ALMA Science Advisory Committee
Draft Agenda for ASAC Telecon Weds 14 March 2001
Conference Date: March 14-2001 (Wednesday)
Here's the dial-in information for the telecon:
Conference Date: 14 Mar 2001
Conference Time: 10:15 AM EASTERN TIME=15:15 UT
Conference Duration: 1 hr
Service Level: STANDARD
Call Type: MEET ME/DIALOUT
USA Toll Free Number: 1-888-323-2716
Non-USA Number: +1-712-271-0629
PASSCODE: ALMA
Conference Leader: Mr Al Wootten
Agenda items so far include: Old Business
(1) Please approve the minutes of the February Meeting.
New Business
(1) NSF budget and consequences for US-ALMA project (Bob Brown)
(2) - Report from configuration PDR (S. Guilloteau, Al Wootten, P. CoxGreat. )
(3)- Draft of ASAC report (Welch) (report should be ready to circulate before Wednesday)
Report of the ASAC Face-to-Face meeting.
The text is here.
However, I would like to make a point of
clarification. At the meeting, I reported that investigations by
Ray Escoffier and John Webber show that reprogramming of the correlator will
allow more efficient correlation. There was some question as to exactly what
this efficiency would be. In this mode the 3 bit initial quantization
and 7-bit requantization, with 3 or 4 bits passed along to the correlator after
FIR filtering, results in a gain in efficiency from 88% for the nominal
baseline correlator to 95% in this more efficient mode, according to
calculations by Brent Carlson. Some tradeoff in correlator capacity must
be made to run in this very efficient mode; these are detailed in the
table at
this location.
Note that this results in a correlator satisfying
point b(i) of the recommendations, of achieving full 3bit efficiency.
Detail: Brent Carlson performed some simulations to quantify
these efficiencies. He has some
preliminary numbers that he emailed Webber before incorporation into a
memo:
1. 3-bit quantization and correlation; no FIR filters; threshold step =
0.65 sigma; odd output encoding (-7,-5,-3,-1,1,3,5,7); expected
correlation coefficient=0.5; 10^8 samples; correlator output converted
to normalized correlation coefficients by dividing by the geometric mean
of lag 0 autopower. Efficiency is 0.962 (3.8% SNR loss).
2. Same as 1. only 4-bit/15 level with integer encoding and threshold
step = 0.374 sigma. Efficiency is 0.987 (1.3% SNR loss).
3. 3-bit initial quantization (8-level; odd encoding; 0.65 sigma thresh.
step) and 4-bit requantization (15-level; integer encoding; 0.374 sigma
thresh. step); WIDAR filtering/processing with -12.5 dB cutoff filters
(for minimum additional SNR loss) and filter gain compensation; 12-bit
LUT output in the FIRs; 10^8 samples; lag 0 power normalization;
expected correlation coefficient=0.5. Efficiency is 0.943 (5.7% SNR
loss).
The efficiency in 1. x the efficiency in 2. is 0.962 * 0.987 = 0.949
(5.1% SNR loss) which agrees reasonably well with the efficiency found
in 3. (5.7% SNR loss).
I am also running tests with a correlation coefficient of 0.1 and the
results thus far are in good agreement with the above results.
In the ALMA correlator case where you are using 4 x FPT n=3, 2-bit
multipliers to do 4-bit correlation, the 4-bit output encoding must be
odd encoding--which may give a slight (<~0.1%) efficiency increase
(according to a curve in Hovey's master's thesis on 14-level correlation
we have here).
Perhaps that largest source of error in these results is that I just
produce normalized correlator coefficients by dividing the raw
correlator data by the lag 0 autopower. This does not take into account
the slight up-turn in the curve of raw coefficients vs normalized
coefficients. This, however, should make the above efficiencies
slightly better than presented.
Other dates during the next month's timeframe:
April 5-6 ACC meeting in Tokyo
(4) Tuning Range Question.
The AEC asks:
The AEC asks the ASAC to comment on the following issue: If it is not
feasible to achieve adequate receiver performance over the full RF
bandwidth for the initial receivers what is the tuning range that
*necessarily* must be covered. This applies to all four initial ALMA
bands, not just band 7.
There are three important points here:
(1) The firm plan is to achieve the full RF bandwidth in the Project
Book for all receiver bands, and do so by the end of construction;
(2) But the first few (~10?, 20?) frequency cartridges delivered for
each band may have more limited performance. It is for these cartridges
that it would be helpful to have the ASAC comment on the more restricted
frequency range.
(3) If it is necessary that the initial cartridges have somewhat limited
performance, those cartridges will be retrofited with cartridges that do
achieve the specs as soon as the retrofits can be made available.
Details:
Dear Wolfgang and John,
During yesterday's telecon, the subject of the frequency coverage in
band 7 was discussed.
As we understand it, it appears that in at
least one combination, i.e. single-ended mixers with a 4-8GHz IF, it
would not be possible to provide enough LO power on a band wide
enough to cover the RF range 275-370GHz.
If we understand well, John Webber stated that he felt it might be
possible to provide adequate LO power for balanced mixers over a LO
range equal to RF-2*8GHz.
Maybe he could supply enough LO power for single-ended mixers over a
range equal to RF-2*12GHz (to be confirmed).
>From the mixer side, this means either balanced mixers that _may_
have a 4-8GHz IF, or single-ended mixers with a 8-12 (or 4-12) GHz
IF. Of course, the combination balanced mixers, plus IF extending to
12GHz would be even better. However, either of these (balanced,
higher IF) increases the technical difficulty of mixer design and
prototyping, in a context where the technical ambition conflicts with
the schedule.
>From the LO side, and if we remember well, John Webber stated that
the projected performance was at the limit of the simulation results,
i.e., it will be hard to reach.
This being said, we may, within the FE subsystem, discuss the way to
meet the specs with least technical risk, but we know that is a
difficult goal to reach. So, we propose to re-examine the scientific
rationale behind the specs (a point raised by the PDR reviewers).
As a first step, we have performed a simple exercise, trying to
estimate the relative priority in various parts of band 7. Using
the data in:
http://physics.nist.gov/cgi-bin/micro/table5/start.pl
we plotted the line intensities versus frequency, overlaying the
atmospheric transmission curve for 1mm PWV. Two plots are
attached, with the second one having an expanded vertical scale. As
far as we can see, there are only rather weak lines (most intensities
are recorded in Orion) between the 275GHz lower edge of band 7 and at
least 290GHz, and, to a lesser degree, up to 300GHz. The absence of
lines on our plot between 363 and 370 GHz is probably due to lack of
coverage of the observations, so we won't discuss it.
Given that the coverage of the specified RF band is at best
difficult, and may be uncertain, should not the technical efforts be
targeted at regions of maximum scientific interest. In other words,
should we not propose to the Science group to define a "first
priority" coverage of band 7, not excluding full coverage as
development progress might allow. Please note that even the more
modest option of dropping 275-290 GHz from first priority yields a
reduction of LO range (all other things equal) of 15 GHz, almost
twice the reduction provided by raising the upper edge of the IF band
from 8 to 12GHz (2*4=8GHz).
We provide the enclosed arguments and data so that you may use them
as you see fit to approach the ASAC and maybe obtain a reasonable
compromise between requirements and technical difficulty. You might
also consider to distribute the present message to the JRDG for
consideration at the next telecon.
With our best regards,
S.Claude & B.Lazareff
(5) ALMA Enhancements - Hasegawa
Before the Florence meeting, ASAC has been asked by ALG on the priority
among the enhancements enabled by Japanese participation under a 10% cut
(see the attached pdf document). At the beginning of the Florence meeting,
we are told that ACC now wants ASAC to think about the priorities under 20%
cut. We tried to do so in the course of the meeting, with a very miserable
mood, and managed to squeeze out some choice. We did not have time and
energy in Florence to think of the 10% case. At the same time, ASAC have
sent a letter to ACC to think the 20% cut over again.
The above sequence of events means that we still need to make a
prioritization for the 10% cut constraint. I would like to present a model
supported by Japanese ASAC members as follows. We believe that it greatly
enhances the scientific capability of ALMA in many ways while staying within
the budgetary constraints set by the 10% cut. We want to hear the ASAC
thoughts on Wednesday.
Net new resource assumed $140M (under 10% cut, overhead of ALG-estimated
$50M removed; see the pdf document)
(6) Next Teleconference -
According to one calculation, the time zones relevant are now (1515 UT):