[mmaimcal]Re: Comment in ALMA memo 489

[Al Wootten]

Mark Holdaway writes:
 > 
 > > > Mark's level
 > > > I suggest that you have an effective algorithm for dealing with the +ACA
 > > > data case, and that you are dealing with the homogeneous array (ALMA+SD)
 > > > case less effectively.  I am calling for a fair comparison.
 > > 
 > > Koh-Ichiro's level
 > > Do you mean that ALMA+SD mosaic should use OTF scan same
 > > as ACA?  However, the dump rate of the baseline correlator
 > > for cross-correlation is different (much slower) from that
 > > for auto-correlation.  So, I do not know how to realize
 > > an interferometric observation and an OTF single dish scan
 > > simultaneously with the ALMA 64-element array.
 > > Can you give me your detailed idear?
 > 
 > Mark's new level
 > Yes, I am suggesting that to be a fair comparison, we need to use the
 > same total power data.
 > 
 > I have been informed by Larry that all the 12m (the ALMA 64 and the ACA 4)
 > antennas have the the same electronics for total power.  There are square
 > law analog detectors that can be used for fast total power continuum, or
 > the autocorrelation spectra from the correlators, which can still work
 > pretty fast.  The 4 antennas in the ACA are special in that they are the 
 > only ones which will have nutators.  
 > 
 > Also, if we have to use the correlator (ie, for spectral line), it is not
 > all that slow, and if we are doing spectral line we don't need to be
 > zooming across the source so fast as atmospheric stuff won't be a big
 > problem for spectral line.
To my knowledge we can get:
total power autocorrelation data on 1ms timescales (some data rate issues)
total power from the back end square law detector on 2ms timescales
interferometric data on 16ms timescales (data rate issues)
Interferometric OTF data will be restricted by data volume.
Interferometric data through new tunable filters 512 ms timescales except that
there is a 'bypass' mode allowing 16ms w/no use of tunable filters (data rate 
issues)
Interferometric data with 90 d phase switching ~2s timescales (data rate issues)

The problem with getting back end square law detector data on 1ms timescales
is one of network speed.  The System Design Review committee found no
compelling science reason why the software ipt should change their plans
to incorporate faster communication to accommodate <2ms integration times
for the square law detector system.  That system will use 4 2 GHz samplers
and associated electronics.  There will be limitations on both of these
systems (square law detector and correlator) from system stability 
(including 1/f noise), bandpass limitations and other effects.  In theory,
one could process e.g. all 32 GHz of data from a B6 receiver with 8 GHz
bandpass and 4 IFs with the total power square law detector system, whereas
only 16 GHz can be passed through the correlator.  Given the limitations,
there is doubt any gain could actually be achieved, and at present no other
band has adopted this configuration (and even for B6, the present plan is
to only bring 2 IFs out, for cost reasons--comment invited on this also).  But
if there are good science reasons to try to get data on shorter timescales
for the total power square law detector system, almasci folks should type their
case in now, as the System Design Review report is in its last stage of
editing (it recommends no change to the 2ms square law detector readout time,
and it recommends using 4 IFs for B6, but is only adivisory to the JAO).

Clear skies,
Al