[alma-sw-ssr] response to Lucas' comments

[Tony Willis <twillis@drao.nrc.ca>]

 Hello all,
 
 Here is my response to Robert Lucas' initial comments
 (see http://iram.fr/%7Elucas/almassr/meetings/canadaProposalComments.html)
 on our real-time imaging proposal.

 Tony
======================================
 
 
 >Comments on the Canadian Software Proposal (as discussed at last ASAC
 >meeting).
  
 >General comments:
 
 >   * It is at least surprising that this proposal be issued without explicit
 >     reference to the considerable amount of work that has been pursued
 >     within the ALMA project to define the science based software
 >     requirements, in the last 18 months or so. The Science Software
 >     Requirements Committee
 >     (http://www.alma.nrao.edu/development/computing/workinggroups/ssr/index.html)
>     has widely distributed reports (Alma Memo 293, now superseded by
 >     ALMA-Software Memo 11 (at
 >     http://www.alma.nrao.edu/development/computing/docs/memos/index.html);
 >     they have been submitted to a review process. Requirements have been
 >     presented to the ASAC Florence meeting in February. See also ALMA Memo
 >     367.
   
 Our apologies for not referencing these documents more carefully. We
 do refer to 'ALMA planners' in our proposal, but we should have explicitly 
 given these documents their due credit. We certainly did not mean to imply
 that the general concept of a pipeline was particularly original on our part.  
 What is different in our proposal is the concept of how the pipeline
 operates and what software components it contains.
  
 >   * This work is continuing by a detailed report on the requirements for
 >     the pipeline and for off-line data analysis. A first draft of this is
 >     to be discussed at the SSR meeting on July 16/17.
 >
  
 We shall be interested to see the details of this document.
  
 >   * The software group is taking the stepwise approach to specify the
 >     requirements in some detail before making actual implementation
 >     choices. Naturally however we have a good feeling on what can be done
 >     and what cannot be done in a quasi-real time pipeline, in order not to
 >     overspecify the system.
 > 
 
 I'm not sure what is meant here. How have you obtained
 'a good feeling' on what can be done and what cannot be done in 
 a quasi-real time pipeline? 
 
 >   * Given that discussion on pipeline requirements is already taking place
 >     within the SSR, we feel that it would be more efficient for ASAC to
 >     review our conclusions rather than each committee having parallel
 >     discussions that would somehow have to be later integrated.
 
 We were unsure of the best route through which to submit this proposal.
 
 >
 >More specific technical comments will be made by the SSR Committee, let me
 >give a few comments of my own:
 >
 >   * Note that in the requirements document ALMA-Software Memo 11 (Req.
 >     1.0-R7, 6.3-R1) the SSR states that a fundamental requirement for ALMA
 >     is that scientifically analyzable images will be automatically produced
 >     for observers.
 > 
 
 OK. We agree on this point. Our proposal suggests a route
 to getting a scientifically analyzable image with the minimum amount of
 human effort or intervention.
 
  
 >   * Our opinion is that the first order feedback on the scheduling should
 >     be based on the calibration quality, not on the final images. Hence the
 >     separation between the calibration pipeline, the quick look pipeline,
 >     and the science data pipeline. Therefore the image processing time
 >     scale is not a few minutes, but a few hours at the end of the observing
 >     session.
 > 
 
 How does one evaluate calibration quality? Essentially we
 are proposing that as much as possible a 'smart' system should evaluate
 the calibration quality, amongst other things, and provide appropriate
 feedback to telescope operators and astronomers. Evaluation of 
 calibration quality can well include evaluation of images of calibration
 sources produced with the current calibration parameters.

 >   * There is too much emphasis on self-calibration, which will not be
 >     feasible for many projects, probably a majority.
 
 I was unaware that we were giving exceptional emphasis to 
 self-calibration in this document. As far as I can see the only place the 
 words 'self-calibration' appear are in the feedback loop of Figure 1b between
 the Deconvolution and Calibration oparations. If one replaces the 
 words 'self-calibration' with 'VIrtual Synthesis Expert (VISE) Feedback'
 then the diagram reflects the intent of our proposal even more closely.
 
 However, in our minds, self-calibration is important. It could be a routine
 part of processing the calibrator images to evaluate the quality of
 phase-referencing. Also, is the entire ALMA universe expected to 
 consist of images of less than 100:1 dynamic range? What about using 
 self-calibration parameters obtained for a continuum image and applying 
 them to line observations to improve quality of the line images?
   
 >   * I doubt whether an automatic expert system can be made before any human
 >     expertise is obtained on submillimeter image synthesis, especially at
 >     the high frequencies. One has to go by steps.
 
 And indeed we intend to. We do not assume that the VISE will solve all
 problems. However we believe that the time has come to begin development
 of such a system because of the large data rate anticipated in future
 synthesis instruments. Even if a VISE initially is only useful at the
 lower frequencies, that would be a benefit since it would free up
 astronomer time to tackle the really tough higher frequency problems. 
 
 Note that we see our proposed simulator as a repository of knowledge
 about the telescope and atmosphere. As our knowledge about various
 facets of ALMA increase we would add them to the simulator. Our simulator
 would observe proper 'strawman' skies through 'strawman' atmospheres. 
 This should allow us to eventually expand the VISE capability to
 handle high frequency problems. Also our simulator would produce data
 at a rate similar to that of the actual telescope correlator, so that
 very early on, the VISE would be subjected to data rates similar to
 what will actually be produced by the final instrument.
 
 >Robert Lucas
 >(Chair of the ALMA Science Software Requirements Committee)
 > 
 
-- 
Tony Willis

Internet  :   Tony.Willis@hia.nrc.ca  
Snailnet  :   Dominion Radio Astrophysical Observatory
              P.O. Box 248, Penticton, BC, Canada V2A 6K3
BC Tel net:   (250) 493-2277    Faxnet    :   (250) 493-7767
voicemailnet: (250) 490-4343    Localnet  :   ext 343