Axis configuration going tonight. I am assuming that I use 2.8 Pre with a RIP install like my present configuration. I am not sure about the servo tuning right now because I can't use the new Z axis drive until I pass the point of no return with the install (very different mechanics and electronics).

I am going to test my new Z axis mechanics with my present servo motor and electronics today and see if it will work. The code changes in the micro will be less than a couple of hours for me. This will make the transition much easier if it works.

John

islander261 wrote:

Rod
Do you know where Dewey has the requirements for the HAL Scope files defined? I was actually planning on making some HAL Scope recordings in the next few days using my present external THC.

The requirements are in an email to me so I'll go digging.

islander261 wrote:

I may have a go at getting an Axis configuration going tonight. I am assuming that I use 2.8 Pre with a RIP install like my present configuration. I am not sure about the servo tuning right now because I can't use the new Z axis drive until I pass the point of no return with the install (very different mechanics and electronics).

You need to reinstall the external offset branch using the steps I outlined earlier before you start (actually, they are Dewey's instructions). Sounds like you need to get your new Z axis servo tuned before you start with this. I can't help there but I know Tommylight added a good tutorial on the forum here.

Having a servo might be a good thing. Dewey worked out you need to be able to accelerate at about 2120 mm/sec/sec to cut corrugated iron at book speed. I think acceleration and velocity will be pretty important tuning parameters. The current eoffset_pid component allows them to be changed on the fly.

Look out for an email from me.

I think linuxcnc will work just fine for a thc... (using only linuxcnc and a mesa THCAD)
This is just a hal mockup/proof of concept - I am currently looking at setting up a offset branch config.

200ipm - it was peaking at 10v (.040") error at the top of the corrugations. I don't think that is bad for a first (well 3rd) try. the tip may have been touching the metal - It all happened so fast



Probably take a bit.

Thanks for the replies.

I couldn't manage to get a newer version of the EO branch downloaded and installed using git. I did manage to download the .zip file containing it from the github site. Kept getting an error message about needing a valid repository, nothing I tried would work. Saved my old linuxcnc-dev directory as a backup and created a new one by using
git clone git://github.com/linuxcnc/linuxcnc.git linuxcnc-dev . Now I think I just need to extract the .zip file to the correct location?

Old servo motor does drive the new mechanics just fine. Hit a major snag in that local hardware store doesn't have fine thread SHCS so I am stalled until I get to the mainland next week. Still getting the rest of the fitting done today.

John

Rod

This what happens when I try and follow your instructions:
I've tried many ways to address the repository but evidently haven't found the correct one yet.

What I have done for now is to extract the down loaded .zip file into a directory named linuxcnc-dev.

John

so -
if you did
git clone git://github.com/linuxcnc/linuxcnc.git linuxcnc-dev

cd linuxcnc-dev
git checkout dgarr/external_offsets
cd src
./autogen
./configure
make
sudo make setuid

linuxcnc.org/docs/devel/html/code/buildi...cnc.html#Quick-Start

sam

John, I think that Skunkworks was right. you were not in the repository directory (linuxcnc-dev) before running the commands so git told you that Also I think the reset requires that the branch has already been checked out (which it should have been as you had an earlier version installed).

Thanks for the replies. Just blindly following instructions here.

What I did that appears to work.

Downloaded .zip file from external_offsets branch on github.com
Extracted files to linuxcnc-dev preserving directory structure
cd linuxcnc-dev/src
./autogen.sh
./configure --with-realtime=uspace
make clean
make -j2 [the number of cpu cores you have- speeds up compiling-Parallel compilation]
sudo make setuid

As expected when I copied over my last working configuration it is broken with this version. Interestingly the errors don't seem to be related to external-offsets but are in the pendant code. Now do I need a full working Axis configuration to test the Z axis or just a sim configuration?

John