Graham Sutherland / PolynomialNov 5, 2022

fun MLCC fact: the crystalline structure of the dielectric ceramic (usually barium titanate) deforms over time, which modifies the material's dielectric constant, causing a capacitance derating effect that follows a roughly log10 curve vs. time in hours.

this effect is essentially innate and continues regardless of most operating parameters, except heat. if you heat the capacitor to the curie point (~150C will do it) the curve resets to t=0.

you can do this as many times as you like.

Show threadTimon 🛠Nov 5, 2022
@gsuberland huh! Now I wonder if that might be one the things that made the odd "throw your GPU in the oven" tips work
Show threadGraham Sutherland / PolynomialNov 5, 2022
@timonsku yeah, I've pondered that before, too.
Show threadGraham Sutherland / Polynomial
@timonsku I'm definitely not sold on the whole "cracked solder joint BGA reflowing" repair thing - I'm certain there's more to it than just the reflow, otherwise the failure wouldn't reoccur after such a short time. my guess is that a combination of thermal effects coincide to make post-reflow operation more reliable for a short time. wouldn't be surprised if there's some oxide layer reformation in aluminium/alupoly/tantalum caps for reduced leakage current.
Nov 5, 2022 at 2:45amWeb
Show threadTimon 🛠Nov 5, 2022
@gsuberland Yea the reflowing part I don't really believe, most these guides recommended temps far below reflow temps and thermal cycling I can't imagine fixing a cold joint for more than a few hours.
Show threadGraham Sutherland / PolynomialNov 5, 2022
@timonsku if it weren't so obscenely expensive and time consuming I'd be very tempted to buy a bunch of YLOD consoles with the same problem, reflow them, wait for them to fail again, and then carefully cut the parts off the board and have someone with a fancy microscope abrade them down and image the internals for microfissures and oxide layer formation.