GC5316 Datasheet, PDF (34/75 Pages) Texas Instruments – HIGH-DENSITY DIGITAL DOWNCONVERTER AND UPCONVERTER

GC5316
SLWS154A − JANUARY 2004 − REVISED MARCH 2004
www.ti.com
Table 8. Programming
VARIABLE
crastsrttap_cfir(4:0)
symmetric_cfir
cfir_gain
DESCRIPTION
These bits define the number of taps that CFIR uses for the filtering. DUC CFIR: (2 x
crastarttap_cfir) +1, Note: crastarttap_cfir must be odd.
Set to 1 if filter is symmetric. Saves a bit of power.
CFIR gain adjustment.
The PFIR filter’s 18-bit coefficients are loaded by the software cmd5316. The user must
provide a coefficient file with one integer coefficient per line. Note that the PFIR filter
coefficients are shared by the A and B signals in CDMA mode.
3.1.5.3 Transmit CIC Filter
Force 0
Z ^−m1
Z ^−m2
Z ^−m3
Z ^−m4
Z ^−m5
Z ^−m6
N
In
18
19
20
21
22
23
24 Zero Pad
by 3−31
m1, m2, m3, m4, m5, m6 = 1 or 2
24 Shift
Z ^−1
Z ^{−1 29}
Z ^{−1 34}
Z ^{−1 39}
Z ^{−1 44}
50
Z ^−1
Shift
0− 31
&
18
O ut
R ound
Figure 28. Transmit CIC Filter
The transmit six stage CIC filter interpolates over a programmable range from 4 to 32. The filter is made up of six
banks of 24-bit subtractor sections followed by six banks of integrator sections. Each of the six subtractor sections
can be independently programmed with a differential delay of one or two. A shift block follows the last integration stage
and can shift the 50-bit accumulated data down by 31−TCIC_SHIFT bits yielding 18-bit output data.
The CIC filter exhibits a droop across its frequency response. Usually the preceding CFIR filter precompensates for
the CIC droop with a gradually rising frequency response. However, it is also possible to provide the precompensation
in the PFIR filter.
CIC interpolation filters can become unstable if an external event (such as a cosmic particle) disturbs a storage node
in the CIC integrator section. This can add a bias which subsequently integrates out of control. The GC5316 transmit
CIC employs a patented method to detect and then automatically flush and reset the filter. Register bits are available
to disable and to test this auto-flush feature. A maskable interrupt becomes active if a CIC error occurred.
The gain of the CIC filter is: Ncic⁵ x 2^{(number ofstages where M=2)} x 2^{(CIC_SCALE − 31)} where CIC_SCALE is 0 to 31. Ncic
is the interpolation ratio and is programmed as cic_interp_decim + 1.
Since the CIC output is full rate for both UMTS and CDMA, a complete hardware path is required for each of the
signals A and B from this point on in the transmit signal path. For CDMA, there are four independent CIC filters (I/Q
for signal A and I/Q for signal B). For UMTS, the two signal B CIC filters are disabled.
34