CERN Accelerating science

thanks for the minutes from your meeting. It appears to have been a very
productive one.

I believe strongly that it is best to have the front-end publish a
spectrum and do the fitting and analysis outside (eg in a Java box),
which also seems to be the conclusion of the meeting.

By separating the generation of the spectra from the analysis code, we
would get around the problem Ralph has pointed out several times: that
the hardware debugging here at FNAL sometimes have been hampered by the
limited functionality of our combined aqcuisition/analysis software.

Essentially, in this case the hardware commissioning can be done
completely independently of the fitting application, by directly
controlling the "knobs" associated to the front end (gain, timings,
etc), and looking at the published spectra. I believe there is already a
kind of "browser" available at CERN that can do this.

Furthermore, I understand there is a built-in "hardware emulation" in
FESA, which means the Java/C++ coding can be done independently of the
status of the hardware, as long as the interface is properly defined.
Hence, we should work at defining this as soon as possible.

Fritz concerns about the averaging is a pure front-end issue, so you
don't need to worry about that. However, he is certainly right that we
need to provide a fit quality parameter (chi-square?).

I disagree somewhat with Ralph in that I think we at Fermilab (you?) can
certainly take a leading role in the development (porting) of the
fitting software (to the extent that LARP funds can support it).
However, it is important that there is a person identified at CERN to
take over long term responsibility once the code is ready, since we can
not support it indefinitely. It is also in CERNs interested to have a
local expert. It would be nice if Mike could appoint such a person now,
so that he/she can be involved from the start. Perhaps you can talk to
him about that?

Cheers,

/Andreas

Suzanne Panacek wrote:
> Schottky meeting 1-9-06 4:30 PM
> Attending: Rhodri Jones, Fritz Caspers, Jean-Jaques Gras, Suzanne Gysin
>
> The Schottky fitting software can be written by one of two groups:
> 1) the Instrumentation Group (lead by Jean-Jaques Gras)
> 2) the Applications/Controls Group (lead by Mike Lamont)
>
> Who will write it is determined by the language and who has the time. The Instrumentation group writes software that resides in the front end and is written in C and C++. The App/C Group writes application software in Java. The GUI is definitly written in Java, the fitting and extraction of parameters could be in either C or Java. The FNAL fitting code is written in C++ with Root. Jean-Jaques will not accept an interface to Root in the front end. We at FNAL already decided to rewrite the fitting in Java.
>
> Jean-Jaques Gras prefers the fitting to be written in Java outside of the front end. And written in a way that it could easily be ported to C/C++ if performance becomes an issue. So for the rest of the discussion we assumed the fitting as well as teh GUI will be written outside of the front end in Java. As a side note: FNAL would also like to replace the Root fitting with a Java one.
>
> Jean-Jaques has a person assigned to writing the front end code, his name is Stefan Bart-Peterson (sp?), if the fit is written in Java, Mike Lamont will have to assign somone in the longterm to maintain it, even if I were to write it.
>
> There will be an expert interface to tune the instrument while testing it. This will be done by Jean-Jaques group independent of all other sw.
>
> Fritz Caspers was concerned that the spectrum is different than that from Fermi, and that the fit may have to be modified. His remarked the noise peaks are wider and closer to the signal. He did agree that starting with the FNAL fit is ok.
>
> Fritz is also concerned about vector averaging the signal before applying the FFT and sending the spectrum out. This is between Jean-Jaques and Fritz, since this will be done in the front end.
>
> Fritz's last concern is that he wants the quality of the fit to be known and maybe even a visual of the signal and fit with the option to modify it. I know FNAL has implemented something on this line. I will have to see if this is sufficient.
>
> In the end, we decided that the front end will 'publish' a spectrum. The Schottky application will 'subscribe' to the spectrum and do the fitting and publish the relevant parameters. The parameters are written to the database via the middle wear and the subscribe mechanism.
>
> The publish/subscribe mechanism replaces the FNAL database and Acnet APIs. This P/S mechanism is written by Mike Lamont's group and needs to be thouroughly understood by the developer (me). I am working with the LSA group on the remote operations cener and this project is a nice way to learn the interfaces.
>
> Schedule and Planning:
> The fitting software can be done independent of the hardware with simulated data from FNAL. Jean-Jaques said with 2 weeks notice he can make simulated data available.
>
> The hardware with an FFT will be available in summer '06. The fitting sofware can be written anytime, the GUI is not time critical either.
>
> While I am still at CERN I will learn as much as possible about the Publish/Subscribe mechanism.
>
> Any questions/comment?
>
> Suzanne
>

--
_____________________________________________
Andreas Jansson Tel: 1-630-840-2824
Fermilab, MS 341 Pager: 1-630-266-0911
Batavia, IL 60510 Fax: 1-630-840-8737