AUDIO DEMONSTRATIONS

 Performance of Nonlinear MMSE Interference Cancellation


Multiaccess interference (MAI) is known to be the principal performance-limiting factor in a multiuser CDMA system.  The conventional matched filter receiver treats MAI as additive noise, and MAI mitigation is limited to that provided by the spreading gain.  All currently deployed commercial CDMA receivers use matched filter receivers.  Interference cancelers suppress MAI by removing estimated interference from each user's signal with the help of tentative decisions made on the transmitted signals of all users.  Parallel interference canceler (PIC) is an interference cancellation architecture which consists of a sequence of concatenated stages of tentative-decoding and interference cancellation for all users simultaneously.  Each stage consists of a bank of matched filters for all users followed by signal reconstruction and interference subtraction for each user.  The Conventional PIC (CPIC) hardlimits the matched filter outputs in each stage to estimate the transmitted signals of all users.  Therefore, interference is added--instead of being subtracted--whenever an incorrect tentative decision is made on any user's signal.  This residual interference propagates through the stages of the PIC and results in possible final decoding errors.

                    A Nonlinear MMSE Interference Canceler (NMIC), developed by the ASPECT group, minimizes errors in signal estimation by estimating the transmitted signals subject to a nonlinear MMSE criterion, in place of hardlimiting the matched filter outputs.  We have designed NMIC receivers for CDMA systems that employ BPSK or M-ary orthogonal modulation, and Rake reception.  Implementation of the NMIC receiver requires estimation of effective complex channel gains for all the stages, which is achieved using a low-complexity adaptation scheme.  Effective channel gains are different from the actual channel gains due to the correlation that exists between MAI and desired signal in stages after the first stage.

                    The following audio demonstration compares the performance of a NMIC with that of a matched filter (MF) receiver and a conventional PIC.  The demonstration is the result of passing speech data through a simulated synchronous (worst-case) CDMA system with BPSK modulation and quadrature spreading with random (aperiodic) PN sequences.  Eb/No (without interference) is chosen to be 20 dB.  Speech data are compressed using the state-of-the-art EVRC (Enhanced Variable Rate Codec) speech coder and demultiplexed to all the transmitters in the CDMA system.  The demodulated signals of all users from the receiver are then multiplexed and decompressed back to speech data.  Since the objective of this demonstration is to show the benefits of interference cancellation, error-control coding is not included in the simulations.

                    Click on the   icon in the following tables to listen to the corresponding audio.  The first table gives the transmitted speech data.  The received speech data are listed in the second table.  The first column of the second table lists the number of users in the CDMA system, the second column lists the number of cancellation stages of the PIC receivers (not applicable to the `MF' column), the third column lists the spreading gain (number of PN chips per bit), and the last three columns list the clickable audio outputs.  Since the EVRC decoder erases an entire frame (184 bits) whenever a decoding error is detected, the audio output in a high bit-error rate scenario consists of just silence.
 

Transmitted Audio
MF: Matched Filter (Conventional CDMA Receiver)
CPIC: Conventional PIC
NMIC/AVG: Nonlinear MMSE PIC with Conventional Averaging Channel Estimation
NMIC/OBS: Nonlinear MMSE PIC with New Adaptive Filtering Channel Estimation using OBS Algorithm
                                                              
# USERS # STAGES  SPREADING GAIN   MF  CPIC  NMIC/AVG NMIC/OBS
15 3 60
40
 3
60
60
 3
60
80
 3
60