Quality assurance for assembled boards

[AndreyB]

Ten assembled boards arrived - I guess packaging was not great, couple of buttons were squeezed into pieces and one of the huge capacitors was knocked off it's position.

Which brings me to the question - what kind of QA can I run on these boards?

Right now the plan is to drive my red miata around the block with each one of these before shipping out. What else can I implement reasonably easily?

Should I try two car batteries in series and run 24 volts for 5 minutes? Does Frankenso support 24 volts for 5 minutes? Would it test anything?

Any cheap/DIY vibrations test beds? I know the boards would not survive washer nor drier.

[stefanst]

How are you going to make the connection? That would be my biggest problem. If we know what car they're going to go into, you could solder the right connection, but that seems too much effort in anything but an NA Miata.

How about reviving the stimulator idea and using a board with pogo-pins to connect?

[kb1gtt]

Hmmm, traditionally people make a test fixture, then smack it within controlled limits.

Perhaps electrical connections could be made with spring pins. Basically make a board with a bunch of spring pins which you can bolt / screw to the board which is being tested. The go drive it around the block.

[AndreyB]

There is probably no problem with a connector - boards are coming with all W* jumper pins populated. Miata boards would have couple of pins redirected while the rest of the boards (well, at least most) would have 64 pin connector - so far there was no interest in boards without the connector.

I am mostly interested in putting some stress (mechanical or electrical) on the boards - not just functional test under ideal parameters.

[stefanst]

[...]
I am mostly interested in putting some stress (mechanical or electrical) on the boards - not just functional test under ideal parameters.

My guess would be that >90% of all problems can be found with a simple functionality test- potentiometers for the analog inputs and LEDs for the ouputs, maybe, if you want to go fancy, a VR or two hooked up to a spinning wheel. And that would already go beyond what a real-life Miata tests, since there we're essentially dealing with a digital signal on a Miata.

If you want to go more extreme, we could use 8 cheap-o used injectors and hook them up to the lo-side drivers and some beefy resistors hooked up to the hi-los.

Throw some LEDs in the mix to make sure if things work, run the board overnight, rattling the injectors and see the next morning if it still works. Operate directly under the smoke detector for increased safety.

Maybe immobilize the pintles on the injectors, so we don't make so much noise, which tends to aggravate better halves after a few nights.

A variable power supply (let's say 7-16V) would be nice too. If you really want to mess with the board, you could throw a small resistor (Let's say 0.1 Ohms or so) in the GND supply to make sure we can handle fluctuations in the supply voltage.

I really don't think we need a driving test for this task.

[kb1gtt]

Seems there are many Orbital shakers on ebay for less than $50.

I would think some injectors in a bucket of water would be safe if run over night. Remember that water is not going to conduct much current with 12V.

Looks like these supply's are reasonably low cost. They include a volt meter and current meter. You could get like 5 of them. Set each one for a low voltage, like 8V, 12V, 14V, 18V, then measure the current from each. If you see an odd current you would get an indication that something has gone wrong.
http://www.ebay.com/itm/5A-LED-DC-Step-down-3V-3-3V-5V-12V-19V-24V-Adjustable-Voltage-Regulator-Module-/172519887917?

[Abricos]

Best test 24 v fixing all problems ...)))

[kb1gtt]

Frankenso uses a BZX79C20 which is only rated to 20V. I would expect a sustained 24V would damage Frankenso R0.4. Also I expect some level of conduction below 20V, so might be good to keep the 18V test reasonably short.

[stefanst]

[...]
Looks like these supply's are reasonably low cost. They include a volt meter and current meter. You could get like 5 of them. Set each one for a low voltage, like 8V, 12V, 14V, 18V, then measure the current from each. If you see an odd current you would get an indication that something has gone wrong.
http://www.ebay.com/itm/5A-LED-DC-Step-down-3V-3-3V-5V-12V-19V-24V-Adjustable-Voltage-Regulator-Module-/172519887917?

I like those little guys. I have one or two in a box here. So you're thinking just pre-setting a bunch of them and do an automated switch from one to the next with relays or something?

[kb1gtt]

Hmmm, we are drawing less than 0.125A and we are at room temperature not elevated temperatures. We could use the hi / lo driver to change the supplies.

Why can't I find one of these that does serial programming instead of POT programming? Seems the price gets steep for anything with a serial port, or some kind of communications.

Also those supplies require a 24VDC supply, as they are a buck style supply.

[stefanst]

Replace pot with serial or spi controlled pot?

I have a 15v adjustable (pot) power supply that goes from ~10V to ~18V is I remember right. That would probably be a good start...

[kb1gtt]

serial POT would remove the need for multiple devices. Any way to get the current readings into a PC or some kind of jig?

[Rhinoman]

For a production test you only need to prove that the board is correctly assembled. I would replace the firmware with something that just reads the inputs and exercises the outputs. Connect up some simulated input signals and connect some output loads, maybe even just connect the outputs to the inputs.

[stefanst]

Any more input on this? Are we OK with a test of assembly that does not involve taking the board for a ride in a real car?