Kam, Tight bounds on the bit-error probabilities of 2DPSK and 4DPSK in nonselective Rician fading,IEEE Trans. Versions of Matlab/toolboxes), appropriate modifications to be done. For other spectrum (I read in the Mathworks website that other spectrums can be configured in later In this only Jakes spectrum can be compared. It would be nice if Mathworks can modify (may be in the next release) berfading such that, the theoretical BER expected take care of such fading characteristics (in this case Doppler). The latter result is in close unison with the simulation as well. The pink (Pb_doppler_theory) is the approximate and more correct bound, for Jakes spectrum (See, , ). Blue curve is what Matlab ‘berfading’ gives. The BER curves are shown in the figure, attached with. *(1+ (EbN0.*(1-besselj(0,2*pi*Ts*fd)) )) The simulation results closely match this result, as against the berfading result “berfading(SNR,’dpsk’,M,1)” For DBPSK, the exact BER, considering maximum Doppler frequency, for a uniform scattering model (Jakes spectrum) can be better approximated to Pb=(0.5./(1+EbN0) ). Just to illustrate the point, I am enclosing here, a snapshot of the comparative BER results. Closer approximations of the average bit error probability are derived in closed form for some of the differential schemes (say DBPSK, DQPSK etc). There are theoretical results (not closed form) to reflect such impacts. In the presence of Doppler, there may be an irreducible error floor for certain modulation schemes, especially differential modulation schemes. Currently, for example, the berfading gives out the BER for Rayleigh (or Rice) fading channel, without considering the impact of Doppler spread. I believe the theoretical BER under fading scenarios, provided by the ‘berfading’ function in Communication toolbox is not quite accurate, for frequency flat fading, when Doppler shift to be considered. Unfortunately, I couldn’t trace their Email reply since it was sent to my earlier official ID, which has changed! Hopefully, I will ping Mathworks again and get this up. I wonder why! Anyway, here is the Email I had sent to them. I did get an acknowledgment from them saying that they will incorporate this into the future version. It was in 2006, I guess, I had written a mail to Mathworks on this topic. If you ever need to grab a copy, please be aware that, the ownership of this is with them. Since I may need it once in a while, I’ll keep a copy here for my own selfish quick reference. The frequency domain characteristic of the popular window functions are now in here for reference, thanks to Marcel Müller. These days, doing a Wikipedia search to find a temporary fix and then move on is the adopted, yet not entirely satisfying strategy. What window function is more suitable is something that I’ve never mastered. Usually, some windowing as well is used to adjust the mean square versus resolution. For ages, I was using the pwelch (now elch) function (pwelch method to estimate the spectrum) to compute/plot the spectrum in Matlab. I hit upon this yet again while trying to map the theoretical spectrum to spectrum computed from discrete samples (for an OFDM modulated signal) and then to analog spectrum measured by an instrument (By the way, I figured out that, the closest discrete technique which maps to the spectrum computation in spectrum analyzer is the Danielle method, which for some reason is not there in Matlab!). The hardest one was to remember what is what and how they looked in frequency domain (the time domain view was something thankfully I recollect from the names). Windowing techniques have always offered confusion to me.
0 kommentar(er)
