Manipulating welder output with Arduino

[joshuab]

I see what you mean about autorouters being kind of bogus. With a little practice, I think I did a much better job, although there's surely still room for improvement. I wanted to focus on keeping as much of the board as possible available to the copper pour, which is why some of the traces go out of their way when going directly would be possible.

Top trace.



Bottom trace.



Top, with copper pour.



Bottom, with copper pour.



Also, I realized that, as long as I'm using the Boarduino on headers (vs. incorporating it into the PCB), the least I could do is use some of the space underneath it, so I stuck the resistors for the LEDs under there. I think that some of the remaining components, like the optoisolators, will also be low-profile enough to stick under the Boarduino if I need to.

[joshuab]

A question: when I designed my board, I put bypass capacitors on all the Vcc pins of my ICs--three of them, to be exact: an analog mux and two solid-state relays. When I got to PCB layout, I realized that it made sense to put those chips right next to each other, and looking at the ratsnest, I realized how dumb it seemed to have three capacitors, all within 150 mils of each other, all going from +5v to GND. I took them out and settled on just one capacitor.

I know the rule of thumb is, "one bypass cap per Vcc pin," but does this become redundant if the Vcc pins are close enough together?

[Rambozo]

It all depends on the signals involved. Typically you want the bypass caps as close to the pins as possible, say 250 to 500 mils or thereabouts. Things get crazy as the frequency goes up.

[joshuab]

... but if I could keep them within 250-500 mils, then it's fine?

[Ron66]

The effectiveness of the caps are reduced by trace inductance. That is why it is generally preferred to use one bypass per pin. I assume you are using the bottom pour for ground. You may even want to consider using 1206 surface mount capacitors going from the vcc pins to the ground pour for best noise performance. I would also recommend adding thermal reliefs to your ground and vcc pins otherwise soldering becomes difficult on those pins since you have to heat up the entire plane. If there are any left over pins and you have room add extra connectors with signals, power and ground. Maybe one for some a/d inputs, an extra usart line, and some general purpose I/O. Copper is free!

[Ron66]

Also I know it's a little late but they make these devices called digital isolators. Maxim has a six channel MAX14850 and TI has single channel parts like ISO721. They offer galvanic isolation of around 600V but can function as high as 150MHZ. TI even has parts that offer isolated power for the isolated side that comes from the main power. Check them out!

[joshuab]

Also I know it's a little late but they make these devices called digital isolators. Maxim has a six channel MAX14850 and TI has single channel parts like ISO721. They offer galvanic isolation of around 600V but can function as high as 150MHZ. TI even has parts that offer isolated power for the isolated side that comes from the main power. Check them out!

Yup. I'm using opto-isolators on all of the interface circuits now. The challenge is that I'm feeding the output-control circuit with a PWM signal, and the opto-isolators have to handle high enough frequency to eat that. You can get high-speed opto-isolators, but they require 5v and up supply on the load side, and when the welder is running on 110v power, there is only 2.7v supplied on the remote circuit. I haven't been able to find any high-speed opto-isolators that will also run on 2.7 volts supply.

But it all may be moot, because I have been playing around with some modifications to the circuit tonight, and I think I have found something that will work. A guy on another forum gave me some tips for resistor values and locations to change to increase the responsiveness of the opto-isolators. At the moment, it seems to be working really well. Fingers crossed!

[joshuab]

... and thanks for the tip on the thermal relief. I hadn't noticed that the program hadn't added those automatically.

[joshuab]

Okay! The changes to the circuit to increase the responsiveness seem to be working. What used to be happening was that the welder would skip straight from about 80 amps to 96 amps (max on 110v power) or straight from about 25 amps to 9 amps (min on 110v power). With these changes to the circuit, I get sub-1-amp output resolution across the entire output range, and on top of that, I seem to be able to use the max PWM frequency (32 kHz) without any negative effect at all--whereas before, excessively high PWM frequencies made the "skip-over" range larger. I still haven't welded with it, of course, so that's the ultimate test. Nevertheless, this is very encouraging. Bear in mind that the circuit I'm using is NOT driving the wiper pin directly, but is switching resistors in and out of circuit to create a potentiometer-like effect on the signal voltage provided by the welder. Also, there is total galvanic isolation between the welder and the controller box. So this is very encouraging... We are at the point in development now where the basic development is, in some respects, almost done, and every last percent absolutely matters, but the easy stuff has already been done, so every last percent is also more difficult to achieve.



Here's an updated board layout, that includes the interface circuit. After seeing how expensive PCBs are to have printed, I made every effort to reduce the size of the board as much as I could. Unfortunately, the Boarduino and the terminal blocks really limit the minimum size of the board. If I was willing to solder on the wires instead of using terminal blocks, I could probably peel off another three square inches, but I don't think it's worth it. And the Boarduino itself is about as compact as it can possibly be, thanks to the professional PCB designers at Adafruit, so it's not like I'm going to improve on that unless I go to a different board (like the Pro Mini, which is something like 1x2". But even if I did that, it'd save me one square inch,and the board house I'm looking at charges $5 / sq.in. for three copies of a board. So really, that one square inch only costs me $1.67. Screw it.

[Rambozo]

You could switch to pin headers and ribbon cables to hook up the I/O. Not only will they save space, they don't put the stress on the board joints, that spring terminals or direct soldered terminals, do. You can get plugs with screw terminals if you really think you will be making lots of changes.
Something like this.

[joshuab]

You know, the reason I didn't go with ribbon cables in the first place was cost, but three square inches of board saved equals $5, so maybe it evens out. I like the screw-terminal connector you linked to. Do you have a part number for that? If I could use something like that, it would basically cut the width of the terminal blocks in half, saving maybe 1.5 square inches of board space.

My main reason for wanting to use terminals is that I want to be able to disconnect the board from the wires without having to resolder it. Especially at this point in development, I am needing to do that pretty regularly. More than that, the solder joint is fragile enough that the stranded wires end up breaking off. At least with terminals, if that happens, I can just snip the wire and stick it back in.

[Rambozo]

Digikey and Mouser have tons of options for them, that one was from AllElectronics, because they had a nice picture. I see variations on those all the time for I/O interfaces for all kinds of things.

Thanks to old floppy, IDE, SCSI, serial, and parallel case cables, I don't think I've bought any ribbon cable or ends in years. Combine that with some pin headers and you can disconnect the board easily and the only solder joints that might fatigue will be at your pots and switches.

[joshuab]

Let's take the discussion a different direction. When I have the Box working 100% with lift-arc, then I will consider it to be basically rev 1.0, but my ultimate goal, of course, is to get it working with HF start too, and that means noise protection. One thing that I've done is buy a die-cast aluminum project box with no vent-holes or anything, to limit the amount of EMI that gets into the box. But that won't do me any good if the EMI comes into the box on wires, and of course it will. So I took the cover off my welder and looked inside.



The photo above shows the back side of the remote jack. A pair of wires coming from the jack is routed to this board, which contains an inductor coil and two what appear to be capacitors, but they could easily be some other component in a similar form factor. I couldn't really get at the board to get a good look at them. I assume that these two wires are the torch switch, because the only other thing they could be is the remote-detect circuit, but since that circuit has no wires external to the welder, I bet it is less in need of noise protection than the torch switch, which runs alongside the welding lead.



A bundle of three wires, which I assume to be the potentiometer connection, goes to a second coil, pictured above. This bundle doesn't appear to get its own special capacitors (or whatever)--just the coil.

So I guess I am going to eventually want to put an inductance coil on the wires where they enter the box. It would be mighty convenient if I could simply wrap all the wires around the same toroid, but there's probably a reason why the Everlast engineers didn't do it that way. Any advice from the smart folks here is welcome, as usual.

[Rambozo]

The noise may turn out to be a non-issue for you. The welder will have HF filtering for the control signals that need it. Your big problem will be not crashing your micro from EMI. That will be coming from the arc itself, not from the control circuit. However your control line will be acting as an antenna to catch that and bring it inside your shielding. It might take nothing more than a ferrite bead to control that, or you might need to add a different common mode choke depending on what you see. Shielded cables can help, when nothing else does, but just making sure that your enclosure is well shielded might be all you need to worry about. Large plastic switches might be something to avoid in your final design. Metal pots keep your shield integrity pretty good.

Unless you want the kind of isolation that will let you strike an HF arc to the outside of your enclosure. That might take a lil more work.

[joshuab]

The noise may turn out to be a non-issue for you. The welder will have HF filtering for the control signals that need it. Your big problem will be not crashing your micro from EMI.

Well, yes, exactly! I tested this once, months ago, and about every third time I struck an arc with HF, the Arduino crashed and had to be reset. Of course, it wasn't inside an enclosure at the time.

However your control line will be acting as an antenna to catch that and bring it inside your shielding.

Bingo. And, of course, the torch's welding lead, and the two wires going to the torch switch that run parallel to the welding lead for 12 feet or so.

I wonder if the shielding on the welder's end of the Box/Welder line will also protect the Box. After all, an inductor resists AC current flow, and current flow along a wire is equal, right? So then the location of the inductor on the line shouldn't matter, right?

Unless you want the kind of isolation that will let you strike an HF arc to the outside of your enclosure. That might take a lil more work.

Naahhhh... just not having the dang thing crash while I'm using it will be fine. But I think you're right. I think I will table this issue for now and see if just the plain metal box does the trick.

[everlastsupport]

Any wires going to you circuit should be treated/considered as an antenna. I would scope them Also note the number of turns of wire on that toroid. De-coupling caps will not solve the RF issues. Also, maybe scope the AC input on the circuits PS. Did you try a Faraday cage? That was brought up many posts ago I think.

[joshuab]

Any wires going to you circuit should be treated/considered as an antenna. I would scope them Also note the number of turns of wire on that toroid.

I can't tell if you're saying that the exact number of turns is important, or if it's just important to have a lot of turns.

De-coupling caps will not solve the RF issues.

Sure. For one thing, the signal is likely induced equally in both the signal and the return line (unless they're twisted-pair), so there's no potential difference for a cap to shunt.

Did you try a Faraday cage? That was brought up many posts ago I think.

I've got an aluminum project box, but I haven't got around to installing in it yet.

[Ron66]

PCB Express usually runs a $50 special.

[everlastsupport]

Exact number of turns. You can wind and unwine with a scope and maybe dial it in better too.

[joshuab]

Going to be a bit of a lull in the project reporting. For one thing, I'm going out of town on business for a week. Then I have a big family vacation coming up around the 16th. So it'll probably be close to the 21st before I'm back thinking about The Box. But the timing is good. I ordered my PCBs, and they've got about a 2-week turnaround, so this means I won't be looking at the calendar every day while I'm waiting for them to come back. Like I said earlier, I've got a nice die-cast aluminum box that I'm going to put the board in, and I'm going to properly mount it on standoffs this time so it won't be flopping around and such. Basically going to really try to get the thing fit-and-finished like a real Product instead of somebody's Hobby Project.

I still have some fine-tuning to do on the welder interface circuit, but it's good enough for the time being. I pulled the appropriate output pin from the Arduino out to one of the unused pins on the header so that if I want to, I can build a different interface circuit on a small board and substitute it for the one that's printed on the PCB. I thought I was real clever for thinking of that. I decided to go with the spring-loaded terminals for the time being, since I already had ordered them, and had th footprint all drawn up and stuff, but in the future, I think I will go with pluggable terminals so I can connect and disconnect without having to individually sort out each wire. The other change that is still outstanding is that I still haven't sourced a good 7-conductor wire for the plug to the welder. I'm still using an Ethernet patch cable that I cut the ends off of. Problem is, those conductors are so small-gauge that they break off pretty easy, and are hard to strip without cutting some of the conductors. Hopefully the spring terminals will help a bit with that, but I'd really like to find something closer to 22-gauge or even a little thicker. If I ever decide to make more than one of this thing, I'll pony up for a 100' roll of something, but in the mean time, that's excessive.

Here's a render of what the boards are expected to look like, top and bottom. I ended up getting a little stuck with the routing, and so I ran the auto-router. I realized what I could do is run the auto-router so that it "interleaved" the traces and figured out where the vias should go, then I could manually slide the traces up and down as needed, without changing their "interleaving", to clean up the pattern. For example, moving traces to they ran neatly next to each other, instead of sort of randomly spaced out. I think it was a pretty good compromise between pure hand-routing and pure automatic. I put some effort into getting a ground-plane copper pour over as much of the board as possible.