Transcript Memory & Special ckts

11/14
1
Latch Circuit
Using Cross coupled inverters
Figure 11.1 (a) Basic latch. (b) The latch with the feedback loop opened. (c) Determining the operating
point(s) of the latch.
11/14
Memory_special.2
SR latch
Set – Reset using NOR gates
Reset
Set
Figure 11.2 (a) The set/reset (SR) flip-flop and (b) its truth table.
11/14
Memory_special.3
CMOS - Clocked SR Latch
Clock f
Set
Clock f
Reset
Figure 11.3 CMOS implementation of a clocked SR flip-flop. The clock signal is denoted by f.
11/14
Memory_special.4
Clocked SR Latch - analysis
Path used for
Simplified Circuit
Analysis
clock
Set
Pull Q! node to
Gnd
To flip latch to Hi
Figure 11.4 The relevant portion of the flip-flop circuit of Fig. 11.3 for determining the minimum W/L ratios of Q5 and
Q6 needed to ensure that the flip-flop will switch.
11/14
Memory_special.5
Simpler CMOS Clocked SR Latch
clock
clock
Set
reset
S – R inputs sampled by clock
Figure 11.5 A simpler CMOS implementation of the clocked SR flip-flop. This circuit is popular as the
basic cell in the design of static random-access memory (SRAM) chips.
11/14
Memory_special.6
D latch
Clocked two phase – non-overlapping clocks
? When
is output
Q valid
?
Figure 11.7 A simple implementation of the D flip-flop. The circuit in (a) utilizes the two-phase
nonoverlapping clock whose waveforms are shown in (b).
11/14
Memory_special.7
Master Slave
D flip flop
DFF
implementation
When is output Q
valid ?
Edge triggered
Figure 11.8 (a) A master–slave D flip-flop. The switches can be, and usually are, implemented with
CMOS transmission gates. (b) Waveforms of the two-phase nonoverlapping clock required.
11/14
Memory_special.8
Clocking Methodology
Setup – Hold time ?
Clk
Setup Hold
Setup
Hold
.
.
.
.
.
.
Don’t Care
.
.
.
•
•
•
•
•
.
.
.
All storage elements are clocked by the same clock edge
Cycle Time = CLK-to-Q + Longest Delay Path + Setup + Clock Skew
(CLK-to-Q + Shortest Delay Path - Clock Skew) > Hold Time
Setup time: minimum time data signal held steady before clock edge so data
reliably sampled.
Hold time: minimum time data signal held steady after the clock edge so data
reliably sampled
Memory_special.9
11/14
Latch & FF timings
11/14
10
Memory_special.10
Monostable Circuit a.k.a. one shot –
Multivibrator
Creates a single pulse –
width = T
from a trigger pulse
Figure 11.10 A monostable circuit using CMOS NOR gates. Signal source vI supplies positive trigger
pulses.
11/14
Memory_special.11
Monostable timing diagram - Analysis
G1 output
input pulse
G2 :Output pulse
11/14
Figure 11.13 Timing diagram for the monostable circuit in Fig. 11.10.
Memory_special.12
Multivibrator Circuit
= Oscillator
As C charges up (eg VO2= VDD;
Vt1 decays until G2 trip and goes
high
Figure 11.15 (a) A simple astable multivibrator circuit using CMOS gates. (b) Waveforms for the astable circuit in (a).
The diodes at the gate input are assumed to be ideal and thus to limit the voltage vI1 to 0 and VDD.
11/14
Memory_special.13
Ring
Oscillator
A better
Oscillator
3 – 5 inverters
In cascade
Frequency is function of inverter prop delay
Figure 11.16 (a) A ring oscillator formed by connecting three inverters in cascade. (Normally at least five inverters are
used.) (b) The resulting waveform. Observe that the circuit oscillates with frequency 1/6tP.
11/14
Memory_special.14
Main Memory Background
• Performance of Main Memory:
– Latency: Cache Miss Penalty
• Access Time: time between request and word arrives
• Cycle Time: time between requests
– Bandwidth: I/O & Large Block Miss Penalty (L2)
• Main Memory : DRAM: Dynamic Random Access Memory
– Dynamic: must be refreshed periodically (8 ms, 1% time)
– Addresses divided into 2 halves (Memory as a 2D matrix):
• RAS or Row Address Strobe
• CAS or Column Address Strobe
• Cache :: SRAM: Static Random Access Memory
– No refresh (6 transistors/bit vs. 1 transistor
Size: DRAM/SRAM 4-8,
Cost/Cycle time: SRAM/DRAM 8-16
11/14
Memory_special.15
Magnetic Core Memories (50s & 60s)
IBM 405 Alphabetical
Accounting Machine.
First magnetic core memory
• Core Memory stored data as magnetization in iron rings
– Iron “cores” woven into a 2-dimensional mesh of wires
11/14
Memory_special.16
Semiconductor Memory, DRAM
• Semiconductor memory  competitive in early 1970s
– Intel formed to exploit market for semiconductor memory
• First commercial DRAM  Intel 1103
– 1Kbit of storage on single chip
– charge on a capacitor used to hold value
• Semiconductor memory quickly replaced core in 1970s
• September 2007 , 1GB DRAM < $30
– Individuals can easily afford to fill a 32-bit address space with DRAM
(4GB)
11/14
Memory_special.17
IC memory classification
Volatile memories
Non-volatile memories
Lose data when power down
Keep data without power supply
SRAM
ROM
DRAM
PROM
EPROM
EEPROM
FLASH EEPROM
11/14
Memory_special.18
SRAM
Organization
square array
• Array  2M rows ´ 2N
columns
• can be organized as 1, 4, 8
or 16-bit wide
• 64M-bit  64M x 1
• 226 bit address
• Cells  square array
•
design / layout efficiency
Figure 11.17 A 2M+N-bit memory chip organized as an array of 2M rows ´ 2N columns.
11/14
Memory_special.19
1Mb SRAM
organization
• 1M-bit  1024 rows,
1024 columns
• Each cell connected to
• a row line (word line)
AND
• A column line (bit line)
• Cell selected by activating
word & bit lines
• Row & column decoders
used
• Sense amps. Detect low
bit line voltages
Figure 11.17 A 2M+N-bit memory chip organized as an array of 2M rows ´ 2N columns.
11/14
Memory_special.20
SRAM Cell
Stores one bit
Word Line
Bit Line
Bit Line!
Cell looks like ???
Figure 11.18 A CMOS SRAM memory cell.
11/14
Memory_special.21
SRAM Cell
Circuit analysis –
READ
cell contains “1”
1
_
B 0
Q1 ON
Figure 11.19 Relevant parts of the SRAM cell circuit during a read operation when the cell is storing a logic 1. Note
that initially vQ = VDD and vQ = 0. Also note that the B and B lines are usually precharged to a voltage of about VDD/2.
However, in Example 11.2, it is assumed for simplicity that the precharge voltage is VDD.
11/14
Memory_special.22
SRAM Cell
Circuit analysis – WRITE “0” operation
Figure 11.20 Relevant parts of the SRAM circuit during a write operation. Initially, the SRAM has a
stored 1 and a 0 is being written. These equivalent circuits apply before switching takes place. (a) The
circuit is pulling node Q up toward VDD/2. (b) The circuit is pulling node Q down toward VDD/2.
11/14
Memory_special.23
SRAM
organization
• Sense amps. Detect bit
line voltages
Figure
11/14 11.17 A 2M+N-bit memory chip organized as an array of 2M rows ´ 2N columns.
Memory_special.24
SRAM
Sense Amps detect “sense”
bit line voltage
small voltage swing for
speed
Amplified by sense amp
Figure 11.23 A differential sense amplifier connected to the bit lines of a particular column. This arrangement can be
used directly for SRAMs (which utilize both the B and B lines). DRAMs can be turned into differential circuits by
using the “dummy cell” arrangement shown in Fig. 11.25.
11/14
Memory_special.25
SRAM Cell
Bit line voltage sensing waveforms
Small voltage
swings
Figure 11.24 Waveforms of vB before and after the activation of the sense amplifier. In a read-1
operation, the sense amplifier causes the initial small increment DV(1) to grow exponentially to VDD. In
a read-0 operation, the negative DV(0) grows to 0. Complementary signal waveforms develop on the B
line.
11/14
Memory_special.26
SRAM Row decode
distributed NOR
Vdd
Dynamic
precharge
000
001
010
011
111
Figure 11.26 A NOR address decoder in array
form. One out of eight lines (row lines) is
selected using a 3-bit address.
11/14
_
A2
A2
_
A1
A1
_
A0
A
Memory_special.27
0
SRAM Column decode
Figure 11.27 A column decoder realized by a combination of a NOR decoder and a pass-transistor
multiplexer.
11/14
Memory_special.28
Register File Cell
2 Read port1 & 1 write port
Read port 0
Read port 1
Write
port
Write “0”
11/14
Write “1”
Memory_special.29
DRAM
latch (static) replaced by cell capacitor
Capacitive cell must be
periodically recharged
To retain value stored in cell
Figure 11.21 The one-transistor dynamic RAM cell.
11/14
Memory_special.30
One Transistor DRAM - FYI
1-T DRAM Cell
word
access
transistor
VREF
bit
TiN top electrode (VREF)
Ta2O5 dielectric
Storage
capacitor (FET
gate, trench,
stack)
poly
word
line
11/14
W bottom
electrode
access
transistor
Memory_special.31
DRAM Architecture
bit lines
Col.
2M
Col.
1
N+M
Row 1
Row Address
Decoder
N
M
word lines
Row 2N
Column Decoder &
Sense Amplifiers
Data
Memory cell
(one bit)
D
• Bits stored in 2-dimensional arrays on chip
• DRAM chips have 4 or more logical banks on each chip
– each logical bank physically implemented as smaller arrays
11/14
Memory_special.32
DRAM  1-Transistor Memory
Cell
row select
• Write:
– 1. Drive bit line
– 2. Select row
• Read:
– 1. Precharge bit line to Vdd/2
– 2. Select row
– 3. Cell and bit line share charges
bit
• Very small voltage changes on the bit line
– 4. Sense (fancy sense amp)
• Can detect changes of ~1 million electrons
– 5. Write: restore the value
• Refresh
– dummy read to every cell. – on a row basis
11/14
Memory_special.33
DRAM Operation Read & Write
Three steps to read/write
• Row access (RAS)
–
–
–
–
decode row address, large number bits / row
Bit lines share charge with storage cell
Small voltage detected by sense amplifiers - latch whole row of bits
sense amplifiers drive bit lines full rail to recharge storage cells
• Column access (CAS)
– decode column address to select small number of sense amplifier latches (4, 8,
16, or 32 bits depending on DRAM package)
– read,  send latched bits out to pins
– Write  change sense amplifier latches, storage cells charged to required
value
– can perform multiple column accesses on same row without another row access
(burst mode)
• Precharge
– charges bit lines to known value, required before next row access
• DRAM standards (DDR, DDR2, DDR4, GDDR6, RDRAM)
11/14
Memory_special.34
DRAM Read Timing
• DRAM access:
RAS_L
– assertion of RAS_L
– Followed by CAS
CAS_L
A
DRAM Read Cycle Time
WE_L
256K x 8
DRAM
9
OE_L
D
8
RAS_L
CAS_L
A
Row Address
Col Address
Junk
Row Address
Col Address
Junk
WE_L
OE_L
D
High Z
Junk
Data Out
Read Access
Time
Early Read Cycle: OE_L asserted before CAS_L
11/14
High Z
Data Out
Output Enable
Delay
Late Read Cycle: OE_L asserted after CAS_L
Memory_special.35
SDRAM timing (Single Data Rate)
CAS
Precharge
RAS
x
CAS Latency
Burst
READ
• Micron 128M-bit dram (using 2Meg16bit4bank ver)
11/14
– Row (12 bits), bank (2 bits), column (9 bits)
Memory_special.36
Double-Data Rate (DDR2) DRAM
200MHz
Row
Column
Precharge
Row’
Data
[ Micron, 256Mb DDR2 SDRAM datasheet ]
11/14
400Mb/s
DataMemory_special.37
Rate
DRAM has very low voltage swings
Differential sensing used FYI
Figure 11.25 An arrangement for obtaining differential operation from the single-ended DRAM cell.
Note the dummy cells at the far right and far left.
11/14
Memory_special.38
Non-volatile memory
Floating gate memories
11/14
Memory_special.39
USB Flash Stick
11/14
Memory_special.40
ROM Cell
Read only memory
Programmed by
presence / absence
of transistor
Figure 11.29 A simple MOS ROM organized as 8 words ´ 4 bits.
11/14
Only one word line active at a time
Memory_special.41
NMOS transistor review
Source (S), Drain (D), Gate (G)
L channel length, W width of transistor
NMOS transistor cross-section
11/14
Memory_special.42
EPROM Cell
Additional floating gate
Stores “1” / “0”
NMOS
Figure 11.30 (a) Cross section and (b) circuit symbol of the floating-gate transistor used as an EPROM
cell.
11/14
Memory_special.43
“programming” a floating gate cell
changes VT
• MOS transistor: 1 fixed VT
• Flash memory cell: VT can be changed by
program/erase
MOS transistor
Id
Floating gate transistor
Id
programming
erasing
Vgs
Vgs
VT
11/14
Memory_special.44
Why Flash Memory?
• read/written – doesn’t need POWER for data retention.
• fast.
• durable up to alimit
Memory_special.45
Applications
many, many, more!
Memory_special.46
Device Structure
• Similar to MOSFET
• Added Floating Gate (FG)
between Control Gate (CG) and
inversion layer
• FG surrounded by insulators
• FG traps electrons (~50 years)
• Charged FG disrupts / affects
inversion layer
• Current flows from the drain to
source via inversion layer
Write Function – Logic 0
• Electrons injected into the
FG via hot-electron
injection
• Vt is modified, changing
current flow (less)
• Reduced current flow in
inversion layer  logic 0
Effect of “programming”
floating gate
Figure 11.31 Illustrating the shift in the iD–vGS characteristic of a floating-gate transistor as a result of programming.
11/14
Memory_special.49
Lifespan
• Hot electron injection or tunneling results in
device deterioration
• Electrons get trapped in oxide layer
• Trapped electrons in oxide disrupt Vt
• Flash Memory can “wear out”
• Between 1,000 10,000 and up to 100,000 write
cycles
Memory_special.50
Major Players in Flash Market
•
•
•
•
•
•
•
•
Samsung
Toshiba
AMD (Spansion)
Intel & Micron  IMI Flash Technologies
Hynix
SST
ST Microelectronics
Sharp
Memory_special.51
Motivations
• Big & growing $$$$ mkt
• All handheld / mobile
products
• Optimists predict NAND
Flash will replace CD/DVD
• NAND Flash prices
continue to fall
• Storage densities following
Moore’s Law
Flash program and erase methods
11/14
Memory_special.53
EEPROM, Flash, NAND, NOR
• EPROM – original ; complete erase before program
• EEPROM – electrically erasable & programmable
 Word control
• NOR Flash ; finer erase / program; word control
• NAND flash for use in memory, SSD,..
11/14
Memory_special.54