We know what you’re thinking. It’s a bad power supply, of course it was capacitors to blame. But even if we all intuitively know at this point that bad caps are almost always the culprit when a PSU gives up the ghost, it’s not always easy to figure out which one is to blame. Which is why this deep dive into a failed ETK450AWT by [eigma] is worth a look.
The first sign of trouble was when the computer would unexpectedly reboot with nothing in the system logs to indicate a problem. Eventually, [eigma] noticed a restart before the operating system even loaded, which confirmed the hardware was to blame. A quick look at the PSU output with a voltmeter showed things weren’t too far out of spec, but putting an oscilloscope on the 12 V line uncovered a nasty waveform that demanded further investigation.
By carefully following traces and comparing with common PSU diagrams, [eigma] was able to identify the SG5616 IC that checks the various voltages being produced by the PSU and generates the PWR_OK signal which tells the motherboard that everything is working normally. As before, all of the DC voltages at this chip seemed reasonable enough, but the pin that was measuring AC voltage from the transformer was showing the same ripple visible on the 12 VDC line.
Even more digging uncovered that the transformer itself had a control IC nestled away. The 13 VDC required by this chip to operate is pulled off the standby transformer by way of a Zener diode and a couple capacitors, but as [eigma] soon found, the circuit was producing another nasty ripple. Throwing a few new capacitors into the mix smoothed things out and got the PSU to kick on, but that’s not quite the end of the story.
Pulling the capacitors from the board and checking their values with the meter, [eigma] found they too appeared to be within reasonable enough limits. They even looked in good shape physically. But with the help of a signal generator, he was able to determine their equivalent series resistance (ESR) was way too high. Case closed.
Fans of retro computers from the 8-bit and 16-bit eras will be well aware of the green death that eats these machines from the inside out. A common cause is leaking electrolytic capacitors, with RTC batteries being an even more vicious scourge when it comes to corrosion that destroys motherboards. Of course, time rolls on, and new generations of machines are now prone to this risk. [MattKC] has explored the issue on Microsoft’s original Xbox, built from 2001 to 2009.
The original Xbox does include a real-time clock, however, it doesn’t rely on a battery. Due to the RTC hardware being included in the bigger NVIDA MCPX X3 sound chip, the current draw on standby was too high to use a standard coin cell as a backup battery. Instead, a fancy high-value capacitor was used, allowing the clock to be maintained for a few hours away from AC power. The problem is that these capacitors were made during the Capacitor Plague in the early 2000s. Over time they leak and deposit corrosive material on the motherboard, which can easily kill the Xbox.
The solution? Removing the capacitor and cleaning off any goop that may have already been left on the board. The fastidious can replace the part, though the Xbox will work just fine without the capacitor in place; you’ll just have to reset the clock every time you unplug the console. [MattKC] also points out that this is a good time to inspect other caps on the board for harmful leakage.
Anyone who pokes around old electronics knows that age is not kind to capacitors. If you’ve got a gadget with a few decades on the clock, there’s an excellent chance that some of its capacitors are either on the verge of failure or have already given up the ghost. Preemptively swapping them out is common in retrocomputing circles, but what do you do if your precious computer has already fallen victim to a troublesome electrolytic?
That’s the situation that [Ronan Gaillard] recently found himself in when he booted up his Mac SE/30 and was greeted with a zebra-like pattern on the screen. The collected wisdom of the Internet told him that some bad caps were almost certainly to blame, though a visual inspection failed to turn up anything too suspicious. Knowing the clock was ticking either way, he replaced all the capacitors on the Mac’s board and gave the whole thing a good cleaning.
Unfortunately, nothing changed. This caught [Ronan] a bit by surprise, and he took another trip down the rabbit hole to try and find more information. Armed with schematics for the machine, he started manually checking the continuity of all the traces between the ROM and CPU. But again, he came up empty handed. He continued the process for the RAM and Glue Chip, and eventually discovered that trace A24 wasn’t connected. Following the course it took across the board, he realized it ran right under the C11 axial capacitor he’d replaced earlier.
Suddenly, it all made sense. The capacitor must have leaked, corroded the trace underneath in a nearly imperceptible way, and cut off a vital link between the computer’s components. To confirm his suspicions, [Ronan] used a bodge wire to connect both ends of A24, which brought the 30+ year old computer roaring back to life. Well, not so much a roar since it turns out the floppy drive was also shot…but that’s a fix for another day.
There’s a reason we often use the phrase “It ain’t Rocket Science”. Because real rocket science IS difficult. It is dangerous and complicated, and a lot of things can and do go wrong, often with disastrous consequences. It is imperative that the lessons learned from past failures must be documented and disseminated to prevent future mishaps. This is much easier said than done. There’s a large number of agencies and laboratories working on multiple projects over long periods of time. Which is why NASA has set up NASA Lessons Learned — a central, online database of issues documented by contributors from within NASA as well as other organizations.
Unfortunately, all of this body of past knowledge is sometimes still not enough to prevent problems. Case in point is a recently discovered issue on the ISS — a completely avoidable power supply mistake. Science payloads attach to the ISS via holders called the ExPRESS logistics carriers. These provide mechanical anchoring, electrical power and data links. Inside the carriers, the power supply meant to supply 28V to the payloads was found to have a few capacitors mounted the other way around. This has forced the payloads to use the 120V supply instead, requiring them to have an additional 120V to 28V converter retrofit. This means modifying the existing hardware and factoring in additional weight, volume, heat, cost and other issues when adding the extra converter. If you’d like to dig into the details, check out this article about NASA’s power supply fail.