Timestamp is relative (distance between consecutive events), but the time controller is able to recovery from protocol induced delays.There are two FIFOS (input and output) that can save the event information and its timestamp for up to 128 events (output) and 256 events (input).It has one AER output bus and one AER input bus.It is designed around a Spartan II 200 FPGA, with a PCI interface developed in VHDL by RTC, which makes it cheaper and faster.This interface, shown in figure 1, has the following technical features:
#Aer switch Pc
This PCI-AER interface is a very fast communication channel (up to 10 Mev/sec vs 1Mev/sec of Rome board) between an AER bus and the PC software. The CAVIAR PCI-AER is not a communications centre and does not have a mapper or a splitter/merger included.Our interface differs from the Rome PCI-AER in several characteristics:
To obtain the maximum bandwidth the PCI is nowadays one of the best candidates. To inject and read events to an AER system, using a PC, a bridge between the AER bus and one PC buses is necessary.
#Aer switch drivers
All of them with Linux and XP drivers and Matlab interfaces. Our group (Robotics and Computer Technology – RTC) has developed a PCI to AER interface, taking as a starting point the Rome PCI-AER developed by Dante two USB to AER interfaces and an AER to AER interface. These AER-Tools are divided into 4 different PCBs with 8 different functionalities, depending on the firmware downloaded to them.ĬAVIAR PCI-AER USB-AER AER-Switch Mini-USB-AER To make all this vision system possible, a set of AER-Tools or debugging and interconnection purposes are not only useful, but also necessary.
This chain is composed by a 128×128 retina that spikes with temporal and contrast changes, four convolution chips to detect a ball at different distances from the retina, four object chips to filter the convolution activity and a learning stage composed by two chips: delay line and learning. CAVIAR connects the biggest AER chain at the moment. We are four different partners working together in the design of a neuromorphic vision system totally based on AER. This milestone was the start point for the development of a set of Address-Event-Representation (AER) Tools under the European Project CAVIAR. Complex systems developed by Neuromorphic Engineers require interfaces to interconnect them and to connect them to PCs for debugging or other purposes.
0 kommentar(er)
