Design Flaw in Gate Valve and Interface Boards

[TerryJC]

You may recall that I said that I'd blown up a couple of Pis earlier last week. I've been a bit distracted recently, but I finally worked out what I think went wrong. See the circuit below:

Gate_Valve_Circuit(50%).png (114.03 KiB) Viewed 1151 times

What happened was that the copper became detached from the Veroboard under the outer Pin H (this is the one nearest the edge of the Veroboard and was connected to GND on the A/D). The result of this is that 3.3 V from the Pi was still going to the Actuator circuit and a voltage of (at the time) around 1.5 V was coming back from the pot in the motor. The A/D is not specified to work when the analogue input voltage is higher than Vdd and in this case the circuit for Vdd was incomplete and the A/D failed. This in turn put strain on the Pi, probably through the I2C path and it also failed.

As indicated above, this is slightly theoretical, because I know what physically failed in my circuit, but I can only surmise what happened inside the devices. However, during the diagnostic process, I discovered another flaw in the design of the circuit; the 'Chinese copy' DuPont contacts are quite unreliable. Both female and male contacts are poor, but for different reasons (maybe it's the same reason (**** metal)), but the end result is different:

1. Male contacts - I've been using male contacts to screw into the terminations on the Motor Drive Board, but the metal is quite brittle and the slightest movement once they are clamped down causes the tip to break off. The bought-in contacts were not like that, but I have to crimp a female contact onto the other end.....
2. Female contacts. These are OK the first time they are used, but once the cable end has been removed (for access for example), they are no longer reliably retained on the pin (they just fall off). The bought-in contacts were not like that, but I have to crimp something to the other end at the moment.

This is beyond annoying and I'm amazed that I hadn't realised it before. The problem is that if my theory regarding the failure of the Pis is correct, then vibration from the Gate Valve motor, or something could also cause any of the current wiring to 'drop-off' with the loss of another Pi and A/D if it happens to be the 3.3 V or 0 V connection to the A/D.

There are obviously alternative approaches:

1. I could only use ends that have been made up by the manufacturer of the bought-in wires. This would make the wires super long, like the first Interface was.
2. As 1 above, but solder the other end to the Header Pin on the Veroboard. This can't work where both devices have header pins as part of their construction, although I could solder to the pins in the DC/DC Converter. I wouldn't want to solder to the pins on the A/D, Motor Drive Board inputs or the Pi.
3. Something else.

Whatever we do, I don't want to carry on building Gate Valves and Interfaces, until we have a more robust solution.

All ideas welcome.

[hamishmb]

I can't say I fully understand, and consequently I have nothing very helpful to say, but I am at least glad you figured it out.

[Penri]

Terry

This is very frustrating, even more so for you doing the hard work.

Do you think you could bring a "complete" gate valve to WMT one day this coming week so we can take a look at the issue together?

I think I can visualise the situation but without the hardware in front of me it's hard to appreciate fully, leave alone suggest any alternative courses of action.

[TerryJC]

Yes. As I said earlier in my email, I can do most of next week apart from the usual Monday / Thursday morning shopping (although that can be worked round) and Tuesday morning.

[hamishmb]

Is our existing, installed gate valve vulnerable to this? If so, do we want to unplug it in case we fry the pi connected to it?

[Penri]

Interesting question this, so far the Gate Valve has "mechanically" appeared to worked reliably. When things have gone wrong, resetting appears to have restored it to working condition. For me I'd say it's ok to leave in place but happy to bow to Terry's view as the designer.

[TerryJC]

Theoretically it is vulnerable, because the vibration might shake one of the header connectors out (or a partially detached Veroboard track may become fully detached). I have no reason to believe that this would happen over the next few hours (especially the Veroboard issue) and I was intending to talk to you about this later today when I come in.