Board bring up of an early prototype is one of the most important steps for a design team. The first boards must pass through a battery of tests to demonstrate that the hardware is rock-solid. Non-intrusive technologies can be used to accelerate this process.
Do I Smell Smoke?
When the first prototype comes in from the factory, engineers are tempted to immediately power it up to see what happens. But, it’s worthwhile doing some visual spot checks first. Loose or partly soldered parts, bent pins, small metal fragments and other faults can damage the board. Later, during volume production, these spot checks can be replaced with solder paste inspection (SPI), automatic optical inspection (AOI), and MDA (unpowered manufacturing defect analyzer) systems.
Who’s Got the Power?
Once a level of confidence has been achieved that the board won’t start to smoke, it’s important to check that its various voltage regulators are operating at their rated voltage values. This can be done during bring-up with a multi-meter or, during volume production, with an MDA or in-circuit tester (ICT). Causes of failures at this step can be bad parts, shorts, opens, or enable faults (when the enable line to a voltage regulator doesn’t trigger).
In addition to operating at the correct voltage values, system power also typically has to come up in the correct order. Usually an oscilloscope is used for this. If the ‘scope detects wrong sequencing, firmware for the power controller usually needs to be debugged.
What Time is it?
CPUs, chipsets, and other components need reference clocks in order to start up. A quick check with an oscilloscope will indicate if these are up and running. Everything else can be working perfectly, but if there’s an issue here, the board may never function.
A Defect by any other Name…
Now comes the tricky part. Power is on, and clocks are running, but the board might not yet work. Symptoms can be varied: it might not initialize, or it boots and doesn’t work, or it boots and fails intermittently. Is the problem hardware, firmware or software? It’s best to start at the bottom, and check first that the hardware is OK – specifically, that it had been assembled correctly.
Boundary scan is a good first candidate to use to detect structural defects like shorts and opens. It requires a minimum of the board to be “up” and provides net and pin-level diagnostics comparable to MDA and ICT. And it’s ideal for prototypes since it’s the lowest-cost structural tester that doesn’t require an expensive fixture.
But boundary scan, being a digital test technology, usually doesn’t provide total structural test coverage of a board. So next in line as a bring-up technology is processor-controlled test (PCT). As with boundary scan, it does not require the board to boot to run the test. And because it is an at-speed functional test technology, PCT will provide both structural and functional test coverage on a design.
At this point, if the boundary scan and PCT tests pass, there’s a high confidence that gross structural defects have been detected. There could still be some firmware, software or just plain board design functionality issues that need to be debugged. But the final step before firing up the firmware and software debuggers is to catch defects that may have been missed at the earlier steps, especially on higher-speed buses. IBIST/HSIO on Intel designs, or IJTAG-based solutions like PLX’s visionPAK and FPGA I/O test instruments on other designs, is used for this. Intel’s IBIST/HSIO technology, for example, will detect manufacturing and process variances like solder voids, micro-cracks, head-in-pillow defects, and other such issues.
So finally, you’ve used inspection, multi-meters, MDAs, ‘scopes, boundary scan, PCT and I/O instruments on your design, and it’s still not working. What next? Well, it’s probably a design issue, with a chip, or the board or system hardware, firmware or software. Debugging these is both an art and a science, and I’ll cover some new techniques in a future Blog.