1. RICE UNIVERSITY On the architecture design of a 3G W-CDMA/W-LAN receiver Sridhar Rajagopal and Joseph R. Cavallaro Rice University Center for Multimedia Communication http://www.cmc.rice.edu sridhar@rice.edu This work is supported by Nokia, TI, TATP and NSF

2. RICE UNIVERSITY Introduction  A baseband communications processor  Wireless LAN  Wideband CDMA  RENE

3. RICE UNIVERSITY Motivation  No architecture developed yet to meet real- time requirements of 3G systems.  2 - 8 Mbps range for wideband CDMA  100 Mbps range for wireless LAN  Design factors that makes the problem harder  Low power  Flexibility

4. RICE UNIVERSITY Previous Work  Designing algorithms from an implementation perspective  algorithms with high degree of parallelism  fixed-point computations  simple operations - multiplications/additions  Example: multiuser channel estimation & detection  Real-time implementation on DSPs/FPGAs/ASICs  area-time tradeoffs

5. RICE UNIVERSITY Possible contributions of this work  A real-time low-power VLSI architecture design  using on-line arithmetic  A real-time programmable architecture design  using a media processor simulator -- IMAGINE  Integrating these two architectures in one.

6. RICE UNIVERSITY Contents  Low-power VLSI architecture design using on- line arithmetic  Programmable architecture design using the IMAGINE simulator  Conclusions

7. RICE UNIVERSITY On-line arithmetic  Uses a redundant number representation.  Pipelined digit-serial arithmetic with MSDF computations.  Successive computations as soon as  inputs available ( = 1..4, typically).  Algorithms available for various operations (+,*,/,sqrt) and for fixed-point computations. z5z5 …z4z4 z3z3 z2z2 z1z1    Output z …y5y5 y4y4 y3y3 y2y2 y1y1 Input y …x5x5 x4x4 x3x3 x2x2 x1x1 Input x

8. RICE UNIVERSITY Why is on-line arithmetic useful?  Conventional operations in 3G wireless systems  high precision operations (16-32 bits) but with low precision outputs.  Only most significant digits (1-3 bits) needed.  Use MSDF computation to find the needed digits and avoid computation of the successive digits.  Digit-serial computations and hence, low power Detection

9. RICE UNIVERSITY Redundant number systems  Radix -r number system:  digit has |r| values: 0,1,2…..,r-1  Redundant number system:  digit has q >|r| values  r+2  q  2r-1  Example: each digit in the number has a sign associated with it.  10(-1)2 = 992 has 2 equivalent representations.  Redundancy helps in carry-free additions - MSDF

10. RICE UNIVERSITY Adder Implementation t conv – conventional adder time per bit t OL – online delay time per digit d – bit-precision

11. RICE UNIVERSITY On-line radix-4 adder Digit serial inputs Digit serial outputs Digit selection Carry Save Adders Residual feedback

12. RICE UNIVERSITY Comparison with regular adders  Addition time and area independent of digit precision (X area dependent on precision)  Savings in time obtained by chaining operations as successive operations can start as soon as MSD is obtained.

13. RICE UNIVERSITY -0.500.51 0 1 1.5 2 2.5 3 3.5 4 4.5 5 Signal Amplitude Time taken for addition On-line addition Conventional addition Dependency of execution time for on-line addition on SNR

14. RICE UNIVERSITY Detection Example Multi- user Single user Detector 3.00m* t OL =8t CMF =24Throughput +m* S*t OL =94 +2*t CMF =168 1.79 t MF +S*t PIC (2*S-1)*t CPIC Latency log 2 (d)*t conv = 21 2.63 m* t OL =8(log 2 (N)+2)* Throughput t OL +t stop = 14log 2 (d)*t conv = 21 1.50 (log 2 (N)+2)* Latency SpeedupOn-lineConventional

15. RICE UNIVERSITY Low power VLSI design  Power savings due to 2 reasons  eliminating unwanted computations  digit-serial hardware  Real-time requirements met by proper pipelining of computations and exploiting parallelism in the algorithms.

16. RICE UNIVERSITY Contents  Low-power VLSI architecture design using on- line arithmetic  Programmable architecture design using the IMAGINE simulator  Conclusions

17. RICE UNIVERSITY A programmable architecture simulator  Flexibility in the algorithm requirements  channel dependent computations  changing algorithms on-the-fly  seamless switching between wireless LAN and wideband CDMA.  Simulator needed to test performance of algorithms  extensions/modifications for critical operations

18. RICE UNIVERSITY The IMAGINE architecture and simulator  IMAGINE is a media signal processor, built at Stanford.  Many common workload features  Good starting point to explore.  Local expertise - Dr. Scott Rixner (rixner@rice.edu)

19. RICE UNIVERSITY IMAGINE architecture  Great for media processing algorithms  1024 pt FFT in 7.4  s on a 500 MHz processor with a 8-cluster (48 units)  3.8W of power  Great for parallel, vector and streaming computations  Performance/extensions to sequential computation kernels such as Viterbi traceback needs to be investigated.

20. RICE UNIVERSITY Conclusions  On-line arithmetic useful for a low power real- time implementation  A programmable real-time architecture is being investigated using the IMAGINE simulator  Aim is to then integrate these two features