From sschmitt@mail.desy.de Fri Jul  2 16:51:29 2004
Date: Fri, 4 Jun 2004 13:14:15 +0200 (MEST)
From: Stefan Schmitt <sschmitt@mail.desy.de>
To: Osamu Ota <ota@hakone.phys.metro-u.ac.jp>
Cc: David South <southd@mail.desy.de>
Subject: Re: Offline new method

Hallo Osamu,

I understand that the documentation is very poor, please apologize.
The program was written for testing purposes, and there is a lot of
redundant information, stored for convenience.

Most values are stored for colliding and non-colliding. These variables
end with "2". For example: status2(2) is the fit status for non-colliding
and colliding bunches.

If the variable has an error, the error is stored with a "d" in front.
For example:
  pol2(2)    online polarisation, colliding bunches
  dpol2(2)   error on onlien polarisation, colliding bunches

> 1)There are some values which mean "polarisation",pol2,pol,offpol,s3py2,
>   plpol,lpol2. As far as I understand,
>   pol2   : online value, in two arrays,(pol2[0],pol2[1]) colliding
>            non-colliding value are stored respectively. is it right?
>   pol    : 100*pol2, so is this same to pol2?
>   offpol : result from offline fit,averaged L and R ? if so, no
>            colliding and non-colling value stored?
>   s3py2  : have two arrays(s3py2[0],s3py2[1]), L and R value are stored
>            respectively?
>   plpol  : online LPOL value, and I can not understand the difference
>            between plpol and lpol2.

pol2(1),focus2(1) : non-colliding bunches, online method
pol2(2),focus2(2) : colliding bunches, online method
pol,size          : polarisation and beam size (focus), averaged over all
                    bunches, online method

pol and size are calculated as the error-weighted average. The formula is
  dpol = 100./(1/dpol2(1)**2+1/dpol2(2)**2)
  pol = (pol2(1)/dpol2(1)**2 + pol2(2)/dpol2(2)**2)*dpol
or something similar.

"pol" is calculated (with the factor 100 in front), because it can be
easily compared to "plpol".

For the "new" offline analysis, things are a bit more complicated:

S3PYL2(2) : polarisation for left helicity state, non-colliding and
             colliding bunches from offline fit
S3PYR2(2) : same for right helicity state

These values are averaged (taking into account the different sign and the
correlation matrix):

s3py2(2)  : polarisation from offline fit, averaged over left and right
             helicity state

Then the average is calculated of colliding and non-colliding bunches:

offpol    : polarisation from offline fit, average of colliding and
             non-colliding bunches. For comparison with plpol

>From the LPOL the following numbers are stored:

tlpol : time stamp of latest lpol per-minute measurement known at the time
            the data was taken
plpol : latest lpol per-minute measurement known at the time the data was
           taken

For detailed comparison, the colliding and non-colliding polarisation of
the LPOL was calculated in addition:

tlpol2 : time stamp of latest lpol 5-minute measurement of single bunches
plpol2(2) : latest 5-minute LPOL average polarisation for colliding and
              non-colliding bunches. Calculated from single-bunch
              5-minute average

polarisation     LPOL     online     offline    offline(L)  offline (R)
=======================================================================
all bunches:     plpol     pol        offpol       ---        ---
non-colliding:   lpol2(1)  pol2(1)    s3py2(1)   s3pyl2(1)   s3pyr2(1)
colliding:       lpol2(2)  pol2(2)    s3py2(2)   s3pyl2(2)   s3pyr2(2)

Unfortunately there is no consistent treatment of the factor 100 and the
sign...


> 2)About beamsize and focus.
>   size   : focus from online?
>   focus2 : focus on colling and non-colliding respectively?
>   I could not find the value correspond to beamsize from offline fit.

beam size        online     offline
=====================================
all bunches      size        ----
non-colliding    focus2(1)   SIGMAY2(1)
colliding        focus2(2)   SIGMAY2(2)

> 3)I executed new method as follows:
>   daq>linux/testff -d /acs/tpol/03/04/Jan/ 26568 26568 -t +2
>   in that case, I understand colliding bunch only.
>   but, some are stored in both ntuple "focus2[0],focus[1]".

All variables are stored regardless of teh command line options.
If You select -t 2 the offline fit is not executed for the non-colliding
bunches. All variables from the offline fit with index (1) will be
meaningless.

I would recommend to start without the -t option, then You get
results for colliding and non-colliding bunches.
The non-colliding fit does not take so much time anyway, because it has a
smaller number of parameters.

Stefan