14
\$\begingroup\$

I've ordered fabrication and assembly of small PCB batches (~100) a couple of times already. Each time, the assembly factory asked me if I wanted to do testing on the boards. Since I'm not a professional, I didn't know what to do about it, so I tested the boards myself when I got them back... finding unacceptable percentages (>10~20%) of soldering problems. Although my boards were fairy complex (+100 components), I hadn't designed any testing procedures on them or even know how to do it. So I'm wondering, what are the usual test capabilities for a low volume production like this, and what kinds of test can be asked to the factory without going overboard? Is that usually costly? I'm talking about testing the assembled board, not the bare PCB. Thanks in advance.

\$\endgroup\$
  • 7
    \$\begingroup\$ No amount of testing will "fix" an assembly process that is so fundamentally flawed that it produces 10-20% of bad assemblies. Even if a board passes a visual or functional test at one point in time, it will probably have long-term reliability issues. You really need to find a better assembly house. \$\endgroup\$ – Dave Tweed Apr 6 '13 at 1:59
  • 1
    \$\begingroup\$ Simplest answer : ask them to quote for "shorts and opens" testing and see what they say. These will catch virtually all soldering defects which are your big problem. They may ask for test pads on the board or information you can supply. Ask them what you'd need to do to an existing board to make it testable. \$\endgroup\$ – Brian Drummond Apr 6 '13 at 9:52
  • 1
    \$\begingroup\$ Thanks, guys. It's true, I need to find a new assembly house. However they were different each time. I thought it was kind of normal to have such errors! The errors were becasuse of insufficient heat (unmelted solder) in parts of the boards. They were tin-lead reflow soldered, so I don't know how could they end up that way. I blamed myself for not knowing how to place the components properly for an even distribution of the heat (it's a dense board), but I guess it's their fault too. The board has some 0.4 mm pitch components I couldn't get rid of. \$\endgroup\$ – Guillermo Prandi Apr 6 '13 at 12:45
  • \$\begingroup\$ This is NOT normal. In fact getting them to test and hand-fix what they find may mask their crappy assembly and you WILL have many field failures. Maybe your design is partly to blame, but I would really find another assembly house. They should have been able to see the bad soldering by visual inspection! \$\endgroup\$ – Spehro Pefhany Nov 15 '16 at 16:10
5
\$\begingroup\$

I am working for such a company but am mainly situated in the development, so I have some insight but there are probably people out there who know much more:

When it comes to testing, we follow various approaches. Basically we start with AOI (Automated Optical Inspection). This is rather cheap and reveals many errors before any further steps are done which require the board to be powered.

The next step is checking electrical connections. We do this in a few different ways, it usually depends on the amount of test points available and if the board was designed for testing (yes, usually no one cares for that in advance). The methods we use most are:

  • Flying Probe (basically an automated way of contacting pins and checking if the resistance meets expected values). This also comes relatively cheap since the programs can be created from netlists which have to be specified by the customer.
  • Boundary Scan: The Flying Probe can mainly contact test points or larger points such as resistors, ... if there is no available point to touch it on the PCB, the Flying Probe is of no use. To test Inter-IC connections we mainly rely on Boundary Scan Tests if the Controller supports these. But they have their limits as well. Also, programs can be "written" automatically but need to be adapted.
  • In-Circuit-Tests: This is probably the most extensive test method we use (and also the most expensive). Basically we build an adapter which will host the DUT and contacts the various testpoints. Using integrated Boundary Scan techniques and stimulating digital and analog signals nearly every mode of testing is possible. E.g. it is also possible to boot the board to some bootloader and execute tests available in the bootloader, test Ethernet connections, test USB connections, ... No need to say that this comes at a cost.

I am sure there are even more possibilites of testing available but these cover the requirements we get from our customers quite well. Still, 100% testing is not possible.

\$\endgroup\$
  • \$\begingroup\$ Thank you, Tom, for your answer. My next question here is, what happens if tests fail? For example (as in my case) a 136-pin 0.4 mm connector not completely soldered? Does the fab take care of the repairs? Let's say other tests fail... would they simply discard the boards? In my case, the board have pretty expensive components... I'd rather get the boards anyway and see what I can do about the errors. Any comments about this will be appreciated. \$\endgroup\$ – Guillermo Prandi Apr 7 '13 at 21:29
  • \$\begingroup\$ Again, that depends on our customer: If tests fail and we have some knowledge about the device (schematics e.g.) we will try to repair it (especially if it's just things like a not soldered component or a bad solder or maybe some shifted component). If we do not have this knowledge we usually deliver it to the customer but we will charge some amount even if the tests fail (if the customer also doesn't give us a chance to repair). During production, a log is generated which details the detected errors so in case the customer will produce the same or a similar part, those errors can be avoided \$\endgroup\$ – Tom L. Apr 8 '13 at 21:35
  • \$\begingroup\$ in advance. This is especially useful to detect problems during soldering or placement of components. We usually do not discard any boards with components since many times the customer will try to repair them. Of course, if there is a really serious problem we will try to correct it in-house. 0.4mm sounds like a rather small pitch, I do not know if this error could have been detected by AOI or FP, especially since it's a connector (I guess AOI would have seen at least some). \$\endgroup\$ – Tom L. Apr 8 '13 at 21:36
  • \$\begingroup\$ But again, usually we talk to our customer also before production actually starts since soldering and placement departments usually try to eliminate most errors before the first production and therefore take a closer look at the provided materials. --- DAMN comment limit. \$\endgroup\$ – Tom L. Apr 8 '13 at 21:39
  • \$\begingroup\$ To make it short: We usually talk to our customers before production and find some agreement :-). \$\endgroup\$ – Tom L. Apr 9 '13 at 5:48
1
\$\begingroup\$

There are a few things to consider ...

Cost Of Testing

  • Initial cost of test fixtures.
  • Possible PCB changes to make this easier.
  • Cost of test (Usually the cost of the person doing the test is the largest factor)

All these are affected by what the tests are. In this case you have production faults after assembly then functional tests would be needed.

The choice is Manual testing, Fully automated or a mixture of both.

Manual testing is more expensive in time but cheap in initial costs normally (Meters and a bag of wires)

Fully automated is expensive on initial costs but usually cheap in production (Test fixtures connecting to automated test equipment + writing the test programs).

In exstrem cases Manual test time could be 8 hours, Automated test time could be 10 minutes

For the production of 100's per year a simple manual fixture can save a huge amount of time. The sort of thing I thinking is where you have 30 wires to connect by hand. Replace that with a plug in connector and you will get quicker test time and the screwdriver hand doesn't get worn out. This can make testing unpleasant for the person testing. (100 x 30 wires x 2 seconds per wire = 100 minutes)

If your testing digital then connecting to a buffer chip with LED's in the test equipment makes visual comparisons easy.

For Analogue signals (Any Voltage , current or waveform) I check at 2 input values unless there is a reason for 3 or more values.

How far you go in to detail testing depends on what's critical. An output needs to be +/- 0.001V then test it with a meter. If it is +/- 0.1V and your high volume you might use comparators.

Often I test broadly and coarsely and refine tests when they are critical and when failures are not caught. Of course if the circuit is critical in some manor you spend more time ensuring it functions as required.

Also remember some connectors wear out quickly. Look for 10,000 cycle connectors for test equipment and not the 500 or 1000 cycle connectors used on the product if they are compatible. If you can only use low cycle connectors then add then to the cost of test.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.