How do you write a test plan

[TheUnnamedNewbie]

Hi guys,

As a hobbyist, with our own testequipment, we can generally just probe and measure to our hearts content, take a measurement, and have a think about what it means for our circuits or if it is what we were expecting. After that, we can look at what we should measure or tweak next, to narrow down the problem, or verify that the system works within spec.

Unfortunately, when you work in a Uni lab or a company, this is not always possible. In my case, I need to use quite expensive measurement equipment such as a VNA, high frequency synths, 6.5 digit multimeters, etc. Since my research group has only one or two of those, we have limited time to use them, and we really need to think through what we are doing before we connect are devices to the worth-more-than-a-house measurement equipment. We can't just go "oh I wonder what happens if I probe this point".

On top of that, high frequency measurements need some time to prepare - I can't just poke a probe at a LNA's output and look what is going on - doing so will likely disturb the measurement significantly. You would need some time and preparation to do this (by breaking the signal path and soldering a sacrificial coax cable or see if you can use a GSG probe station ...)

So, I wanted to set up a lab plan, and prepare what we should measure, what we expect to measure, and what we should look for when it doesn't do what we expect. I don't know how best to approach this and how far to go with this. I can start writing down the core measurements and steps (Terminate everything in 50ohms, put 7.5V on input of circuit, verify that output of voltage regulators and bias circuits are at the voltages you need), but how far do you take this?

Do you guys have any resources or suggestions to look at/consider? Any experience and tips to share?

EDIT: I would like to add that every time I try to research this topic, all I seem to find is legal documents describing the "test plan" in terms of a "How will we measure the specifications to prove we meet the contractual requirements of our customer" with abstracts and legal notices and definitions.... So not at all what I'm looking for.

[tszaboo]

Depends on the type of testing. There is component test, which will just make sure the right component is in the right place, and it is soldered in. This is usually done by a bed of nails type fixture.There are two ways of doing ICT, one powers up the circuit, and measures eg. power supply rails. The other type measures just the components, and this is much more expensive and thorough.
Then there is functional test, where you connect input and output signal to the intended connector, and you see if the device functions properly. Both should be done to a high quality product.

From what you wrote, you want to do a bit of both, which might be a problem. First, those thorough test plans take time to prepare. Doing functional test is easy. Just see if the device works. Now, do this for a bunch of devices, say 100. Identify which points to measure, which way. All the devices that passed, do the ICT, calculate minimum, maximum values. Then write a test plan. It is tedious, it takes time, ultimately it is an art, like any other engineering task.
There is usually an entire team of test engineers working on it at bigger companies. So asking for a simple guide, is like asking for a 10 minute crash course for electronics.

[TheUnnamedNewbie]

Depends on the type of testing. There is component test, which will just make sure the right component is in the right place, and it is soldered in. This is usually done by a bed of nails type fixture.There are two ways of doing ICT, one powers up the circuit, and measures eg. power supply rails. The other type measures just the components, and this is much more expensive and thorough.
Then there is functional test, where you connect input and output signal to the intended connector, and you see if the device functions properly. Both should be done to a high quality product.

From what you wrote, you want to do a bit of both, which might be a problem. First, those thorough test plans take time to prepare. Doing functional test is easy. Just see if the device works. Now, do this for a bunch of devices, say 100. Identify which points to measure, which way. All the devices that passed, do the ICT, calculate minimum, maximum values. Then write a test plan. It is tedious, it takes time, ultimately it is an art, like any other engineering task.
There is usually an entire team of test engineers working on it at bigger companies. So asking for a simple guide, is like asking for a 10 minute crash course for electronics.

Perhaps I should clarify what I mean by "test". I'm not after production test and verification. I'm fully aware this takes a fleet of specialist test and verification engineers.

I'm still working in a prototype stage. I've just populated the first run of our board by hand, and now we need to go and verify that our board works as we expected, and how close our real values match those of the simulations/calculations. I know from experience that if we don't think about this ahead of time (And check that the equipment can actually do what we want it to) we will forget things, fail to measure things (example of this: we wanted to measure the output noise level of one of our devices, but we didn't take into account that all the gear we had to measure it with had a far worse noise floor than the noise we were trying to measure...).

This is also a test plan for me and my team, to give us a thing to go by once we are in the lab, not to hand out to a client to show that we know what we are doing or to prove that every of the devices that we roll of the assembly belt works as advertised.

[Brumby]

I've not been involved in setting out formal test plans for this sort of thing, but I would start by taking the circuit you have designed and add the measurement equipment's equivalent circuit to the testing points in turn and run your simulation to see what effect it has - both in how the DUT performs and what is sent down the line to the test gear (so you don't fry it).

I might also suggest you look for some simple "sanity tests" - checks that can identify, possibly indirectly, that a measurement of an operating circuit is as expected, before connecting the house priced test gear.

[tszaboo]

I see. So it is testing of a prototype.
If you have the specifications, that is a good start. I would measure power supply voltages first though.
It is kind of hard to tell, becuase bringing a prototype to life also takes a lot of work. Most of the time any electronics device will have a programmable part on it, so instead of solder and done, you have days-weeks-years of work ahead before it is fully functional.

[Brumby]

The apprehension I would have with a prototype, especially in the very first incarnations, is that the parameters in the specification may be rather different in the constructed device - due to some unexpected coupling/parasitic/oscillation/(insert your demon).

I have a friend who was commissioned to work out why a new motherboard design was having problems.  Tested au naturel it worked fine, but when installed, it would go haywire.  After some time exploring various ideas, it was discovered that if the side panel were removed, things played along nicely, but when refitted, the problems returned - and that this simple process provided a reliable means to cause and alleviate the problems.  The customer was not convinced.  After much wracking of brain, he had a thought, spawned by his RF interests (he designed high gain antennae) and followed through on it.  It turned out that the layout of the circuitry projected an RF field and with the distances involved and the frequencies used, the signal reflected off the case panel back onto the PCB and affecting signal integrity.

Not all prototypes will have such bizarre issues - but they tend to have some.  As I see it, part of the concern here is that they not be unknowns that may risk high value test equipment.