UniBone: Linux-to-DEC-UNIBUS-bridge, year #1

[Jörg Hoppe]

 >> Well, I was expecting to have to do all of the work myself. There’s 
still the problem of the disk Unibus itself to solve
 > -the disk UBA doesn’t terminate into a normal Unibus.
 > It goes into the disk RH11 directly, and the bus is terminated on the 
far end of the RH11. I’d either have to buy another Unibus backplane to 
plug the Unibone into, or find a way to plug the cables from the UBA 
directly into the Unibone.
 >This still leaves the issue of terminating the bus.
 > The ideal scenario would be if the first slot of a RH11 (where the 
bus jumper comes in) can accommodate the (quad card)
 > Unibone without issues, the rest of the RH11 boards can simply be 
pulled without breaking bus continuity,
 > and the normal terminator in the far slot can be used. I haven’t 
looked at any prints or anything yet.
I gave UniBone a set of pinheaders for all UNIBUS signals in parallel to 
the gold fingers.
So an adapter board can be designed, which plugs onto the pinheaders, 
contains some provision for the UBA connection and contains the 
terminator array.
The UBA-UniBone adapter may consist of two parts coupled via flat cable, 
with flipchip plugs on one end if necessary.
All this is only mildly annoying, did similar before, for example 
http://www.retrocmp.com/tools/uniprobe

 > Right, there were two unibus ports on a 2020: The first one went to the
 > RH11-C and was very odd in that "Hog Mode DMA" was enabled to allow the
 > device to just stream data as much as it wanted to the controller. This
 > would mean that other devices on the bus would time out and not have
 > their interrupts serviced, but since the RH11 was the only thing it
 > didn't matter (and I think this is why you could use RM03's instead of
 > RM02's: The whole track could be read and buffered to the 2020's UBA
 > controller in one shot.
 > That would have to be programmed into the BBB software to ignore the 
16 word
 > DMA limits and go as fast as the drive can go).

As the disk drives are also emulated, they are not putting any 
constraints on the DMA logic: give them the speed and DMA length you prefer.

Joerg