Advanced VLSI Design Memory CMPE 640
Transcript of Advanced VLSI Design Memory CMPE 640
1
Adv CMPE 640
Me
ses dynamic DRAM hysical location of data.ter) and Content Access
h larger than read time.
tion.ata. More difficult to
anced VLSI Design Memory
mory
Can be categorized into: Read Write Memory (RWM)
Random Access Memory (RAM): static SRAM (faster) ver(smaller) structures possible. Access time independent of p
Non-RAM: Serial Access Memory (FIFO, LIFO, Shift regisMemory (CAM). Non-uniform access time.
Non-volatile Read Write Memory (NVRWM): write time muc
EPROM, E2PROM, FLASH Read Only Memory (ROM)
A second classification for RAMs and ROMs: Static-load: no clock required. Synchronous: require a clock edge to enable memory opera Asynchronous: recognize address changes and output new d
build.
2
Adv CMPE 640
Meide (typically 1, 4 or 8
g exactly, high.
or small memoriesge memories
rd (where word = 8 bits) lines, provided by someice.
anced VLSI Design Memory
mory ArchitectureIn order to build an N-word memory where each word is M bits wbits), a straightforward approach is to stack memory:
This approach is not practical.What can we do?
S0
S1
S2
SN-2
SN-1
N w
ords
Word 0
Word 1
Word 2 Storage cell
Word N-2
Word N-1
Input-Output(M bits)
A word is selected by settinone of the select bits, Sx
This approach works well fbut has problems for lar
For example, to build a 1Mwo memory, requires 1M select
off-chip dev
3
Adv CMPE 640
Me
since the vertical wires
l
es the number of ex
ss pins from
to 20.
anced VLSI Design Memory
mory ArchitectureAdd a decoder to solve the package problem:
This does not address the memory aspect ratio problem:
The memory is 128,000 time higher than wide (220/23) !Besides the bizarre shape factor, the design is extremely slow
are VERY long (delay is at least linear to length).
S0
S1
S2
SN-2SN-1
Word 0Word 1Word 2 Storage cel
Word N-2
Word N-1
Input-Output(M bits)
Dec
oder
A0A1
A2
AK-1
K = log2N
one-hot
Bin
ary
enco
ded
addr
ess
This reduc
addre
1M
4
Adv CMPE 640
Mer an aspect ratio of unity.:
Storage cell
Word line
Sense amps and drivers not shown
anced VLSI Design Memory
mory ArchitectureThe vertical and horizontal dimensions are usually very similar, foMultiple words are stored in each row and selected simultaneously
S0
S1
S2
SN-2SN-1
Input-Output(M bits)
AKAK+1
AK+2
AL-1
Column address = A0
AK-1
Bit line
A0 to AK-1
Row address = AK to AL-1
A column decoder is added to select the desired word from a row.
Column decoder
Row
Dec
oder
5
Adv CMPE 640
Mes.nes.blem:
RA
CA
BA
ck P-1
A
32 blocks withblock.ock: 1024 rowslumns.
anced VLSI Design Memory
mory ArchitectureThis strategy works well for memories up to 64 Kbits to 256 KbitLarger memories start to suffer excess delay along bit and word liA third dimension is added to the address space to solve this pro
Global Data bus
owddress
olumnddress
lockddress
Block 0 Block i Blo
Globalamplifier/driver
I/O
ddress: [Row][Block][Col]
Block selector
4 Mbit: P =128Kbits/128Kbit bland 128 co
6
Adv CMPE 640
Me
te memory.
r word = 28 = 256 words.
e buffers
2n-k bits
anced VLSI Design Memory
mory: Architecture
An example:
For example: Let N = 1,048,576 and M = 8 bits for a 1 million byn = log2N = 20, k = 8 and m = log2M = 3.
Then there are 2n-k rows = 212 = 4096 and 2k+m columns/23 bits pe
Row decoder
Row decoder
Row decoder
Row decoder
Column decoder column mux, sense amp, writ
2m+k bits
A0
Ak
Ak+1
An-1
Ak-1[An-1..Ak][Ak-1..A0]
7
Adv CMPE 640
RO
BL
BL
S used to holdh. n-MOSes pull-down
S
fe
anced VLSI Design Memory
MROM cells are permanently fixed: Several possibilities:
BLWL
1
0
BLWL
Diode supplies currentto raise BL (bitline) forall cells on the row.
BLWL
BJT supplies currentto raise BL for eachcell on the row. RequiresVDD to be routed.
WL
WL
WL
p-MOBL higprovidpath.
psuedo n-MONOR gate.
Resistance on/p should bat least 4.
8
Adv CMPE 640
No
h modifications to
in create high electric
leaving it negatively
ate does not permit the
in
20V
anced VLSI Design Memory
n-volatile Read-Write MemoriesVirtually identical in structure to ROMs.Selective enabling/disabling of transistors is accomplished througthreshold voltage. This is accomplished through a floating gate.
Applying a high voltage (15 to 20 V) between source and gate-drafield and causes avalanche injection to occur.
Hot electrons traverse first oxide and get trapped on floating gate,charged.
This increases the threshold voltage to ~7V. Applying 5V to the gdevice to turn on.
tox
tox
Source Dra
Substrate
n+ n+
Gate
Floating Gate- - - - - - - -
-5V afterprogrammingthis device off
-
20V
9
Adv CMPE 640
No
he various classes of
ing and access device.
anced VLSI Design Memory
n-volatile Read-Write Memories
The method of erasing is the main differentiating factor between treprogrammable nonvolatile memories.
EPROM:UV light renders oxide slightly conductive.Erase is slow (seconds to several minutes).Programming is slow (5-10 microsecs per word).Limited number of programming cycles - about 1000.Very dense - single transistor functions as both the programm
10
Adv CMPE 640
No
via Fowler-Nordheim
riting.in
10V
ults in a
or.
oblem:
turned off.
anced VLSI Design Memory
n-volatile Read-Write Memories
EEPROM or E2PROM:Very thin oxide allows electrons to flow to and from the gatetunneling with VGD applied.
Erasure is achieved by reversing the voltage applied during w
tox
tox
Source Dra
Substrate
n+ n+
Gate
Floating Gate- - - - - - - -
thin tunneling ox
BLWL
VDD
Removing too much charge res
Remedy: Add an access transist
Threshold control becomes a pr
depletion device that cannot be
11
Adv CMPE 640
No
M.
ld during erasure - mak-ent device.
in
12V
anced VLSI Design Memory
n-volatile Read-Write Memories Flash EEPROM:
Combines density adv. of EPROM with versatility of EEPROUses avalanche hot-electron-injection approach to program.Erasure performed using Fowler-Nordheim tunneling.Monitoring control hardware checks the value of the threshoing sure the unprogrammed transistor remains an enhancem
Programming performed by applying 12V to gate and drain.Erasure performed with gate grounded and source at 12V.
tox
Source Dra
Substrate
n+ n+
Gate
Floating Gate- - - - - - - -
thin tunneling ox
-
erasure
- programming
12V
12V
12
Adv CMPE 640
Re
w
anced VLSI Design Memory
ad-Write Memories (RAM)SRAM:
ord line
VDD
bit bit
13
Adv CMPE 640
Re
anced VLSI Design Memory
ad-Write Memories (RAM)Generic RAM circuit:
Bit LineConditioning
clocks
RAM cell
Sense AmpColumn MuxWrite Buffers
n-1;k
k-1;0
read-data write-dataAddress
bit bit
word linerow decoder
column decoder
14
Adv CMPE 640
Re
ine improves perfor-
w
ize speed, annels as
ge devices.
precharge
bit, bit
word
data
anced VLSI Design Memory
ad-Write Memories (RAM)SRAM: Read Operation
Precharging bit and bit_bar to 5V before enabling the word lmance.
precharge
ord
bit bit
data
To optimuse n-chprechar
VDD
15
Adv CMPE 640
Re
w
write-data
bit, bit
word
w
w
write
cell, cell
ginally.o pull Pbit below.
anced VLSI Design Memory
ad-Write Memories (RAM)SRAM: Write Operation:
ord
bit bit
N5 N6
N3 N4
N1 N2
rite-data
rite
cell cell
Zero stored in cell ori
Nd
Nd, N1, and N3 have tthe inverter threshold
0->11->0 Pbit
1
0
16
Adv CMPE 640
Re
read-data1
a
-read-port
Adv: Nomatterwhat theload, cellcannot beflipped.
d
anced VLSI Design Memory
ad-Write Memories (RAM)Register files:
4/1
2/2
2/3
2/1
8/1
4/1
4/1
4/1
4/1
write-data read-data0
ddr<3:0>
Single-write-port, doubleOverpowersweak feedbackinverter
Biased towardVSS to help write.
ecode
17
Adv CMPE 640
Re
g the cell contents.
SRAM cell.
anced VLSI Design Memory
ad-Write Memories (RAM)DRAM:
Refresh: Compensate for charge loss by periodically rewritinRead followed by a write operation.Typical refresh cycles occur every 1 to 4 milliseconds.
4 transistor DRAM created by simply eliminating the p tree in an
Logic 1 values are, of course, a threshold below VDD.
word line
bit bit
18
Adv CMPE 640
Re
n bit1 and assert write.rent).licity in both design and
write
bit1
read
bit2
X
V
anced VLSI Design Memory
ad-Write Memories (RAM)3T DRAM:
Most common method of refresh is to read bit2, place its inverse oPrecharge method of 'setting' bit2 is preferred (no steady-state curMemory structure of choice in ASICs because of its relative simpoperation.
write
readbit1 bit2
X
VDD-VTbit2 is either clamped to VDD or is precharged to either VDD or VDD-VT.
No device ratioing necessary here !
Note that this cell is inverting
19
Adv CMPE 640
Re
node X and node bit. the delta-V value is typ-
word-line
bit
X-VT
ng
V
e)Cx
(Cx + Cbit)
anced VLSI Design Memory
ad-Write Memories (RAM)1T DRAM
During read operation, charge redistribution occurs between Cx is typically 1 or 2 orders of magnitude smaller than Cbit so
ically 250 mV.
Most pervasive DRAM cell in commercial memory design.
word-line
bit
VDD
write read
X
Vpre = VDD/2
VDD
sensi
V = Vbit - Vpre = (Vx - Vpr
Cx Cbit
20
Adv CMPE 640
Re
ry in order for the cell to
operation.mposed onto the bit line
tance). Capacitance must
ts VT loss on storage
anced VLSI Design Memory
ad-Write Memories (RAM)1T DRAM observations:
Amplification of delta-V (through a sense amplifier) is necessabe functional.
Other cell designs used sense amps only to speed up the read The read-out operation is destructive ! Output of sense amp is iwith word-line high during read-out.
1T transistor requires an explicit capacitor (3T used gate capacibe large (~30fF) but area small - key challenge in design. Bootstrapping word-line to a value larger than VDD circumven
capacitor.
Vpre
V(1)Sense amp activated
Word-line activated
V(0)
21
Adv CMPE 640
Re
ed word.
t of the wordo the match line.
MOS ble.
ANY SRAM celll or bit/cell.
anced VLSI Design Memory
ad-Write Memories (RAM)Content Access Memory (CAM):
Determines if a match exists between a data word with a storUsed in Translation-look-aside buffers.
word line
VDDbit bit
cellcell
matchXOR function.
SRAMwith extran-channels
Each biis tied t
Dynamic or Pseudo n-implementations possi
to implement
Match is 0 ifhas bit/celequal to 1