[mmaimcal] ALMA Correlator

Brian Glendenning bglenden at cv3.cv.nrao.edu
Thu May 24 16:00:22 EDT 2001


> The datarates
> expected of ALMA, 6 MB per second normal, will accommodate the 512
channels
> of the baseline correlator while it is not at all clear that this will
> be the case for the future correlator.

People get confused by the data rates - I think you should note that the
6MB/s is not fundamental to the correlator but is imposed by assumed
computing resources and will thus increase with time as commodity computing
gets cheaper (and even so we say we will support 60MB/s peak sustained). I
believe the correlator is fundamentally capable of dumping at ~3GB/s which
will suffice for some time! So, for the large number of channel case the
data rate is essentially a spec on the minimum dump time, e.g. for 60MB/s
32k channels 2 polarizations all baselines the minimum dump time is <15s if
I've done my sums properly.

Put another way, data rate should not be an issue since any plausible output
data rate can be supported with either correlator by spending the same
amount of money on post-correlator computers and archives. As far as I can
tell, the issues should only relate to the scientific utility of improved
spectral resolution at the wider bandwidths and the number of bits of
correlation (with no change in spectral resolution).

Anyway, I'm sure everyone knows all this - sorry for cluttering your
mailbox.

Cheers,
Brian





More information about the mmaimcal mailing list