matlab - How many iterations should you make for the simulation to be a 'Monte Carlo simulation' for BER calculations? -

edited question

how many iterations should make simulation accurate 'monte carlo simulation' bit error rate calculations?

what minimum value? if want repeat simulation exponentially growing number 5 times? should start 1e2 thus>> iterations = [1e2 1e3 1e4 1e5 1e6] or 1e3 >> [1e3 1e4 1e5 1e6 1e7]? or else? common practice?

additional info: used [8e3 1e4 3e4 5e4 8e4 1e5] before not enough according prof. because result not satisfactory.

simulations take long time on computer cannot keep changing iterations based on result. if there common practice this, please let me know.

thanks @billbokeey helping me edit question.

what professor propose strikes me qualitative, not quantitative way estimate convergence of simulation.

frankly, don't know how ber computed, deal lot integral calculations mc.

in such case sample x_i on interval , compute f_mc = s_i f_i / n, s denotes summation. know f_mc converge true value variance of sigma²/n (or std.deviation of sigma/sqrt(n)). then, compute in same simulation estimation of sigma, assume large enough n approximation of sigma , simulation error plotted. in practical terms alongside f_mc compute second momentum sum , average f²_mc = s_i f²_i / n, , @ end s=sqrt(f²_mc - (f_mc)²)/sqrt(n) estimated error of mc simulation (it bit biased though).

thus plot on same graph value of ber , statistical error of simulation. better - ask user input required statistical error (say, in %, meaning user enters s/f*100), , continue simulation in bunches till reach required precision.

then judge if 10⁹ points enough or not...