95 lines
2.9 KiB
Python
95 lines
2.9 KiB
Python
import numpy as np
|
|
from scipy.spatial.distance import euclidean
|
|
|
|
pilot_value = 1 + 1j
|
|
|
|
qam_mapping_table = {
|
|
0 : 1 + 1j,
|
|
1 : -1 + 1j,
|
|
2 : -1 - 1j,
|
|
3 : 1 - 1j
|
|
}
|
|
|
|
qam_demapping_table = { x : y for y, x in qam_mapping_table.items() }
|
|
|
|
def modulate(in_data, pilots=0):
|
|
"""
|
|
Modulates into 4-QAM encoding, might change that number later.
|
|
|
|
This turns input data into a "constellation" of complex numbers,
|
|
ready to be fed into an IFFT. This constellation could also be used
|
|
directly with an IQ modulator.
|
|
|
|
Parameters:
|
|
in_data - m X n array, m symbols to run on
|
|
pilots (optional) - number of pilot signals to intersperse into carriers
|
|
|
|
Output:
|
|
data
|
|
"""
|
|
|
|
num_data_carriers = len(in_data[0])
|
|
|
|
all_carriers = np.arange(num_data_carriers + pilots, dtype=int)
|
|
|
|
if pilots > 0:
|
|
pilot_carriers = all_carriers[::(num_data_carriers + pilots)//pilots]
|
|
pilot_carriers = np.delete(pilot_carriers, 0) # not sure how to not have this line
|
|
data_carriers = np.delete(all_carriers, pilot_carriers)
|
|
else:
|
|
data_carriers = all_carriers
|
|
|
|
|
|
#initialise output array with additional pilot carriers as well
|
|
out_data = np.ndarray((len(in_data), num_data_carriers + pilots), dtype=np.csingle)
|
|
|
|
for i in range(len(in_data)):
|
|
data_index = 0
|
|
for carrier_index in data_carriers:
|
|
out_data[i][carrier_index] = qam_mapping_table[in_data[i][data_index]]
|
|
data_index += 1
|
|
|
|
if pilots > 0:
|
|
for j in pilot_carriers:
|
|
# Value for pilot carriers
|
|
out_data[i][j] = pilot_value
|
|
|
|
|
|
return out_data
|
|
|
|
def demodulate(in_data, pilots=0):
|
|
"""
|
|
Demodulates incoming signal.
|
|
"""
|
|
|
|
all_carriers = np.arange(len(in_data[0]), dtype=int)
|
|
|
|
if pilots > 0:
|
|
pilot_carriers = all_carriers[::(len(all_carriers)) // pilots]
|
|
pilot_carriers = np.delete(pilot_carriers, 0) # not sure how to not have this line
|
|
data_carriers = np.delete(all_carriers, pilot_carriers)
|
|
else:
|
|
data_carriers = all_carriers
|
|
|
|
out_data = np.ndarray((len(in_data), len(data_carriers)), dtype=np.uint8)
|
|
|
|
# Just pull the constellation array data out
|
|
constellation = [ x for x in qam_demapping_table.keys() ]
|
|
|
|
for i in range(len(in_data)):
|
|
data_index = 0
|
|
for carrier_index in data_carriers:
|
|
distances = np.ndarray((len(constellation)), dtype=np.single)
|
|
|
|
# Here we have to map to the closest constellation point,
|
|
# because floating point error
|
|
for k in range(len(constellation)):
|
|
distances[k] = euclidean(in_data[i][carrier_index], constellation[k])
|
|
|
|
# output is the index of the constellation, essentially
|
|
# this may have to change if I want to generalise
|
|
out_data[i][data_index] = np.argmin(distances)
|
|
data_index += 1
|
|
|
|
return out_data
|