On Thu, 1 Jul 2004 00:45:56 +2500 (BST)(!) Tony Duell wrote: Bad solder joints in power supplies bring back memories... cooked ones... My Superboard 2 (to be exhibited at VCF East) came without a power supply, so this intrepid owner got to build his own. Any bad solder joints are the responsibility of exactly one party... I kludged a STD-Bus 32K CMOS RAM card (using 16 of 6116 2kx8 parts) onto the Superboard, and used a free set of 4 size-F NiCad cells as a battery backup. Size F is used to make the 6-volt lantern batteries, by the way, so the current capacity of my backup was calculated to keep the board running without power for roughly 7 years, disregarding internal battery leakage. 7 Amp-hour NiCads can also start a fire in milliseconds: with that in mind, I placed a service switch in series with the battery so I could take it out of circuit if I had to fix something. Clever, right? One day (about 1985) the power supply became intermittent and I had to dive in to the Superboard's case to identify the problem and correct it. Awake and alert at the time, I pulled the AC plug, then switched off the battery. Disassembling the machine, finding the solder joint and reassembling took me longer than expected, so it was around midnight when I was ready to power everything back up. Here's where circuit details caught me: my main charge circuit for the NiCads was limited to about 5.1 volts - perfect for a 5-volt RAM card. However, NiCads should have a "trickle" charge of 0.1-1% of capacity to keep them healthy: for a 7Ah battery, that's 7mA (drawn from a 12-volt supply via a resistor). This is important, since the RAM card, idle, draws 0.1mA. Without the battery to absorb the trickle current, the RAM voltage goes (way) out of spec. (Fortunately, the Superboard had a completely independent regulator...) I plug in power, then realize with horror that the battery's out of circuit and the trickle charge is feeding 12 volts to the RAM. Yank cord. Check everything out to see what damage was done - couldn't find anything seriously hurt. Whew. About 2AM, I plug the cord in again - Moron! The battery's STILL off! This time I found 2 RAM chips shorted (and snipped the trickle charge resistor out of circuit). It took me another week to find the last victim, a 74HC245 bus buffer with its ground bond wire blown internally - no physical evidence - and CMOS protection diodes on its I/O pins kept the chip working as long as it had enough logic-0 inputs to work the logic-0 outputs. When the Superboard READ 0FFh from the RAM board, the '245 starved and I got errors... In hindsight, there were so many ways to prevent that... Bob Maxwell