DIY Arduino THC - Pass 2

[EmptyNester]

The software I've written to date (fairly complete), is running pretty solid. My issues have been with the voltage divider and hardware filter.

The first pass of the op-amp based filter circuit did a good job filtering, but always had about 1.2 volts output with zero volts input, and the output was limited to about 3.8 volts. This was coupled with a poor choice of voltage divider values. While it ended up working, the voltage resolution wasn't as good as I wanted. (It would servo properly and "seemed" to work. I didn't test extensively and am still on the CNC learning curve.)

That first filter was creating using "FilterPro Desktop". It's a free filter design application from TI.

For pass 2 - I used "Filter CAD", a free filter design application from Linear Technologies. This approach used a passive filter chip instead of an op-amp.

The differences for the pass 2 design were:
1) new filter circuit
2) new voltage divider
3) added opto isolators on serial output
4) added opto isolators on the CNC signals (Torch On, Arc Good, Torch Up, Torch Down)



The picture above shows the Arduino Mega with a prototype board mounted on it containing the LTC1062 analog filter circuit and some opto-isolators. It's sitting on top of the schematic for the new filter.

The filter circuit did a good job filtering, but the response was sooo slow that it can't be used to servo on. The capture below shows the trace.



The yellow trace is the voltage just coming off the CNC port. The red trace is the voltage coming out of the filter. (The noise shown on the red trace isn't seen by the A/D. I think that's common mode noise the scope is picking up. The analog read of the value is pretty stable.)

On the trace, each square horizontally is half a second. You can see that its taking well over a second for the red trace to reflect voltage changes in the yellow trace. So, this filter doesn't seem to do the trick.

The values I used for the new voltage divider were better, but still not as good as I'd like. The divided voltage for the pilot arc (as high as I expect the voltage to be) is about 3.12 volts. So, it's improved, but I want to crank it up a little more to increase the resolution of the voltage reading.

I added the opto isolators on the serial output. Then I connected that to a TTL to USB interface (couple bucks on ebay). This allowed me to capture data (print out torch voltages so I can graph them) and debug (print debug messages based on behavior). In the future, it could be used to download presets based on the material being cut and the feed speed.

The circuit I found on the web uses a lower speed opto isolator. It is supposed to work at up to 19.2K baud. I didn't mess with it too much, but had no problem running it at 9600 baud. That's really sufficient for what I need, so I probably won't change the circuit to get higher serial port baud rates.

When I was building the board, I started wiring the opto-isolator for the CNC interface. After I got half the parts on, I ran out of space. Since the serial port works, I know the CNC signals will work. So, I won't do any more work on that until I do a custom PCB.

Next Steps........

I couldn't do a Spice circuit simulation of the filter chip since no model was available of the chip. However, using the model of the first circuit I built, I was able to accomplish two things. They are:
1) figure out the exact voltage output of the CNC port and ideal voltage divider values (taking into account the 100K resistors internal to the CNC port on both the positive and negative side).
2) found an op amp that will plug into my first circuit and should give me the best possible filtering (If the simulation matches reality)

I just ordered the new op amp and voltage divider resistors from Digikey and hope to be able to try again on the weekend.

If the next hardware update works - I will probably make a (prototype) custom Arduino shield board to make assembly really easy. After that and a little more debugging, I'll post all the information online (circuit and software).

--

The Christmas present my wife "ordered for me" showed up on Monday. A 200DX. So, that will be a whole new set of learning challenges (never did TIG before). Of couse, I told her I had to open it to "inspect for damage". Its a thing of beauty I just won't start using it until after she "gives it to me" I'll be on the forum a lot for help when I get started with that.

[sportbike]

Looks like you are making progress, or at least learning a hell of alot more about circuits then you knwe a year ago...

The 200DX is a good choice. I would suggest performing a multitude of testing with it to ensure all is in working order. Remember to test multiple times as well, adn to test for a duration long enough to make sure it is working. Let it cool down overnight, then test again. Repeat about 10 times, just to be sure...

Once you get all of this working, I just may have to invite myself over to check it out.

[EmptyNester]

I just finished some pilot arc voltage capture testing.

This is the "third" hardware configuration. It uses the same op amp circuit as the first pass, but a different op amp. The first op amp never read less than 1/3 scale and never more than 2/3 scale. I've replaced the old op-amp (TL972) with a new one (LT1632) and I changed the voltage divider to get a higher input voltage (and that gives better voltage resolution).

This is a capture of serial output from the Arduino. I had a lot of data corruption using TeraTerm to capture. I had to delete a lot of corrupted data, but I think I got a reasonable representation of the filter output.

The capture below shows the analog reading that was filtered by the hardware and then by the software using a sliding window average of 5 samples. The averaged value was then converted to volts input (analog count / 7).



The x-axis numbers are sample numbers not time. From the time the voltage initially hits 110V until it drops is about 3.3 seconds of elapsed time.

You can see that the reading has about 2 to 3 volts noise in it. But, because there are so many samples, I think that can be smoothed either with a hardware filter tweak or a larger software filter sample size.

I'm confident that this can be made clean enough for decent control.

My next step is to draw out all the schematics and layout an Arduino shield. I'd like to make a couple prototype boards to have a cleaner setup to test with. I will continue to test, but I want to do the proto boards first since it will take about a month to get them.

Again, the end goal is to publish everything (schematic, board layout, source code, design description) so that people could easily build a THC. I've had a couple people PM me about making enough boards to sell. If you're interested, send me a message. I'd like to get a count of people that would be interested in buying a bare PCB. Ideally, if there's enough interest, I could get one of the Arduino parts companies to make the bare PCB's and sell them.

[SeanMurphy265]

You should do what got trikes did with his bender and sell the plan. That way you have no worries about parts or warranties.

[EmptyNester]

You should do what got trikes did with his bender and sell the plan. That way you have no worries about parts or warranties.

My goal was to make all the plans and software avaialble free. The biggest thing to make it easy to build is to have a custom printed circuit board (PCB), then its just buying all the parts and soldering it. I'd prefer not to sell anything. I did find a site that will let you upload PCB designs and allow anyone to buy them. It might be more work though because they only accept specific file formats for PCB's.

But it would definintely all be "at your own risk".

[EmptyNester]

Here's the schematic for where things are today. I just sent out boards to be built and then realized I forgot the overvoltage protection for the torch voltage input.

This copy of the schematic has that update. (Use at your own risk. I'm not an electrical engineer, but I stayed at a Holiday Inn Express last night.)



The top block of the schematic takes the full voltage input from the CNC port and divides it down. It is scaled to cutting voltage instead of max voltage. This gives better resolution on the analog to digital conversion. There is a zener diode to protect the circuit from over voltage. The rest of the block is the filter using an LT1632 op amp. C4 is a spare that I haven't needed yet, but wanted space for on the board.

The middle block of the schematic shows two things. First is wiring of the Torch On line from the CNC controller and the Arc Okay from the PP50 to the digital inputs that support interrupts. The second part of that block is optional and is optical isolation for a serial port.

The bottom block of the schematic shows optical isolation for the CNC Signals (Torch On, Arc Ok, Torch Up, Torch Down) and the connectors for the buttons.

When I uploaded the file, the web site gave me a sample image of what the board would look like.



It shows board under the pin connectors on the top, bottom and right side, but that shouldn't be there. All the pins stick out past the edge of the board.

This board fits on top of an Arduino Mega with a 2 line LCD display shield next to it.

I haven't had time to play with the CNC table or the THC software. Just worked on this every now and then when I had time. I had hoped to work on all of it over the Christmas break, but my vacation just got cancelled due to work.

[Gerry]

Very nice design, I know at one time Radio Shack sold board kits that you could design and etch what you needed. But now days they don't have much electronic stuff for the diy people. Just cell phones and misc stuff.

[EmptyNester]

Now that I have the boards and am debugging them, I found a number of errors in the schematic.

A number of them are just signal name errors, there are some resistor value errors but so far only one layout error in the serial port opto-isolator. I've switched the Tx and Rx lines between the Arduino UART and the opto-isolator.

I'll post an updated schematic later when I've finished debugging and tuning everything.

[bigbopper]

Now that I have the boards and am debugging them, I found a number of errors in the schematic.

A number of them are just signal name errors, there are some resistor value errors but so far only one layout error in the serial port opto-isolator. I've switched the Tx and Rx lines between the Arduino UART and the opto-isolator.

I'll post an updated schematic later when I've finished debugging and tuning everything.

I am changing my cnc flame cutter (which I converted many years ago to cnc), to plasma , I have changed many mills and lathes to updated controlers , but over the hill now , my last project ,any way I need a thc control , can you help : george

[joshuab]

Are you familiar with this tool?

http://sim.okawa-denshi.jp/en/CRlowkeisan.htm

You can tweak R and C values for your filter to get the desired frequency response and, most importantly, settling time.

[EmptyNester]

I am changing my cnc flame cutter (which I converted many years ago to cnc), to plasma , I have changed many mills and lathes to updated controlers , but over the hill now , my last project ,any way I need a thc control , can you help : george

If you want to build a THC, I still have bare boards, PM me. You can find design and build info on github, the project is regeg/ArduinoTHC.

[EmptyNester]

Are you familiar with this tool?

http://sim.okawa-denshi.jp/en/CRlowkeisan.htm

You can tweak R and C values for your filter to get the desired frequency response and, most importantly, settling time.

Yes, that is one of the many I used.

[bigbopper]

If you want to build a THC, I still have bare boards, PM me. You can find design and build info on github, the project is regeg/ArduinoTHC.

i'm trying to pm you ,I want board : george

[bigbopper]

i'm trying to pm you ,I want board : george (bigbopper)

[EmptyNester]

George,

I did get all of your PM's. I've been working 12+ hours at my day job, so I'm not online much.

Based on the PM where you said you'd need help with the software - I'm concerned that this probably isn't the best path for you. You'd need to build the hardware and debug it. (I've built a number of different versions of the board and had to debug the hardware on almost every one due to construction errors on my part or bad components. If you'd never done that with electronics it can be frustrating and a challenge.) Additionally, you need to be comfortable doing Arduino software (minor changes, building, downloading).

The final thing I forgot to bring up earlier (but is in the documentation on GitHub) is that this was designed specifically for the Power Plasma 50.

If you want to be able to get support, a commercial system (like C and CNC) may be the best path for you.

Rege