def THDns(x, fs):
"""
We take in rows of data and return a column of THDN measurements.
"""
# first, zero-mean
rowcnt = len(x)
meanxr = x.mean(1)
meanxr.shape = (rowcnt, 1)
xr = x - meanxr
out = empty(len(x), Float)
thdlist = []
for i, xrow in enumerate(xr):
thdn = 0.0
enob = 0.0
(thdnpy, A, B, C, w) = pysinlesq.computeTHDN(xrow, int(fs))
(thdnc, A, B, C, w) = csinlesq.computeTHDN(xrow, int(fs))
if thdnc < thdnpy:
out[i] = thdnc
else:
out[i] = thdnpy
return out
def compWave(filename):
f = tables.openFile(filename)
num = 2
t1 = f.root.A1.gain100.hpf1.sine
r1 = [ (row['frequency'], row['data']) for row in
t1.where(t1.cols.sourcevpp > 0 )]
x1= r1[num][1][:]
fs = 32000
N = len(x1)
t = r_[0.0:N]/ fs;
(thdn, A1, B1, C1, w1) = csinlesq.computeTHDN(x1, fs)
x1m = (A1*cos(t*w1) + B1 * sin(t*w1) + C1)
print "x1 thdn = ", thdn
pylab.figure(1)
pylab.plot(x1-x1m, label = "g=1")
pylab.legend()
pylab.grid()
pylab.figure(2)
pylab.plot(x1, 'r')
pylab.plot(x1m, 'g')
pylab.legend()
pylab.grid()
pylab.grid()
test = []
for i in range(100):
test.append(mean(x1[(i*1000):((i+1)*1000)]))
t = r_[0:1:(1/fs)]
n = array(randn(len(t))/2**14)
fc = 300
(b, a) = signal.bessel(1, fc/fs, btype='high')
print b, a
bits = 16
thdnf = []
thdn = []
wl = r_[20:1000:50]
for w in wl:
x = sin(w*t * pi * 2)
y = x + n
q = signal.lfilter(b, a, y)
z = quant(q, bits)
zu = quant(y, bits)
thdn.append(sinlesq.computeTHDN(zu[2**10:], fs))
thdnf.append(sinlesq.computeTHDN(z[2**10:], fs))
pylab.plot(wl, thdn, label = "unfiltered")
pylab.plot(wl, thdnf, label = 'filtered')
pylab.grid(True)
pylab.legend()
pylab.show()
def plotmanyTHDns():
f = tables.openFile(sys.argv[1])
h = io.read_array(file('HPF-as-fir.dat'))
ns = [2**13, 2**11, 2**9]
plots = [2, 10, 19]
for pnum, i in enumerate(plots):
pylab.subplot( len(plots), 1, pnum+1)
cl = list(colorlist)
for n in ns:
t = f.root.A3.gain10000.hpf0.sine
x = array(t[i][0], dtype=Float64)
y = (x - mean(x)) /2**15
fs = 32000
#print "before", t[i][1], max(y), min(y),
#y = signal.convolve(h, y, mode='same')
#print "after", max(y), min(y)
N = len(y) / n
y.shape = (N, -1)
errs = empty_like(y)
thdns = []
params = []
for j, yrow in enumerate(y):
thdn = 0.0
enob = 0.0
(thdn, A, B, C, w) = csinlesq.computeTHDN(yrow, fs)
errs[j] = thdnMeasure.getSineError(yrow, A, B, C, w, fs)
thdns.append(thdn)
params.append((A, B, C, w))
worst = argmax(array(thdns))
best = argmin(array(thdns))
b = errs[best]
w = errs[worst]
print n, len(thdns)
pylab.plot(arange(N, dtype=float)/(N-1), thdns,
color = cl[0], label = "length %d segments " % n)
pylab.axhline(mean(thdns), color = cl[0], linestyle = '--',
label = "_nolegend_")
cl.pop(0)
pylab.grid()
pylab.title("Freq = %d " % t[i][1])
pylab.legend()
leg = pylab.gca().get_legend()
ltext = leg.get_texts() # all the text.Text instance the legend
pylab.ylabel('THD + n')
pylab.setp(ltext, fontsize='xx-small') # the legendfontsize
pylab.show()