Create ifo_test.m

matlab/updated_scripts/ifo_test.m

@@ -0,0 +1,93 @@
+% This script is for testing how to detect integer frequeny offsets (IFO) in OFDM
+
+clear all;
+
+%% Parameters
+sample_rate = 15.36e6;
+fft_size = get_fft_size(sample_rate);
+carrier_spacing = sample_rate / fft_size;
+data_carriers = 600;
+starting_zeros = 100;
+cp_len = 80;
+frac_frequency_offset = 0e3;
+carrier_offset = -1;
+freq_offset = carrier_spacing * carrier_offset;
+cp_backoff = 0;
+max_allowed_int_offset = 5;
+enable_plots = 0;
+snr = 15;
+
+if (abs(carrier_offset) > max_allowed_int_offset)
+    warning("TEST WILL FAIL!!  OFFSET TOO HIGH");
+end
+
+%% Get Data Carrier Indices
+left_carriers = (fft_size / 2) - (data_carriers / 2);
+right_carriers = left_carriers - 1;
+carrier_indices = (left_carriers + 1):(left_carriers + 1 + data_carriers);
+
+%% Create OFDM Symbol
+
+% Generate ZC sequence and zero the middle sample (will become FFT DC)
+zc_seq = zadoffChuSeq(600, 601);
+zc_seq((data_carriers / 2) + 1) = 0;
+
+% Build the FFT freq domain by placing the ZC sequence in the middle with
+% the nulled out ZC sample at DC
+golden_freq_domain = reshape([zeros(left_carriers, 1); zc_seq; zeros(right_carriers, 1)], 1, []);
+
+% Convert to time domain
+golden_time_domain = ifft(golden_freq_domain);
+
+% Create and prepend the cyclic prefix
+cp = golden_time_domain(end-cp_len+1:end);
+ofdm_symbol = [cp, golden_time_domain];
+
+%% Create Sample Vector with Burst
+
+% Create a "burst" with zeros on either side
+samples = [zeros(1, starting_zeros), ofdm_symbol, zeros(1, starting_zeros)];
+
+% Add some noise
+samples = awgn(samples, snr, 'measured');
+
+if (enable_plots)
+    figure(3);
+    plot(10 * log10(abs(samples).^2));
+end
+
+%% Apply Integer Frequency Offset
+
+% Generate the vector that will be used to rotate each sample
+freq_offset_vector = exp(1j * 2 * pi * freq_offset / sample_rate * [1:length(samples)]);
+samples = samples .* freq_offset_vector;
+
+%% Apply Fractional Frequency Offset
+frac_offset_vector = exp(1j * 2 * pi * frac_frequency_offset / sample_rate * [1:length(samples)]);
+samples = samples .* frac_offset_vector;
+
+%% Find Starting Sample Index
+
+% Use autocorrelation to search for the start of the ZC sequence.  This
+% takes advantage of the fact that the ZC sequence is symmetrical about the
+% center in both the time and frequency domains.  Each iteration of the
+% loop below will look at one `fft_size` window of samples sliding to the
+% right by one sample on each iteration.  The window is then split in two
+% with the second half fliped so that it's backwards.  These two windows
+% are then correlated with each other, and the score saved off
+sto_search_window = fft_size;
+sto_scores = zeros(1, starting_zeros * 2);
+for td_idx=1:length(sto_scores)
+    sto_window = samples(td_idx:td_idx + fft_size - 1);
+    sto_left_window = sto_window(1:(fft_size / 2));
+    sto_right_window = sto_window((fft_size / 2) + 1:end);
+    sto_scores(td_idx) = xcorr(sto_left_window, fliplr(sto_right_window), 0, 'normalized');
+end
+
+if (enable_plots)
+    figure(1);
+    plot(abs(sto_scores)); title('STO Scores');
+end
+
+% Find the index of the maximum value from the autocorrelation above
+[~, sto_est] = max(abs(sto_scores));