The FIR Compiler core provides support for processing multiple input sample streams using the same implementation. Each input stream is filtered using the same filter configuration (rate change, etc.) using the currently selected filter coefficient set.
In many applications, the same filter must be applied to several data streams. A common example is the simple digital down converter shown in
This Figure
. Here a complex base-band signal
is applied to a matched filter M(z). The in-phase and quadrature components are processed by the same filter.
One solution to this issue is to employ two separate filters; however, this can waste logic resources. A more efficient design can be realized using a filter architecture that shares logic resources between multiple time division multiplexed (TDM) sample streams. As more channels are processed by the core, the sample throughput is commensurately reduced. For example, if the sample rate for a single-channel filter is f s , a two-channel version of the same filter processes two sample streams, each with a sample rate of f s /2. A three-channel version of the filter processes three data streams and supports a sample rate of f s /3 for each of the streams.
A multichannel filter implementation is very efficient in resource utilization. A filter with two or more channels can be realized using a similar amount of logic resources to a single-channel version of the same filter, with proportionate increase in data memory requirements. The trade-off that needs to be addressed when using multichannel filters is one of sample rate versus logic requirements. As the number of channels is increased, the logic area remains approximately constant, but the sample rate for an individual input stream decreases. The number of channels supported by a filter core is specified in the filter Customize IP dialog box. The FIR Compiler supports two multichannel implementation: Basic and Advanced.