def getdata(fname, maxv):
print "Processing ", fname
dat, attribs, cf = bu.getdata(os.path.join(path, fname))
if (len(attribs) > 0):
fsamp = attribs["Fsamp"]
xdat = dat[:, data_columns[1]]
lentrace = len(xdat)
## zero pad one cycle
corr_full = bu.corr_func(dat[:, drive_column], xdat, fsamp, fdrive)
return corr_full[0], np.max(corr_full)
def getdata(fname, maxv):
print "Processing ", fname
dat, attribs, cf = bu.getdata(os.path.join(path, fname))
if( len(attribs) > 0 ):
fsamp = attribs["Fsamp"]
xdat = dat[:,data_columns[1]]
lentrace = len(xdat)
## zero pad one cycle
corr_full = bu.corr_func( dat[:,drive_column], xdat, fsamp, fdrive)
return corr_full[0], np.max(corr_full)
def getdata(fname, maxv):
print "Processing ", fname
dat, attribs, cf = bu.getdata(os.path.join(path, fname))
if (len(attribs) > 0):
fsamp = attribs["Fsamp"]
xdat = dat[:, data_columns[1]]
lentrace = len(xdat)
## zero pad one cycle
MinusDC = dat[:, drive_column]
corr_full = bu.corr_func(MinusDC - np.median(MinusDC), xdat, fsamp, fdrive)
#plt.figure()
#plt.plot( xdat)
#plt.plot(dat[:,drive_column])
#plt.show()
return corr_full[0], np.max(corr_full)
path = "/data/20140620/xtalk"
channels = np.array([1, 2, 3, 4, 5, 6])
drive_column = -1
dat, attribs, f = bu.getdata(os.path.join(path, fname))
corrs = []
corrmaxs = []
for i in range(7):
if i != drive_column:
corrmax = np.max(
bu.corr_func(dat[:, drive_column],
dat[:, i],
attribs['Fsamp'],
attribs['drive_freq'],
filt=True))
sub_corrs = bu.corr_blocks(dat[:, drive_column],
dat[:, i],
attribs['Fsamp'],
attribs['drive_freq'],
filt=True,
N_blocks=3)
corrs.append(sub_corrs)
corrmaxs.append(corrmax)
plt.figure()
corrs = np.array(corrs)
drive -= np.mean(drive)
drive2 = drive**2
drive2 -= np.mean(drive2)
if (False):
plt.figure()
plt.plot(drive)
#plt.plot(response)
b, a = sig.butter(3, (2.0 * drive_freq + 5) / (Fs / 2.0))
filt_resp = sig.filtfilt(b, a, response)
plt.plot(filt_resp, 'c')
plt.plot(drive2)
plt.show()
dummy_freq = 41 ## freq doesn't matter since we do 0 offset only
corr_dr = bu.corr_func(drive, response, Fs, dummy_freq)[0]
corr_dr2 = bu.corr_func(drive2, response, Fs, dummy_freq)[0]
tot_corr_dr.append(corr_dr)
tot_corr_dr2.append(corr_dr2)
tot_dat[fidx, 0] = cpos
tot_dat[fidx, 1] = np.mean(tot_corr_dr)
tot_dat[fidx, 2] = np.mean(tot_corr_dr2)
tot_dat[fidx, 3] = np.std(tot_corr_dr) / np.sqrt(len(tot_corr_dr))
tot_dat[fidx, 4] = np.std(tot_corr_dr2) / np.sqrt(len(tot_corr_dr2))
tot_dat_pts.append([cpos, tot_corr_dr, tot_corr_dr2])
if (False and elec == 0):
plt.figure()
plt.plot(drive)
fname = r"allon_drive_pxi6259_chan0_10000mV_41Hz.h5"
path = "/data/20140620/xtalk"
channels = np.array([1, 2, 3, 4, 5, 6])
drive_column = -1
dat, attribs, f = bu.getdata(os.path.join(path, fname))
corrs = []
corrmaxs = []
for i in range(7):
if i != drive_column:
corrmax = np.max(bu.corr_func(dat[:, drive_column], dat[:, i], attribs['Fsamp'], attribs['drive_freq'], filt = True))
sub_corrs = bu.corr_blocks(dat[:, drive_column], dat[:, i], attribs['Fsamp'], attribs['drive_freq'], filt = True, N_blocks = 3)
corrs.append(sub_corrs)
corrmaxs.append(corrmax)
plt.figure()
corrs = np.array(corrs)
plt.errorbar(range(len(corrs[:, 0])), corrs[:, 0], yerr = corrs[:, 1], fmt = 'o')
plt.errorbar(range(len(corrs[:, 0])), np.array(corrmaxs), yerr = corrs[:, 1], fmt = 'o' )
plt.show()
## take correlation with drive and drive^2
response = cdat[skip_dat:-skip_dat, dcol]
response -= np.mean(response)
if (elec < 3):
drive = cdat[skip_dat:-skip_dat, elec + 8]
else:
drive = cdat[skip_dat:-skip_dat, elec + 9]
if (cpos < 0 and gain == 0.2): drive *= -1
drive -= np.mean(drive)
drive2 = drive**2
drive2 -= np.mean(drive2)
dummy_freq = 38 ## freq doesn't matter since we do 0 offset only
corr_dr = bu.corr_func(drive, response, Fs, dummy_freq)[0]
corr_dr2 = bu.corr_func(drive2, response, Fs, dummy_freq)[0]
## find the error on the correlation from a few dummy frequencies
tvec = np.linspace(0, (len(response) - 1.0) / Fs, len(response))
rand_phase = np.random.rand() * 2. * np.pi
s_m5 = np.sin(2 * np.pi * tvec * (drive_freq - 5.0) + rand_phase)
s_p5 = np.sin(2 * np.pi * tvec * (drive_freq + 5.0) + rand_phase)
corr_dr_m5 = bu.corr_func(s_m5, response, Fs, dummy_freq)[0]
corr_dr_p5 = bu.corr_func(s_p5, response, Fs, dummy_freq)[0]
## upper bound on the error
if (False and elec == 3):
plt.figure()
plt.plot(drive)
#plt.plot(response)
def getdata(fname, maxv, ang, gain):
print "Processing ", fname
dat, attribs, cf = bu.getdata(os.path.join(path, fname))
## make sure file opened correctly
if( len(dat) == 0 ):
return {}
dat[:, drive_column] *= gain
if( len(attribs) > 0 ):
fsamp = attribs["Fsamp"]
drive_amplitude = attribs["drive_amplitude"]
xdat, ydat = rotate_data(dat[:,data_columns[0]], dat[:,data_columns[1]], ang)
drive_amp = np.sqrt(2)*np.std(dat[:,drive_column])
if( remove_laser_noise ):
laser_good = bu.laser_reject(dat[:, laser_column], 60., 90., 4e-6, 100, fsamp, False)
else:
laser_good = np.ones_like(dat[:, laser_column]) > 0
corr_full = bu.corr_func(dat[:,drive_column], xdat, fsamp, fdrive, good_pts=laser_good)
corr = corr_full[ maxv ]
corr_max = np.max(corr_full)
corr_max_pos = np.argmax(corr_full)
xpsd, freqs = matplotlib.mlab.psd(xdat, Fs = fsamp, NFFT = NFFT)
#ypsd, freqs = matplotlib.mlab.psd(ydat, Fs = fsamp, NFFT = NFFT)
max_bin = np.argmin( np.abs( freqs - fdrive ) )
ref_bin = np.argmin( np.abs( freqs - fref ) )
## also correlate signal with drive squared
dsq = dat[:,drive_column]**2
dsq -= np.mean(dsq)
sq_amp = np.sqrt(2)*np.std( dsq )
## only normalize by one factor of the squared amplitude
corr_sq_full = bu.corr_func(dsq*sq_amp, xdat, fsamp, fdrive)
corr_sq_max = np.max(corr_sq_full)
corr_sq_max_pos = np.argmax(corr_sq_full)
xoff = sp.filtfilt(boff, aoff, xdat)
if(False):
plt.figure()
plt.plot( xdat )
plt.plot( dat[:, drive_column] )
plt.figure()
plt.plot( corr_full )
plt.show()
ctime = attribs["time"]
## is this a calibration file?
cdir,_ = os.path.split(fname)
is_cal = cdir == cal_path
curr_scale = 1.0
## make a dictionary containing the various calculations
out_dict = {"corr_t0": corr,
"max_corr": [corr_max, corr_max_pos],
"max_corr_sq": [corr_sq_max, corr_sq_max_pos],
"psd": np.sqrt(xpsd[max_bin]),
"ref_psd": np.sqrt(xpsd[ref_bin]),
"temps": attribs["temps"],
"time": bu.labview_time_to_datetime(ctime),
"num_flashes": attribs["num_flashes"],
"is_cal": is_cal,
"drive_amp": drive_amp}
cf.close()
return out_dict
def getdata(fname, maxv, ang, gain):
print "Processing ", fname
dat, attribs, cf = bu.getdata(os.path.join(path, fname))
## make sure file opened correctly
if (len(dat) == 0):
return {}
dat[:, drive_column] *= gain
if (len(attribs) > 0):
fsamp = attribs["Fsamp"]
drive_amplitude = attribs["drive_amplitude"]
xdat, ydat = rotate_data(dat[:, data_columns[0]], dat[:, data_columns[1]],
ang)
drive_amp = np.sqrt(2) * np.std(dat[:, drive_column])
if (remove_laser_noise):
laser_good = bu.laser_reject(dat[:, laser_column], 60., 90., 4e-6, 100,
fsamp, False)
else:
laser_good = np.ones_like(dat[:, laser_column]) > 0
corr_full = bu.corr_func(dat[:, drive_column],
xdat,
fsamp,
fdrive,
good_pts=laser_good)
corr = corr_full[maxv]
corr_max = np.max(corr_full)
corr_max_pos = np.argmax(corr_full)
xpsd, freqs = matplotlib.mlab.psd(xdat, Fs=fsamp, NFFT=NFFT)
#ypsd, freqs = matplotlib.mlab.psd(ydat, Fs = fsamp, NFFT = NFFT)
max_bin = np.argmin(np.abs(freqs - fdrive))
ref_bin = np.argmin(np.abs(freqs - fref))
## also correlate signal with drive squared
dsq = dat[:, drive_column]**2
dsq -= np.mean(dsq)
sq_amp = np.sqrt(2) * np.std(dsq)
## only normalize by one factor of the squared amplitude
corr_sq_full = bu.corr_func(dsq * sq_amp, xdat, fsamp, fdrive)
corr_sq_max = np.max(corr_sq_full)
corr_sq_max_pos = np.argmax(corr_sq_full)
xoff = sp.filtfilt(boff, aoff, xdat)
if (False):
plt.figure()
plt.plot(xdat)
plt.plot(dat[:, drive_column])
plt.figure()
plt.plot(corr_full)
plt.show()
ctime = attribs["time"]
## is this a calibration file?
cdir, _ = os.path.split(fname)
is_cal = cdir == cal_path
curr_scale = 1.0
## make a dictionary containing the various calculations
out_dict = {
"corr_t0": corr,
"max_corr": [corr_max, corr_max_pos],
"max_corr_sq": [corr_sq_max, corr_sq_max_pos],
"psd": np.sqrt(xpsd[max_bin]),
"ref_psd": np.sqrt(xpsd[ref_bin]),
"temps": attribs["temps"],
"time": bu.labview_time_to_datetime(ctime),
"num_flashes": attribs["num_flashes"],
"is_cal": is_cal,
"drive_amp": drive_amp
}
cf.close()
return out_dict
plt.loglog(cfreqs, cpsd, color=col_list[j])
plt.plot(cfreqs[cidx],
cpsd[cidx],
'o',
color=col_list[j],
linewidth=1.5)
if (plot_freqs): plt.show()
drive_freq_list = np.array(drive_freq_list)
print "Drive freqs are: ", drive_freq_list[:, 0]
for didx, dcol in enumerate(data_cols):
for eidx, ecol in enumerate(electrode_cols):
corr_full = bu.corr_func(dat[:, ecol], dat[:, dcol], fsamp,
drive_freq_list[eidx][0])
dc_val = np.mean(dat[:, ecol])
corr_mat[fidx, didx, eidx, :] = [dc_val, corr_full[0]]
## now plot total correlation
for eidx, ecol in enumerate(electrode_cols):
plt.figure()
plt.subplot(3, 1, 1)
plt.plot(corr_mat[:, 0, eidx, 0],
corr_mat[:, 0, eidx, 1],
color=col_list[eidx])
plt.title("Electrode %d, X" % eidx)
plt.subplot(3, 1, 2)
