process_tech
Differences
This shows you the differences between two versions of the page.
Both sides previous revisionPrevious revisionNext revision | Previous revision | ||
process_tech [2013/01/16 17:53] – [Cheesing / bamboo structures] azonenberg | process_tech [2015/12/17 22:52] (current) – [Eight or more layers] azonenberg | ||
---|---|---|---|
Line 32: | Line 32: | ||
===== Cheesing / bamboo structures ===== | ===== Cheesing / bamboo structures ===== | ||
- | Copper | + | Modern |
- | Xilinx CoolRunner-II CPLD (180nm | + | Xilinx CoolRunner-II CPLD (180nm |
{{:: | {{:: | ||
Line 44: | Line 44: | ||
===== Single layer ===== | ===== Single layer ===== | ||
- | Very simple chips have only one metal layer and (usually) one polysilicon layer. These are normally very simple devices like 7400-series logic up to i4004 era and are largely hand laid out, as standard logic cells are infeasible in such a process. Geometry is greater than 5 μm in most cases. Almost always | + | Very simple chips have only one metal layer and (usually) one polysilicon layer. These are normally very simple devices like 7400-series logic up to i4004 era and are largely hand laid out, as standard logic cells are infeasible in such a process. Geometry is greater than 5 μm in most cases. Almost always |
FIXME: get a photo | FIXME: get a photo | ||
Line 50: | Line 50: | ||
===== Two layers ===== | ===== Two layers ===== | ||
- | Large geometry (> | + | Large geometry (> |
Standard cells are commonly used in 2-metal processes, with large gaps between individual rows of cells. Horizontal routing runs between rows of cells on M1 (M1 on top of the cells is used by the cell library and for power routing) and vertical routing runs on top of both on M2. | Standard cells are commonly used in 2-metal processes, with large gaps between individual rows of cells. Horizontal routing runs between rows of cells on M1 (M1 on top of the cells is used by the cell library and for power routing) and vertical routing runs on top of both on M2. | ||
Line 70: | Line 70: | ||
{{: | {{: | ||
- | ===== Five and seven layers ===== | + | ===== Four to seven layers ===== |
- | Feature sizes range from 250 to 180nm. | + | Feature sizes range from 350 to 180nm. |
Typically one layer for cells and two or three interconnect layers in each direction. Power distribution is similar to 3-metal designs. | Typically one layer for cells and two or three interconnect layers in each direction. Power distribution is similar to 3-metal designs. | ||
Line 81: | Line 81: | ||
{{: | {{: | ||
+ | |||
+ | TODO: Cross section example | ||
===== Eight or more layers ===== | ===== Eight or more layers ===== | ||
Line 86: | Line 88: | ||
Feature sizes are 130nm and below. | Feature sizes are 130nm and below. | ||
- | Many dense interconnect layers, as with 5/7 layer devices, but with the addition of a dedicated power layer (or layers) on top metal for improved power distribution performance and to avoid cluttering high-density interconnect layers with power busing. | + | Many dense interconnect layers, as with 4-7 layer devices, but with the addition of a dedicated power layer (or layers) on top metal for improved power distribution performance and to avoid cluttering high-density interconnect layers with power busing. |
- | Example image (Xilinx XC6SLX4, 45nm Samsung process, | + | Example image (Xilinx XC6SLX4, 45nm Samsung process, |
{{: | {{: | ||
+ | |||
+ | ====== Typical layer dimensions ====== | ||
+ | |||
+ | This section prepared initially to better understand typical layer thickness to understand how to improve wet etching processes and to better understand limitations of lapping. | ||
+ | |||
+ | |||
+ | ===== Bipolar ===== | ||
+ | |||
+ | |||
+ | ===== NMOS/PMOS ===== | ||
+ | |||
+ | |||
+ | ===== CMOS ===== | ||
+ | |||
+ | Samsung KM44C4000J-7 16 Megabit DRAM, Report Number: SCA 9311-300I vertical dimensions: | ||
+ | * Die thickness: 0.3 mm | ||
+ | * Die coat: 9.5 um | ||
+ | * Die coat: "A patterned (to clear bond pads) polyimide die coat was present to protect against alpha particle-induced leakage. Coverage was good." | ||
+ | * Passivation 3: 0.55 um | ||
+ | * Passivation 2: 0.3 um | ||
+ | * Passivation 1: 0.1 um | ||
+ | * M2 (Al): 0.9 um | ||
+ | * Glass 3 ILD: 0.4 um | ||
+ | * Glass 2 ILD: 0.4 um | ||
+ | * Glass 1 ILD: ~0.08 um | ||
+ | * M1 cap: 0.04 um | ||
+ | * M1 (Al): 0.55 um | ||
+ | * M1 barrier: 0.15 um | ||
+ | * Intermediate glass 2: 0.5 um | ||
+ | * Polycide-silicide: | ||
+ | * Polycide-silicide poly 4: ~0.05 um | ||
+ | * Intermediate glass 1: 0.2 um | ||
+ | * Oxide on poly 3: 0.1 um | ||
+ | * Poly 3: 0.1 um | ||
+ | * Capacitor dielectric: ~0.015 um | ||
+ | * Poly 2: 0.15 um | ||
+ | * Interpoly oxide total: 0.35 um | ||
+ | * Interpoly oxide nitride: ~0.04 um | ||
+ | * Poly 1: 0.2 um | ||
+ | * Local oxide (under poly 1): 0.3 um | ||
+ | * Oxide on N+: ~0.08 um | ||
+ | * Oxide on P+: ~0.06 um | ||
+ | * N+ source/ | ||
+ | * P+ source/ | ||
+ | * (likely) N-well: 4.5 um | ||
+ | |||
+ | |||
+ | [Interconnect Scaling] says that ITRS '99 " | ||
+ | * 0.18 um process | ||
+ | * Global (ie long) height: ~1.16 um | ||
+ | * Semiglobal height: ~0.64 um | ||
+ | * Local height: ~0.36 um | ||
+ | |||
+ | TODO: Add extracted design rules from XC2C32A | ||
+ | |||
+ | ==== Thick metal process ==== | ||
+ | |||
+ | [Thick Metal] has: | ||
+ | * 0.8 um process | ||
+ | * Intended use: maybe RF ICs | ||
+ | * Inter-Metal Dielectric (IMD) thickness: 1.1 um | ||
+ | * Metal thickness: 2.1 um or 2.1 um | ||
+ | |||
+ | ===== BiCMOS ===== | ||
+ | |||
+ | |||
+ | ====== References ====== | ||
+ | |||
+ | * Thick Metal CMOS Technology on High Resistivity Substrate and Its Application to Monolithic L-Band CMOS LNAs (" | ||
+ | * Interconnect Scaling: http:// |
process_tech.1358358824.txt.gz · Last modified: 2013/10/20 14:59 (external edit)