sram a-factors for simple 6t sram cell using microprocessor logic cmos process technology
Post on 01-Jan-2016
48 Views
Preview:
DESCRIPTION
TRANSCRIPT
SRAM A-Factors for Simple 6T SRAM Cell using Microprocessor Logic CMOS Process Technology
020406080
100120140160180200
0.13micron (Intel, IEDM
2000.p.567))
0.13micron (M
otorola, IEDM2000,p.571))
0.18micron (Intel, IEDM
1998,p.197)
0.25micron (Intel, IEDM
1996,p.847)
0.35micron (Intel, IEDM
1994)
0.3micron (M
otorola, IEDM1994)
A-F
ac
tor
for
SR
AM
Ce
ll S
ize
(s
qu
are
fe
atu
re s
ize
)Average A-Factor = 161.67
DRAM half-pitch (F) A-Factor (A*F2)0.13micron (Intel, IEDM2000.p.567)) 143.70.13micron (Motorola, IEDM2000,p.571)) 146.740.18micron (Intel, IEDM1998,p.197) 172.530.25micron (Intel, IEDM1996,p.847) 164.160.35micron (Intel, IEDM1994) 167.30.3micron (Motorola, IEDM1994) 175.6
0.15, TSMC, VLSI00 152
0.13, Toshiba, VLSI00 148
0.18 (0.13 poly), Motorola, VLSI00, embedded!
85 (or 162 using 0.13)
0.13, IBM, SOI, VLSI00 128
0.13, IBM, bulk, VLSI00 147
0.18, TSMC, VLSI99 136
0.18, IBM, VLSI99 119
0.18, IBM, ISSCC00 131
0.25, UMC, IEDM97 101
0.25, Samsung, VLSI98 102
0.13, Fujitsu, VLSI98
0.25, Motorola, VLSI98
147
150
F, Company, Reference A factor
Virtual Silicon libraries based on United Microelectronics (UMC) processes
A-factors:
0.25 m, high-performance (10 tracks):2-in NAND/NOR: 371INV: 248MUX2: 867DFF: 2106
0.18 m, high-performance (11 tracks), quoted max density = 93.5K gates/mm2, translating to 10.7 m2/gate or 330F2 :
2-in NAND/NOR: 377INV: 251MUX2: 878DFF: 2133
0.15 m, high-density (8 tracks), about 20% smaller than high-performance, quoted max density = 173K gates/mm2, translating to 5.8 m2/gate or 258F2.
2-in NAND/NOR: 307INV: 205MUX2: 717DFF: 1638
If we assume contacted metal pitch = 2.5*F (e.g. MP = 0.625 m for 0.25 m), this gives ~60 MP2 for 2-in NAND/NOR, which is inline with BACPAC calcs
Current recommendations:
SRAM cell size = 150-160F2
Std. Cell size = 375F2??
SRAM overhead: use factor of 1.6 (60% overhead penalty)
These areas don’t include any white-space consideration so the actual packing density should be lower
top related