Current status of the IOBridge
- Very late compared to initial plan:
PC interface incorrectly specified
Pin count problem
Specification modifications
- Current status:
IOBus interface being redesigned:
Master complete, being tested
Slave not finished
DynaBus interface correct
PBus interface designed, not tested
Application schematic complete (update needed)
- Planning:
Not ready for a run before 1 month min.
New IOBus specs
- PC/AT bus now understood
- PC bus extended to 32-bits addressing (hack)
- Address mapping supported (2 maps!):
64K in [0..1M) mapped by 4K pages
14M in [1M..15M) mapped by 1M pages
- Back-to-back cycles for 32-bit access
- Full complement of byte-swapping methods:
byte, INT16, INT32 orders
- Estimated round-trip for PC access:
DynaBus -> IOBus : 7 DynaBus CK
IOBus: min. 4 PCLK+4 PCLK/16 bits
IOBus -> DynaBus: 3 DynaBus CK
--> + 2ms for a 32-bit transfer
IOBridge pinout
- Very close to 300-pin PGA limit:
PGA allows 268 signal pins
- IOBridge pins:
DynaBus: 139/143
DBus: 10 (8 if slave only)
PBus: 44/47
IOBus: 62
-----------------------
Total: 255/262
- Very tight fit
- Pad-frame implementation may reveal topological problems
DBus status
- DBus now asynchronous with DynaBus:
Allows clock tuning through DBus
Almost no change in chip design (hidden in standard DBus interface)
Allows non-DynaBus synchronous scan-paths
- Multiple DBus scan-path per chip (8 max):
Allows more flexibility
1 path for chip ID
1 path for clock tuning
1 path for DynaBus-clocked signals
path(s) for other clocks
- Negative aspects:
DFreeze now asynchronous -> synchronous system start problematic
Adds one pin to all chips (8 pins for slave DBus interface)
- New DBus document to be released soon