A
2V
CMOS
Programmable Pipelined Digital Differential
Matched Filter for DS-CDMA System
-
A T +
She-Hwa Yen and Chorng-Kuang Wang
I/P - b, b, .... b,
Department of Electrical Engineering National Taiwan University, Taiwan. R.0.C
bn+l Abstract
This paper presents a 2V DS-CDMA programmable
digital matched filter with a differential and pipelined
structure. Differential PN (Pseudo-Noise) code scheme is
adopted to reduce the number of multiplication and summations (M&S). The PDDMF (Pipelined Digital Differential Matched Filter) not only saves the hardware and the power, but also improves the operation speed, which makes PDDMF more suitable, for personal communication at high speed and low power requirements than the conventional approach. The PDDMF implemented in a
0.6um CMOS technology is clocked at 2.5 MHz and
consumes 1.6mW from a single power supply 2V. Introduction
Direct Sequence Code Division Multiple Access (DS-
CDMA) has been successfully applied to many low
bandwidth wireless communication systems. In these systems, matched filter is usually employed for code acquisition since it offers significant advantages in the speed of synchronization. However, the primary limitation of the conventional digital filters is the number of stages that is
greatly associated with the hardware implementation of multiplication and accumulation. This paper describes a CMOS low-voltage low-power programmable pipelined digital matched filter that employs a differential PN code
scheme [I] in order to reduce hardware complexity and hence
the chip size and power.
Pipelined Digital Differential Matched Filter The total number of multiplication and summation in conventional matched filter M, is 2'-1, where r is the order of the PN code polynomial generator. Obviously as the number
of stages increases, the number of M&S significantly
increases. A differential PN code scheme (Fig.1) [l] is employed to avoid this problem. The output hnction of the
matched filter can be described as :
f~(T)=[b~+lXo+b~X,+b~-lX2 . . . + b 2 X ~ - l + b l X ~ ] + f ~ ( T - l ) =[-aNxO+( aN-aN. I)x I+(aN- I-aN-2)+.
. . .
.+(a2-aJx~.+ a , x ~ ] + f ~ ( T - 1)
where b, for i from 0 to N + l is the multiplication coefficients
of the PDDMF, and a, is the PN code coefficient of the
conventional matched filter. Since the coefficients a, of the
U
The pipelined structure, simplifies the summation circuit of the multiple input, and improves the speed of the PDDMF. Furthermore, the throughput of the PDDMF is independent of the PN code length.
Circuit Architecture
Fig.2 is a block diagram of the PDMMF for DS-CDMA receiver. The 9.6kHz input data are spread into a signal band 1.2288MHz in IS-95 based CDMA systems. Because the
SNR of the spreading signal
is
very low, the 4bits soft... !
*-I-
F i g 2 The programmable PN coefficients of PDDMF
decision matched filter is adopted to increase the reliability of data detection. The multiplication coefficients of PDDMF are user programmable by the PN code generator. The PN code generated by the PN generator is first stored in the PN code register. Then the differential encoder generates the multiplication coefficients according to the differential PN code scheme. It controls the M&S cell for coefficients setting. The new coefficient will be 0 when adjacent Coefficients are the same value.
1 l l l l
reset
cin
Fig.3 shows the circuit of two's complementary M&S with coefficient 2 and -2. The input data d l , d2, d3, d4 multiply the coefficient first. If the multiplication coefficient bi is -2 the pins neg and cin are set to logic 1 ; otherwise they will be set to logic 0 for bi=2.
Power Supply Technology Tap Length SamplesKhip PN Code Rate Clock Rate Power Consumption Chip Area
z
m
GNDr-L
2v 0.6 CMOS 16 2 1.25MHz 2.5MHZ 1.6mw 1500um*
1500um6
r
l
Differential Encoder9 Bits Ripple Adder
U I I U
Fig.3 Circuit design for M&S cell circuit of 2 and -2
A 9 bits ripple adder performs the summation in the final.
Ripple adder is chosen for low power rather than for high speed. The chip-rate of the PN code used is 1.2288MHz, the tap length is 16 and the number of multiplication coefficients for bi= ai-ai-l = 0 is 7. Only 8 coefficients need to be multiplied, rather than 32 coefficients in the conventional 16
taps length matched
filter
with 2 over-sampling per chip-time. The number of M&S is reduced to 1/4. Fig.4 is the die photo of the programmable 16 taps length PDDMF. It is fabricated in a 0.6um CMOS technology. The chip area is
1500um by 1500um.
Experimental Results
Fig.5 shows the testing results of the PDDMF by IMS logic master, The testing mode can be programmed either automatic self-test mode or normal mode. In self-test mode, the input PN code and coefficients are generated automatically. Under normal test mode whose input code is fed externally, while filter coefficients are chosen by user. In both modes, the auto-correlation results can track the PN sequence correctly.
, / I