Combined FRM and GDFT filter bank designs for improved nonuniform DSA channelisation

FPGA based FRM GDFT filter banks

“…In [6,7] the FRM technique was used to reduce the prototype filter length of the GDFT-FB. The prototype filter () Hz of the GDFT-FB is first expressed in FRM form as:…”

“…In the alternative FRM configuration, the base filter can be moved to the output side of the GDFT-FB in order to operate at the lower sub-band output sample rate [6]. Consequently, the FRM interpolation factor L will be applied in a decimated form and this reduces the zero padding and sample delay required.…”

“…The most efficient form is the alternative narrowband FRM GDFT-FB in which only the positive branch (the upper branch in Figure 2) is required. This form must use an oversampled configuration however [6], and the resulting design is shown in Figure 4. Every sub-band filter is given by:…”

“…These can be difficult to design and implement in practice. Potential solutions to this problem include multi-stage filtering [5] and a number of Frequency Response Masking (FRM) based designs [6,7]. The FRM solutions in particular replace a single long prototype filter in a GDFT-FB with a cascade of interpolated filters and masking filters, each of which is shorter than the original filter and appear to provide comparable or better computation loading than the basic GDFT-FB designs.…”

“…However, for polyphase filter-bank, where a large number of sub-bands are required, a very long prototype filter with a large number of coefficients will typically be needed. For example, [6] found that a 256-channel DFT-FB required a prototype filter with 8085 coefficients to meet the TETRA Voice and Data 25 kHz standard. FRM techniques can be used to overcome this problem.…”