DIY Arduino THC - Video

[Rambozo]

Very nice. I'm sure you can smooth out some of the steps with a little tuning. Maybe tighten up the hysteresis values a little so the steps are smaller. Any digital system is going to have some steps, and from what I can see the cuts look fantastic, so maybe you don't need to touch it. The kerf detection is a great feature, and I would love to see a video of actually crossing a kerf. I just found a video clip of the exact problem I was describing before where someone was restarting a CNC in a hole that was cut. This was with a commercial THC and it was not able to detect the kerf and crashed the torch. Looks like you can beat that.

[EmptyNester]

I'm still waiting for the Rev B PCB's to come in (expecting them on Monday). However, I wanted to get everything posted as promised.

The Rev B hardware design (eagle files), Arduino THC code and Windows data capture code are now all uploaded. They are on "github"
http://github.com

They can be found by searching on "regeg/Arduino"

Work will continue. I also have to work on the "users manual", software build instructions and the hardware build instructions.

If Rev B works as hoped, then I'll have a higher speed serial link and may try to add control from the Mach interface. Also thinking about moving to the Freescale Freedom platform, a 32 bit arm processor that's only $12 for a board.

[sportbike]

Sounds good. Once you have the logic figured out swapping to another platform should be easier.

It is amazing how much you learn when doing a project like this.

[EmptyNester]

I've finished building the Rev B board. The initial testing has been on the serial port. With the old hardware, it would only go 38,400 baud. With the new design it will do 115,200 (I haven't tried faster). I still have to finish testing the rest of the functionality (but there were no other functional changes).

I started looking into what it would take to integrate the THC right into Mach3 with a plug-in. Seems do-able, but more work than I'd like to do right now. But for easier testing I've been updating my Windows interface app to allow control and monitoring of the THC.

I've found that with the THC below the table (between the PP50 and my motor controllers), it's not convenient to access it, or even see it, while cutting.

Dropping support for the LCD would cut the cost by about $20 to $25 (cost would be about $20 for Arduino Mega, $2 for relay, $5 for USB to serial, $10 for THC shield PCB, $15 to $20 for THC shield parts, $10 for a wall wart - so about $60 total). It may also improve the performance of the THC and allow you to just watch the screen w/ Mach and the control app.

While it needs more work, it currently looks like:




HERE'S THE QUESTION:

For those with any interest in this project, would you prefer a THC with it's own LCD and push-button interface, or a PC software interface app?

[sportbike]

I have interest..not sure that I will ever proceed with building a table or not though.

I'd say the software interface is fine. Maybe an LED on the box to show it is in fact on and / or arc ok, if that is doable. Even that may not be of much benefit really.

The motion controller is solely PC based through Mach 3, I don;t see any issues with the THC having only a PC interface as well.

[acourtjester]

From what you say having the display is not the much more money and I seems simpler to look at the Arduino display for voltage reading and making changes there. I would rather have the LCD and not the Mach3 screen with another page to look at. The movement of the Z axis display of Mach3 will show what’s going on there.

My 2 cents

Tom

[EmptyNester]

From what you say having the display is not the much more money and I seems simpler to look at the Arduino display for voltage reading and making changes there. I would rather have the LCD and not the Mach3 screen with another page to look at. The movement of the Z axis display of Mach3 will show what’s going on there.

My 2 cents

Tom

Tom,

I'd like to understand how you use it. If the control of the THC was on the "default" Mach screen - would that make a difference?

I've found the Mach screens are overwhelming. So I was thinking about doing a custom screen with just the stuff I care about. Then later, including the THC control interface on the "simplified" screen. (Right now the interface is in its own application, but that sucks because Mach takes up the whole screen and you keep having to put one in front of the other.)

I like the idea of the stand-alone display. But then placement of the controller becomes important. So, while it was handy while I was developing the software, I've found that it's not convenient for me when I'm using the table. But, after all this work, I'd like to end up having a system that would work well for others.

Thanks,
Rege

[acourtjester]

HI Empty
I was under the impression that the voltage control would be on the arduino screen and adjustment could be make independent of the Mach 3 control. I though that Mach 3 controlled the Z axis up/down due to the changing voltage sensed by the THC unit. I have seen custom screens for Mach that the guy put just what he wanted on it. Or maybe a separate screen could be developed and selecting a different screen while it is operating could be used too. Like changing between the program run screen to the tool path screen.
I have the same problem with Mach 3 hogging the monitor screen.
I may be way off base with how I am thinking don’t allow me to throw you off by my lack of knowledge about the operation of the THC. And I do not know how rework Mach 3 screens.

[bfleming]

Don't mean to just jump in here, but I have been looking for a DIY THC for some time now. I certainly want to help, but it will be limited until I get my hands on a compatible plasma.

One thing I noticed about your sawing problem... It probably a very similar problem to what we do in the aerospace world which is to zero the error (difference between ideal and filtered reading) you do so by integrating over time, not just instantaneous changes. In other words, if you see the voltage drifting high the start ramping the z-axis in response. The key is to make the changes over time, not instantly, which is all about tuning the reaction velocities and accelerations. One thing that helps here is prediction of where the voltage is going to go the next sample. Then you target your velocities to intersect the predicted voltage at some point in the near future (100-200 ms). This is an area that i'm fairly familiar with and like to help tune the algorithms.

Since I don't have a machine... I'm going to have to write some simulation code based off your waveforms to see what kind of reactions I get. The one thing that will hurt is the 60hz oscillation your getting. You mentioned 2 voltage sources, are both voltages in phase with each other? If not then we can use the difference of the two to cancel out the ringing. Just a thought.

[EmptyNester]

Don't mean to just jump in here, but I have been looking for a DIY THC for some time now. I certainly want to help, but it will be limited until I get my hands on a compatible plasma.

One thing I noticed about your sawing problem... It probably a very similar problem to what we do in the aerospace world which is to zero the error (difference between ideal and filtered reading) you do so by integrating over time, not just instantaneous changes. In other words, if you see the voltage drifting high the start ramping the z-axis in response. The key is to make the changes over time, not instantly, which is all about tuning the reaction velocities and accelerations. One thing that helps here is prediction of where the voltage is going to go the next sample. Then you target your velocities to intersect the predicted voltage at some point in the near future (100-200 ms). This is an area that i'm fairly familiar with and like to help tune the algorithms.

Since I don't have a machine... I'm going to have to write some simulation code based off your waveforms to see what kind of reactions I get. The one thing that will hurt is the 60hz oscillation your getting. You mentioned 2 voltage sources, are both voltages in phase with each other? If not then we can use the difference of the two to cancel out the ringing. Just a thought.

Feel free to jump in - that's what its all about.

I reached the point with the first pass of hardware where I have filtered the 60 Hz out. The signal is very clean an only has +/- 1 A/D count jitter (where 7 counts = 1 volt). It think the "sawing" has been pretty much eliminated. What I see now if more discrete stepping of the height. But, I think that is more a factor of the speed of the Z-axis.

I've been working on the Rev B hardware, but made a change to the way I did the ground plane on the board and the signal totally sucks. The one advantage of Rev B over Rev A is that the serial port now runs at 119.2K vs. 38.4K. It's fast enough to do all control via the serial port rather than the LCD and button interface. (I'd eventually like to integrate the interface into Mach.)

I have a couple more things to test and then I'll start work on the Rev C board to try to make the best of both hardware configurations.

Doing filtering over 100-200 ms is way to slow. I'm currently doing kerf detection based on large instantaneous change over 6 or 8 ms.

Everything is available on GitHub and it you PM me with your email address, I can send you some data files.

[EmptyNester]

After building two sets of Rev B hardware and doing extensive testing, I am going to declare it a fail. With the same single pole filter that got rev A to +/- 1 count (1/7 volt), rev B hardware is greater than +/- 1 VOLT and intermittent spikes that are worse.

There are minor hardware changes to rev B, but I think the changes that did it are either or all of:
- split ground area on board (torch voltage filter split from the rest of the board except for a couple thin traces)
- no +5 volt plane on the torch voltage filter portion of the board
- filter circuit spread out to a slightly larger area (crossing all the analog input pins on the Arduino)

So, I will finish up Rev C. The planned changes for Rev C from Rev B are:
- power inlet adapter to feed the shield instead of the Arduino
- put a +5 voltage regulator on the board
- shield will be the size and shape of the Arduino board (versus a vertical shield that leaves room for the LDC)
- Use uninterrupted ground plane on bottom of board
- Use uninterrupted +5 volt plane on top of board
- Use solder-on right angle DB-9 connectors for connecting CNC and torch (so you don't have to make wiring harnesses)

Since it seems pretty straight forward to do an integrated THC inteface in Mach - I'm going to go that route and abandon the LCD display. It shouldn't be a big deal to add it back later if I change my mind.

I'm thinking of making the board jumper selectable to run on +5 or +3.3. This would allow it to work with an Arduino or a Freescale Freedom. But, I want to talk to my EE "consultant" before I do that.

So, depending on how quickly I can complete the changes, it will probably be about 3 or 4 weeks until I can do testing. In the mean time, I've put the Rev A hardware back on the table since it is a solid performer.

[gerrymoore]

Hi all...I've been following this link, with great interest, for a long while! I have built myself a cnc machine for my plasma cutter but I really do need THC and can't afford any of the commercial ones...spent up making the CNC! I would love to have a go at building this but would like to ask a couple of questions....feel a bit cheeky as I've had no input but hope it's OK. First the hardware...I have seen an ad on EBAY for the Sainsmart UNO R3 Mega 328P-AU ATMega 16U2 board together with the Sainsmart LCD1602 LCD and a USB cable for £20...sounds a bargain but is this the right Arduino board? Second question...I have logged into the site on GitHub but don't seem to be able to download any files...is there a trick to it? Third and last...If I do manage to down the files..are there details of what components I need to interface with my plasma (which doesn't have a volt divide output). Look forward to answers and hopefully building myself a THC :-))

[bfleming]

Hi all...I've been following this link, with great interest, for a long while! I have built myself a cnc machine for my plasma cutter but I really do need THC and can't afford any of the commercial ones...spent up making the CNC! I would love to have a go at building this but would like to ask a couple of questions....feel a bit cheeky as I've had no input but hope it's OK. First the hardware...I have seen an ad on EBAY for the Sainsmart UNO R3 Mega 328P-AU ATMega 16U2 board together with the Sainsmart LCD1602 LCD and a USB cable for £20...sounds a bargain but is this the right Arduino board? Second question...I have logged into the site on GitHub but don't seem to be able to download any files...is there a trick to it? Third and last...If I do manage to down the files..are there details of what components I need to interface with my plasma (which doesn't have a volt divide output). Look forward to answers and hopefully building myself a THC :-))

No 100% sure on the arduino, yours is a 16Mhz CPU which should be fast enough but it's only 16KB of flash which is awfully low. EmptyNester was using an Mega, which has 256KB of flash. So I'd look for a Mega256 at a minimum.

To download from github, just go here and click the zip button. If you don't see it try this link.

Hardware is still TBD, EmptyNester is still working it. Course the best for everyone is to jump in a prototype his 3rd gen electronics on a breadboard and see how it works for your plasma cutter.

[EmptyNester]

No 100% sure on the arduino, yours is a 16Mhz CPU which should be fast enough but it's only 16KB of flash which is awfully low. EmptyNester was using an Mega, which has 256KB of flash. So I'd look for a Mega256 at a minimum.

To download from github, just go here and click the zip button. If you don't see it try this link.

Hardware is still TBD, EmptyNester is still working it. Course the best for everyone is to jump in a prototype his 3rd gen electronics on a breadboard and see how it works for your plasma cutter.

You are right - the flash is an issue. I currently use 18k+ (UNO only has 16k) for code and 1.6k of RAM. The Uno only has 2K RAM, so that is cutting it close too. Also, if you use an LCD shield, you have to have a Mega to have enough digital inputs. The final thing is that there is no second serial port on the Mega which I'm currently using for grabbing data and a control interface.

For general directions - once I get final hardware I plan on writing an overall guide.

The big thing to keep in mind is that I developed the hardware and software for an Everlast Power Plasma 50. I did that because that's the torch I have and it has a CNC interface. This should be easy to use with any torch that has a CNC interface with just a couple resistor value changes (in the voltage divider).

For non-CNC interface units, I've not tried to connect raw voltage. You would need at least a couple of higher power resistors and may need some hardware design changes.

Another factor that I think is important is that the PP50 also has pilot arc start. I'm under the impression that without pilot arc, it would be an RF start and that could cause significant interference problems.

I'm not an Everlast employee - but I'd highly recommend the PP50 is you are starting a CNC build. It has a great feature set, good tech support and a great community of forum contributors for help.

[bfleming]

Another factor that I think is important is that the PP50 also has pilot arc start. I'm under the impression that without pilot arc, it would be an RF start and that could cause significant interference problems.

I'm not an Everlast employee - but I'd highly recommend the PP50 is you are starting a CNC build. It has a great feature set, good tech support and a great community of forum contributors for help.

Now that's interesting that you brought that up. I'm helping a friend setup his plasma table (no z, so I can't use him to help tune the ArduinoTHC) and we noticed motors jumping.... Let me explain.

We were doing test cut circles to get everything tuned, and once one of the cuts went out over the space where another circle was cut, so essentially air cutting of the water table. As soon as the torch wasn't cutting steel, the carriage jumped across the void and continued cutting on the other side. So somehow cutting nothing caused the motors or driver to be all sorts of unhappy. And I do believe its a High Frequency start, and not a Pilot Arc.

[EmptyNester]

Now that's interesting that you brought that up. I'm helping a friend setup his plasma table (no z, so I can't use him to help tune the ArduinoTHC) and we noticed motors jumping.... Let me explain.

We were doing test cut circles to get everything tuned, and once one of the cuts went out over the space where another circle was cut, so essentially air cutting of the water table. As soon as the torch wasn't cutting steel, the carriage jumped across the void and continued cutting on the other side. So somehow cutting nothing caused the motors or driver to be all sorts of unhappy. And I do believe its a High Frequency start, and not a Pilot Arc.

Was the kerf it was crossing part of that G-code file? If so, a "smart" processor could know the torch is crossing a kerf and accelerate it (while possible to do, I'd be surprised if there was non-commercial software that did that).

To see if it is noise related, do the same cut with the torch off and see if you get the same jump. If you do, you can rule out noise from the torch.

[bfleming]

Was the kerf it was crossing part of that G-code file? If so, a "smart" processor could know the torch is crossing a kerf and accelerate it (while possible to do, I'd be surprised if there was non-commercial software that did that).

To see if it is noise related, do the same cut with the torch off and see if you get the same jump. If you do, you can rule out noise from the torch.

Wasn't a kerf jump, it was one of the jumps that made me go "Whoops, that isn't right". Imagine that we cut a 4" circle starting on the left side. The second circle test was cut 2" more left (by mistake) also starting left. As soon as it hit the empty space from the 1st circle it did a rapid move, in X only, about 4-5 times faster than we had rapids set for.

Couldn't see how it was anything other than interference when the torch entered HF mode to try to keep the torch lit. But I'm open to ideas.

[Rambozo]

Wasn't a kerf jump, it was one of the jumps that made me go "Whoops, that isn't right". Imagine that we cut a 4" circle starting on the left side. The second circle test was cut 2" more left (by mistake) also starting left. As soon as it hit the empty space from the 1st circle it did a rapid move, in X only, about 4-5 times faster than we had rapids set for.

Couldn't see how it was anything other than interference when the torch entered HF mode to try to keep the torch lit. But I'm open to ideas.

Sounds like it to me. You might need to do some shielding of the CNC. Complete metal chassis and shielded cables. Ferrite chokes can also help filter out HF. Also be sure not to create any ground loops as that will play havoc with digital inputs. Setup a star grounding system. Noise can enter in the strangest places. Even things like limit switches often need shielded cables and ferrite beads. Otherwise they act as nice long antenna to pull in the HF.

[gerrymoore]

If it's any help....my plasma cutter does have pilot arc start but when I was trying to get it to work with my CNC, everytime the plasma was switched on the Mach3 stopped with a limit switch warning. It took a long time to figure out what was causing it.....

First I rewired every cable - to the steppers, limit switches, estop and plasma torch - with new, good quality, shielded cable...result was exactly the same! Then I tried .1uf caps on all the BOB inputs...this helped a little but still not right. Then I changed the input signal debounce in Mach3 general settings to 100...better but still not right.

After much deliberation and sleepless nights I thought about the plasma torch...the 'beam' seemed much brighter with the machine torch I had fitted in place of the original hand torch....mmmm light bulb came on!

The torch had been fitted with a 1.6mm nozzle instead of the 0.8mm nozzle suitable for my plasma machine...changed this and voila it all worked perfectly......

Hope this helps in debugging your problem .....but I have read that HF start is notoriously bad for 'electronics' and only pilot arc is recommended for CNC.

[everlastsupport]

GerryMoore, all good info. I have heard by many HF is a pain to deal with. Decoupling and filtering is required. Common ground as well.

Not to change the subject, but at $35 for the base board, has anyone thought about the Raspberry Pi for the main controller? I have 3 now and use one to open and close my garage from my android phone. Something to look at for sure. Runs Debian and Android I know, probably many other OS's. Anything smaller I would just make by hand. But it is a lot of bang for the buck.