Publicação
Study of neuromorphic properties of circuits based on resonant tunneling diodes
| Resumo: | RTDs, with their negative differential conductance, small size, capability of high frequency of operation and excitable response have the potential to be implemented as basic components in spiking neuromorphic circuits, i.e as nodes that produce and detect spikes. The present study addresses the neuromorphic properties of resonant tunneling diodes (RTDs), namely their spike generation and detection dynamics. The layout consists firstly on the basic requirements for neuromorphic implementations, followed by a description of the physics behind the RTDs non-linear current voltage characteristic that creates the conditions for spike generation. The mechanisms and conditions by which the RTDs create an excitable response are then put-forward. After this, a series of experimental measurements were carried out for different sized RTDs provided by the ChipAi project. First the measurements of the RTDs I-V curves, followed by a characterisation on their operation as a VCO, concluded with a study on their excitable all-or-nothing response to electrical perturbations. Finally, several experimental activities were carried out to infer the properties and requirements of these devices to work as potential neuromorphic devices. These parameters were the voltage thresholds, resting potentials and refractory times. these were then discussed for all of the different sized RTDs and operation points. On top of this experimental study, tools of simulation were also designed and constructed for these RTDs devices followed by a verification study on not only on their VCO mode but also their neuromorphic properties i.e. excitable response. It was found that the RTDs can indeed function as spike generators when an appropriate disturbance is applied in their bias voltage. The prospect for the scaling down of these devices to work as nanoneuromorphic devices seems promising due to the fact that smaller RTDs presented higher frequencies of operation leading to shorter periods of oscillation and refractory times. The simulation tool, based on the RLC model of the RTDs, developed using Matlab Simulink/Simscape and the I-V curves of these devices, was able to emulate the oscillatory behaviour of the RTDs when biased in their NDC regions and the neuromorphic property of excitability, again it was found that the scaling down of these devices can lead to faster communication rates. |
|---|---|
| Autores principais: | Neves, Luís Pedro Teixeira das |
| Assunto: | Díodo de túnel ressonante Resistência diferencial negativa Excitabilidade Redes neuronais de spikes Dispositivos neuromórficos Teses de mestrado - 2023 |
| Ano: | 2023 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório da Universidade de Lisboa |
| Resumo: | RTDs, with their negative differential conductance, small size, capability of high frequency of operation and excitable response have the potential to be implemented as basic components in spiking neuromorphic circuits, i.e as nodes that produce and detect spikes. The present study addresses the neuromorphic properties of resonant tunneling diodes (RTDs), namely their spike generation and detection dynamics. The layout consists firstly on the basic requirements for neuromorphic implementations, followed by a description of the physics behind the RTDs non-linear current voltage characteristic that creates the conditions for spike generation. The mechanisms and conditions by which the RTDs create an excitable response are then put-forward. After this, a series of experimental measurements were carried out for different sized RTDs provided by the ChipAi project. First the measurements of the RTDs I-V curves, followed by a characterisation on their operation as a VCO, concluded with a study on their excitable all-or-nothing response to electrical perturbations. Finally, several experimental activities were carried out to infer the properties and requirements of these devices to work as potential neuromorphic devices. These parameters were the voltage thresholds, resting potentials and refractory times. these were then discussed for all of the different sized RTDs and operation points. On top of this experimental study, tools of simulation were also designed and constructed for these RTDs devices followed by a verification study on not only on their VCO mode but also their neuromorphic properties i.e. excitable response. It was found that the RTDs can indeed function as spike generators when an appropriate disturbance is applied in their bias voltage. The prospect for the scaling down of these devices to work as nanoneuromorphic devices seems promising due to the fact that smaller RTDs presented higher frequencies of operation leading to shorter periods of oscillation and refractory times. The simulation tool, based on the RLC model of the RTDs, developed using Matlab Simulink/Simscape and the I-V curves of these devices, was able to emulate the oscillatory behaviour of the RTDs when biased in their NDC regions and the neuromorphic property of excitability, again it was found that the scaling down of these devices can lead to faster communication rates. |
|---|