Chapter 6: Turning on

As we’ve seen in the previous chapter, there are lots of those bright orange electrolytic capacitors all over the place. Some are on the 5.1 V supply rail, others on 12 V⁠; a few near the CPU are on 3.3–3.45 V⁠. Should I get them reformed as well? Not the 3.3-volt ones, perhaps — they should be fine; but what about the others?

That would mean bringing up the 5 and 12-volt rails slowly, gradually increasing the voltages. However, power sequencing in complex computer electronics is a delicate matter. The Alpha processor, itself a 3.3-volt CMOS device with 5V-tolerant I/O, is surrounded by 5-volt logic, which was commonplace at the time. This arrangement comes with warnings like this one, for example:

Obviously, I want to stay away from messing with power supply rails. I’d rather take my chances with the caps. The I/O module is simpler, though, so I briefly send 5 volts down both its 5 and 12-volt rails before putting the module back into the system.

The power supply is connected; everything is back together, except for the memory. For better visibility, I thought I’d leave it aside for the first power-up.

Something blew up. Rolling back the footage, I found it. It’s one of those orange electrolytics. Let’s take a closer look.

The video recording points at C31. There is hardly any visible trace of the violent ejection. Is the dot in the middle of its square terminal our only clue? Good job the camera was on! I would never have found the culprit otherwise.

A continuity test reveals that this is a 12 V supply decoupling capacitor. Given that it’s got absolutely no markings, will I be able to identify it? An image search on the internet eventually brings up tantalum caps that were made by Sprague (now Vishay), which have this characteristic shape and colour.

As I keep in touch with Chris — the lovely gentleman who sold me this Alpha machine — with general updates on this project, I let him know what happened. But I forgot to mention that this was a surface mount part rather than one of those leaded tantalum beads, which were ubiquitous at the time.

Chris advises against attempting to replace it, as there is the danger of damaging the multilayer PCB, whereas losing one or two decoupling caps should make no difference. ‘The caps often connected to mid layer ground or supply planes within pcb layers and it’s difficult to get enough heat into the joint to unsolder safely,’ he said. This is a very sound advice when working with leaded beads, which is unfortunately not applicable to C31, as I couldn’t just cut it off. On the other hand, unsoldering a surface mount part is safer in that respect, as it’s got no leads that penetrate the PCB.

Having removed the popped cap, I decide to unsolder another one like it to take measurements (making sure it’s on the same power rail) in case I’d want to get a replacement at some point. The measurements show 33 µF with ESR of around 0.2 Ω, and I solder the second cap back in.

Unable to find this exact part, I settle on a slightly larger one that would still fit onto the same footprint. It’s a 47 µF, 0.11 Ω Vishay part no. 594D476X0016C2T. I’ll order it later when I know what else I need.