***Archived: This is a past project, it is not on offer at the moment!***
Project Description:
Memory constitutes a basic component of any non-trivial modern computing system. Several types of volatile, such as DRAM, SRAM and non-volatile memory, such as ROM and FLASH exist. Dynamic memory (DRAM) is preferred over other types of memory (e.g., static RAM (SRAM)), since this type of memory supports particularly high densities; a fundamental prerequisite for high-performance computing systems. This situation is due to the simple structure of the elemental memory cell consisting of a transistor and a capacitor. The charge stored in this capacitor corresponds to a logic value of ‘0’ or ‘1’ for a binary digit (bit).
Although the integration of DRAM memory cells can be extremely dense, charge leaking mechanisms complicate the design and correct operation of DRAM memory. In addition, the increasing size of the memory modules further hinders the ability to sense large capacitive lines and accurately render the bit information. Consequently, to guarantee the correct operation of the DRAM, the stored information is periodically refreshed. The frequency of the “refresh” operation depends both on the design and manufacturing technology of the memory. Due to these issues, a variety of DRAM circuit designs exists related to the memory architecture, the sensing circuitry, and the addressing scheme as well as the DRAM fabrication process. Furthermore, several design and technology parameters determine the performance and, specifically, the timing of a DRAM module.
The development of appropriate timing models can assist the exploration of different memory organizations and designs. Thus, the goal of this semester project is to develop timing models that combine an as broad (and disparate) as possible set of parameters that primarily determine the speed of a DRAM memory. These parameters can include, for example, the different organization of the memory blocks, technological parameters as provided by different memory vendors, and others. The timing models can be either analytic or to combine analytic expressions and simulations. The developed timing models will subsequently be integrated to a CAD tool that will support the exploration of diverse tradeoffs resulting from these parameters that determine the overall timing of the DRAM memory.
Tasks:
- Exhaustive study of existing literature on modeling the timing of a DRAM memory. Further information may be obtained from the websites of memory vendors
- Development of timing models of DRAM memories that include primary physical and design parameters
- Verification of the models through existing “hard” DRAM macros provided by different design kits
- Integration of the models within a computer aided design tool
Requirements:
- Basic knowledge of memory circuits and their operation
- Basic knowledge of Cadence and/or SPICE design tools
- Sufficient programming skills in C, SystemC, and/or HDL languages (Verilog, VHDL)
This project was supervised by Vasilis Pavlidis.
***Archived: This is a past project, it is not on offer at the moment!***