Foot pedal and thumb control for TIG welder?

It seems the Arc Captain TIG welder doesn't want to respond to the amperage control input on the front panel, whether connected to the torch they supplied with the machine or the foot pedal I sourced independently.

There's a few Facebook posts on the subject, such as this one that shows the pinouts on the so-called "aviation" connector that is beloved of the welding manufacturers.

You may sometimes need to hold down the pedal / torch button and wait for 2 beeps to "enroll" the remote input but this doesn't help.

Opening the machine allows me to check the wiring isn't damaged or missing. 

No, it's all present and correct.

Looks to me as if the firmware isn't recognising that input. Bollocks. I've put in a request for support via their AliExpress store front and am (so far) still waiting for a reply.

Back to the original thought:

Here's the current control (and arc enable) switch of the supplied torch when disassembled: 

The tiny PCBA contains nothing more than a tiny preset pot and a surface mount switch.

There's a 5 way edge connector that takes the 5 wires back to the machine via the "aviation" connector.

It all plugs and clicks into place, so on the face of it it looks as if the assembly could be simply replaced with a cunningly redesigned PSR-based circuit. 

Let's get into this latest rabbit hole, then:

Can Claude create a schematic for me? I'm imagining a double opamp for the force-to-voltage stage and a comparator stage (perhaps using a 3rd opamp within the same package, driving a MOSFET), to trigger the arc enable signal. Let's see what can be done.

Create an opamp-based circuit that does the following:

• runs off 5V single supply
• uses a pressure sensitive resistor (PSR) https://www.amazon.co.uk/sourcing-map-Pressure-Accuracy-Instruments/dp/B0FWCNF4G5?th=1 to convert finger pressure to an output voltage and hence the amperage setting of a TIG welder.
• the output voltage increases with pressure.
• the output is a fairly linear function of the pressure.
• when the pressure exceeds a lower limit, a digital output is generated that can be used to enable a TIG welder.
• the output sensitivity (volts per N of force) can be adjusted with a potentiometer.
• Create a schematic in EasyEDA, if possible, so that I can create a PCB for manufacture.

It looks plausible that there's a credible circuit going on here in Claude's response - but the interconnections are all made by declaring net names for each component, rather than by connecting them together using wires, as we would normally expect. It doesn't make a lot of sense to human eyes.

It's difficult to review and modify the circuit if we can't easily visualise the circuit. I suppose I could blindly go straight to layout and simply hope it all worked out when I finally got the PCBAs back but - call me a cynic - I don't think that would end well.

It imported ok. the main issue is that the compts are connected together by net names, not wires. can you not create those wires yourself?

That looks a bit better:

But of course, there need to be LCSC part numbers (and hence footprints) associated with each component before this can be taken into layout. I'd like to do that so I can see what the compts look like. Obvs for stuff like 0805 resistors and caps, that's no mystery but the relay may be another matter. 

I can't get Claude to drive EasyEDA directly but he can see the LCSC parts list (LCSC is JLCPCB's contract manufacturer), so it makes sense to get "him" to allocate parts from their inventory, rather than expect The Stupid Fat Bloke to do it without fucking it up.

Here's what I've ended up with after making a few changes:

• Change to TSSOP package for the quad opamp.
• Change from 2N7002 to AO3400A (a proper switching MOSFET).
• Change to a smaller relay as it was the biggest compt on the board.
• Change the pots to 5mm SMT devices.
• Change caps and resistors from 0805 to 0603.
• Remove the series Schottky from the supply input and replace it with a parallel one (will protect my circuit without losing supply voltage).
• The external connections will be fingers, not a pin header.
• The PSR will be soldered directly into the PCB.

I can now define the order of the finger contacts:

• 1 - relay NO
• 2 - relay NO
• 3 - Vcontrol
• 4 - GND
• 5 - +5V

The BOM looks half reasonable:

But I will need to check a few things, including some handy design hints from Claude when it comes to layout:

Swapping parts: tick Keep current symbol wherever EasyEDA won't swap a part in directly, as with U1. K1 is the exception: its symbol already uses the relay's real pin numbers (1, 3, 5, 7), so its footprint's pads need to be numbered the same way.

Diode polarity: D1, D2 and D3 all have the cathode on pin 1. Make sure pad 1 is the cathode end on each footprint.

Edge fingers and PSR holes: you'll need to make your own footprints for these. Set the finger pitch to match your mating edge socket, and the two PSR holes to suit the sensor's legs.

Layout:

• Keep the PSR holes, C5 and U1 pin 2 close together, with short tracks, because that part of the input picks up interference easily.
• Route the relay contact tracks (fingers 1 and 2) away from it.
• Put a ground pour on the back of the board.
• Place C2 right next to U1 pin 4.

It seems to have imported Claude's Kicad files OK:

And Claude claims that it simulates correctly:

I ran the circuit in ngspice and put the plots in a PDF. I've also made LTspice versions of the same tests for you to run. The simulation backs up the design, with one limitation at the highest sensitivity setting.

What the simulation shows:

1. Linearity is good. The output is a straight line against sensor conductance (roughly proportional to force), within 0.7% of full scale. The sensor's own non-linearity will be far bigger than that.
2. The sensitivity trimmer RV1 works as intended. Gain runs from 1.1 to 11.1. With R3 = 4k7, full output (4.95 V) is reached at:
   • about 290 Ω sensor resistance at the lowest setting
   • about 2.8 kΩ at the highest setting
3. The enable switching is clean. The relay switches on at the threshold and off about 48 mV lower, so it won't chatter. The trip point stays the same whatever RV1 is set to. The threshold range (0.20–0.88 V on the force signal) corresponds to a sensor resistance of roughly 5.9 kΩ down to 1.4 kΩ.
4. Response time suits a pedal. After a sudden press, the output settles within about 20 ms, with no overshoot. That's set by C4 and R3; a smaller C4 would make it faster but let more noise through.
5. The relay drive works. Q1 switches fully on and the coil draws about 10 mA. At switch-off, D3 limits the voltage spike to about 5.7 V.

Two small things it revealed:

• The output doesn't reach 0 V at high sensitivity. With no force applied, the output is about 20 mV at low gain but about 0.22 V at maximum gain. That's because the op-amp can't quite swing to 0 V, and stage 3 amplifies what's left. That's about 5% of full scale, which may mean a small minimum current at the welder. Keep RV1 lower if you can. For an exact zero, a small change to stage 3 would fix it; I can make it if you want.
• A small linearity tweak is available. Changing R4–R7 from 100k to 1M improves linearity from 0.7% to 0.07%. It's optional and not worth changing unless you want it.

Caveats:

• Op-amp model: these results use a generic rail-to-rail op-amp model, not Microchip's MCP6004 model. The README explains how to swap Microchip's model in.
• Other parts: the MOSFET, diodes and relay coil are approximate models too.
• LTspice files: they use the same circuit file as the ngspice runs, but I couldn't run LTspice here, so they haven't been tested in LTspice itself.

Next:

• Check / allocate footprints to compts.
• Create the PCB outline.
• Take the schematic and BOM into layout and see if I can get it all to fit in the space available.