From: VAXN::RICE
Date:  2-MAR-1994  10:38:09
Description: MSE Code Modifications                                  

Some minor modifications have been made recently to the MSE code called by EFIT
(stark_multi.for and related subroutines).  In summary, we have changed the MSE
data uncertainty calculation, reorganized some of the library routines and MSE
directories, and fixed an error trapping bug which prevented reading neutral
beam data during the discharges around 80072. The new uncertainty calculation
may, in some cases, result in significant differences in the equilibrium
obtained using EFIT.  These modifications are described in more detail below.
 
MSE Uncertainty Calculation:
 
The MSE data suffers from three main sources of error: correlated calibration
errors (misalignment or Faraday rotation in optics), uncorrelated errors (drift
in calibration of individual lock-in amplifiers),  and statistical noise due to
low signal levels due to poor beam penetration.  The correlated calibration
error, or offset, is measured as accurately as possible using beam-into-gas and
shrinking plasma discharges.
 
The remaining uncorrelated errors (sgamma in EFIT) were previously calculated
to be the larger of 0.2 degrees or the standard deviation of the measured pitch
angle in the time bin of interest (usually 10 ms).  This results in an
uncertainty that is too large because the averaging of the signal in the time
bin is not taken into account.  During high density discharges such as VH mode
or high beta, the noise on the central MSE channels can become quite large (1-5
degrees), resulting in a poor fit in the core and an incorrect estimate for q0.
 
We have modified the uncorrelated uncertainty calculation to be:
 
	sgamma = sqrt(sigma_calib^2 +sigma_signal^2/n)
 
where sigma_calib is the estimate of channel-to-channel calibration differences
which is now set to 0.35 degrees, sigma_signal is the standard deviation of the
raw signal in the integration time window, and n is the number of data points
in the integration time window.  We divide by n here to give the uncertainty in
the signal mean. Since the MSE signals are digitized at 2 kHz, there are 20
data points in a 10 ms window, so the signal uncertainty can be reduced by
up to sqrt(20) in cases where sigma_signal is large.
 
In general, low density shots will not be significantly affected by this change
since sigma_signal is small anyway.  High density discharges, however, should
note an improved fit in the core, and a more consistent value of q0.  q0 should
never be below about 0.8 during sawtoothing discharges and obviously should be
> 1 during nonsawtoothing discharges (if you find cases where this is not true
then come see me).  Furthermore, no matter how noisy the MSE data is,
improvements in uncertainty can always be obtained (at least down to 0.35 deg)
by increasing the averaging time bin (iavem) in EFIT.
 
 
Neutral Beam Error Problems:
 
It was noted that the MSE code was not reading the neutral beam signals
correctly for shots around 80072 (beam cables were swapped during run).  This
has been fixed.
 
New MSE directories on phys_data:
 
The various MSE codes and calibration files have been consolidated into some
new directories on the phys_data disk.  The source code in in
PHYS_DATA:[EQDSK_D3.MSE.SOURCE] while the calibration files are in
PHYS_DATA:[EQDSK_D3.MSE.CALIB].  Two logicals mse_source_dir and mse_calib_dir
have been set up for these areas.  The code in these areas are the master
versions which should be used with EFIT on Vaxs or workstations, Review, and
IDL.
 
In addition, there are two subdirectories which users may find useful.
PHYS_DATA:[EQDSK_D3.MSE.SOURCE.TESTS] contains test routines (try running
test_stark_multi) which will print the MSE data to a terminal just like you see
when EFIT is run.  This is useful for screening MSE data for many shots or time
slices without having to run EFIT.  Also, there is a directory
PHYS_DATA:[EQDSK_D3.MSE.SOURCE.IDL3] that contains IDL routines (stark_mutli,
msep_multi, and mse_geometry) which can be used to view MSE data and geometry
parameters.  Just run these in IDL and information on usage will be given.