Professional Documents
Culture Documents
S. A. Ibrahim
Ain Shams University
ICL
Outline
S. A. Ibrahim
2
Dynamic CMOS
S. A. Ibrahim
3
Dynamic Gate Example
off
Clk Mp Clk Mp on
1
Out Out
In1 CL ((AB)+C)
A
In2 PDN
C
In3
B
Clk Me
off
Clk Me on
S. A. Ibrahim
5
Properties of Dynamic Gates (1)
S. A. Ibrahim
6
Properties of Dynamic Gates (2)
Overall power dissipation usually higher than
static CMOS.
No static current path ever exists between VDD and
GND (including Psc).
No glitching
Higher transition probabilities 01 P( Out 0 )
Extra load on Clk
PDN starts to work as soon as the input
signals exceed VTn, so VM, VIH and VIL equal to
VTn.
Low noise margin (NML)
Needs a precharge/evaluate clock.
S. A. Ibrahim
7
Logical Effort of Dynamic Gates
Clk Mp Clk Mp
Out Out
A 3
IN 2
B 3
Clk Me 2 Clk Me 3
Cgate,IN= 2 Cgate,A= 3
LEIN= 2/3 LEA= 1
S. A. Ibrahim
8
P-type Dynamic Gates
off
Clk Me Clk Me on
In1
A
In2 PUN
C
In3
Out B
0
Clk Mp Out
CL off
Clk Mp on ((A+B)C)
Two phase operation
Pre-discharge (Clk = 1)
Evaluate (Clk = 0) OK but Slower
S. A. Ibrahim
9
Example
Implement the logic function F=AB+C using n-type
dynamic logic.
Repeat using p-type dynamic logic.
Clk Mp Clk Me
Out
A
B
A B C
C
Out
Clk Me Clk Mp
S. A. Ibrahim
10
Cascading Dynamic Gates
V
Out2
S. A. Ibrahim
11
Clock Delaying (1)
• Input has to be stable during evaluate.
• Use an appropriately-delayed clock.
• Note that there is a “race” during pre-charge.
• Not a functional problem but wastes power.
• May limit the maximum delay permitted.
In
Clk2
Clk1 Mp delay Mp Clk1
Out2
Out1
Clk2
In
Out1
Clk Me Clk Me
Out2
Possible race
t
S. A. Ibrahim
12
Clock Delaying (2)
• Delay the clock for each stage so that the inputs are “levelized”
before the stage EVALUATES.
• Must use the longest delay.
• Or match each element independently.
• Goal is to track the delay of the circuit.
• Under environmental variations.
• Use dummy logic gates that look like the circuit.
S. A. Ibrahim
13
Domino Logic
Clk Mp Clk Mp
11
Out1 Out2
10
00
In1 01
In2 PDN In4 PDN
In3 In5
Clk Me Clk Me
S. A. Ibrahim
14
Why Domino?
Clk
Clk
S. A. Ibrahim
15
Properties of Domino Logic
S. A. Ibrahim
16
Designing with Domino Logic
S. A. Ibrahim
17
Logical Effort of Domino Gates
AND Gate OR Gate
Clk Mp 2 Clk Mp 2
Out Out
A 3 1 B 2A 2 1
B 3
Clk Me 2
Clk Me 3
S. A. Ibrahim
18
LE of Domino Gates with Skewed Inverters
AND Gate OR Gate
Clk Mp
Why not larger? Clk
4 Mp 4
Out Out
A 3 1 B 2A 2 1
B 3
Clk Me 2
Clk Me 3
S. A. Ibrahim
19
Average Logical Effort
Clk Mp 2
Out
A 3 1
LEavg LEavg
B 3
Clk Me 3 𝐿𝐸𝑎𝑣𝑔 = 𝐿𝐸
S. A. Ibrahim
20
Footless Domino
VDD VDD
VDD
Clk Mp Clk Mp Mr
Out1
Out2
In1
In2 PDN In4 PDN
In3
Can be eliminated!
Clk Me Clk Me
Inputs = 0
during precharge
• The first gate in the chain needs a foot switch, others don’t.
• Precharge is rippling. short-circuit current
• A solution is to delay the clock for each stage.
S. A. Ibrahim
21
LE of Footless Domino
AND Gate OR Gate
Clk Mp 4 Clk Mp 4
Out Out
A 2 1 B 1A 1 1
B 2
S. A. Ibrahim
22
Parasitic Delay
LE LE LE
LE LE LE
S. A. Ibrahim
23
Pre-Charge Properties
Many domino gates can evaluate in one half-cycle, so it should be
easy to pre-charge a single domino gate in the other half-cycle. But…
The domino gate must pre-charge enough to flip the high skew gate,
then the high skew gate must fall below Vt by sufficient noise margin
before evaluation starts again.
To speed up domino evaluation, we want a small pre-charge transistor
(small diffusion parasitic capacitances).
• Makes pre-charge slow.
• High skew gate falls very slowly.
Delaying the clock to avoid pre-charge contention in un-clocked pull-
down stacks reduces pre-charge time for domino gates with delayed
clocks.
Cycles are getting shorter.
Advanced domino methodologies are stretching the length of
evaluation phase at the expense of pre-charge time.
Bottom line: pre-charge time is becoming an important issue. Size for
roughly equal pre-charge and evaluate times.
S. A. Ibrahim
24
Outline
S. A. Ibrahim
25
Design Issues: Charge Leakage
CLK
Clk Mp
Out
A CL
VOut Evaluate
Clk Me
Precharge
Leakage sources
S. A. Ibrahim
26
Design Issues: Coupling & Ground Bounce
S. A. Ibrahim
28
Delayed Keepers
Weakened keepers are not as effective at restoring the degraded
voltage.
To avoid fighting, we can turn on a stronger keeper after a small
delay.
Can make it only work for HIGH output. [Alvandpour, JSSC02]
Key is to not delay by too much.
Restore before too much charge is gone.
But not start the keeper before all the inputs have arrived.
S. A. Ibrahim
29
Design Issues: Charge Sharing
Clk Me CB
S. A. Ibrahim
30
Charge Sharing
Two cases:
VDD • Ma turns off – incomplete charge sharing
• Ma stays on – complete charge sharing
Clk Mp
VDD case 1) if Vout < VTn C < C VTn Vx
a L
Out VDD − VTn Vx
Mp CL CLVDD = CLVout t + Ca VDD – VTn VX
A Ma Out
or
X
CL Ca
A Ma Ca Vout = Vout t – VDD = – -------- VDD – VTn VX
CL
B0 Mb X
Ca VTn Vx
B=0 Mb case 2) if Vout > VTn Ca > CL
Cb VDD − VTn Vx
Clk Me
Ca
---------------------
Cb Vout = –VDD -
Me C
a + C L
S. A. Ibrahim
31
Charge Sharing Example
What is the worst case drop?
Clk
Out
A A CL=50fF
Ca=15fF B B B B Cb=15fF
Cc=15fF C C Cd=10fF
Clk
30
ABC or ABC with ∆𝑉𝑜𝑢𝑡 = × 2.5 = 0.94 𝑉
30+50
S. A. Ibrahim
32
Solution to Charge Redistribution
Clk Me
S. A. Ibrahim
33
Pre-Charging Internal Nodes
Normally, internal nodes are pre-charged with small PMOS
devices.
Not crucial to get node to 100% of Vdd, just reduce noise.
Gates actually run faster when some charge sharing occurs.
Less capacitance needs to be pulled all the way down.
Sometimes pre-charge an internal node to Vdd-Vt with an NMOS
device.
Maybe even pre-discharge an internal node to speed it up.
Worst case for speed is with node high, worst case for noise is with
node low.
If we can tolerate the noise with node low, we might improve the
speed by guaranteeing the node is low.
Use small NMOS device (make sure it is off during evaluation)
Only can pre-discharge a node if no path to Vdd possibly exists.
Must be sure that noise is tolerable for all cases when doing
this.
S. A. Ibrahim
34
Outline
S. A. Ibrahim
35
Differential (Dual Rail) Domino
off on
Clk Mp Mkp Mkp Mp Clk
Out = AB Out = AB
1 0 1 0
A
A B
B
Clk Me
S. A. Ibrahim
36
Other Forms of Domino Logic
Clk Mp Clk Me
11
Out1
10
In1 In4 PUN
In2 PDN In5
00
In3 01
Out2
(to PDN)
Clk Me Clk Mp
Clk Mp Clk Me
11
Out1
10
In1 In4 PUN
In2 PDN In5
00
In3 01
Out2
(to PDN)
Clk Me Clk Mp
to other to other
PDN’s PUN’s
NORA = NO RAce
S. A. Ibrahim
39
Dynamic CML (DyCML)
DyCML circuits combine the advantages of MOS current mode logic (MCML)
circuits with those of dynamic logic families to achieve high performance at
a low-supply voltage with low-power dissipation. [Allam, JSSC01]
Q2, Q3, and Q4 are ON during pre-charge phase, pre-charging outputs to VDD,
and discharging C1.
During Evaluation, Q1 acts as a current source. Q5 and Q6 amplifies the
difference between a and b.
Size C1 so that the dynamic charge is sufficient to discharge output
capacitance.
S. A. Ibrahim
40
Combinational Logic Summary
Static Dynamic
Robust in the presence Fast and small complex
of noise gates
Ease of design Suffers from parasitic
No static power effects
consumption
Large area and bad
performance for large
fan-in
PT logic is attractive for
multiplexers and XOR-
dominated logic.
S. A. Ibrahim
41
Reading
Rabaey: Sections 6.3
Weste: Sections 9.2.4
A. Alvandpour, R. Krishnamurthy, K. Soumyanath, and
S. Borkar, "A sub-130-nm conditional keeper
technique," JSSC, vol. 37, no. 5, May 2002, pp. 633-638.
M. Allam and M. Elmasry, "Dynamic current mode logic
(DyCML): a new low-power high-performance logic
style," JSSC, vol. 36, no. 3, March 2001, pp. 550-558.
S. A. Ibrahim
42