Conclusions and recommendations
Conclusions
In this thesis a frame differential autocorrelation receiver (FD-ACR) for UWB communi-cations has been studied. More precisely we analyzed its behavior when it operates in a single-user and multi-user environment: the analysis produced results in terms of perfor-mance and underlined the main advantages and drawbacks of the system.
In particular the statistical analysis, derived for the AWGN channel and a generic monocy-cle, produced an exact equation for the BER performances. The bit rate equations for the single-user and multi-user case were also determined. An accurate study of these equations pointed out the relations existing between performances (bit rate and BER) and the system parameters.
Single-user case
In the single-user case, we found out that both the bit rate and the BER are defined by the processing gain of the system. More precisely we saw that the BER is influenced also by the particular ratio Np
Nf and that we should choose the highest Np
Nf to obtain the best
results. Actually this dependence becomes negligible for high processing gain. Moreover the decision about the ratio Np
Nf should take into consideration some technological aspects that are
influenced by that term. In comparison with an analogous receiver, the symbol differential autocorrelation receiver (SD-ACR), the FD-ACR has a lower complexity, but also a worse BER performance because of an increased signal-on-noise term which cannot be controlled just with the use of a proper polarity code. Moreover we saw that, in order to decrease complexity, it is possible to consider a simple scheme of receiver with only one branch (and obviously cardinality of Dk equal to one). This does not cause a worsening in the BER
performance. In the end we concluded that the use of a noisy template causes a significative deterioration of the BER therefore we should find a solution to this problem in order to obtain better results.
Multi-user case
The analysis of the multi-user case pointed out that performances are deeply influenced by the presence of more users in the same system. In particular we found out that bit rate and BER are influenced by the processing gain and by the number of users. Similarly to the single-user case we observed a slight dependence of the BER on the particular Np
Nf
and we found out that a high Np
Nf guarantees the best BER performance. Actually this
2
dependence becomes quickly negligible when we increase the processing gain. Moreover, similar to the single-user case, the decision about the ratio Np
Nf should take into consideration
some realization aspects that are influenced by that term. The analysis of the BER under different conditions shows the strong effect due to interference which causes a floor in the BER curve. In particular accidental collisions between pulses from different signals represent the main cause of deterioration while the unexpected autocorrelations of interfering signals produce a lower disturbance. Therefore solutions to the interference problems have to be investigated in order to improve BER performance and obtain a feasible multi-user system using a FD-ACR. In the end it was shown that such a system has scaling problems: a doubling in the number of users causes a degradation in the BER performances which is not compensated by doubling the processing gain.
Recommendations
Single-user case
We proposed the use of a device that produces an averaged template in the receiver in order to face the problems arising because of the use of a noisy template. A proper analysis of it should be made in order to design a scheme whose complexity is not too high. An easier solution to obtain a better BER is represented by the use of a proper polarity code instead of a simple block repetition code: notice that such a solution would guarantee an improved BER at an acceptable cost and without modifying the bit rate.
Multi-user case
First of all we can observe that the causes of degradation of the BER are various therefore a few possibilities have to be considered. In order to address the disturbance due to the use of a noisy template we proposed, in the same way as in the single-user case, the use of an averaged template. Unfortunately this does not produce any positive effect against the problem of interference. In order to address that problem we proposed the use of orthogonal codes. This could represent an easy solution to cancel the effect due to accidental autocorrelations. In order to face the problem of collisions, which represents the stronger disturbing cause, we proposed the design of a code with a low crosscorrelation between its sequences. Finally we have to say that the use of more efficient schemes like decorrelating detectors or minimum mean square detectors could lead both to a better BER performance and more favorable scaling features without a strong increase in complexity.