CMX980AL7 CML Microcircuits, CMX980AL7 Datasheet - Page 88

• Manufacturer Part Number: CMX980AL7
• Description: TETRA Baseband Processor
• Manufacturer: CML Microcircuits

TETRA Baseband Processor
1.6.4.3 General Procedure for Reconfiguring the CMX980A FIR Filters
1.6.5
1.6.6
1999 Consumer Microcircuits Limited
1. Obtain or design the required filter characteristic(s), either in s- or z-transform format. Note that all
programmable FIR filters are sampled at a frequency of MCLK/64. This is equivalent to 144kHz for
MCLK = 9.216MHz and 128kHz for MCLK = 8.192MHz.
The task may be made slightly easier if the desired filter specification can be split into two filters of
roughly similar complexity, but if this is not possible, or if it is preferable to treat the desired filter as a
single entity, then a single impulse response can be considered as a polynomial in Z and factorised
into two polynomials of appropriate length.
The Tx 79-tap filter MUST have symmetrical (hence linear phase) coefficients, but all of the 63-tap
filters allow asymmetrical coefficients. Hence any prototype filter function aimed at the Tx 79-tap filter
must be linear phase.
2. Obtain the impulse response of the desired filter shape, either by simulation of the filter response to
an impulse, or by using an inverse Fourier transform. Convolve the impulse response with the impulse
response of the appropriate compensation filter described above. This may be accomplished either by
concatenating the filters and simulating the impulse response of the combined system, or by
multiplying the Z transfer function polynomials together.
3. Use a standard windowing function (e.g. Blackman, Hamming) to limit the impulse response to 141
samples for the Tx, 125 samples for the Rx, or 79 and 63 samples (Tx), 63 and 63 samples (Rx), if
designing the filters separately.
coefficients to zero.
4. Scale the FIR coefficients to appropriate values. To maximise the use of dynamic range, scale the
dc gain to give values similar to those given by the default coefficients (approximately 4.5x for Tx and
1.6x for Rx). Note that 16-bit coefficients are used in Rx filters, but only 12-bit for Tx filters, while the
use of appropriate dc gains within the system may result in one or two of the most significant bits
being redundant.
5. Load in your new coefficients, run the chip with maximum sized signals and check for internal
overflows (see TxErrorStatus and RxErrorStatus Registers). Optimise individual filter gains to a
level at least 5% below those which cause overflows.
TBD
Internal Symbol-Clock Phase Adjustment
TBD
Generating a Transmit Frame Sequence with optimal use of ramping features
Alternatively, it may be possible to approximate small outer
88
CMX980A
D/980A/3