Envision Peripherals NV3128 - Manual (Page 20)

NV3128 RS-422A Machine-Control Data Switch 1-10
CROSSPOINT ARCHITECTURE
Figure 1.5 is a simplified representation of the interface between the I/O
connectors, the dynamic ports, and the crosspoint matrix. The crosspoint
matrix itself consists of four LSI 64x64 crosspoint chips connected in a
unidirectional 128x128 “m by n” architecture. In this non-blocking scheme,
any of the 128 inputs can be switched to any of the 128 outputs. Each output
is controlled by a double-buffered register with a load and an active segment.
Upon receipt of an XY take command, the command interpreter fills the load
segment with the address of the input port to which the output will be
connected. The take is consummated when the crosspoint strobe dumps the
contents of the load registers into the active registers. The entire process of
mapping the switch in this fashion occurs within one video vertical retrace
time.
At the crosspoint level, the NV3128 is an X-Y matrix switch. But from the
vantage point of the user, the semantics of routing architecture break down: the
straightforward concepts of input and output, sufficient to describe program
route matrices, no longer have meaning. Despite the complexity, inherent in the
dynamic ports is an expanded capability. Unlike a program router, which has
a fixed I/O dimension, each of the 128 ports can be connected to any other
port. Each of the ports can be either source or destination.
For each requested machine connection, the router uses the input command
information to ensure that the controlling machine automatically sees a controlled
router port, and that the controlled machine looks back into a controlling port.
In practice, this means that the NV3128 can dynamically implement the
connections of Fig. 1 and Fig 3 in successive sessions. In Fig. 1.1, the editor
controls VTR A, which is a controlled device. In a further session, VTR A
dubs down to VTR B, with VTR A now a controlling device, as shown in Fig.
1.3. In a conventional routing switch, this arrangement would require physical
re-connecting of the machines involved. Although there are routing systems
with high-impedance outputs that can be forced into such a mapping, the cost
is a doubling of the number of connections, a 4-times expansion of the number
of crosspoints, and associated increments in operating costs.