Spectrum Analysis

For deriving spectrum, it is better to integrate longer than default for getting better S/N ratio. In case integrating 300 points (30 sec),
IDL> mint=300 $<$CR$>$
IDL> norp_mkint,mint,mvd,tim,fi,fv,mvdav,timav,fiav,fvav $<$CR$>$
Next subtract pre-flare flux
IDL> for m=0,6 do fiav(m,*)=fiav(m,*)-fiavg(m) $<$CR$>$
Using this flux, fit the spectrum
IDL> norp_alpha,freq,fiav,mvdav,mvdfit,alpha_tk,alpha_tn,freqpk,fluxpk $<$CR$>$
Assuming gyrosynchrotron emission, the power-law index of non-thermal electrons is obtained
IDL> norp_alpha2delta,alpha_tn,delta $<$CR$>$

Sometimes fitting will fail if the observation data include some points extremely out of the fitting function. One reason is that the emission mechanism is different in the low frequency range ( not gyrosynchrotron but plasma emission ). Another reason is bad S/N ratio in high frequency observation, especially in 80 GHz. In this case, remove such data points from fitting as follwos:
IDL> mvfreq=[2,3,4,5,6] $<$CR$>$
IDL> norp_alpha,freq(mvfreq),fiav(mvfreq,*),mvdav(mvfreq,*) $ $<$CR$>$
IDL> ,mvdfit,alpha_tk,alpha_tn,freqpk,fluxpk $<$CR$>$

In order to confirm that the fitting is well-done, plot the fitting function over the observation data. For example, for n-th data
IDL> norp_funcp,alpha_tk(n),alpha_tn(n),freqpk(n) ,fluxpk(n),freqfit,fifit $<$CR$>$
IDL> plot,freq,fiav(*,n),/xlog,/ylog,psym=5 $<$CR$>$
IDL> oplot,freqfit,fifit $<$CR$>$