Quote:
Originally Posted by
JD-EE 
RR_NDB....I bet the RF was basically contained on one card
I remember nifty bits of thin coax that could be wire-wrapped, and that kept the "RF" (bus and clock frequencies) more or less where they belonged.
Also common for higher frequency/critical nets was the use of twisted pair with one side wrapped to nearby ground pin. This worked fairly well up to 30Mhz or so.
I did manage to get a 66Mhz clock line to behave by slipping a tiny ferrite core over the driver pin before connecting the wire but that was definatly at the limits.
Another reason wire wrap worked fairly well was that the boards had solid ground and power planes that provided good decoupling and return paths. A common "newby" mistake was to carefully lay all the wires in neat XY rows, this resulted in really bad crosstalk, the best bet was to use direct routing and avoid parrallel runs.
As I recall the original Cray super computers used WW twisted pairs for a lot of interconnect, all of carefully specified lengths so signals would line up as desired.
Thread No 4 continues Here.