[evla-sw-discuss] Mark6 at the VLA

Ken Sowinski ksowinsk at nrao.edu
Fri Jun 14 12:55:49 EDT 2019


Now that a Mark6 recorder has been installed at the VLA it seems
useful to set down the interactions of the correlator and recorder
limitations.  The data rates provided below are expected, not achieved.
Testing over the next few weeks will verify or correct these expectations.


Using 2 bit samples we can supply VDIF data at 4 Gb/sec for 128, 64, or
32 MHz subbands.  With 4 bit samples the maximum data rate that can be
supplied is 3.3 Gb/sec.  The limiting factor in the latter case is the
number of packets which can be sent before the next set of packets is
ready to be transmitted.  The supplied rate for 4 bit samples could be
doubled if a dual NIC were installed in the Mark6 box, and CM taught to
calculate VDIF packet transmission delays taking that into account.

Assumptions/Constraints
1.  1Gb ethernet is 10^9 bits/sec; 10Gb is 10^10 bits/sec
2.  Maximum Mark6 record rate is 4096*1024*1024 bits/sec > 4x10^9 bits/sec
3.  Maximum packet VDIF delay from Widar is 255 usec
4.  VDIF packet payload is 5000 bytes
5.  VDIF packet delay is a multiple of 6 usec to avoid collisions
6.  Both polarizations are recorded from each subband

The tables below describe the maximum VDIF data rate the correlator
can supply for 2 bit and 4 bit samples.  The left half of the table
gives the capability of the correlator; the right half shows what
bandwidth can be recorded when the recording rate is limited to 4Gb/sec.
The first five columns show the subband BW, repetition rate of VDIF
packets in microseconds, maximum number of stream that can be supplied,
total bandwidth that can be supplied, and aggregate VDIF data rate.
The rightmost three columns repeat columns 3, 4 and 5 but with the
recorder data rate limit applied.

For 2 bit samples:
SBBW RepRate  Streams  BW      Gb       Recorder 4 Gb limit
128   78.125   2x6    2x768  6.144      2x4  2x512  4.096
  64  156.25    2x13   2x832  6.656      2x8  2x512  4.096
  32  312.5     2x21   2x672  5.376      2x16 2x512  4.096
  16  625.0     2x21   2x336  2.688
   8 1250.0     2x21   2x168  1.344
   4 2500.0     2x21   2x84   0.672


And, for 4 bit samples:
128 -----      ---    -----  -----
  64   78.125   2x6    2x384  3.072
  32  156.25    2x13   2x416  3.328
  16  312.5     2x21   2x336  2.688
   8  625.0     2x21   2x168  1.344
   4 1250.0     2x21   2x84   0.672



More information about the evla-sw-discuss mailing list