FilterLength=2

(ddc0_fir1_ncoeff

1)+1forODD

´

-

FilterLength=2

ddc0_fir1_ncoeffforEVEN

´

FilterLength=4

(ddc0_fir1_ncoeff

1)+1forHALFBAND

´

-

FilterLength=ddc0_fir1_ncoeffforARBITRARY

(6)

ddc0_fir1_ncoeff

2

ddc0_cic_dec_rate

£

´

(7)

GAIN=

h

+

NCOEFF

å

NCOEFF

1

-

n=1

2h

n

2

1

-

15

(8)

GAIN=

å

NCOEFF

n=1

2h

n

2

1

-

15

(9)

GAIN=

å

NCOEFF

n=1

h

n

2

1

-

15

(10)

Second FIR Filters

ddc0_fir2a_ncoeff

4

ddc0_cic_dec_rate

£

´

(11)

ddc0_fir2b_ncoeff

4

ddc0_cic_dec_rate

£

´

(12)

AFE8220-Q1

The maximum filter length that can be realized is limited by two factors. First, the number of clock cycles

between successive filter outputs limits the number of coefficients that can be processed, as shown in

Equation 7.

where ddc0_cic_dec_rate is the decimation ration of the CIC filter.

Second, the size of the data memory (which stores incoming data samples) limits filter length to 62 taps.

The dc gain of the FIR filter depends on the coefficient values and the filter mode. For ODD mode and

HALFBAND mode, the dc gain is given by Equation 8:

For EVEN mode the, dc gain is shown by Equation 9:

while for ARBITRARY mode the gain is shown by Equation 10:

The first FIR filter is followed by two parallel second FIR filters, FIR2A and FIR2B. Duplicate filters allow the

output of two I and Q output streams with different bandwidths. For example, the bandwidth of FIR2A may be set

wide to accommodate reception of digital broadcasts, while FIR2B may be set narrower to receive an analog

broadcast sharing the same band. Coefficients for FIR2A are stored in memory bank 1 (MEM = 1) and

coefficients for FIR2B are stored in memory bank 2 (MEM = 2).

The operation of the second FIR filter is similar to the first FIR filter with several notable exceptions. First, the

depths of the coefficient and data memories are doubled to 128. This size increase allows for filters up to 126

taps to be realized without running out of data memory. It also allows longer sets of filter coefficients to be stored

in coefficient memory.

Second, because of the additional decimation by two from the first FIR filter, twice as many MCLK cycles are

available to process coefficients, increasing the maximum allowable value of NCOEFF, as shown in Equation 11

and Equation 12.

through the filter. With longer filters (and therefore, smaller coefficients), frequency response errors may be

introduced as a result of coefficient truncation. A Shift parameter has been added to the second FIR filter to

through FIR2A (similarly for ddc0_fir2b_shift and FIR2B). Note that shift values for FIR2A and FIR2B can be set

separately.

Copyright © 2009, Texas Instruments Incorporated

Submit Documentation Feedback

21

Product Folder Link(s): AFE8220-Q1