Oki’s 0.3 5 µm ASIC products deliver ultra-high performance and high density at low power dissipation.
The MSM13Q0000/14Q0000 series devices (referred to as “MSM13Q/14Q”) are implemented with the
industry-standard Cell-Based Array (CBA) architecture in a Sea-of-Gates (SOG) structure. Built in a
0.35 µm drawn CMOS technology (with an L-Effective of 0.27 µm), these SOG devices are available in
three layers (MSM13Q) and four layers (MSM14Q) of metal. The semiconductor process is adapted from
Oki’s production-proven 64-Mbit DRAM manufacturing process.
The MSM13Q/14Q Series contains 6 arrays each, offering over 1 million raw gates and 352 I/O pads. Up
to 66% and 90% of the raw gates can be used for the 3-layer and 4-layer arrays, respectively. Oki’s 0.35
µm family is optimized for 3-V core operation with optimized 3-V I/O buffers and 5-V tolerant 3-V buffers. These SOG products are designed to fit the most popular plastic quad flat packs (QFPs), thin QFPs
(TQFPs) , and plastic ball grid array (PBGA) packages.
The MSM13Q/14Q Series uses the popular CBA architecture from Silicon Architects of Synopsys which
mixes two types of cells (8-transistor compute cells and 4-transistor drive cells) on the same die to deliver
high gate density and high drives. The CBA is supported by a rich macro library, optimized for synthesis.
Memory blocks are efficiently created by Oki’s memory compilers to generate single- and dual-port
RAM’s in high-density and low-power configurations with synchronous RAM options.
As such, the MSM13Q/14Q series is well suited to memory-intensive designs with high production volumes approaching the real estate and cost savings of standard cells. At the same time, its SOG architecture allows rapid prototyping turnaround times. Thus, Oki’s MSM13Q/14Q family offers the best of two
worlds: quick prototyping of a gate array and low production cost of a standard cell.
Oki’s 0.35 µm ASIC products are supported by leading-edge CAD tools including a synthesis-linked
floorplanner, motive static timing analyzer, and H-clock tree methodology. They are further supported
by specialized macrocells including phase-locked loop (PLL), pseudo-emitter coupled logic (PECL),
peripheral component interconnect (PCI), universal synchronous receiver/transmitter (UART) cells, and
ARM7TDMI RISC cores.
FEATURES
• 0.35 µm drawn 3- and 4-layer metal CMOS
• Optimized 3.3-V core
• Optimized 3-V I/O and 3-V I/O that is 5-V
tolerant
• CBA SOG architecture
• Over 1.0M raw gates and 352 pads
• User-configurable I/O with V
state, and 1- to 24-mA options
• Slew-rate-controlled outputs for low-radiated
noise
• H-clock tree cells which reduce the maximum
skew for clock signals
, VDD, TTL, 3-
SS
• User-configurable single and dual-port;
synchronous or asynchronous memories
• Specialized macrocells including PLL, PECL,
PCI, UART, and ARM7TDMI
• Floorplanning for front-end simulation, backend layout controls, and link to synthesis
• Joint Test Action Group (JTAG) boundary scan
and scan-path ATPG
• Support for popular CAE systems, including
Cadence, IKOS, Mentor Graphics, Synopsys,
Viewlogic, and Zycad
The primary components of a 0.35 µm MSM13Q/14Q circuit include:
• I/O base cells
• Configurable I/O pads for VDD, VSS, or I/O (optimized 3-V I/O and 3-V I/O that is 5-V tolerant)
•V
and VSS pads dedicated to wafer probing
DD
• Separate power bus for output buffers
• Separate power bus for internal core logic and input buffers
• Core base modules containing three compute cells for each drive cell
• Isolated gate structure for reduced input capacitance and increased routing flexibility
Each array has 24 dedicated corner pads for power and ground use during wafer probing, with 4 pads
per corner. The arrays also have separate power rings for the internal core functions (V
and output drive transistors (V
DDO
and V
SSO
).
The array architecture uses optimally sized transistors to efficiently implement logic and memory in a
metal programmable technology. CBA uses two types of cells: compute cells and drive cells. The compute cell employs four PMOS and four NMOS trasnsistors whose sizes are optimized for logic and memory implementations as shown in
Figure 1
. The quantity and size of the transistors in a compute cell are
carefully selected to maximize the efficiency of most commonly used functions in VLSI design. The drive
cell consists of two large PMOS pull-up transistors and two large pull-down transistors. The compute
and drive cells are tiled to create a channelless core array, with three comput cells for each drive cell as
shown in
Figure 2
. The 3:1 ratio of compute to drive cells was selected for optimal implementation of
emerging applications. Macrocells are created using either compute cells, drive cells, or combinations of
compute and drive cells.