Contemplating Mach 3 to LinuxCNC

[xnaron]

I'm putting this thread in general as there isn't a LinuxCNC forum I can see that isn't related to the CandCnc controllers.

I've been using Mach 3 to control cnc machines like my CNC router for over 10 years. I bought a used langmuir crossfire and have done many mods to it. I added a floating z, Limits, estop, inputs, proximity probing and a magnetic collision detector. I ended up going with a proma SD for torch height which turned out to be a mistake. I was planning to move to a price or mini thc but this would mean that I need to move away from the controller in the crossfire https://buildyourcnc.com/item/electroni ... -USB-Board as it does not have enough inputs. I do have an ethernet smooth stepper and a BOB I could retrofit in the controller but it would require a rebuild and I don't think I have enough room in the existing enclosure to fit it. So looks like I need to build a new controller.

I had another thread for troubleshooting my Proma SD and mentioned the above and RodW suggested I might want to consider LinuxCNC. Having worked with Linux for the last 20 years professionally and personally this intrigued me and started me down the path... Here is the hardware I am thinking of using

Mesa 7i76e - Controller. Could go with the 7i96 but it is out of stock
THCAD-5 - monitor the 50:1 arc voltage from the Powermax 65 (is this the right THCAD for this or should I use a THCAD-10?)
THCAD-5 - ohmic sensing

My Hardware:
-24 x 30" table capable of 300ipm with water table
-Powermax 65 with machine torch and 50:1 and CPC port
-Custom floating z axis with magnetic breakaway collision detection system
-Proximity probe sensors (I think I will be ditching these)
-Planning to add ohmic sensing with the water switch. I know the water table will cause issues. I am hoping that maybe the Hypersensing that RodW worked on will make this a viable option
- Thinkpad touchscreen i5 laptop or i5 lenovo mini pc


Questions:
- The plasmaC docs point to installing this version of LinuxCNC http://www.linuxcnc.org/testing-stretch-rtpreempt/ Is this the latest install ISO? I heard someone refer to version 2.9. The path to 2.9 (if it exists) might just be to upgrade from 2.7. Maybe available in the installer? I need to investigate more,
- Ohmic probe with Hypersensing workable?
- any other pieces I should but from Mesa to make this work?
- Is the software in LinuxCNC and PlasmaC in a workable state? Still being developed? I know RodW has made a bunch of custom stuff and not sure what and what hasn't been integrated into LinuxCNC and PlasmaC.
- are there many people on this forum using LinuxCNC or is it still very niche?
- ideally I would like to install LinuxCNC onto my Touchscreen thinkpad. I am hoping that it will support the touchscreen.
- more questions to follow

Thanks,
Brendin

[SeanP]

Don't know about the rest sorry, but I had enough with ohmic sensing and went back to floating head, I found ohmic a pain in the butt to be honest, not really with water problems but failed touchoffs due to small deposits on the shield, dodgy connections on the shield tab, I find a well balanced floating head gives less trouble.

[robertspark]

Questions:
- The plasmaC docs point to installing this version of LinuxCNC http://www.linuxcnc.org/testing-stretch-rtpreempt/ Is this the latest install ISO? I heard someone refer to version 2.9. The path to 2.9 (if it exists) might just be to upgrade from 2.7. Maybe available in the installer? I need to investigate more,

You update it to 2.9 / master once you get the ISO working

- Ohmic probe with Hypersensing workable?

Rod uses it, if you have an ohmic sensor then it should not be necessary or the cheap alternative which is the water switch https://www.google.com/search?q=kemo+wa ... 24&bih=529

my ohmic sensor is adjustable via a pot on it.

- any other pieces I should but from Mesa to make this work?

your list looks complete

- Is the software in LinuxCNC and PlasmaC in a workable state? Still being developed? I know RodW has made a bunch of custom stuff and not sure what and what hasn't been integrated into LinuxCNC and PlasmaC.

it is workable and it is still being developed / honed

- are there many people on this forum using LinuxCNC or is it still very niche?

its a bit niche on here.... most seem to use Mach3 or CandCNC or plasmacam or some other proprietary software that came with their table (a [very] few of us seem to use uccnc)

- ideally I would like to install LinuxCNC onto my Touchscreen thinkpad. I am hoping that it will support the touchscreen.

no idea about the touchscreen, but thinkpad implies laptop which are frowned upon by linuxcnc due to power saving systems that are added to save power affecting latency / consistent latency. but you are using a Mesa motion controller card so its only the 1mSec Servo thread which is used for communication & timing with the mesa ethernet card

Really you are best to ask your questions on the Linuxcnc + Plasmac forum thread / section.
https://forum.linuxcnc.org/plasmac

[Rodw]

I'll see if I can answer some questions.

Linuxcnc's Plasmac config is now stable but its still being actively developed. Whilst it may be a niche system on this forum, its rapidly gathering a head of steam. There is no doubt in my mind its superior to anything on the market. Its not a THC, its a complete plasma controller with THC (on steroids?). You'd have to go to Hypertherm Pronest to get anything close to it (so I've been told).

Earlier this week another Aussie sent me a photo of a commercial system I know he spent AUD $12k on that he was ratting for parts. He wanted to use the toroids for another project. He said he could not bring himself to flog it off on ebay to some unsuspecting victim! He says adopting Plasmac was the best decision he's ever made. His machine is running in a major manufacturing operation and he worked on the RS485 connection to the Hypertherm. This was pioneered by Pedro in Mexico who retrofits and build plasma tables as part of his welding repair business. The also have offices in the US. So now Pedro's RS485 code is enhanced and incorporated into the Plasmac core.

I think the 7i76e is a better choice over the 7i96 becasue it supports MPG's and has a lot more inputs. The 7i96 is really nice too as it has built in relays and can be configured to be active high or active low where as the 7i76e needs a seperate version to do active low. Normally we would use an Active high config on a new machine but active low may be useful when retrofitting a machine to use Linuxcnc. The 7i76e needs external relays on outputs but the 7i96 doesn't. The only relay I use with hypersensing is for the torch on. You can use the 7i76e's built in spindle relay to trigger the torch but I did not know that when I started...

Plasmac has always had a couple of cool features requested by islander261 on this forum for water tabales. One is a button in the GUI to pulse the torch so he could get the air postflow going before starting a job. The other is an Ohmic test button he could press to check if the tip was shorted internally (if the ohmic sense light came on, there was a short).

Islander261 also contributed a robust ohmic sensing circuit using a couple of opto isloated relays and some diodes which is included in the plasma primer. This would be cheaper than the German made water switch which is not available in all countries (mine included). He is another guy we don't hear from much becasue Plasmac has solved all of his problems...

Hypersensing came about because Mesa offered to send somebody a free THCAD-5 if they would do some experiments so I held up my hand. The feeling was it would be much more robust as it could be left powered on all the time and 100% isolate the logic side from the plasma side. Peter from Mesa wanted to know if we could see a change in resistance with the THCAD as the torch touched the material. To use a voltage signal for a probe input, I needed to write a component. That is now in the Linuxcnc core (ohmic.comp) and instructions on how to use it is in the Plasma Primer I wrote which is accessible from the Linuxcnc docs for V 2.8 & 2.9. We found that we could see the probe voltage change over a 0.04mm distance. At that time I did not have a water table. the component works as an adjustable relay where the on and off thresholds can be varied to maximise signal hysteresis.

Afterwards, I added a water table and ran a job at 120 amps (which uses a much larger hole in the torch tip). I had some of the problems seanp mentions with water in the nozzle with the bigger orifice. Then a couple of guys had bought the THCAD-5 to use Hypersensing with Hypertherms (which seem much more susceptible to the water. So with a few other testers, I revised the ohmic component to become ohmic2.comp. Its not in the Linuxcnc mainstream but shared on the Linuxcnc forum. It combines a 25k 5 W pull up (or is it pulldown?) resistor on the high voltage inputs suggested by Peter at Mesa to increase the voltage differential when the torch was wet. The new component computes a moving average of the probe voltage so a combination of hardware and software have made it possible to probe reliably on a water table. I'm not 100% happy with it and want to add a couple more features before putting it forward to be included in the code. So right now official Hypersensing has the problems of conventional probing as mentioned by SeanP but unofficially, there is a more robust component that has been tested on several Hypertherms and my Thermal Dynamics that works reliably.

So onto Linuxcnc versions. Linuxcnc requires a real time Linux kernel so it can guarantee things happen exactly on time. The current official release version is 2.7 on a very obsolete version of Linux Debian Wheezy that is now end of life. It uses the RTAI kernel. Mesa ethernet cards require the PREEMPT_RT kernel. A year or so ago, they hived off V 2.8 with a view to doing a release upgrade. That means the unstable development branch is V 2.9. There has been a lot of problems with changes to Linux, the deprecation of Python 2.7 requiring a move to Python 3 that have made this a massive job. I think the dev team is just about done as their attention seems now to be on the documentation.

The current official path would be to go to the downloads page https://linuxcnc.org/downloads/ install the Debian Stretch distro which will give you a PREEMPT_RT kernel. From there its quite trivial to change the repository settings to point to V 2.8 (recommended) or V 2.9 after which upgrading Linux will grab the verison you are pointing at. You can do this in the GUI in the Synaptic package manager or from the command line per the buildbot. It should only take 10 minutes. Once that is done, you just need to do an upgrade at any time and you will update to the latest verison on the Buildbot which creates new versions whenever a change is made to Linuxcnc.

More recently, some Linux distributions are finally offering patches to install the PREMPT_RT real time kernel so i followed this method the other day which worked a treat. https://forum.linuxcnc.org/9-installing ... stallation
I had an existing USB with Mint 19.3 on it and it worked seamlessly.

Re Laptops. Typically Laptops contain a lot of power saving features in the BIOS which generate significant latency in the hardware (we gotta run on time right?) So they are unsuitable for use with Linuxcnc. On my machine I use a Gigabyte Brix J1900 Celeron attached to the back of the monitor. There are a few tricks in getting the BIOS settings right but it does the job.
The PC I installed on the other day is an Odroid H2+ and it looks like it will have lower latency than my Brix. It has a UART on the GPIO pins so I am hoping we can add a RS485 chip to that to talk to Hypertherms. Or to a Hypertherm Emulator on a Arduino.

Finally if any of the moderators read this far, it would be great if you could create a "Linuxcnc Plasmac" section in the forum and move this thread there as its probably as good a start as you could get.

[Rodw]

Oh and I forgot. The THCAD-5 is fine with a 50:1 divider. That gives you a 5x50 = 250 volt full scale reading. The THCADs can handle overvoltage of 500 volts indefinitely. Some machines can be configured with a 30:1 divider (my TD A120 has this as the default) and on those a THCAD-10 is ideal.

You only really need a full scale of around 200-250 volts as any full scale readings will happen at times we don't care about the torch voltage. Plus lower full scales results in more accurate voltage readings in the range we care about. So with 30:1, my full scale is 300 volts and I do see it go past that on piercing, particularly on 16mm material.
EDIT:
And if you are ordering from Mesa, be sure to get a couple of their common block terminal strips and some of their DIN rail mounts. I think you need 6.

[Rodw]

Here is a 7i96 with all required hardware. Note the Common block (2 x 12 pins commoned) Use 1 for +ve field power and the other for -ve field power. The piece of circuit board was just a place holder to so you might need to add some resistors but I'm told you can get them mounted in DIN rail terminal blocks.
For a 7i76e, you can ditch the smaller 5 volt power supply on the left.

7i96.jpg

[xnaron]

Thanks @robertspark and @Rodw for the infor. @Rodw thanks for providing such detailed info. It is just what I needed to get my order ready.

Mesa has only one thcad-5 left. I received the Kemo waterswitch the other day but have not tried it yet. However I don't think I can use the hypersensing without a thcad-5 for the ohmic. I am in Canada and shipping is expensive from Mesa so I'd like to be sure I can get everything I need in one order.

On another note I was looking at the https://www.cncdrive.com/UC300ETH.html UC300ETH and saw that it has mach 3/4 and uccnc drivers. I searched to see if it also supported linuxCNC but I couldn't find anything saying it was supported. It would be nice if that card did as it would allow me to experiment with all 3. I could also retrofit it into my existing controller easily. That would buy me some more time to build a new custom linux cnc controller.

As for the pc to control it I won't use a laptop then. I do have a mini pc I will use. It is a lenovo thinkcenter with core i5-4570t with 12GB ram that I will use instead.

[robertspark]

I use uccnc on a uc400eth dedicated for a minithc and plasma.

I also have a uc300eth that I use with uccnc for a mill, router and lathe (although uccnc does not support lathes)

uc400eth, uc300eth, uc100 and the axbb-e only run mach3, mach4 and uccnc, no Linux.

I am happy with uccnc.

I run the development release of uccnc. it has a lot of synchronous macros to control thc and turn it on and off in sync with motion.

there are various plugins including modbus so it could be made to communicate with hypertherm rs485 serial / ASCII if desired.

it is a very good upgrade to mach3 and a lot of people have converted from it to uccnc

[Rodw]

Thanks @robertspark and @Rodw for the infor. @Rodw thanks for providing such detailed info. It is just what I needed to get my order ready.

Mesa has only one thcad-5 left. I received the Kemo waterswitch the other day but have not tried it yet. However I don't think I can use the hypersensing without a thcad-5 for the ohmic. I am in Canada and shipping is expensive from Mesa so I'd like to be sure I can get everything I need in one order.

On another note I was looking at the https://www.cncdrive.com/UC300ETH.html UC300ETH and saw that it has mach 3/4 and uccnc drivers. I searched to see if it also supported linuxCNC but I couldn't find anything saying it was supported. It would be nice if that card did as it would allow me to experiment with all 3. I could also retrofit it into my existing controller easily. That would buy me some more time to build a new custom linux cnc controller.

As for the pc to control it I won't use a laptop then. I do have a mini pc I will use. It is a lenovo thinkcenter with core i5-4570t with 12GB ram that I will use instead.

You should be able to use the Kemo switch. More or less follow the relay circuit in the plasma primer when wiring it in. Just pulse the torch to get airflow before starting a job.

The UC300ETH is a motion controller. Linuxcnc is a Motion controller. You must use one or the other. You can't use both. This is where Linuxcnc becomes superior. All motion is done on the PC using a real time operating system. This is not possible in windows as its not real time and latency abounds so the Windows solution has been to move the motion controller to an external board with an emasculated embedded controller that simply cannot match the superior processing power of a full blown PC's CPU. Linuxcnc's architecture allows you to write a few lines of C code and compile it into a component which can be installed via a hal file (simple text file). Once your custom component is installed it is treated as if its part of the Linuxcnc's motion controller so the system is very extensible. Plus being open source, it is possible to modify the internal code. I am in the process of doing this to publish the arc radius direct from the interpreter so that a real time component can alter hole and arc processing without any need for g code. Plasmac already supports something similar by filtering and editing your code files as they are opened to process holes differently.

The mesa cards are fairly dumb devices. Their primary purpose is to offload step generation from the motion controller (linuxcnc). LCNC tells the Mesa stepgens to generate pulses at a specific frequency until its told to stop or change the frequency. Linuxcnc can then change the commanded frequency 1000 times a second. The Mesa cards can generate steps at up to 10 MHz >3x faster than the average Mach external controller.

[xnaron]

Thanks for the clarification. I wasn't sure of the function of the mesa boards. I have a Lenov Tiny PC

Annotation 2020-07-16 100613.png

I could put linux cnc on that to run the controller. It is a 2.9Ghz i5-4570t with 12GB ram. I would like to use my windows 10 touchscreen thinkpad laptop as the control console. I believe you had mentioned using team viewer with LinuxCNC in your build thread. I'll try vnc first.

I've been sidetracked for the last couple of days redesigning my floating z axis. I went from the wheels to Hiwin mg9 rails.


The old

20200712_100521.jpg

The New

20200715_204245.jpg

20200715_204241.jpg

Now back to the LinuxCNC controller build.

[Rodw]

Not Teamviewer. I just installed XRDP on Linux and used the Microsoft RDP client included in Win10 Pro.
Nobody does it but Linuxcnc has a client/server design. There is a config file that says what IP address it should connect to and by default thats 127.0.0.1 (localhost). I've not played with that..

I like your Z axis design. I looked at a side mounted design like that but found it did not really reduce the distance from the mounting surface to the torch tip.

Is your float switch based on the torch sliding up in a sleeve? Thats cool!
Can you tell me about your sensors? Whats the plan there?

[xnaron]

XRDP even better!

This crossfire is pretty small so the goal with the z axis was to keep it small and light. I used the plates I had cut out on the plasma cutter earlier to mount the 8020 to the slides. The ballscrew is keyed into the 8020 piece with a bolt on one side and a bolt with a spring on the other. I used the plates I had cut out on the plasma cutter earlier to mount the 8020 to the slides. The spring is here to offset the weight of the torch. I measured the force required to do a probe and it is around 400 grams. I don't have a strong enough spring in there though. I am sure I can get it to half of that.

The sensors are proximity sensors

https://www.automationdirect.com/adc/sh ... /ksk-ap-4h

They are supposed to sense any metal at around 18mm. They connect to a pnp transistor and then to a probe input. I use a relay to switch the power on and off for the sensors and this will let me use the floating z instead. Also the floating z will trigger if there is a failure on the proximity sensors. A guy named tool junkie on youtube made some videos about them and that is where I got the idea to try them. He seems to have had better luck then I did with them though. One of the things I noticed when it was probing 16 gauge was that if it happened to probe over a slat it seemed to trigger at an incorrect height. To be fair though I have not spent a lot of time trying to debug them. I removed them today as I am going to hookup my ohmic probe to the Kemo water switch. From what I have read I should use an isolated powersupply for the Kemo which I will do. I have the ohmic probe for the hypertherm and will connect one wire from that to the kemo water sensor and the other wire to the metal cutting table. The relay outputs will go to the probe input on my controller.

[Rodw]

Cool. Looks good. Yes keep the probe on a seperate power supply. I would suggest a Meanwell HDR-15-12 which is cheap and is also a Class 3 (or something like that) which means its isolated to handle up to 500 volts.
See https://www.meanwell-web.com/en-gb/ac-d ... dr--15--12

I've got the 24V version and it takes the full plasma arc in the circuit when its cutting and its not a problem at all. I'm not sure if the kemo will like that though! Plasmac supports a probe enable pin so you can use that to power up the probe circuit only when probing. Opto isolated relays are the way to go. The US guys have been using Opto22's but they are pretty expensive down under so I use RS Components 888-6843 slim line Din rail relays (which you have to buy in a pack of 5). If I moved the torch to the 7i76e's spindle relay, I would end up with a relay-less machine!

Can you provide a link to the prox sensors?

[xnaron]

I did put the link to them in the post just above yours. If you order be sure to get the connector cable with it. It is linked in the sensor page.

As for the opto 22... Something like this?

[xnaron]

Oh and are they using two of them to switch both leads of the ohmic?

[Rodw]

No more like this.
https://www.opto22.com/products/dc60mp

There is a mounting board that takes 4 of them and they come in input and output versions.

[robertspark]

No more like this.
https://www.opto22.com/products/dc60mp

There is a mounting board that takes 4 of them and they come in input and output versions.

like these??

[Rodw]

I'm sure it would do.

This is the Opto 22 board
https://www.opto22.com/products/pb4
You can get from Element 14 in the UK.

But I've been using this (In a pack of 5) https://au.rs-online.com/web/p/solid-st ... /?sra=pstk
I've been slowly pulling them out actually. Relays are breeding like coathangers here

[robertspark]

. Relays are breeding like coathangers here

..... like my collection of lathe chucks....

[Rodw]

. Relays are breeding like coathangers here

..... like my collection of lathe chucks....

I'm actually short one after upgrading my lathe so if you have a D1-4 5C collet chuck Send it my way

[simkellar]

XRDP even better!

This crossfire is pretty small so the goal with the z axis was to keep it small and light. I used the plates I had cut out on the plasma cutter earlier to mount the 8020 to the slides. The ballscrew is keyed into the 8020 piece with a bolt on one side and a bolt with a spring on the other. I used the plates I had cut out on the plasma cutter earlier to mount the 8020 to the slides. The spring is here to offset the weight of the torch. I measured the force required to do a probe and it is around 400 grams. I don't have a strong enough spring in there though. I am sure I can get it to half of that.

The sensors are proximity sensors

https://www.automationdirect.com/adc/sh ... /ksk-ap-4h

They are supposed to sense any metal at around 18mm. They connect to a pnp transistor and then to a probe input. I use a relay to switch the power on and off for the sensors and this will let me use the floating z instead. Also the floating z will trigger if there is a failure on the proximity sensors. A guy named tool junkie on youtube made some videos about them and that is where I got the idea to try them. He seems to have had better luck then I did with them though. One of the things I noticed when it was probing 16 gauge was that if it happened to probe over a slat it seemed to trigger at an incorrect height. To be fair though I have not spent a lot of time trying to debug them. I removed them today as I am going to hookup my ohmic probe to the Kemo water switch. From what I have read I should use an isolated powersupply for the Kemo which I will do. I have the ohmic probe for the hypertherm and will connect one wire from that to the kemo water sensor and the other wire to the metal cutting table. The relay outputs will go to the probe input on my controller.

That is a great looking torch holder. Did you make it?? Cheers, Simon.

[xnaron]

Yes. Details here