Several months ago I picked up a Tektronix TDS 640, 4-channel 500mhz oscilloscope at a hamfest for $50.
It was failing several of the power-on self tests with some mysterious errors like "FAIL ++ Acquisition" and "FAIL ++ Attn/Acq interface". After my successes repairing Cricuts and resurrecting my first $20 analog scope with a bunch of Deoxit spray I figured I'd get my money's worth of entertainment out of it.
Googling the subject quickly revealed this as the inevitable consequence of "Capacitor Plague." It seems the original SMD electrolytic capacitors all end up leaking corrosive gases and goo onto the circuit boards, which, if not cleaned up in time, eventually results in assorted broken vias and board traces. The visible result of all this is these failed self-tests.
The Tektronix Community Oscilloscope Technical Support Forum, fortunately, has accumulated quite a few threads over time with information and various troubleshooting & repair tips for this issue.
So after a couple of months of building up my courage, buying a ESD-safe temp-controlled iron, antistatic work mat, and desoldering "hot tweezers", I was ready!
I started out by practicing desoldering and resoldering a bunch of similar SMD caps on a handful of old busted PC graphics cards from the pile on the hack shelf.
Removing all the old caps turned out to be pretty easy. As it turns out, and it's hard to believe this until you try it out, the best way to remove the old parts is actually to press down and gently wiggle the part right off. The copper leads work-harden and snap off without stressing the underlying solder pad one bit. After I learned about this method I practiced it a dozen times to prove to myself that it was actually the LEAST damaging way to get them off there. (The corroded old solder doesn't take heat well).
Fast forward to last week, I'd ordered in and received all the new capacitors and was ready to dig in.
After removing all 100 or so old caps on 3 circuit boards, the next thing was to use a combination of solder sucker, iron, and desoldering braid to remove all the old solder and remnants of broken leads, leaving the solder pads as clean as possible.
Once this is done, you super clean the entire board with a toothbrush and full strength Simple Green, rinse in the sink under super hot water. Bake in the oven at 125 degrees for an hour or two, followed by an Isopropanol rinse, and another overnight drying period in a warm place.
The resoldering process follows by putting some flux on all the solder pads, and melting a little pillow of solder on each one. Since the caps are polarized, you need to take care to orient the new one properly, but on this scope the process is super easy because nearly every single one is oriented in the same direction, and all but one pad is marked with a +.
Soldering the cap goes like this: use the iron to melt the solder pillow on one pad, and, just using your fingers, hold the cap in place. Touch the iron to the lead and stay long enough to make sure it bonds properly (too quick and you'll get a cold-solder joint). The cap will be attached on one side now, and slightly cocked due to the unmelted pillow under the other lead. To finish soldering, just press down lightly with one finger while heating up the lead. When the solder melts, you'll feel the part move downwards until it's fully seated on the board.
Now, just repeat this process 100 times 🙂 🙂 It sounds like a really big job, but after the first few dozen it goes very quickly. The whole process excluding drying time only took a few hours.
At this point, I was eager to fire up the scope and see if those darn errors went away. Annnnnnd... no change! Arrgh!
Going back to the Tek Forum, I learned that the reason for this was almost certainly due to some combination of corroded vias, broken traces, or fried components.
The way to start finding these is to use the voltmeter and, with the machine running, carefully check the voltage across all the capacitors and make sure they are all about 5 or 15 volts. Similarly, check the voltages across the power & ground leads to all the dozen or so opamp chips on the board to make sure they are also similarly valued.
In my case, I found one opamp chip with a bad ground, and one capacitor with only 1 volt across it instead of 5. Both problems ended up being due to a broken via, and were fixable by carefully determining exactly where the broken lead went to and soldering on a tiny little 30-gauge jumper wire to bridge the gap. (tracing the connections can be a little tricky with a multilayer board like this)...
After a couple rounds of probing and probing and installing the two jumper wires, I fired the scope back up... and YES, the self test passes completely now, and the formerly failing Signal Path Compensation routine also passes.