New notebook for si_fir_filter

sim/si_fir_filter/README.md

@@ -1,20 +1,32 @@
 ## Design of the Si TX Spectral Shaping Filter
 
-It's a 17-tap digital FIR filter. No idea what the sample rate is, but
-it could be the IF? That's 400kHz or so..
+The SI contains a 17-tap digital FIR filter for use in GFSK modes.
+
+We use this in aprs mode to filter the gpio squarewave into a sensible
+afsk signal.
+
+There's some notes in a ipython notebook `si_fir_filter.ipynp` on
+this. This notebook is also used to calculate the deviation values.
+
+## Prerequisites
+
+```
+sudo apt-get install ipython libblas-dev liblapack-dev gfortran
+pip install scipy matplotlib
+```
+
+Yes you need a fortran compiler!!!
 
 ## Python Usage
 
-`ipython --pylab`
-
-`%run test.ipy`
+`ipython notebook`
 
 ## Calcuating new coefficients
 
 There's an awesome calculator here
 [here](http://t-filter.appspot.com/fir/index.html)
 
-## Sources
+## So
 
 The python script is from
 [here](http://nbviewer.ipython.org/github/unpingco/Python-for-Signal-Processing/blob/master/Filtering.ipynb)

sim/si_fir_filter/test.ipy

@@ -1,57 +0,0 @@
-from __future__ import division
-from scipy import signal
-
-fig, axs = subplots(2,1,sharex=True)
-subplots_adjust( hspace = .2 )
-fig.set_size_inches((5,5))
-
-t_fir_coeff = [0.01665389596732681,
-  0.07353658710395161,
-  0.10040345335534452,
-  0.04751332314075821,
-  -0.06317967799551721,
-  -0.09143711477971043,
-  0.06319322564548596,
-  0.31257963657965065,
-  0.43524399610103215,
-  0.31257963657965065,
-  0.06319322564548596,
-  -0.09143711477971043,
-  -0.06317967799551721,
-  0.04751332314075821,
-  0.10040345335534452,
-  0.07353658710395161,
-  0.01665389596732681]
-
-# Default si coefficients
-si_default_fir_coeff = [0x1, 0x3, 0x8, 0x11, 0x21, 0x36, 0x4d, 0x60, 0x67]
-
-# Reflects an array of coefficients about the last element
-def sym_fir_coeff(coeff):
-    for i in range(len(coeff)-2, -1, -1):
-        coeff.append(coeff[i])
-
-    return coeff
-# Normalises the si's 8-bit coefficients to 0 - 1
-def si_normalise_coeff(coeff):
-    return [x / 0x300 for x in coeff]
-
-si_default = si_normalise_coeff(sym_fir_coeff(si_default_fir_coeff))
-
-print si_default
-print len(si_default)
-
-fir_coeff = si_default
-
-ax=axs[0]
-w,h=signal.freqz(fir_coeff,1) # Compute impulse response
-ax.plot(w,20*log10(abs(h)))
-ax.set_ylabel(r"$20 \log_{10} |H(\omega)| $",fontsize=18)
-ax.grid()
-
-ax=axs[1]
-ax.plot(w,angle(h)/pi*180)
-ax.set_xlabel(r'$\omega$ (radians/s)',fontsize=18)
-ax.set_ylabel(r"$\phi $ (deg)",fontsize=18)
-ax.set_xlim(xmax = pi)
-ax.grid()