How to resolve reshape error

Hi There,

I previously asked a question about this piece of code with QPSK however, when using 64-QAM i get an error with the reshape value. The code is as follows:

N = 10e6;
M = 64;
k = log2(M);
bit_stream = randi([0,1],[1,N]);
tx_bits = reshape(bit_stream,k,[]);
tx_symbols = sum(tx_bits.*pow2(k-1:-1:0).',1);
qammod_sig = qammod(tx_symbols,M);
snr = -10:10;
Nsnr = numel(snr);
error_rate = zeros(1,Nsnr);
n_symb = N/k;
for ii = 1:Nsnr
    rx_sig = awgn(qammod_sig,snr(ii));
    qamdemod_symb = qamdemod(rx_sig,M);
    error_rate(ii) = nnz(tx_symbols ~= qamdemod_symb) / n_symb;
end
scatterplot(qammod_sig)
figure()
semilogy(snr,error_rate(1,:),'.-')
xlabel ('SNR [dB]')
ylabel('BER')

The error which I am receiving is this:

Error using reshape
Product of known dimensions, 6, not divisible into total
number of elements, 10000000.

Error in coursework_task_1_qam (line 5)
tx_bits = reshape(bit_stream,k,[]);

[Dyuman Joshi]

bit_stream is 1x10e6. You are trying to reshape it into an array with 6 rows and 10e6/6 columns, which is not possible as number of columns must be natural numbers.

What is the idea behind reshaping bit_stream? What is the operation you are trying to perform?

[Liam]

The main aim that I am trying to achieve to observe the bit error rate of a system using QPSK, 16-QAM and 64-QAM. I have had success with QPSK and 16-QAM but not with 64-QAM and I am unsure as to how to get it to work

Answer 1

[Voss]

Pick an N that's a multiple of k, or adjust N to be a multiple of k in the code, or append enough bits onto the end of the bit stream to make it a multiple of k in length before reshaping.

N = 1e6; % pick an N that's a multiple of k
M = 64;
k = log2(M);

N = ceil(N/k)*k; % or make N the first multiple of k >= N

bit_stream = randi([0,1],[1,N]);

m = mod(N,k);
if m
    bit_stream(end+1:end+k-m) = 0; % or append some zeros to the bit stream as necessary
end

tx_bits = reshape(bit_stream,k,[]);
tx_symbols = sum(tx_bits.*pow2(k-1:-1:0).',1);
qammod_sig = qammod(tx_symbols,M);
snr = -10:10;
Nsnr = numel(snr);
error_rate = zeros(1,Nsnr);
n_symb = N/k;
for ii = 1:Nsnr
    rx_sig = awgn(qammod_sig,snr(ii));
    qamdemod_symb = qamdemod(rx_sig,M);
    error_rate(ii) = nnz(tx_symbols ~= qamdemod_symb) / n_symb;
end
scatterplot(qammod_sig)

figure()
semilogy(snr,error_rate(1,:),'.-')
xlabel ('SNR [dB]')
ylabel('BER')

[Liam]

The main issue with this is that the brief that I have been given states that the bitstream needs to be 10^6

[Voss]

Let me know how you plan to make a sequence of symbols that are each 6 bits in length and end up with a million bits in total, and I'll adjust my code.

I've edited my answer to show how to make the number of bits generated the next multiple of k at or above the specified N, and to show how to append bits to the end of the bit stream so that its length is a multiplr of k. You do not need to use more than one of those methods (but it doesn't hurt anything if you do).

[Liam]

Would padding it with zeros assist with this? or does this complicate things?

[Voss]

Padding with zeros works. See my answer for how to do that. Specifically, this part:

m = mod(N,k);
if m
  bit_stream(end+1:end+k-m) = 0; % or append some zeros to the bit stream as necessary
end

[Dyuman Joshi]

If OP had R2023b, resize would have been useful.