|
Address Event Representation (AER) is an emergent neuromorphic interchip communication protocol that allows for real-time virtual massive connectivity between huge number neurons located on different chips. By exploiting high speed digital communication circuits (with nano-seconds timings), synaptic neural connections can be time multiplexed, while neural activity signals (with mili-seconds timings) are sampled at low frequencies. Also, neurons generate 'events' according to their activity levels. More active neurons generate more events per unit time, and access the interchip communication channel more frequently, while neurons with low activity consume less communication bandwidth. In a typical AER transmitter chip, there is an array of neurons that generate events. They send events to a peripheral circuitry (let's call it "AER Generator") that transforms those events to neurons coordinates (addresses) which are put sequentially on an interchip high speed digital bus. This bus includes a parallel multi-bit address word plus a Rqst (request) and Ack (acknowledge) handshaking signals for asynchronous data exchange. There have been two main approaches published in the literature for implementing such "AER Generator" circuits. They differ on the way of handling event collisions coming from the array of neurons. One approach is based on detecting and discarding collisions, while the other incorporates arbitration for sequencing colliding events . The first approach is supposed to be simpler and faster, while the second is able to handle much higher event traffic. In this article we will concentrate on the second arbiter-based approach. Boahen has been publishing several techniques for implementing and improving the arbiter based approach. Originally, he proposed an arbitration squeme by rows, followed by a column arbitration. In this scheme, while one neuron was selected by the arbiters to transmit his event out of the chip, the rest of neurons in the array were freezed to transmit any further events during this time window. This limited the maximum transmission speed. In order to improve this speed, Boahen proposed an improved 'burst mode' scheme. In this scheme after the row arbitration, a complete row of events is pipelined out of the array and arbitered out of the chip at higher speed. During this single row event arbitration, the array is free to generate new events and communicate to the row arbiter, in a pipelined mode. This scheme significantly improves maximum event transmission speed, specially for high traffic situations were speed is more critical. We have analyzed and studied this approach and have detected some shortcomings in the circuits reported by Boahen, which may render some false situations under some statistical conditions. The present paper proposes some improvements to overcome such situations. The improved "AER Generator" has been implemented in an AER transmitter system
|
