Glendenning
Lucas
Mangum
Morita
Schilke
Schwarz
Scott
Shepherd
Testi
Wright
Excused: Richer, Wilson
Minutes
Subsystem Scientist Activities (all)
Jeff Mangum -- Control
Continue with ATF activities, which serve as a good
proving ground for the eventual ALMA control system.
John Richer -- Telescope Calibration
Peter Schilke -- Data Reduction User Interface
Steve Scott -- Correlator
Prepared a Use Case for operation of the correlator in conjunction
with Jim Pisano. Very illuminating to have to write down the details.
Thanks to Debra for all her help.
Worked on datarates, which can affect the correlator data
processing.
Attended the correlator hardware CDR as an internal committee
member. Did my best to represent both the scientific and computing
interests of the SSR and computing IPT. The main news is that the
proposed tunable fir filter enhancement would provide a great increase
in scientific capability, particularly at wider bandwidths. But it
also introduces new challenges in data processing (sub-band
stitching), configuration specification, and data rate management
(data rate increases plus smoothing and decimation options).
Designed a test plan with Jim Pisano (f2f), but it has not yet
been formally written down. Due to other schedule constraints this
won't be completed until the end of the month.
Debra Shepherd -- Off-Line Analysis
Helped to write Use Cases in collaboration with Pat Palmer:
Worked with Scott Rankin to develop Use Cases and a Test Plan for I&T.
Worked with Mel Wright and Allen Farris to draft a test plan and
schedule for the scheduling subsystem. Robert sent in a test plan for
the TelCal subsystem. I drafted a Gantt chart for the TelCal test
schedule.
Wrote a detailed description for the January 2004 Offline test.
Rough summary: testers and datasets will be:
Chris Wilson (SSR, McMaster Univ.): - PdBI carbon star
Munetake Momose (Ibaraki Univ.): - VLA NH3 core
Crystal Brogan (SMA): - VLA NH3 star forming region
Debra Shepherd (SSR, NRAO): - PdBI star forming region
Began working with Francois to develop Use Cases and a test plan for
the Archive subsystem. Still in work.
Attended two AIPS++ meetings/week to track development plans along
with bi-monthly NRAO software testing group meetings. Did some minor
AIPS++ testing on another large NH3 dataset.
Jerome Pety and I submitted the AIPS++/IRAM Phase II report as ALMA
memo 473. Phase I and III reports are memos 472 and 474.
Leonardo Testi -- Observing Program Preparation
In September I hosted the ObsPrep face-to-face meeting in Florence,
most of the time was dedicated to finalizing the R1 delivery, but we
also discussed a number of issues of relevance for the SSR.
We decided to postpone by one month the first user tests (which is
now scheduled for December 2003), the reasons are: for R1 the group
still delivers the OT prototype for integration and we really want to
start the tests on the real OT; installation and startup (and
shutdown) of the whole thing ACS+OT is still cumbersome, we hope to
have an automated startup/shutdown and, possibly, a full Java ACS by
the end of November; some more work is needed to complete some of the
components on which the team would like to have feedback from the user
tests.
Francois Viallefond -- Archive
Chris Wilson -- Pipeline The only activity I've been able to undertake in
the last month was two contact meetings via telecon.
Mel Wright -- Scheduling
I sent Debra a Scheduling test plan, which she put into the ALMA
format.
Two recent memos might also be of interest:
BIMA memo 97 - CARMA Spectral Line Correlator Requirements
We review the expected sensitivity and spatial dynamic range of the
CARMA array during the next 10 years in comparison with the ALMA, ACA,
SMA and IRAM arrays.
BBIMA memo 98 - J-Miriad: Java Wrappers for MIRIAD Methods
Gerry Harp and Melvyn Wright
Debra adds: The draft test plans are due Oct 31st. Four more
are needed, for Archive, Control, Correlatr and Executive. There is
now an Integration and Testing test plan which can provide some input
for subsystems which rely on global testing. The Scheduling Subsystem
Test Plan can be used as an example.
Can we have several correlator setups in one SB?
(Leonardo Testi, Peter Schilke)
(item carried from the last meeting). Discussing with Peter Schilke, it came out that with the baseline
correlator, a line survey project is highly inefficient, unless
one manages to use several correlator setting (for a single tuning)
in the same scheduling block. This, I believe, is not currently
possible. Maybe we should discuss this point.
After discussion it is clear that this is no need to restrict
reconfiguration of the correlator during a Scheduling Block. This is
needed anyway for calibration (observe the phase calibrators with
maximum bandwidth all the time). Retuning the correlator does not
imply recalibrationg the passband (as could be the case for a
retuning of the front ends).
Correlator hardware CDR (Steve Scott)
Steve attended the first hardware CDR (see his report
above). Everything seems on time. The main new item is the proposal
to enhance the baseline correlator (see ALMA Memo 476).
The endorsement of a change to include the tunable fir filter
enhancement means new parameters for specification of sub-band
overlap. The large number of channels (8K) per IF means that we need
to specify ways for channel smoothing and resampling, and start and
end channels to conserve data rate. This is probably an issue with
the baseline correlator plan as well, so it should be dealt
with. The proposed enhancement involves stitching together the 32
sub-bands, so we should consider whether we also might want to
stitch IF bands together (e.g. produce an 8 GHz wide band, minus a
small fraction to remove edge effects). This is proposed as an
agenda item for the f2f meeting in Socorro.
Data Rates (Steve Scott)
Steve's email sent after the last meeting covered the issues in
detail. The proposal can be summarized as:
Recommend setting the data rates using basic assumptions of 8K
channels and integration times of 10 sec for the average and 1sec
for peak.
The 8K channels follows from earlier recommendations by the
ASAC and the likelihood of an enhancement to the correlator.
Using the 1sec time for peak (recording only corrected or
uncorrected data) follows from a self-calibration experiment. The
sample time to avoid fringe smearing on the longest baseline is
2.5sec (recording both, to limit sensitivity loss to 1% at the
edge or the primary beam on the longest baselines), so we end up
at about the same peak rate by several routes.
Fulfilling the above assumptions would require a 4.5 times
increase in the average rate and a 1.62 times increase in the peak
rate. Translating to resource units using previously discussed
factors for recording both atmospheric corrected and uncorrected
data, and some fraction of double sideband receiver use, gives an
average and peak rate of 25.2 and 93.6 MBPS. This is almost
certainly a non-trivial increase from the standpoint of the
archive.
1% sensitivity loss at the edge of the beam at the longest
baselines may be too restrictive as those projects using the most
extended configuration are likely to focus on compact
sources. However the peak rats is more set here by the
self-calibration use.
The cost driver is more likely to be the average rate and it
is difficult to evaluate it without a clearer idea of the spectrum
of projects. As the DRSP becomes soon available it is decided to
make use of it and derive the average data rate. To do this we
set 2003-12-31 as a desadline.
ALMA Design Reference Science Plan (DSRP) (Robert Lucas, and all)
The DSRP will soon become fully available (a draft version is
already available, see Al's email to almasci). We should discuss
how we derive the most important dimensioning parameters for the
software: obviously project integration times and data rates, but
also distribution of observing modes, number of SBs per programme,
number of observing units, i.e. complexity of programmes,specific
pipeline requirements, ...). This is an agenda item for the
face-to-face meeting.
Export data format discussion document (Francois
Viallefond and Robert Lucas)
This available as annaounced by Joe Mc Mullin's email (see
http://tinyurl.com/pf5l. Comments are welcome.
Face-to-face meeting It will be in Socorro on Nov. 14/15. The second meeting day is at
the ATF (Sat 15) and visit the prototypes at this occasion.
Agenda Items:
Status of Test Plans
consequences of Calibration Plan
Digestion of Design Reference Science Plan
Export data format discussion document
Presentation from all subsystems scientist about the status of
each, issues raised, difficult requirements, ... including an
overview of the wh\ole system (by Joe Schwarz).
Data reduction options (smoothing, resampling, etc.) for
correlator data processing (after correlator hardware, before FDS).
