Advanced Micro Devices Corporation

Muhammad M Rahmatullah

Philip Chu Yip

Terry Lynn Cole

Saf M Asghar

Tony E Sawan

A method and apparatus for allowing a modem transmitting data in the data mode to initiate retraining mode due to changing line conditions. The modem inserts a test signal within the data received from a computer system. The data and test signal are encoded and sent over a network to a second modem. The second modem compares the test signal with a standard test signal, and if the test signal received is different from the standard test signal, the second modem initiates retraining of the modems to adjust for the varying line conditions.

Monitoring line conditions in the data transmission mode

A digital signal processor (DSP), hardware module, and shared memory coupled together to perform Viterbi decoding on a sequence of received symbols. Given channel coefficients, the DSP calculates initial data for Viterbi processing: combination values for each possible state and branch product values for each possible symbol. These values are stored in shared memory for access by the hardware module. The DSP further calculates the first few stages of the Viterbi processing so path metrics are well defined for every state. Path metric values are also stored into the shared memory. The hardware module is optimized to perform calculations associated with a single stage of the Viterbi algorithm. The DSP invokes by the hardware module by passing a received sample to the hardware module. The hardware module calculates a survivor state value and minimizing path metric value for each state in the state space.

MLSE implementation using a general purpose DSP and shared hardware for a GSM application

The present invention comprises a system and method for a GSM receiver to perform channel estimation under the assumption that the analog-to-digital (A/D) converter is free-running. A search process is employed, whereby the known GSM training signal is sampled with a plurality of phases which vary incrementally from zero to a full symbol period. For each phase a sample set (of the training signal) is generated. Furthermore, each sample set is used, together with the received samples, in a cross-correlation procedure to obtain a candidate impulse response vector for the transmission channel. Thus a plurality of candidate impulse response vectors are produced, one of which will be selected to be the optimal representative for the transmission channel. The optimal impulse response vector is selected as follows. Each impulse response vector is convolved with the corresponding sample set, thus producing an estimated output vector. The estimated output vectors are compared with the vector of received samples (from the free-running A/D converter), and the estimated output vector which is closest in the Euclidean (square-norm) sense designates the impulse response vector which is optimal.

Growing community of inventors

Austin, TX, United States of America

Muhammad M Rahmatullah