Task description and Expected results
In this task we will develop a RF front-end based on a hybrid Filter Bank (FB) to acquire signals with a large bandwidth and high
dynamic range. A front-end with these characteristics will be a key component to build a Software Defined Radio, where the
whole spectrum of interest must be accessible in the digital format.
Moreover, a C ognitive Radio must be able to deal with a large bandwidth even if it uses only a small part of the converted
spectrum. After sensing the spectrum for free holes, the RF signal can be a sparse set of narrow bands scattered over a large
bandwidth.
This kind of signals will be quite difficult to acquire due to the following reasons:
1- The ADC must work at a high sampling rate. The radio front-end must convert to digital the whole bandwidth starting on the
hole with the lowest frequency and ending in the hole with the largest one.
2-The dynamic range of the ADC must be very high. When the bandwidth of the signal is increased its dynamic range will also
increase. Besides, the interference signals between the holes will also be included in the conversion. These interfere signals can
have higher power causing the well-known near-far problem demanding for higher dynamic range of the ADC .
Building an ADC with the required sampling rate and dynamic range is a challenge far beyond the present technology.
One possible solution to this problem can be a front-end with a large number of ADC s using a lower sampling rate and number
of bits in conjunction with an analog FB. We call this solution the cochlear radio due to the similarity with the solution found in the
Human hearing, where an impressive bandwidth and dynamic range is achievable using analog to discrete time sensors with
poor characteristics.
In such a system, the input signal will be split in many frequency bands using an analog FB of band-pass filters. Each filter output
will be converted to the digital domain using sub-sampling. We can use low performance ADC s and reduce the power
consumption over the one ADC design by a factor equal to the number of channels. Then, the digital signals are aggregated into
one single signal using a multirate digital synthesis FB. This FB will be responsible to cancel the aliasing terms due to the subsampling
and equalize the phase and amplitude distortion introduced by the analog filters.
There are several problems to study in order to build this kind of front-end based on a hybrid FB:
1- The design of the synthesis digital FB is based on the correct measurement of the analog FB. Some kind of measurement
technique or blind equalization must be developed in order to measure the filters in place.
2- On previous research we have identified that the hybrid FB can be very sensitive to measurement errors on the analog filters.
Further research should be performed.
3- With a C ognitive Radio, the received signal can be scattered over several spectrum holes. The Hybrid FB can take advantage
of this fact to reduce the number of samples of the digital synthesis filters by optimizing the frequency response of the FB only
over the bands of the holes. In this case, some of the ADC s could be redundant and turned off, reducing even more the power
consuption.
4- The structure of the hybrid filter bank make it a good candidate to integrate in its design spectrum aggregation. We believe
that some of the problems, such as the high PAPR can be solved by performing the spectrum aggregation using a hybrid filter
bank.
5- It can be advantageous to integrate the synthesis FB with the demodulation block. We intend to study this aspect.
6- The size of the analog FB is a key aspect in order to have a useful hybrid filter bank. This is a challenging problem and new
filter design techniques should be found. C eramic filters are good candidates to achieve the required miniaturization.
Mem bers of the research team in this task: (BI) Bolseiro de Investigação (Mestre) 4; Ana Maria Perfeito Tome; José Manuel Neto Vieira; Teófilo José Marques Monteiro;
