[alma-config] (no subject)

[John Conway]

On Fri, 7 Oct 2005, Frederic Boone wrote:

> Hi John,
>
> Following Al's email and as I promised I will not
> discuss the design anymore.
> I also swear this is my last email ever to this list.
>
> > The
> > only (slightly, for radio astromomy) novel aspect of
> > what I did was
> > explicity to choose weights for each uv point in a
> > snapshot to
> > force explicity the beam to a  gaussian shape.
>


> This is precisly the problem: this is not novel, this
> was done in memo 400 ("compensate the density excess
> with respect to an ideal distribution").
>
> I don't pretend I invented "weights".
> My only contribution at this level was to propose
> weights could be used to force the beam to have a
> given shape. You can say your formulation (i.e. the
> words used and how this is achieved) is different than
> mine but the idea is exactly the same.
>


You incompletely quote me above, which gives a misleading
impression,  the full text in  my email was  .....

'The only (slightly, for radio astromomy) novel aspect of what I did was
explicity to choose weights for each uv point in a snapshot to
force explicity the beam to a  gaussian shape. To control the
ill-conditioned  nature of this problem I used a pseudo-inverse,
selecting a number non-zero eignevalues such that for any given
desired accuracy of  modelling the beam a set of weights is chosen
with mimimum norm (and hence minimum sensitivty loss).  I don't see
any of this  discussion in the appendices to  memo 400.'

i.e. I am claiming that as the very, very slightly new part is the
pseudo-inverse which I didn't see in your memo 400. I took
the general idea of beam forcing to be well known and therefore
was not either claiming to invent beam forcing or the idea of
weights. As you say yourself in your memo 400  'The idea that
sidelobes are partly due to defects in the dirstribution of
samples is well known (Woody 2001) and weighting the data in a way that
compensates for these irregulariries is not new (see e.g,
Briggs et al 2001).'


> This is precisly the problem: this is not novel, this
> was done in memo 400 ("compensate the density excess
> with respect to an ideal distribution").
>
> I don't pretend I invented "weights".
> My only contribution at this level was to propose
> weights could be used to force the beam to have a
> given shape. You can say your formulation (i.e. the
> words used and how this is achieved) is different than
> mine but the idea is exactly the same.
>


> > Most implementations of beam forcing in radio
> astronomy,
> > generally force
> > uniform uv density on a grid, then gaussian taper
> the result (this is
> > I think what is described in memo 400)
>
> First I don't know any reference mentioning this (I
> would be interested to know about all these very
> numerous references you imply with "Most
> implementations"). As far as I know radioastronomers
> use uniform weighting to gain in resolution and
> tapering to increase the brightness sensitivity but I
> don't know any example where both are used together.
>

I think in in both AIPS and difmap you can choose
both 'uvwgt UN' to  get uniform weight and set 'uvtaper' to
 a non zero  value to then apply gaussian wgts. So I think it is
there. The problem I believe is that for the average user the concept of
'beam forcing' is unfortunatly a foreign one. You explictly
wrote the idea down which is great (and was a good
enough scholar to note in memo 400 that it was not new, maybe
I should have wrote also that also  was not new in my document
too, but remeber it was an enginnering document not a paper or
memo).  Briggs wrote it down too, I  think  Ryle did too. Meanwhile off
and on  through the years  many people have talked about it, that
I think is the status.

> Second, this cannot work efficiently because to force
> the shape of the beam it is required to get a sense of
> the actual distribution of weights in the data and
> this is very difficult with the gridding used in the
> method you mention (it requires super-Nyquist sampling
> everywhere to be really efficient).
>

I mention this in my previous email as well. Which is
why for snapshots I used by pseudo-inverse based code.


> I have to acknowledge the appendix of the memo was not
> complete enough, and I always thought I should write a
> memo to describe the method used in more detail (I am
> still planing to do this and you give me more
> motivation now). I gave only few more hints (but still
> not enough details) in my thesis manuscript (online on
> my web page, p39, sorry this is in french).
>

Unfortunatly my French is very bad. All I know is from
memo 400 Appedix B it is hard to work out what the  method is;  Appendix B
mentions weight floors and coupling constants. From what I can see these
have nothing to do with my use of pseudoinverse methods. If we
go back to what started this I was only  reacting to your initial claim
that I had somehow using  your implementation  without crediting you.

My present guess is that your
method makes use of the multiscale uv grid you used for the uv
coverage optimations (its doesn't say that anywhere in the
memo appendix but after  this
exchange of emails and thinking about it  this is my guess). Again as I
mentioned in a previous email I suspect such multi-grid methods
are likley  to be  the most  efficent  method for for long tracks, rather
than tryimg to  expand my idea which works for snapshots.

As Bryan mentions we are always reinventing things, take the idea
of flight for instance, many had all the ideas and solved parts of
the problem before  the Wright brothers, but they put them together into
something that actuallly did the job,  I urge you
indeed to write up your  method, and write  working code  then
'market'  the concept of beam forcing so that it is a mental concept
known to general users, not just experts. If a new useful task
comes out of this discussion that is great.

   John




> Regards,
> Frederic
>
>
>
>
>
>
>
>
> ___________________________________________________________________________
> Appel audio GRATUIT partout dans le monde avec le nouveau Yahoo! Messenger
> Téléchargez cette version sur http://fr.messenger.yahoo.com
> _______________________________________________
> Alma-config mailing list
> Alma-config at listmgr.cv.nrao.edu
> http://listmgr.cv.nrao.edu/mailman/listinfo/alma-config
>