david/python-ofdm
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Why study for finals when you can work on a project that isn't getting anywhere :)

Browse Source 
Started switching to pilot symbols instead of subcarriers.
Still need to reconcile the channel estimation.
This commit is contained in:
David Lenfesty 2019-12-15 19:31:25 -07:00
parent 047b83533f
commit 906763055f
2 changed files with 32 additions and 55 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

73
channel.py
View File

@ -4,16 +4,16 @@ import scipy.interpolate
import matplotlib.pyplot as plt
# Dirac delta function response, no change
#channel_response = np.array([0, 0, 0, 1, 0, 0, 0])
channel_response = np.array([1, 0, 0.3+0.3j])
channel_response = np.array([0, 0, 0, 1, 0, 0, 0])
#channel_response = np.array([1, 0, 0.3+0.3j])
# How do I sync this across two files?
# figure it out later
pilot_value = 5 + 5j
pilot_value = 1 + 1j
# 15dB seems to be around the minimum for error-free transmission
snr_db = 300
snr_db = 70
def sim(in_data):
"""
@ -69,60 +69,29 @@ def estimate(in_data, pilots=0):
Actually maybe not? Looked at a paper, check the drive.
"""
all_carriers = np.arange(len(in_data[0]))
carrier_symbol = in_data[0]
if pilots > 0:
pilot_carriers = all_carriers[::(len(all_carriers)) // pilots]
pilot_carriers = np.delete(pilot_carriers, 0)
H_est = 0
H = np.ndarray((len(carrier_symbol)), dtype=np.csingle)
# Obtain estimate for each sub carrier
for i in range(len(carrier_symbol)):
H[i] = carrier_symbol[i] / np.csingle(1 + 1j)
print(H)
# start averaging
#H_est = 0
# Exact channel response
H_exact = np.fft.fft(channel_response, 64)
# for i in range(len(in_data)):
# # Obtain channel response at pilot carriers
# H_est_pilots = in_data[i][pilot_carriers] / pilot_value
plt.plot(range(64), np.real(H), "b-")
plt.plot(range(64), np.real(H_exact), "r")
plt.show(block=True)
plt.plot(range(64), np.imag(H), "b-")
plt.plot(range(64), np.imag(H_exact), "r")
plt.show(block=True)
# # Interpolate estimates based on what we get from the few pilot values
# H_est_abs = scipy.interpolate.interp1d(pilot_carriers, abs(H_est_pilots), kind='linear', fill_value='extrapolate')(all_carriers)
# H_est_phase = scipy.interpolate.interp1d(pilot_carriers, np.angle(H_est_pilots), kind='linear', fill_value='extrapolate')(all_carriers)
return H
# # Take angular form and turn into rectangular form
# H_est += H_est_abs * np.exp(1j*H_est_phase)
H_est_pilots = in_data[0][pilot_carriers] / pilot_value
# Interpolate estimates based on what we get from the few pilot values
H_est_abs = scipy.interpolate.interp1d(pilot_carriers, abs(H_est_pilots), kind='linear', fill_value='extrapolate')(all_carriers)
H_est_phase = scipy.interpolate.interp1d(pilot_carriers, np.angle(H_est_pilots), kind='linear', fill_value='extrapolate')(all_carriers)
# Take angular form and turn into rectangular form
H_est = H_est_abs * np.exp(1j*H_est_phase)
# for an average channel estimate
# H_est = H_est / len(in_data)
# Exact channel response
H_exact = np.fft.fft(channel_response, len(in_data[0]))
plt.plot(all_carriers, np.real(H_exact), "b")
plt.plot(all_carriers, np.real(H_est), "r")
plt.plot(pilot_carriers, np.real(H_est_pilots), "go")
plt.show(block=True)
plt.plot(all_carriers, np.imag(H_exact), "b")
plt.plot(all_carriers, np.imag(H_est), "r")
plt.plot(pilot_carriers, np.imag(H_est_pilots), "go")
plt.show(block=True)
print(H_est_pilots)
print(H_exact[pilot_carriers])
return H_est
else:
return -1
def equalize(in_data, H_est):
out_data = np.ndarray((len(in_data), len(in_data[0])), dtype=np.csingle)

14
main.py
View File

@ -78,7 +78,15 @@ if __name__ == '__main__':
parallel = parallelise(64, bytes)
# modulate data with a QAM scheme
modulated = qam.modulate(parallel, pilots=10)
modulated = np.array(qam.modulate(parallel, pilots=0))
# Insert known "start" symbol
start_symbol = np.zeros((1,64), dtype=np.csingle)
for i in range(64):
start_symbol[0][i] = 1 + 1j
modulated = np.append(start_symbol, modulated, axis=0)
# Run IFFT to get a time-domain signal to send
ofdm_time = np.fft.ifft(modulated)
@ -98,13 +106,13 @@ if __name__ == '__main__':
to_equalize = np.fft.fft(ofdm_cp_removed)
# Find an estimate for channel effect
H_est = channel.estimate(to_equalize, pilots=10)
H_est = channel.estimate(to_equalize, pilots=0)
# Equalise based on estimated channel
to_decode = channel.equalize(to_equalize, H_est)
# Demodulate symbol into output data
to_serialise = qam.demodulate(to_decode, pilots=10)
to_serialise = qam.demodulate(to_decode, pilots=0)
# Turn data back into string
data = serialise(64, to_serialise)