def test_get_beam_w_fr(self):
#interps = fringe.get_beam_w_fr(self.aa, (1,4), ref_chan=160)
#t,firs, frbins, frspace = fringe.get_fringe_rate_kernels(interps, 42.9, 401)
frp, bins = fringe.aa_to_fr_profile(self.aa, (1,4), 100)
print "the bins",len(bins), len(frp)
timebins, firs = fringe.frp_to_firs(frp, bins, self.aa.get_freqs(), fq0=self.aa.get_freqs()[100])
print len(timebins), len(firs[100])
frp_fit = fringe.fir_to_frp(firs[100])
# print n.ones_like(frspace).sum()
# print n.sum(frp**2)
p.subplot(121)
p.plot(timebins, firs[100])
p.plot(timebins, n.abs(firs[100]))
p.subplot(122)
print len(bins), bins[0], bins[-1]
p.plot(bins, frp, 'k')
p.plot(bins, frp_fit, 'b', label='new')
p.xlim(-.0005,.0020)
p.ylim(0.0,1.0)
p.legend()
p.show()
def test_get_beam_w_fr(self):
#interps = fringe.get_beam_w_fr(self.aa, (1,4), ref_chan=160)
#t,firs, frbins, frspace = fringe.get_fringe_rate_kernels(interps, 42.9, 401)
frp, bins = fringe.aa_to_fr_profile(self.aa, (1, 4), 100)
print "the bins", len(bins), len(frp)
timebins, firs = fringe.frp_to_firs(frp,
bins,
self.aa.get_freqs(),
fq0=self.aa.get_freqs()[100])
print len(timebins), len(firs[100])
frp_fit = fringe.fir_to_frp(firs[100])
# print n.ones_like(frspace).sum()
# print n.sum(frp**2)
p.subplot(121)
p.plot(timebins, firs[100])
p.plot(timebins, n.abs(firs[100]))
p.subplot(122)
print len(bins), bins[0], bins[-1]
p.plot(bins, frp, 'k')
p.plot(bins, frp_fit, 'b', label='new')
p.xlim(-.0005, .0020)
p.ylim(0.0, 1.0)
p.legend()
p.show()
h = a.healpix.HealpixMap(nside=64) #healpix map for the beam xyz = h.px2crd(n.arange( h.npix() ), ncrd=3) tx,ty,tz = n.dot(aa._eq2zen, xyz) #rotate the coordinated system to be centered on the array. This is equatorial centered at the array. _bmx = aa[0].bm_response((tx,ty,tz),pol='x')[0] _bmy = aa[0].bm_response((tx,ty,tz),pol='y')[0] bmI = 0.5 * (_bmx**2 + _bmy**2) bmI = n.where(tz > 0, bmI, 0) # only use beam values above the horizon. bl = aa.get_baseline(0,26,'r') * .151 #baseline length in frequency. print aa.get_baseline(0,26,'r') fng = C.frf_conv.mk_fng(bl, xyz) #get the fringe rate filter in frf_conv. aa only has one channel in it. frp, bins = fringe.aa_to_fr_profile(aa, (1,4), 0) #XXX hard-coded tbins, firs = fringe.frp_to_firs(frp, bins, aa.get_freqs(), fq0=aa.get_freqs()[0],alietal=True) frp = fringe.fir_to_frp(firs) #frf,bins,wgt,(cen,wid) = C.frf_conv.get_optimal_kernel_at_ref(aa, 0, (0,26)) #bwfrs = C.frf_conv.get_beam_w_fr(aa, (0,26), ref_chan=0) #tbins,firs,frbins,frfs = C.frf_conv.get_fringe_rate_kernels(bwfrs,42.9,403) #need to renormalize to get the proper scaling. firs are properly normalized. #firs = firs[0] #frfs = n.fft.fftshift(n.fft.fft(n.fft.ifftshift(firs), axis=-1)) #get weights. wgts = scipy.interpolate.interp1d(bins, frp, kind='linear') fng_wgt = wgts(fng) #gets weightings at fringes on the sky. fng_bm = bmI * fng_wgt #flat weighting determined by the maximum possible fringe rate for a baseline #and 0. skypass = n.where(n.abs(bins) < n.max(n.abs(fng)), 1., 0) # This is the noise level after being filtered and averaged by the application of this filter.
ij,
mychan,
bins=frbins,
pol=opts.pol,
bl_scale=opts.bl_scale)
timebins, firs[sep] = fringe.frp_to_firs(
frp,
bins,
aa.get_afreqs(),
fq0=aa.get_afreqs()[mychan],
bl_scale=opts.bl_scale,
fr_width_scale=opts.fr_width_scale,
alietal=opts.alietal,
maxfr=opts.maxfr)
frp = fringe.fir_to_frp(firs[sep])
if opts.boxcar:
print 'Making Boxcar',
print 'Width {0}s ...'.format(opts.teff)
top_hat = n.zeros_like(firs[sep])
l_hat = len(top_hat[0])
if opts.teff: box_time = opts.teff
else: box_time = 2232.
start = n.round(l_hat / 2. - box_time / data_inttime / 2.)
end = n.round(l_hat / 2. + box_time / data_inttime / 2.)
diff = n.round(box_time / data_inttime - (end - start))
if diff != 0: end += diff
if (end - start) % 2 == 0: end += 1
top_hat[:, start:end] += 1.
if opts.frc:
#DEFAULT_FRBINS = n.arange(-.01+5e-5/2,.01,5e-5) # Hz
firs = {}
for sep in seps:
c = 0
while c != -1:
ij = map(int, sep2ij[sep].split(',')[c].split('_'))
bl = a.miriad.ij2bl(*ij)
if blconj[bl]: c+=1
else: break
print mychan,ij,opts.bl_scale
frp, bins = fringe.aa_to_fr_profile(aa, ij, mychan, bins=frbins,pol=opts.pol,bl_scale=opts.bl_scale)
timebins, firs[sep] = fringe.frp_to_firs(frp, bins, aa.get_afreqs(), fq0=aa.get_afreqs()[mychan],
bl_scale=opts.bl_scale, fr_width_scale = opts.fr_width_scale, alietal = opts.alietal,maxfr=opts.maxfr)
frp = fringe.fir_to_frp(firs[sep])
if opts.boxcar:
print 'Making Boxcar',
print 'Width {0}s ...'.format(opts.teff)
top_hat = n.zeros_like(firs[sep])
l_hat =len(top_hat[0])
if opts.teff: box_time = opts.teff
else: box_time = 2232.
start = n.round(l_hat/2. - box_time/data_inttime/2.)
end = n.round(l_hat/2. + box_time/data_inttime/2.)
diff = n.round(box_time/data_inttime - ( end - start))
if diff != 0: end += diff
if (end-start) % 2 == 0: end +=1
top_hat[:,start:end] += 1.
if opts.frc:
timebins, firs[sep] = fringe.frp_to_firs(frp[scale][frw_scale][sep], bins, aa.get_afreqs(), fq0=aa.get_afreqs()[mychan], limit_xtalk=True, bl_scale = scale, fr_width_scale = frw_scale, maxfr=opts.maxfr)
timebins*= opts.data_inttime/opts.inttime
if False and scale ==1:
delta=prms0[-1]/n.sqrt(1+prms0[-1]**2)
print 'model fit parameters: ',prms0
print 'norm is: ', n.sum(frp)
print 'mean is: ', n.sum(bins*frp)/n.sum(frp)
mn= n.sum(bins*frp)/n.sum(frp)
sq= n.sqrt(n.sum((bins-mn)**2*frp)/n.sum(frp))
sk= n.sum(((bins-mn)/sq)**3*frp)/n.sum(frp)
ftsk= (4-n.pi)/2.* (delta*n.sqrt(2/n.pi))**3/(1-2*delta**2/n.pi)**(1.5)
print 'actual skew is: ', sk
print 'fitted skew is: ', ftsk
frps[sep], frp_freqs = fringe.fir_to_frp(firs[sep],tbins=timebins)
baselines = ''.join(sep2ij[sep] for sep in seps)
if PLOT:
ax_frp.plot(frp_freqs*1e3,frps[sep][mychan]/n.max(frps[sep][mychan]),
label='{0}'.format(scale),color=cmap(cnt))
ax_firs.plot(timebins,n.abs(firs[sep][mychan]),'-',
label='{0}'.format(scale),color=cmap(cnt),
linewidth=2)
ax_firs.plot(timebins,firs[sep][mychan].real,'--',
color=cmap(cnt),alpha=.5)
ax_firs.plot(timebins,firs[sep][mychan].imag,'-.',
color=cmap(cnt),alpha=.5)
ax_firs.set_xlabel('s')
envelope = n.abs(firs[sep][mychan])
envelope /= n.max(envelope)
limit_xtalk=True)
if False and pad == 1:
delta = prms0[-1] / n.sqrt(1 + prms0[-1]**2)
print 'model fit parameters: ', prms0
print 'norm is: ', n.sum(frp)
print 'mean is: ', n.sum(bins * frp) / n.sum(frp)
mn = n.sum(bins * frp) / n.sum(frp)
sq = n.sqrt(n.sum((bins - mn)**2 * frp) / n.sum(frp))
sk = n.sum(((bins - mn) / sq)**3 * frp) / n.sum(frp)
ftsk = (4 - n.pi) / 2. * (delta * n.sqrt(2 / n.pi))**3 / (
1 - 2 * delta**2 / n.pi)**(1.5)
print 'actual skew is: ', sk
print 'fitted skew is: ', ftsk
frps[sep], frp_freqs = fringe.fir_to_frp(firs[sep], tbins=timebins)
baselines = ''.join(sep2ij[sep] for sep in seps)
for sep in seps:
if PLOT:
ax_frp.plot(frp_freqs * 1e3,
frps[sep][mychan] / n.max(frps[sep][mychan]),
label='{0}'.format(pad),
color=cmap(cnt))
ax_firs.plot(timebins,
n.abs(firs[sep][mychan]),
label='{0}'.format(pad),
color=cmap(cnt))
ax_firs.set_xlabel('s')
envelope = n.abs(firs[sep][mychan])
envelope /= n.max(envelope)
if False and pad ==1:
delta=prms0[-1]/n.sqrt(1+prms0[-1]**2)
print 'model fit parameters: ',prms0
print 'norm is: ', n.sum(frp)
print 'mean is: ', n.sum(bins*frp)/n.sum(frp)
mn= n.sum(bins*frp)/n.sum(frp)
sq= n.sqrt(n.sum((bins-mn)**2*frp)/n.sum(frp))
sk= n.sum(((bins-mn)/sq)**3*frp)/n.sum(frp)
ftsk= (4-n.pi)/2.* (delta*n.sqrt(2/n.pi))**3/(1-2*delta**2/n.pi)**(1.5)
print 'actual skew is: ', sk
print 'fitted skew is: ', ftsk
frps[sep], frp_freqs = fringe.fir_to_frp(firs[sep],tbins=timebins)
baselines = ''.join(sep2ij[sep] for sep in seps)
for sep in seps:
if PLOT:
ax_frp.plot(frp_freqs*1e3,frps[sep][mychan]/n.max(frps[sep][mychan]),
label='{0}'.format(pad),color=cmap(cnt))
ax_firs.plot(timebins,n.abs(firs[sep][mychan]),
label='{0}'.format(pad),color=cmap(cnt))
ax_firs.set_xlabel('s')
envelope = n.abs(firs[sep][mychan])
envelope /= n.max(envelope)
dt = n.sqrt(n.sum(envelope*timebins**2)/n.sum(envelope))
dt_50 = (timebins[envelope>0.5].max() - timebins[envelope>0.5].min())
print "pad = ", pad, "variance width = ",sep,
print " [s]:",int(n.round(dt)),
timebins *= opts.data_inttime / opts.inttime
if False and scale == 1:
delta = prms0[-1] / n.sqrt(1 + prms0[-1]**2)
print 'model fit parameters: ', prms0
print 'norm is: ', n.sum(frp)
print 'mean is: ', n.sum(bins * frp) / n.sum(frp)
mn = n.sum(bins * frp) / n.sum(frp)
sq = n.sqrt(n.sum((bins - mn)**2 * frp) / n.sum(frp))
sk = n.sum(((bins - mn) / sq)**3 * frp) / n.sum(frp)
ftsk = (4 - n.pi) / 2. * (delta * n.sqrt(2 / n.pi))**3 / (
1 - 2 * delta**2 / n.pi)**(1.5)
print 'actual skew is: ', sk
print 'fitted skew is: ', ftsk
frps[sep], frp_freqs = fringe.fir_to_frp(firs[sep], tbins=timebins)
baselines = ''.join(sep2ij[sep] for sep in seps)
if PLOT:
ax_frp.plot(frp_freqs * 1e3,
frps[sep][mychan] / n.max(frps[sep][mychan]),
label='{0}'.format(scale),
color=cmap(cnt))
ax_firs.plot(timebins,
n.abs(firs[sep][mychan]),
'-',
label='{0}'.format(scale),
color=cmap(cnt),
linewidth=2)
ax_firs.plot(timebins,
firs[sep][mychan].real,
#firs = n.tile(gaus, (nchan,1))
#firs = n.zeros_like(firs)
#num_t = n.shape(firs)[1]
#firs[:,num_t/2-int(opts.teff/42.9499)/2 : num_t/2+int(opts.teff/42.9499)/2] += 1
#firs /= n.sqrt(n.sum(n.abs(firs)**2,axis=1).reshape(-1,1)) # normalize so that n.sum(abs(fir)**2) = 1
#firs = firs[chans]
top_hat = n.zeros_like(firs)
l_hat = len(top_hat[0])
if opts.teff:
top_hat[:,
l_hat / 2 - n.int(opts.teff / 42.9499) / 2.:l_hat / 2 +
n.int(opts.teff / 42.9499) / 2.] += 1
else:
top_hat = n.ones(n.int(2232 / 42.9499))
junk = fringe.fir_to_frp(top_hat)
#junk = n.roll(junk, frp[0].argmax() - junk[0].argmax() ,axis=-1)
firs = fringe.frp_to_fir(junk)
firs /= n.sqrt(n.sum(n.abs(top_hat)**2, axis=-1)).reshape(-1, 1)
sys.stdout.flush()
for k in days:
noise_array[k] = {}
un_filt_noise_array[k] = {}
for bl in x[k]:
noise_shape = list(x[k][bls_master[0]].shape)
d_size = noise_shape[1]
n_mult = 50.
noise_shape[1] *= n_mult
ns = noise(noise_shape) * NOISE
ns1 = n.copy(ns)
#wij = n.transpose( f[days[0]][bls_master[0]], [1,0]) #flags (time,freq)
h = a.healpix.HealpixMap(nside=64) #healpix map for the beam xyz = h.px2crd(n.arange( h.npix() ), ncrd=3) tx,ty,tz = n.dot(aa._eq2zen, xyz) #rotate the coordinated system to be centered on the array. This is equatorial centered at the array. _bmx = aa[0].bm_response((tx,ty,tz),pol='x')[0] _bmy = aa[0].bm_response((tx,ty,tz),pol='y')[0] bmI = 0.5 * (_bmx**2 + _bmy**2) bmI = n.where(tz > 0, bmI, 0) # only use beam values above the horizon. bl = aa.get_baseline(0,26,'r') * .151 #baseline length in frequency. fng = C.frf_conv.mk_fng(bl, xyz) #get the fringe rate filter in frf_conv. aa only has one channel in it. frp, bins = fringe.aa_to_fr_profile(aa, (0,26), 0) tbins, firs = fringe.frp_to_firs(frp, bins, aa.get_freqs(), fq0=aa.get_freqs()[0]) frp = fringe.fir_to_frp(firs) #frf,bins,wgt,(cen,wid) = C.frf_conv.get_optimal_kernel_at_ref(aa, 0, (0,26)) #bwfrs = C.frf_conv.get_beam_w_fr(aa, (0,26), ref_chan=0) #tbins,firs,frbins,frfs = C.frf_conv.get_fringe_rate_kernels(bwfrs,42.9,403) #need to renormalize to get the proper scaling. firs are properly normalized. #firs = firs[0] #frfs = n.fft.fftshift(n.fft.fft(n.fft.ifftshift(firs), axis=-1)) #get weights. wgts = scipy.interpolate.interp1d(bins, frp, kind='linear') fng_wgt = wgts(fng) #gets weightings at fringes on the sky. fng_bm = bmI * fng_wgt #flat weighting determined by the maximum possible fringe rate for a baseline #and 0. skypass = n.where(n.abs(bins) < n.max(n.abs(fng)), 1., 0) # This is the noise level after being filtered and averaged by the application of this filter.
timebins, firs[sep] = fringe.frp_to_firs(frp, bins, aa.get_afreqs(), fq0=aa.get_afreqs()[mychan], limit_xtalk=True,frpad=pad)
if False and pad ==1:
delta=prms0[-1]/n.sqrt(1+prms0[-1]**2)
print 'model fit parameters: ',prms0
print 'norm is: ', n.sum(frp)
print 'mean is: ', n.sum(bins*frp)/n.sum(frp)
mn= n.sum(bins*frp)/n.sum(frp)
sq= n.sqrt(n.sum((bins-mn)**2*frp)/n.sum(frp))
sk= n.sum(((bins-mn)/sq)**3*frp)/n.sum(frp)
ftsk= (4-n.pi)/2.* (delta*n.sqrt(2/n.pi))**3/(1-2*delta**2/n.pi)**(1.5)
print 'actual skew is: ', sk
print 'fitted skew is: ', ftsk
frps[sep], frp_freqs = fringe.fir_to_frp(firs[sep],
tbins=timebins*opts.lstbintime/opts.inttime)
baselines = ''.join(sep2ij[sep] for sep in seps)
#print the effective integration time for each seperation
for sep in seps:
print "sep",sep
print " NEBW T_eff = ",
print 1.2/noise_equivalent_bandwidth(frp_freqs,frps[sep][mychan])
if NOISE > 0.: # Create a fake EoR signal to inject
print 'FILTERING WHITE NOSIE at {0} Jy ...'.format(NOISE) ,
sys.stdout.flush()
## this last term is to make the power spectrum equal
## to expected noise line if 21cmSense noise only no filter is run
ed = noise((nchan,711)) * NOISE
ed1 = n.copy(ed)