Transcript Document

Low Power Wireless Design
Dr. Ahmad Bahai
National Semiconductor
New paradigm in Wireless
Power Efficiency
J/Bits/s/Hz
Bits/s/Hz
Configurability
Design
for worst
case
Configurable
Design
Architecture
Centralized
Hybrid
Power efficiency
Distance to IP
Network
TX Power Data Rate
Cellular
Miles
100
mW
100s
kbps
WLAN
Yards
10
mW
10s
mbps
UWB
Feet
mW
100s
mbps
Pervasive IP
Tx power ~ Circuit
power
(1nJ/bit transmission energy- 10 m distance)
Power = Tx power +
Circuit power
Comm Theory, Asym Values
Eb
2C / W  1
 lim
 ln 2  1.59dB
C
/
W

0
N0
C /W
Absolute minimum energy for reliable transmission
of 1 bit of information
Eb  691017.4 J
Min switching energy for digital gate
(1 electron @100mV): 1.6X10-20
C 
Pav
Pav
log2 e 
N0
N 0 ln 2
b/s
Transmission vs. Circuit Energy
Communication Theory usually
considers Transmission energy
only!
Transmission Energy

Spectral Efficiency
R
L
CE  
B BTon
But
Ec  Ton
Optimal Bandwidth-time pair?
Total Energy (MQAM)
Platform
Phy
Tx/RX
MAC layer
including
ARM and PCI
PCI
interface
RF/Analog
supporting up
to 4 radios
Power profile in WLAN (TX)
DSP+MAC
DAC
BBFE
LO
RF FE
PA
261, 22%
530, 44%
40, 3%
36, 3%
47, 4%
288, 24%
Power profile in WLAN (RX)
DSP+MAC
ADC
174, 24%
BBFE
LO
RF FE
176, 24%
60, 8%
47, 6%
200, 28%
72, 10%
FEC
Channel Effect
IMEC
Collaboration
Comm Theory Approach
Bandwidth
SiNR
Power Mask
Dynamic Range
Modulation
Interference
Margin
Coding
Data rate
Noise figure
Statistical
performance
Synchronization
MAC State machines
BER
Channel
Energy
QoS
Adaptive design
Energy and Throughput
Common Approach:
Define SINR and capacity as

Pl Gll
Rl  log
  Pj Glj  nl
 j l

, l  1,, L


Assume BPSK with BER target of 10e-q, bandwidth W and
target data rate of R>C; then we can show that minimal
power vector supporting network topology for low SIR can
be derived as:
P  ( I  F ) 1 b
P  ( P1 ,  , PL )
Cl nl
bl 
GllW / 2 q log 2 10
F
 Glk Cl
 G W
ll
 0
if k  l
if k  l
Design Strategy
System level approach to low
power communication design
Case study: ZigBee
•Profile the power consumption
•Study effect of multi-layer optimization
•A new design strategy
IEEE 802.15.4 PHY
BAND
2.4 GHz
ISM
868 MHz
915 MHz






ISM
COVERAGE
DATA RATE
CHANNEL(S)
Worldwide
250 kbps
16
Europe
20 kbps
1
Americas
40 kbps
10
Direct Sequence Spread Spectrum (DSSS) radio
2Mchip/s OQPSK modulation
1 symbol = 32-chip PN sequence
1 symbol = 4 bits
PHY data rate: 250kbps
Transmit power up to 0 dBm
802.15.4 spec. summary
 Symbol rate and Tx RF accuracy: +/- 40 ppm
 Packet Error Rate (PER)

Defined for PSDU of 20x8 bits
 Sensitivity: -85 dBm (PER < 1%)
 RSSI: sens. level +10 dB, 40 dB range (+/- 6dB)
 Max input level: -20 dBm
 Jamming resistance (interference performance)



0 dB for adjacent channels (ref: -82 dBm)
30 dB for alternate channels (ref: -82 dBm)
Interferer is 802.15.4 compliant interferer
 Tx Error Vector Magnitude : < 35% for 1000 chips
 Tx PSD: -20 dB or –30 dBm |f-Fc| > 3.5 MHz (rbw 100kHz)
 Output power: > -3 dBm (@ max power setting)
 Rx-Tx turnaround time: 12 Symbols (192 ms)
ZBIC, one-chip solution
ZBIC
Power Profile
4-state/Transition Energy Profile
VDD = 1.8V
Shutdown
80 nA
970 us
691pJ
194 us
6.63 uJ
Idle
396 uA
194 us
6.63 uJ
RX
19.6 mA
Transition Energy

T(transition) x I(target state) x VDD
TX
-25 dBm: 8.42 mA
-15 dBm: 9.71 mA
-10 dBm: 10.9 mA
-7 dBm: 12.17 mA
-5 dBm: 12.27 mA
-3 dBm: 14.63 mA
-1 dBm: 15.785 mA
0 dBm: 17.04 mA
IMEC/MIT
Observations
Efficiency (energy/bit) changes
with:
Larger packet size
Transmit power control
Network Load
Contention
Channel Coding
Link layer performance
Power Breakdown
Breakdown between the states
In high load, the node spends more time in RX than in TX mode!
IMEC/MIT
More comprehensive Energy model
Energy efficiency metric:
E[ Energytotal ]

E[ Payload]
E[ Energytotal ]  k [ Energyt 0tal | statek ]P[statek ]
TX, RX, Collision, sensing, Transitions, ramp up
New model for total energy was used to optimize
back off strategy in an ad-hoc network.
Energy Efficient Backoff

r  W  1 L  1  W  1

1 N 1
L
1  W 
Standard
backoff
Proposed
backoff
Resetting back-off is more energy efficient than
DCF backoff due to carrier sensing overhead.
Summary
Statistical Performance Analysis:
New design paradigm in communication
Configurable and low power design:
Key Design objectives
Multimode/Multi-layer Optimization
Analog/mixed signal:
critical in power consumption
Mixed signal processing and
cross layer optimization