RRAM-based ADC design for neuromorphic computing

Neuromorphic computing is an exciting field of research in electrical engineering, which aims to emulate the functionality of biological neurons and synapses in silicon-based electronic systems. One crucial component of such systems is the analog-to-digital converter (ADC), which is responsible for converting continuous analog signals into digital form that can be processed by digital circuits.

The design of ADCs for neuromorphic computing poses unique challenges due to the specific requirements of the system. These requirements include high resolution, low power consumption, and low latency, all of which are essential for achieving the high efficiency and accuracy necessary for realtime processing of neural signals.

RRAM-based (Resistive Random Access Memory) ADCs have been proposed and studied in recent years. RRAM is a type of non-volatile memory that stores information by changing the resistance of a material. RRAM-based ADCs take advantage of this property to perform analog-to-digital conversion.

In RRAM-based ADCs, the input signal is applied to an RRAM array, and the output is the digital representation of the resistance values of the RRAM cells. The resistance values can be read out using voltage or current sensing techniques, and a digital converter is used to generate the final digital output.

RRAM-based ADCs have several advantages over traditional ADCs. They offer high linearity, low power consumption, and high density, making them suitable for applications such as image and speech recognition. However, there are also challenges associated with RRAM-based ADCs, including issues with device variability, endurance, and write-read disturbance.

Overall, RRAM-based ADCs are an exciting area of research that has the potential to revolutionize the field of analog-to-digital conversion. As an electrical engineering student, it is essential to be aware of this emerging technology and its potential applications in future electronic systems. And the aim of this project is to design higher performance ADC circuits which will be integrated into RRAM crossbars.

Goals for RRAM-based ADC design:

  1. High precision and accuracy: In-memory computing requires high precision and accuracy for reliable computation. Therefore, RRAM-based ADCs should be designed to achieve high resolution and accuracy.
  2. Low power consumption: In-memory computing requires low-power devices to minimize energy consumption. RRAM-based ADCs should be designed to consume low power to enable efficient inmemory computing.
  3. High speed: In-memory computing also requires fast devices to enable real-time processing of data. RRAM-based ADCs should be designed to operate at high speeds to facilitate efficient computation within the memory array.

[1] DAC '19: Proceedings of the 56th Annual Design Automation Conference 140, pp 1–6 (2019)
[2] IEEE Transactions on Electron Devices 70, 4, (2023)