DIC-Lec13 [email protected] 1
Digital Integrated CircuitsLecture 13: SRAM
Chih-Wei Liu
VLSI Signal Processing LAB
National Chiao Tung University
DIC-Lec13 [email protected] 2
Outline
Memory Arrays
SRAM Architecture
SRAM Cell
Decoders
Column Circuitry
Multiple Ports
Serial Access Memories
DIC-Lec13 [email protected] 3
Memory Arrays
Memory Arrays
Random Access Memory Serial Access Memory Content Addressable Memory(CAM)
Read/Write Memory(RAM)
(Volatile)
Read Only Memory(ROM)
(Nonvolatile)
Static RAM(SRAM)
Dynamic RAM(DRAM)
Shift Registers Queues
First InFirst Out(FIFO)
Last InFirst Out(LIFO)
Serial InParallel Out
(SIPO)
Parallel InSerial Out
(PISO)
Mask ROM ProgrammableROM
(PROM)
ErasableProgrammable
ROM(EPROM)
ElectricallyErasable
ProgrammableROM
(EEPROM)
Flash ROM
DIC-Lec13 [email protected] 4
Array Architecture
2n words of 2m bits eachIf n >> m, fold by 2k into fewer rows of more columns
Good regularity – easy to designVery high density if good cells are used
row decoder
columndecoder
n
n-kk
2m bits
columncircuitry
bitline conditioning
memory cells:2n-k rows x2m+k columns
bitlines
wordlines
DIC-Lec13 [email protected] 5
12T SRAM Cell
Basic building block: SRAM CellHolds one bit of information, like a latchMust be read and written
12-transistor (12T) SRAM cellUse a simple latch connected to bitline46 x 75 λ unit cell
bit
write
write_b
read
read_b
DIC-Lec13 [email protected] 6
6T SRAM Cell
Cell size accounts for most of array sizeReduce cell size at expense of complexity
6T SRAM CellUsed in most commercial chipsData stored in cross-coupled inverters
Read:Precharge bit, bit_bRaise wordline
Write:Drive data onto bit, bit_bRaise wordline
bit bit_b
word
DIC-Lec13 [email protected] 7
SRAM Read
Precharge both bitlines highThen turn on wordlineOne of the two bitlines will be pulled down by the cellEx: A = 0, A_b = 1
bit discharges, bit_b stays highBut A bumps up slightly
Read stabilityA must not flip
bit bit_b
N1
N2P1
A
P2
N3
N4
A_b
word
0.0
0.5
1.0
1.5
0 100 200 300 400 500 600time (ps)
word bit
A
A_b bit_b
DIC-Lec13 [email protected] 8
SRAM Read
Precharge both bitlines highThen turn on wordlineOne of the two bitlines will be pulled down by the cellEx: A = 0, A_b = 1
bit discharges, bit_b stays highBut A bumps up slightly
Read stabilityA must not flipN1 >> N2
bit bit_b
N1
N2P1
A
P2
N3
N4
A_b
word
0.0
0.5
1.0
1.5
0 100 200 300 400 500 600time (ps)
word bit
A
A_b bit_b
DIC-Lec13 [email protected] 9
SRAM Write
Drive one bitline high, the other lowThen turn on wordlineBitlines overpower cell with new valueEx: A = 0, A_b = 1, bit = 1, bit_b = 0
Force A_b low, then A rises highWritability
Must overpower feedback inverter
time (ps)
word
A
A_b
bit_b
0.0
0.5
1.0
1.5
0 100 200 300 400 500 600 700
bit bit_b
N1
N2P1
A
P2
N3
N4
A_b
word
DIC-Lec13 [email protected] 10
SRAM Write
Drive one bitline high, the other lowThen turn on wordlineBitlines overpower cell with new valueEx: A = 0, A_b = 1, bit = 1, bit_b = 0
Force A_b low, then A rises highWritability
Must overpower feedback inverterN2 >> P1
time (ps)
word
A
A_b
bit_b
0.0
0.5
1.0
1.5
0 100 200 300 400 500 600 700
bit bit_b
N1
N2P1
A
P2
N3
N4
A_b
word
DIC-Lec13 [email protected] 11
SRAM Sizing
High bitlines must not overpower inverters during readsBut low bitlines must write new value into cell
bit bit_b
med
A
weak
strong
med
A_b
word
DIC-Lec13 [email protected] 12
SRAM Column Example
Read Write
H H
SRAM Cell
word_q1
bit_v1f
bit_b_v1f
out_v1rout_b_v1r
φ1
φ2
word_q1
bit_v1f
out_v1r
φ2
MoreCells
Bitline Conditioning
φ2
MoreCells
SRAM Cell
word_q1
bit_v1f
bit_b_v1f
data_s1
write_q1
Bitline Conditioning
DIC-Lec13 [email protected] 13
SRAM Layout
Cell size is critical: 26 x 45 λ (even smaller in industry)Tile cells sharing VDD, GND, bitline contacts
VDD
GND GNDBIT BIT_B
WORD
Cell boundary
DIC-Lec13 [email protected] 14
Decoders
n:2n decoder consists of 2n n-input AND gatesOne needed for each row of memoryBuild AND from NAND or NOR gates
Static CMOS Pseudo-nMOS
word0
word1
word2
word3
A0A1
A1word
A0 1 1
1/2
2
4
8
16word
A0
A1
11
11
4
8word0
word1
word2
word3
A0A1
DIC-Lec13 [email protected] 15
Decoder Layout
Decoders must be pitch-matched to SRAM cellRequires very skinny gates
GND
VDD
word
buffer inverterNAND gate
A0A0A1A2A3 A2A3 A1
DIC-Lec13 [email protected] 16
Large Decoders
For n > 4, NAND gates become slowBreak large gates into multiple smaller gates
word0
word1
word2
word3
word15
A0A1A2A3
DIC-Lec13 [email protected] 17
Predecoding
Many of these gates are redundantFactor out commongates into predecoderSaves areaSame path effort
A0
A1
A2
A3
word1
word2
word3
word15
word0
1 of 4 hotpredecoded lines
predecoders
DIC-Lec13 [email protected] 18
Column Circuitry
Some circuitry is required for each columnBitline conditioningSense amplifiersColumn multiplexing
DIC-Lec13 [email protected] 19
Bitline Conditioning
Precharge bitlines high before reads
Equalize bitlines to minimize voltage difference when using sense amplifiers
φbit bit_b
φ
bit bit_b
DIC-Lec13 [email protected] 20
Sense Amplifiers
Bitlines have many cells attachedEx: 32-kbit SRAM has 128 rows x 256 cols
128 cells on each bitline
tpd ∝ (C/I) ΔVEven with shared diffusion contacts, 64C of diffusion capacitance (big C)
Discharged slowly through small transistors (small I)
Sense amplifiers are triggered on small voltage swing (reduce ΔV)
DIC-Lec13 [email protected] 21
Differential Pair Amp
Differential pair requires no clockBut always dissipates static power
bit bit_bsense_b sense
N1 N2
N3
P1 P2
DIC-Lec13 [email protected] 22
Clocked Sense Amp
Clocked sense amp saves powerRequires sense_clk after enough bitline swingIsolation transistors cut off large bitline capacitance
bit_bbit
sense sense_b
sense_clk isolationtransistors
regenerativefeedback
DIC-Lec13 [email protected] 23
Twisted Bitlines
Sense amplifiers also amplify noiseCoupling noise is severe in modern processesTry to couple equally onto bit and bit_bDone by twisting bitlines
b0 b0_b b1 b1_b b2 b2_b b3 b3_b
DIC-Lec13 [email protected] 24
Column Multiplexing
Recall that array may be folded for good aspect ratioEx: 2k word x 16 folded into 256 rows x 128 columns
Must select 16 output bits from the 128 columnsRequires 16 8:1 column multiplexers
DIC-Lec13 [email protected] 25
Tree Decoder Mux
Column mux can use pass transistorsUse nMOS only, precharge outputs
One design is to use k series transistors for 2k:1 muxNo external decoder logic needed
B0 B1 B2 B3 B4 B5 B6 B7 B0 B1 B2 B3 B4 B5 B6 B7A0
A0
A1
A1
A2
A2
Y Yto sense amps and write circuits
DIC-Lec13 [email protected] 26
Single Pass-Gate Mux
Or eliminate series transistors with separate decoder
A0A1
B0 B1 B2 B3
Y
DIC-Lec13 [email protected] 27
Ex: 2-way Muxed SRAM
MoreCells
word_q1
write0_q1
φ2
MoreCells
A0
A0
φ2
data_v1
write1_q1
DIC-Lec13 [email protected] 28
Multiple Ports
We have considered single-ported SRAMOne read or one write on each cycle
Multiported SRAM are needed for register files
Examples:Multicycle MIPS must read two sources or write a result on some cycles
Pipelined MIPS must read two sources and write a third result each cycle
Superscalar MIPS must read and write many sources and results each cycle
DIC-Lec13 [email protected] 29
Dual-Ported SRAM
Simple dual-ported SRAMTwo independent single-ended readsOr one differential write
Do two reads and one write by time multiplexingRead during ph1, write during ph2
bit bit_b
wordBwordA
DIC-Lec13 [email protected] 30
Multi-Ported SRAM
Adding more access transistors hurts read stabilityMultiported SRAM isolates reads from state nodeSingle-ended design minimizes number of bitlines
bA
wordBwordA
wordDwordC
wordFwordE
wordG
bB bC
writecircuits
readcircuits
bD bE bF bG
DIC-Lec13 [email protected] 31
Serial Access Memories
Serial access memories do not use an addressShift RegistersTapped Delay LinesSerial In Parallel Out (SIPO)Parallel In Serial Out (PISO)Queues (FIFO, LIFO)
DIC-Lec13 [email protected] 32
Shift Register
Shift registers store and delay dataSimple design: cascade of registers
Watch your hold times!
clk
Din Dout8
DIC-Lec13 [email protected] 33
Denser Shift Registers
Flip-flops aren’t very area-efficientFor large shift registers, keep data in SRAM insteadMove read/write pointers to RAM rather than data
Initialize read address to first entry, write to lastIncrement address on each cycle
Din
Dout
clk
counter counterreset
00...00
11...11
readaddr
writeaddr
dual-portedSRAM
DIC-Lec13 [email protected] 34
Tapped Delay Line
A tapped delay line is a shift register with a programmable number of stagesSet number of stages with delay controls to mux
Ex: 0 – 63 stages of delay
SR
32
clk
Din
delay5
SR
16
delay4
SR
8
delay3S
R4
delay2
SR
2
delay1
SR
1
delay0
Dout
DIC-Lec13 [email protected] 35
Serial In Parallel Out
1-bit shift register reads in serial dataAfter N steps, presents N-bit parallel output
clk
P0 P1 P2 P3
Sin
DIC-Lec13 [email protected] 36
Parallel In Serial Out
Load all N bits in parallel when shift = 0Then shift one bit out per cycle
clkshift/load
P0 P1 P2 P3
Sout
DIC-Lec13 [email protected] 37
Queues
Queues allow data to be read and written at different rates.Read and write each use their own clock, dataQueue indicates whether it is full or emptyBuild with SRAM and read/write counters (pointers)
Queue
WriteClk
WriteData
FULL
ReadClk
ReadData
EMPTY
DIC-Lec13 [email protected] 38
FIFO, LIFO Queues
First In First Out (FIFO)Initialize read and write pointers to first element
Queue is EMPTY
On write, increment write pointer
If write almost catches read, Queue is FULL
On read, increment read pointer
Last In First Out (LIFO)Also called a stack
Use a single stack pointer for read and write