SRAM boards (that is what these are, isn't it?) are usually simple enough to troubleshoot that you can get by with just a 'scope and an hour or so of time, depending on how inclined you are to work on the hardware. Each RAM device will have a write strobe, a select strobe, (maybe a separate enable as well) and some data inputs and outputs. Because it's a 16K board, I assume you have 4kx1 or 1kx4 SRAMs. (similar to 2147 or 2148 types, respectively) The 1Kx4 parts have common I/O, and both types are 18-pin DIPs. These 16K boards use 2147's don't they? Does your diagnostic (whatever it is) tell you which bit is failing? If it can't read the bits, perhaps it's a (1) a failed decoder or (2) it's a failed data bus buffer, or (3) it's a roached address buffer. The only apparent alternative would be a failed RAM. It does smack of a failed on-board decoder, however, since only the one 4K block is failing. If you can't read/write even a single bit in that 4K block of RAM, it's undoubedly a buffer or a decoder. If only one bit fails, it's probably not a decoder, but then the RAM itself is suspect. Since the failure is in the "middle" of your board, it's probably not difficult to compose a loop which cycles through the ram board in 1K blocks, doing a write to one location in each 1K block, first the lsb, then left-shift and do it again, etc. This enables you to trace the device select, and write strobes through the circuit. A 16K board in a 64K memory map ought to be pretty simple to decode. There should be jumpers/switches to select which 16K block in the memory map the board occupies, and there should be a 2-line to 4-line decoder to select the individual bytes of memory. If the organization is as 16 pairs of the 2148-types, then there's probably either a 4 line-16 line decoder (e.g. 74154) or a pair of 2-to-8 types (e.g. 74138) to select a byte of RAM for each memory cycle. Assuming you have the equipment at your disposal, you can determine whether each RAM device is receiving the write and select strobes at the correct times, and likewise, whether they get their enable and select when being read. IF those are both correct, you then write a loop which writes all zeroes to the rams and then having finished writing the zeroes, cycles infinitely on reading them back. You then look to see where the zeroes no longera appear as zeroes, bit by bit, until you have it narrowed down to one data bit and replace the device which isn't transmitting the zero. If they all transmit zeroes OK, then you write all ones out to the ram board and then look for where the zeroes coming from the ram are no longer zeroes. It's not a bad idea to monitor the write operation as well to see that the data makes it TO the RAM in the correct sense, inasmuch as the S-100 has separate in and out busses. This whole process, aside from the associated repair shouldn't take too long and will fix the problem once you isolate and replace the failing component. If, however, you're determined simply to move the failing block of memory to the top where it's easy to ingore, you can switch decoder outputs by tracing back from the failing block of RAM to the decoder and then cutting the traces and soldering on the jumpers which will accomplish what you want. Dick -----Original Message----- From: Mike Noel <Mike-Noel(a)GCI.net> To: Discussion re-collecting of classic computers <classiccmp(a)u.washington.edu> Date: Friday, December 03, 1999 10:21 AM Subject: Processor Technology S100 memory boards