KIP-Veröffentlichungen

This thesis addresses the characterization and calibration of a full-custom mixed-signal Leaky Integrate and Fire (LIF) neuron circuit for highly accelerated neuromorphic hardware.
Using a self-developed software environment for the acquisition and storage of measurement data, various sub-circuits of overall 96 neurons on three HICANN-DLS prototype chips are characterized.

In particular, the function of the integrated capacitive parameter memory array is verified as well as the analog parts of all neuron circuits; namely its analog readout chain, leak- and reset circuitry and both synaptic inputs.
From the acquired data, achievable ranges of LIF model parameters are derived as well as possible improvements for future prototypes.

As final outcome, a mapping between hardware- and model parameters is established for each neuron, making them ready for practical use.