Comments by the Space Geodynamics Laboratory (SGL)
on the
Draft Proposal for a VLBI Standard Interface (VSI) Specification
Cannon, W.H., Feil, G., Feir, B., Newby, P., Novikov, S.
March 5 1999
1. Electrical interface:
--------------------
The proposed electrical interface refers to individual lines
of balanced ecl, requiring as many as 64 separate lines
each at 128 Mb/s to achieve 8192 Mbit/sec. An approach which
promises to be more efficient and easier to implement is to
serialize the data into smaller numbers of lines for the purposes
of transmission from the Data Acquisition System (DAS) to the
Data Input Box (DIB) and from the Data Output Box (DOB) to the
Correlator. For example, Low-Voltage Differential
Signaling (LVDS) devices are now available for data
transmission at rates of 1.84 Gbit/sec (460 Mbit/sec on four
lines), with devices under development for rates of
5.37 Gbit/sec (671 Mbit/sec on eight lines). Another possibility
is HP's HDMP-1022/HDMP-1024 1.25 GB/s transmitter/receiver.
In view of the need for long-term continuity of the
VSI specification, selection of one of these options or
a similar alternate should be based on careful consideration
of both technical suitability and market strength.
If a serialized non-ecl method of transmission is used,
compatibility with hardware having balanced ecl interfaces
would require external conversion hardware.
2. Data Validity:
-------------
The proposal calls for data validity to be transmitted
from the Data Output Box (DOB) to the Correlator, indicating
"whether the data are believed to be valid or invalid
on a bit-bit-bit basis."
Currently it appears that various Correlators deal
with data validity differently. Some carry the
validity bit through the signal path at full resolution,
while others ignore it completely.
In view of the additional hardware complexity involved
in transporting data validity from the Data Output
Box (DOB) through the Correlator, alternative methods
of communicating data validity (for example, at
at low bandwidth via a network connection) might be
considered.
If a full-bandwidth hardware implementation is indicated,
it would be desirable to have data and data validity
carried on separate cables in order to facilitate
direct connection of the Data Acquisition System (DAS) and
Correlator. This would also simplify connections to
Correlators which have no data validity input.
3. Data Output Box (DOB) Synchronization:
-------------------------------------
Possible methods for synchronizing DOBs are:
1. Hardware high-speed clock (e.g. 32 MHz) and hardware 1-Hz tick.
2. Hardware 1-Hz tick only. Software syncs up internal high-speed clock
using VCO. Offsets to the 1-Hz tick can be specified over the
communications interface to avoid having to generate different
moving 1-Hz ticks in hardware.
3. No hardware signals at all, Correlator sends synchronization command
over communications interface once per second, software syncs up
internal high-speed clock using VCO.
Methods 2) and 3) would also presumably require a specification
for the size of the memory buffer on the Correlator in order to
set a limit on the maximum steady-state timing mismatch between
the Correlator and DOB.
Our preference would be to implement timing control using method
3), but others may require hardware support for methods 1) or 2).