Slightly OT ceramic caps - not in DECs

This was p[osted on the Tekscopes bulletin board , I thought it may be of interest here Geoff. Hello All, Thanks to all that provided info on the 2221A. The problem was the scope was totally dead. The failed part ended up being C945, a 0.1uF/50V monolithic cap that went to about 5 ohms resistance. This appears to be a decoupling cap on the power supply inverter transformer center-tap drive circuit. Very unusual that a ceramic cap went bad instead of a tantalum or electrolytic. Regards, Jim That is not quite as unusual for monolithic caps of that vintage. What you are seeing is silver dendrite growth through cracks in ceramic. When cap is abused, mechanically or thermally, the cap body (ceramic) cracks and in presence of humidity and electrical field, silver (out of ink which forms capacitor plates) leaches out along cracks and forms an irregular structure that somewhat looks like a net and shorts the cap. Decoupling caps usually clear up those shorts, but when not used often enough, or power supply folds down at a relatively low currents (what is frequently a case on power up) those shorts stay and just get beefed up over time. I have seen 50 A power supply being held in fold-down mode by those dendrites. In early 80s, when monolithic caps where new technology, they were accepted quite enthusiastically because of higher capacitance density and low ESR, both required for that new fungled thing called DRAM. Then came a rude surprise, discovery that few months after assembly, mysterious shorts would develop. Things looked so bad that there were even articles predicting that monolithic caps would never make a viable product. In days of through-the-hole components, radial and axial style monolithic caps were usually encapsulated into a rectangular (cylindrical for axial style) epoxy body, the same material as used for IC housing. The problem with epoxy body was, and still is, that water, or rather water vapor, penetrates along leads and supplies moisture for dendrite growth. To mitigate epoxy's package lack of hermeticity, a new type of package was introduced: glass encapsulated capacitor; that package was a variation on the one used for diodes. It was very expensive, but some companies would pay any price to get reliable decoupling cap; DEC was one of them. After around mid 80s, all DEC CPU and memory boards used only glass encapsulated caps. So do not be surprised that your cap failed, it happens all the time, less frequently now then it used to be, but it still happens. The usual scenario is that damaged caps form dendrites and when those caps are used as decoupling (across power supply lines) shorts get cleared. After few clearing iterations, there is no capacitance left, but no one is wiser, it is decoupling; you just end up with a noisy board. However, if cap is used on a signal line, usually there is not enough energy on such lines to clear shorts, signals end up being shorted to some other signals or ground and that is something that is not frequently suspected, so it is difficult to find. Regards Miroslav Pokorni