SMT techniques (was Re: group buy for homebrew CPUs?)

On 8/13/06, Allison <ajp166 at bellatlantic.net> wrote: Yep... 'though "few" is relative. There are lots of simple and common M-series modules like the M111 or M113 or M216, but there are a handful of specialty modules like the M220 (2-bits of each of the major registers, quantity 6) or the M706/M707 TTY interface, which, while they _could_ be done in SMT parts, are much more complex than the simple logic modules, and would each be individual projects themselves. Since this topic has drifted a bit, let me clarify that my interest is in spare parts to repair existing machines, not trying to implement a PDP-8/L from scratch. I know there are scans of -8/L and 8/i docs, but I am not aware of any ASCII netlists or such. Somewhere, I have a partial netlist on paper from removing the wires from a damaged -8/L (the bar supporting the backplane was broken before I got it, and years of flexing broke several wires and backplane pins - it was a "gimme" that came along for free with the first PDP-8/L I bought in 1982), but I can't seem to find it, and in any case, it's only on paper. Yep... the same hex-height board could be populated (at the time of manufacture) with 4kbit DRAMs as 16K or 32K or with 16kbit DRAMs as 128K (but you needed a KT8A to be able to address more than 32K). 16K is ok for playing around, but by the end of the PDP-8/a lifecycle, there were enough machines in the world with a full 32K that a lot of interesting software came out that really wanted more than 16K (ADVENT, to name a famous example). I would _love_ to have some quad-height semi-conductor memory for the OMNIBUS. I have one or two quad-height and dual height protoboards which should be easy enough to populate with a couple of 62256s, say, and some bus buffering/driving. For a one-off, I don't even mind scavenging my existing boards for proper drivers, but I realize that not everyone has tubes of 8641s and 8881s and such lying around (i.e., it's good to have modern work-alikes as an option). To that end, I keep meaning to make a TTL chip tester to hand-sort the chips as DEC once did (7438s?) - of course, as was just pointed out, with modern configurations not being asked to support the wide range of configurations that DEC once supported (i.e. - 1 or 2 OMNIBUS backplanes (not 3 or 4)... one or two mass storage controllers, etc...), it might not be as necessary to strictly adhere to the published bus specs as it once was, especially since one would be putting one board in for 32K rather than four to eight sets of core boards. -ethan