Transcript 10b-technology

CMOS Layout
diffusion
poly
side view
cuts
top view
CSE 567 - Autumn 1998 - CMOS - 1
metal
A View of Interconnect Layers
CSE 567 - Autumn 1998 - CMOS - 2
Intel 4004 (1971)
CSE 567 - Autumn 1998 - CMOS - 3
Pentium II (1996)
CSE 567 - Autumn 1998 - CMOS - 4
Silicon in 2010
Die Area:
2.5x2.5 cm
Voltage:
0.6 V
Technology: 0.07 m
Density AccessTime
(Gbits/cm2)
(ns)
DRAM
8.5
10
DRAM (Logic)
2.5
10
SRAM (Cache)
0.3
1.5
Density
Max. Ave. Power Clock Rate
(Mgates/cm2)
(W /cm2)
(GHz)
Custom
25
54
3
Std. Cell
10
27
1.5
Gate Array
5
18
1
Single-Mask GA
2.5
12.5
0.7
FPGA
0.4
4.5
0.25
CSE 567 - Autumn 1998 - CMOS - 5
CMOS Inverter Layout
CSE 567 - Autumn 1998 - CMOS - 6
Layout
 4-input NAND gate
VDD
In1
In2
In3
In4
Out
In1
In2
In3
In4
CSE 567 - Autumn 1998 - CMOS - 7
Standard Cell Design Methodology
 Example “stick diagram”
metal1
VDD
Well
VSS
Routing Channel
signals
polysilicon
CSE 567 - Autumn 1998 - CMOS - 8
Different Layouts are Possible
 Different layouts for x =
VDD
VDD
x
x
GND
a
c
b
(a) Input order {a c b}
GND
a
b
c
(b) Input order {a b c}
CSE 567 - Autumn 1998 - CMOS - 9
The Barrel Shifter
A3
B3
Sh1
A2
B2
: Data Wire
Sh2
A1
B1
Sh3
A0
B0
Sh0
Sh1
Sh2
Sh3
CSE 567 - Autumn 1998 - CMOS - 10
: Control Wire
4x4 barrel shifter
A3
A2
A1
A0
Sh0
Sh1
Sh2
Sh3
CSE 567 - Autumn 1998 - CMOS - 11
Buffer
Logarithmic Shifter
Sh1 Sh1
Sh2 Sh2
Sh4 Sh4
A3
B3
A2
B2
A1
B1
A0
B0
CSE 567 - Autumn 1998 - CMOS - 12
0-7 bit Logarithmic Shifter
A
A
A
A
3
Out3
2
Out2
1
Out1
0
Out0
CSE 567 - Autumn 1998 - CMOS - 13
Latches

D
In

D

(b) Non-overlapping clocks

(a) Schematic diagram
Pseudo-static Latch
CSE 567 - Autumn 1998 - CMOS - 14
Edge-Triggered Register (Master/Slave)


D
A
In
B



Overlapping Clocks Can Cause
• Race Conditions
• Undefined Signals
CSE 567 - Autumn 1998 - CMOS - 15

Non-Overlapping Clocks


D
In



t12
CSE 567 - Autumn 1998 - CMOS - 16

Dynamic Register


In
D
Input Sampled


Output Enable
CSE 567 - Autumn 1998 - CMOS - 17
Registers and intervening logic
 Switches and/or gates compute new values to store on next clock cycle
straightforward implementation
CL
2
1
this circuit can use the entire clock cycle – no wasted time - a form of retiming
CL
CL
2
CSE 567 - Autumn 1998 - CMOS - 18
1
Memory – Static Register Cells
 8-transistor cell
bit
bit'
rd or wr
(rd or wr)'
sensing amplifier
rd/wr'
CSE 567 - Autumn 1998 - CMOS - 19
Memory – Dynamic Register Cells
 4-transistor cell
data-in
wr
dedicated busses for
reading and writing
rd
dynamic charge storage
must be refreshed
CSE 567 - Autumn 1998 - CMOS - 20
data-out'
Memory – Dynamic Register Cells (cont'd)
 3-transistor cell
pre-charge data-out' to
generate 1 outputs
data-in
no p-type transistors yield
wr
a very compact cell
rd
CSE 567 - Autumn 1998 - CMOS - 21
+5V
data-out'
Memory – Dynamic Register Cells (cont'd)
 1-transistor cell
precharge to
intermediate
voltage level
storage capacitor
is one end of transistor
charge sharing with
bus capacitance
(Ccell << Cbus)
destructive read
(must immediately
write back)
extra demands on
sense amplifier to
detect small changes
in bus charge
CSE 567 - Autumn 1998 - CMOS - 22
Read-only Memory Cells
 To store constants or other invariant data
 Popular for control implementation
bit1
bit2
bit3
read1
read2
programmable logic array structure
(exploits distributed NOR gate structure)
CSE 567 - Autumn 1998 - CMOS - 23
Multi-ported Register Cells
 Add more input and output transistors (similar for all variations)
 Not usually done for 1-transistor cells
bus2'
bus1
row-bus1
row-bus2
bus2
bus1'
CSE 567 - Autumn 1998 - CMOS - 24
Random access memory
 Decoding logic to select word
 Attempt to make critical lines (data lines) as short as possible
 Square aspect ratio rather than rectangular
2m k-bit words per row
n
address
of word
(n+m bits)
m
d
e
c
o
d
e
r
2n by 2m*k bits
memory
cell array
2n rows
multiplexer ( 2m :1)
k bits wide (k bits/word)
CSE 567 - Autumn 1998 - CMOS - 25
Decoders
 Decoder looks like
AND-plane of ROM
(all input combinations)
 usually replicated
throughout array
+5V
d
e
c
o
d
e
r
memory
cell array
n
rd/wr
(row select)
bit
CSE 567 - Autumn 1998 - CMOS - 26
bit'
Multiplexers
 Decode address into one-hot control signals
 Each bit passes through single n-device or pass gate
note: bits of word in row are interleaved
word1
word2
word3
word4
bit1
from a decoder
CSE 567 - Autumn 1998 - CMOS - 27
bit2
Content-Addressable Memory
 Compares input to all entries in memory in parallel
 applications: pattern recognition, TLBs, etc.
 Require and encoder to indicate where a match occurred
 perform inverse function of decoders
 take a one-hot collection of signals and encodes them
m bits
2n rows
content
addressable
memory
cell array
m
CSE 567 - Autumn 1998 - CMOS - 28
e
n
c
o
d
e
r
n
Content-Addressable Memory Cells
 Read and write like normal 6T memory cell
 Match signal is precharged to 1, pulled to 0 if no match
 send data on bit' and data' on bit for matching
 match remains 1 if and only if all bits in word match
row
select
match
bit
bit'
CSE 567 - Autumn 1998 - CMOS - 29