[mmaimcal] Correlator reconfiguration time
Harvey Liszt
hliszt at nrao.edu
Wed Mar 17 11:51:10 EST 1999
>
> I am less worried about a spec for the correlator dump time. As I understand
> it this time is apt to be ~1ms for autocorrelations and ~10ms for cross
> correlations (I actually thought that it was to be ~1ms for both, however
> Chuck Broadwell told me these numbers yesterday).
>
> However the reconfiguration times (mode changes) will depend on
> architectural decisions which are not yet known. If the new correlator is
> implemented like previous correlators, these changes might take several or
> many seconds. This is the spec that I'm trying to get a handle on.
>
> Brian
>
Yes, I think I answered your mail somewhat clumsily; fast dumping for
1/f noise isn't the same as global reconfiguration, we agree. I think
we need to give the correlator designers good targets and then have some
give and take. Clearly a 1 s setup for the correlator would be great and
a 1 minute setup would seem to be on the long side. But battleships don't
turn on a dime.
Brian, you might want to talk to Dick Thompson. Earlier, when Mike Rupen
tried to spec a fast integration time (< 1 s) ART nixed it, saying you
couldn't do the Walsh function correction that fast. Or something.
We've been down this road before.
regards, Harvey
More information about the mmaimcal
mailing list