I missed your original post, so I'll reply here.
1) the polyphase decimator is a mathematical manipulation to reduce the computational load. From a black box perspective, it provides the same output as you would have using a single rate filter followed by a decimator (dropping output samples). The difference is if you are clever about your filter implementation, you don't have to do all the computation for samples that are dropped. The polyphase filter takes advantage of that fact. It consists of R sub-filters whose outputs are summed. Each subfilter is running at the decimated sample rate instead of the input sample rate, so you get a reduction of the computations per unit time by a factor of R (R is the decimation ratio).
2) Multichannel operation is achieved by doubling the tap delays and interleaving the samples from each channel. The physical filter is operated at C times the sample rate (C is the number of channels). In order to do that, you need a multiplied clock and your logic has to be fast enough to operate at the increased clock rate. The reason for doing this is to reduce the size of the hardware by time-sharing (the classic time-area trade-off).
Where you have multiple bandwidths, you've got an opportunity to change the output sample rate to match the selected bandwidth. The advantage to using a CIC filter is that the filter shape referred to the output sample rate is nearly independent* of the decimation ratio. That means that you can follow the CIC with a filter that does not have to be changed when you change the bandwidth. In your case, you have only two bandwitdh settings, so it is not unreasonable to change filter coefficients to change the bandwidth. In either case, if you have a sample decimation ratio greater than 2, it is usually more efficient in terms of computation and amount of hardware to use a multi-rate approach where you decimate in each stage of a multi-stage filter. My DDC designs typically include a mixer followed by a CIC filter, followed by two stages of decimate by 2 FIR filters, the first of which is a halfband filter with a relatively small number of taps (typically around
15). If using a CIC, you'll want the final bandwidth to be no more than
1/4 the CIC's first null in order to keep the CIC's droop manageable.
- The CIC filter response has a very small component that is dependent on the ratio, which is only minimally noticible at ratios less than about 15.
--
--Ray Andraka, P.E.
President, the Andraka Consulting Group, Inc.
401/884-7930 Fax 401/884-7950
email ray@andraka.com
http://www.andraka.com
"They that give up essential liberty to obtain a little
temporary safety deserve neither liberty nor safety."
-Benjamin Franklin, 1759