The TI dRAM cell, a MOSFET with two dynamically programmable threshold states, is very attractive for VLSI dRAM\´s because of its potential 3× density advantage over the one-transistor and-capacitor (1-T) cell, 10× lower leakage at high temperatures compared to the 1-T cell, and its immunity to soft errors. Linear scaling of the 1-T cell by a factor

reduces the available signal by
3, whereas the charging current for the TI RAM cell is invariant to scaling since the

ratio remains constant allowing it to scale to higher density. An experimental array (64 rows by 8 columns), representing a cross section of a 16K dRAM, with on-chip decoding and sensing has been fabricated using the TI RAM cell as the memory element. Using 4-µm design rules, the cell size was 204 µm
2due to pitch requirements for the decoder and sense amplifier. This compares with 170-200 µm
2for the 1-T cell using 2.5-µm design rules being fabricated in the 64K dRAM\´s today. The array which is compatible with 5-V-only operation was designed to provide diagnostic capability rather than speed and shows the data can be accessed 85-100 ns after the

signal. In this paper, the physics of the TI RAM cell are discussed as well as circuit considerations for its implementation into an array.