Upload
harry-haynes
View
237
Download
3
Embed Size (px)
Citation preview
VLSI Algorithmic Design Automation Lab.
1
성균관대학교 성균관대학교 정보통신공학부정보통신공학부
조준동 교수조준동 교수
Low Power OFDM Low Power OFDM DesignDesign
VLSI Algorithmic Design Automation Lab.
2
Energy managementEnergy management• Be energy aware at all levels of your Be energy aware at all levels of your
system (QoS)system (QoS)• React on the environment: React on the environment:
adaptabilityadaptability• Do just enough and not too much for Do just enough and not too much for
a given task (QoS)a given task (QoS)– do not optimize for ‘worst case’ but do not optimize for ‘worst case’ but
for the ‘current case’for the ‘current case’• Schedule communication (use Schedule communication (use
locality of reference)locality of reference)
VLSI Algorithmic Design Automation Lab.
3
LOW Power MethodsLOW Power Methods감축감축
대상대상 66 大 大 전력 감축 전력 감축
문제들문제들
Key ReferencesKey References
V V
1. 1. 모뎀 모뎀 Link-AdaptationLink-Adaptation
전력 제어전력 제어
K. Leung et al., "Integrated Link Adaptation and Power Control for WirelessK. Leung et al., "Integrated Link Adaptation and Power Control for Wireless
IP Networks", AT&T Labs, NJIP Networks", AT&T Labs, NJ
2. Dynamic Voltage 2. Dynamic Voltage
SchedulingScheduling
L. Benini, G. De Micheli, "System-Level Power Optimization: Techniques andL. Benini, G. De Micheli, "System-Level Power Optimization: Techniques and
Tools," ACM Trans. on Design Automation of Electronic Systems, Apr. 2000Tools," ACM Trans. on Design Automation of Electronic Systems, Apr. 2000
kk
3. PSM3. PSM 을 이용 을 이용 standby standby
전력 최소화전력 최소화
E. Shih, A. Chandrakasan, Physical Layer Driven Protocol and AlgorithmicE. Shih, A. Chandrakasan, Physical Layer Driven Protocol and Algorithmic
Design for Energy Efficient Wireless Sensor Networks, MobiCom ’01Design for Energy Efficient Wireless Sensor Networks, MobiCom ’01
4.4. 스위칭 감축저전력 스위칭 감축저전력
아키텍쳐 아키텍쳐
유형석유형석 , , 조준동조준동 , Low-power design and architecture,, Low-power design and architecture, IEEE Potentials, ’01IEEE Potentials, ’01
FF5.Just-in-Context5.Just-in-Context
에너지 제공 에너지 제공
M. Beigl, Context-awareness in ubiquitous computing, Oct. ’02, Karlsruhe Univ.M. Beigl, Context-awareness in ubiquitous computing, Oct. ’02, Karlsruhe Univ.
L. Smit, Energy efficiency through adaptation to the environment, utwente.nlL. Smit, Energy efficiency through adaptation to the environment, utwente.nl
CC6. All-in-one6. All-in-one 집적화를집적화를 통해 통해
locality of referencelocality of reference 를 를 늘임늘임
T. Simunic, "Managing Power Consumption in Networks on Chip", HP Lab.T. Simunic, "Managing Power Consumption in Networks on Chip", HP Lab.
D. Flynn, “A Combined HW-SW approach for Low Power SoCs”, SoC/ASIC ’03D. Flynn, “A Combined HW-SW approach for Low Power SoCs”, SoC/ASIC ’03
VLSI Algorithmic Design Automation Lab.
4Levels for Low Power Levels for Low Power DesignDesign
Level ofAbstraction Expected Saving
Algorithm
Architecture
Logic Level
Layout Level
Device Level
10 - 100 times
10 - 90%
20 - 40%
10 - 30%
10 - 30%
System
Algorithm
Architecture
Circuit/Logic
Technology
Hardware-software partitioning,
Complexity, Concurrency, Locality,
Parallelism, Pipelining, Signal correlations
Sizing, Logic Style, Logic DesignThreshold Reduction, Scaling, Advanced packaging
Regularity, Data representation
Instruction set selection, Data rep.
SOI
Power down
VLSI Algorithmic Design Automation Lab.
5
Power-distribution in integrated Power-distribution in integrated PicoRadio (total: 100 mW)PicoRadio (total: 100 mW)
Jan M. Rabaey
VLSI Algorithmic Design Automation Lab.
6Clock Network Power Clock Network Power Managements Managements
• 50% of the total power50% of the total power• FIR (massively pipelined circuit): FIR (massively pipelined circuit): video processing: edge detectionvideo processing: edge detection voice-processing (data transmission voice-processing (data transmission
like xDSL) like xDSL) Telephony: 50% (70%/30%) idle, Telephony: 50% (70%/30%) idle, 동시에 이야기하지 않음동시에 이야기하지 않음 ..with every clock cycle, data are loaded with every clock cycle, data are loaded
into the working register banks, even into the working register banks, even if there are no data changes.if there are no data changes.
VLSI Algorithmic Design Automation Lab.
7
Multiprocessing BenefitsMultiprocessing BenefitsLess Power & CostLess Power & Cost
VLSI Algorithmic Design Automation Lab.
8Wireless Interface Power-Wireless Interface Power-
Saving Saving Ronny Krashinsky and Hari BalakrishnanRonny Krashinsky and Hari Balakrishnan
MIT Laboratory for Computer ScienceMIT Laboratory for Computer Science
• Sleep to save energy, periodically wake to check Sleep to save energy, periodically wake to check for pending data for pending data – PSM protocol: when to sleep and when to PSM protocol: when to sleep and when to
wake? wake? • A A PSM-staticPSM-static protocol has a regular sleep/wake protocol has a regular sleep/wake
cyclecycle
pow
er
pow
er
time time
PSM off PSM on
750mW 50mW 100ms
Measurements of Enterasys Networks RoamAbout 802.11 NIC
VLSI Algorithmic Design Automation Lab.
9
SYN
ACKDATA SLEEP
PSM onMobile Device
Access Point
Server
100ms
200ms
0msAWAKE
tim
eMobile Device
Access Point
Server
PSM off
연구 개발 개요연구 개발 개요
Power Saving Mode Power Saving Mode Ronny Krashinsky and Hari Balakrishnan, MIT Ronny Krashinsky and Hari Balakrishnan, MIT
VLSI Algorithmic Design Automation Lab.
10
The PSM-static DilemmaThe PSM-static DilemmaCompromise between performance and energyCompromise between performance and energy
If PSM-static is too coarse-grained, it harms performance by delaying network data
If PSM-static is too fine-grained, it wastes energy by waking unnecessarily
Solution: Solution: dynamicallydynamically adaptadapt to network to network activity to maintain performance while activity to maintain performance while minimizing energyminimizing energy
– Stay awakeStay awake to avoid delaying very fast RTTsto avoid delaying very fast RTTs– Back offBack off (listen to fewer beacons) while idle (listen to fewer beacons) while idle
VLSI Algorithmic Design Automation Lab.
11
Low Power CDMA Low Power CDMA Searcher Searcher
CDMA CDMA 단말기에 사용하기위한 단말기에 사용하기위한 MSM MSM
(Mobile Station Modem) (Mobile Station Modem) 칩의 칩의 Searcher EngineSearcher Engine 에 에 대한 대한 RTLRTL 수준 저전력 설계 구현수준 저전력 설계 구현 . . 동작 주파수 동작 주파수 : 12.5MHz: 12.5MHz
Data flow graphData flow graph 를 사용하여 를 사용하여 rescheduling, pre-rescheduling, pre-computation computation 및 및 strength restrength reductionduction, Synchronous , Synchronous AccumulatorAccumulator 를 이용한 저전력 설를 이용한 저전력 설 , area, area 와 와 powerpower 를 를 각각 최대 각각 최대 67.68%, 41.35% 67.68%, 41.35% 감소 시킴감소 시킴 . San Kim and . San Kim and Jun-Dong Cho, “Low Power CDMA Searcher”, Jun-Dong Cho, “Low Power CDMA Searcher”, CAD and VLSI Workshop, May. 1999. CAD and VLSI Workshop, May. 1999.
• Inki Hwang, San Kim and Jun-Dong Cho, “CDMA Searcher Co-Design”, ASIC Workshop, Sep. Inki Hwang, San Kim and Jun-Dong Cho, “CDMA Searcher Co-Design”, ASIC Workshop, Sep.
19991999..
VLSI Algorithmic Design Automation Lab.
12
CDMA SearcherCDMA Searcher
그림 1). 상세 블록도
VLSI Algorithmic Design Automation Lab.
13
탐색자 탐색자 (Searcher)(Searcher)• IS-95 IS-95 기반의 기반의 DS/CDMA DS/CDMA 시스템에서 기지국에서 시스템에서 기지국에서
전송하는 파일롯 채널을 입력으로 하여전송하는 파일롯 채널을 입력으로 하여 , , 초기 동기를 초기 동기를 획득하는 장치획득하는 장치
• 탐색자 탐색자 (Searcher)(Searcher) 의 종류의 종류– 상관기를 사용하는 방식상관기를 사용하는 방식 , , 정합필터를 응용한 방식정합필터를 응용한 방식– 상관기를 사용한 직렬 탐색 및 상관기를 사용한 직렬 탐색 및 Double Dwell Double Dwell 방식을 사용함방식을 사용함 ..
• 국부 국부 (( 단말기단말기 ) PN ) PN 코드 발생기코드 발생기– 1515 개의 개의 registerregister 를 사용하여 생성를 사용하여 생성 ..– 생성 다항식생성 다항식
VLSI Algorithmic Design Automation Lab.
14
Operation FlowOperation Flow1 기지국에서 전송하는 파일럿 채널을 단말기에서 기지국에서 전송하는 파일럿 채널을 단말기에서
발생된 발생된 PNPN 부호열과 역확산 과정 수행부호열과 역확산 과정 수행 ..
2 역확산된 결과를 동기 누적 횟수 역확산된 결과를 동기 누적 횟수 NcNc 만큼 누적한 만큼 누적한 후 에너지 계산 과정을 거침 후 에너지 계산 과정을 거침 (( 제곱 연산제곱 연산 ).).
3 에너지 계산 결과값들은 첫번째 임계치와 비교하여 에너지 계산 결과값들은 첫번째 임계치와 비교하여 초과할 경우 뒷 단에서 비동기 누적초과할 경우 뒷 단에서 비동기 누적 ((NnNn) ) 수행수행 ..
4 그렇지 못할 경우 그렇지 못할 경우 PNPN 부호열을 한 칩 빨리 부호열을 한 칩 빨리 발생시키고 입력되는 신호에 대하여 앞의 과정을 발생시키고 입력되는 신호에 대하여 앞의 과정을 반복반복 ..
5 비동기 누적을 거친 결과값을 두번째 임계치와 비비동기 누적을 거친 결과값을 두번째 임계치와 비교교 ..
6 초과하면 탐색 과정을 종료하고초과하면 탐색 과정을 종료하고 , , 그렇지 않을 경우 그렇지 않을 경우 PNPN 부호열을 한 칩 빨리 발생시키고 앞의 과정을 부호열을 한 칩 빨리 발생시키고 앞의 과정을 반복반복 ..
VLSI Algorithmic Design Automation Lab.
15
Pre-computation Pre-computation
◈ A comparator example A comparator example : : Shrinivas Devadas, 1994 Shrinivas Devadas, 1994
◈ Precomputation for Precomputation for external idleness : external idleness : M. Alidina, 1994M. Alidina, 1994
VLSI Algorithmic Design Automation Lab.
16
Low Power ComparatorLow Power Comparator
VLSI Algorithmic Design Automation Lab.
17Three Input ALUThree Input ALU ( ( Ovadia Bat-Sheva, 1998 )Ovadia Bat-Sheva, 1998 )
The three input ALU consumes much less power The three input ALU consumes much less power than an ALU and an ASUthan an ALU and an ASU
A drawback of using a 3I-ALU is the added A drawback of using a 3I-ALU is the added complexity in calculating the carry and overflow.complexity in calculating the carry and overflow.
MUL0 MUL1
ALU ALU/ ASU
ac c 0 ac c 1
P0 P1
Two ALUs Struc ture
MUL0 MUL1
P0 P1
3IALU
ac c 1
Three Input ALU Struc ture
VLSI Algorithmic Design Automation Lab.
18Time sharing Time sharing 과 과 Carry Save Carry Save Adder Adder 및 및 Pre-computation Pre-computation 적용 적용
XOR XOR XOR XOR
+ +
+ +
() 2 () 2
>
>
+
>
RX I TXI RXQ TXQ RX I TXQ RXQ -TX I
max 값 선택
θ 1 와 비교
θ 2 와 비교
동기 누적단
비동기 누적단
에너지 계산단
XOR XOR XOR XOR
() 2 () 2
>
>
+
>
RX I TXI RXQ TXQ RX I TXQ RXQ -TX I
max 값 선택
θ 1 와 비교
θ 2 와 비교
동기 누적단
비동기 누적단
에너지 계산단
CSA CSA
VLSI Algorithmic Design Automation Lab.
19
Rescheduled Data Flow Graph Rescheduled Data Flow Graph 동기 누적단동기 누적단
– Carry Save Carry Save Adder (or 3 Adder (or 3 Iinput ALU) Iinput ALU) 사용사용
임계치 비교임계치 비교– Pre-computation Pre-computation
적용적용
에너지 계산단에너지 계산단– Data Flow Data Flow 순서를 순서를
변화하여 곱셈 과정을 변화하여 곱셈 과정을 줄임줄임
XOR XOR XOR XOR
()2
>
>
+
>
RXI TXI RXQ TXQRXI TXQRXQ -TX I
max 값 선택
θ1 와 비교
θ2 와 비교
동기 누적단
비동기 누적단
에너지 계산단
| | | |
CSA CSA
VLSI Algorithmic Design Automation Lab.
20Synchronous AccumulatorSynchronous Accumulator 를를이용한 저전력 설계이용한 저전력 설계
• 동기 누적 횟수 동기 누적 횟수 NcNc 를 를 128128 회로 할 회로 할
경우 경우 12bit adder12bit adder 가 필요가 필요 ..
• 동기누적단의 입력값을 동기누적단의 입력값을 22 의 배수와 의 배수와
나머지 나머지 11 로 표현하면로 표현하면 , 10bit , 10bit
countercounter 와 와 9bit counter9bit counter 로 로
adderadder 를 대치 할 수 있음를 대치 할 수 있음 ..
• 12bit adder12bit adder 를 사용할 경우 를 사용할 경우
하나의 입력에 대해 하나의 입력에 대해 4bit4bit 의 의 data data
bit transitionbit transition 이 발생하며이 발생하며 , ,
countercounter 를 사용할 경우 를 사용할 경우 3bit3bit 의 의
data bit transitiondata bit transition 이 발생함이 발생함 ..
XOR XOR XOR XOR
() 2
>
>
+
>
RX I TXI RXQ TXQ RX I TXQ RXQ -TX I
max 값 선택
θ 1 와 비교
θ 2 와 비교
동기 누적단
비동기 누적단
에너지 계산단
| | | |
Syn_acc_Logic_1 Syn_acc_Logic_2
VLSI Algorithmic Design Automation Lab.
21
Syn_acc_LogicSyn_acc_Logic7
6
5
4
3
2
1
0
-1
-2
-3
-4
-5
-6
-7
-8
0111
0110
0101
0100
0011
0010
0001
0000
1111
1110
1101
1100
1011
1010
1001
1000
(2 × 3) + 1
(2 × 3) + 0
(2 × 2) + 1
(2 × 2) + 0
(2 × 1) + 1
(2 × 1) + 0
(2 × 0) + 1
(2 × 0) + 0
(-2 × 0) + (-1)
(-2 × 1) + 0
(-2 × 1) + (-1)
(-2 × 2) + 0
(-2 × 2) + (-1)
(-2 × 3) + 0
(-2 × 3) + (-1)
(-2 × 4) + 0
SU M = 2 ¡¿M + N 2 ¡¿M : 1bit sh ift
SU M = (-2) ¡¿M + N (-2) ¡¿M : 1b it sh ift
DEC BIN Function
동기누적단동기누적단– Input : 4 Input : 4
bitsbits– Mux,CounterMux,Counter
를 사용하여 를 사용하여 MM과 과 NN 을 계산을 계산
– ShifterShifter 와 와 adderadder 를 를 사용하여 전체 사용하여 전체 누적된 값을 계산누적된 값을 계산
VLSI Algorithmic Design Automation Lab.
22
Syn_acc_Logic_1Syn_acc_Logic_1 의 블록도 의 블록도
TXI
RX I
TXQ
RXQ
4
4
10 bitscounter
9 bitscounter
3shifter
+
10
9
12
12
Y I
11
VLSI Algorithmic Design Automation Lab.
23
PN-CodeGeneration
SynchronousAccumulator
(SW)
SynchronousAccumulator1
(HW)
Cost(Speed,Area,Power)
EnergyEstimate
(SW)
SynchronousAccumulator2
(HW)
Comparator(SW)
AsynchronousAccumulator
(SW)
Comparator(SW)
EnergyEstimate
(HW)
Comparatorwith
precomputation(HW)
AsynchronousAccumulator
(HW)
Comparatorwith
precomputation(HW)
GOAL!
PN-CodeGeneration
SynchronousAccumulator
(SW)
SynchronousAccumulator1
(HW)
Cost(Speed,Area,Power)
EnergyEstimate
(SW)
SynchronousAccumulator2
(HW)
Comparator(SW)
AsynchronousAccumulator
(SW)
Comparator(SW)
EnergyEstimate
(HW)
Comparatorwith
precomputation(HW)
AsynchronousAccumulator
(HW)
Comparatorwith
precomputation(HW)
GOAL!
황인기 , 성균관대황인기 , 성균관대
HW SW Co-DesignHW SW Co-Design
VLSI Algorithmic Design Automation Lab.
24
HW SW Co-DesignHW SW Co-Design+ +
+ +
Y I2 YQ
2
>
>
+
>
RXI TXI RXQ TXQ RXI TXQ RXQ -TX I
max 값 선택
θ 1 와 비교
θ 2 와 비교
동기 누적단
비동기 누적단
에너지 계산단
O I = (RXI * TX I)
+ (RX Q * TX Q ) OQ = (RXI * TX Q )
+ (RX Q * (-TX I))
Y I = ∑O I YQ = ∑O Q
Z = max (Y I2 , Y Q
2)
∑ Z
Search Done !!
Yes
YesSearch_Slew No
No
Control Signal Generator
VLSI Algorithmic Design Automation Lab.
25
참고 문헌참고 문헌• [1] J. Henkel, T. Benner, and R. Ernst, “Hardware Generation and
Partitioning Effects in the COSYMA System”, IEEE Int'1 Workshop on Hardware/Software Codesign, pp. 29-40, 1993.
• [2] Wonho Lee, Wangrok Oh, Teajin Chung, Kyungwhoon Cheun, Byeongchul Ahn, “Design and Implementation of A Wideband CDMA Modem”, Proceedings of ICCE'97, pp. 444-445, June 1997.
• [3] Mazhar Alidina, Jose Monterio, Srinivas Devadas, and Abhijit Ghosh, “Precomputation-Based Sequential Logic Optimization for Low Power”, IEEE Transaction on VLSI Systems, Vol. 2, No. 4, pp. 426-435, December 1994.
• [5] “Wideband CDMA multimedia testbed wireless access physical layer spec. ver. 1.0”, Korea Mobil Telecom, internal report, 1996.
• [6] Seongjoo Lee, Sangyun Hwang, and Jaeseok Kim, “VLSI Architecture of CDMA Rake Receiver with Low Hardware Complexity for PCS”, Proceedings of ICCE'98, pp. 160-161, June 1998.
• [7] M. Alidina, J. Monteiro, S. Devadas, and A. Ghosh, “Pre-computation-Based Logic Optimization for Low Power”, Proceedings of ICCAD, pp. 74-81, Nov 1994.
VLSI Algorithmic Design Automation Lab.
26
Image Image 압축압축
VLSI Algorithmic Design Automation Lab.
27
Low Power Equalizer Low Power Equalizer • Jin Woo Kim, J.D.Cho, 1999, SKKUJin Woo Kim, J.D.Cho, 1999, SKKU• IntroductionIntroduction• EqualizerEqualizer• Low-Power Methodology in Low-Power Methodology in
EqualizersEqualizers• SimulationSimulation• Future WorkFuture Work• ReferenceReference
VLSI Algorithmic Design Automation Lab.
28ADSL Block Diagram ADSL Block Diagram ( DMT )( DMT )
ErrorControlCoding
BitMapping
Pre-coding
IFFTCyclicPrefix
Insertion
DigitalFilter
DAC
AnalogFilter
HybridCircuit
AnalogFilter
AGC
ADC
POTSSplitter
TEQCyclicPrefix
RemovalFFT
EchoCanceller
FEQSymbol
Recovery
ErrorDe-
coding
Twisted PairTelephone Loop
POTS Telephone Setor Voice Band Modem
TransmitBit
ReceiveBit
TEQ : Time Domain Equalizer
FEQ : Frequency Domain Equalizer
AGC : Auto Gain Control
VLSI Algorithmic Design Automation Lab.
29
Equalizer – TEQ Equalizer – TEQ ( Time Domain ( Time Domain
Equalizer )Equalizer )• TEQ with a small number of taps can TEQ with a small number of taps can
reduce the cyclic prefix(CP)reduce the cyclic prefix(CP)
• Power Consumption Critical PartPower Consumption Critical Part
• Adaptive FIR filters (LMS algorithm)Adaptive FIR filters (LMS algorithm)– The Filter Output The Filter Output
– The Filter CoefficientThe Filter Coefficient
– The Error SignalThe Error Signal
1
,0
N
k k n k nn
y w x
1, , 0 1kk n k n k nw w e x for n N
k k ke I y
VLSI Algorithmic Design Automation Lab.
30Equalizer – Example of Equalizer – Example of TEQTEQ
x (n)
w0 w1 w2 w3 w4
x (n)
y (n)
e (n )
VLSI Algorithmic Design Automation Lab.
31Equalizer – Example of Equalizer – Example of FEQFEQ
• LMS(Least Mean Square) Adaptive FilterLMS(Least Mean Square) Adaptive Filter
x y
x̂
Conjugate
x *
c
- +
VLSI Algorithmic Design Automation Lab.
32Low Power Methodology Low Power Methodology in Equalizers – Circuitin Equalizers – Circuit
• FIR FilterFIR Filter– Carry-save adderCarry-save adder– Grouped Multipliers(Wallace Tree Grouped Multipliers(Wallace Tree
Multiplier)Multiplier)– Booth Recoding techniqueBooth Recoding technique
• Updating CircuitsUpdating Circuits– Power-of-two LMS updatingPower-of-two LMS updating– The use of register fileThe use of register file– Coefficient freezingCoefficient freezing– Make Make step sizestep size constant( constant() to power-of-2) to power-of-2
VLSI Algorithmic Design Automation Lab.
33Low Power Methodology Low Power Methodology in Equalizers – Run-timein Equalizers – Run-time
• The Error MonitorThe Error Monitor
• Adaptive Bit PrecisionAdaptive Bit Precision
• Burst-Mode UpdateBurst-Mode Update
• Adaptive Filter LengthAdaptive Filter Length
VLSI Algorithmic Design Automation Lab.
34Low Power Methodology Low Power Methodology in Equalizers – Examplein Equalizers – Example
FIR filterProgrammable
gainSlicer
UpdateError
monitor
Control
x
w A
z y
e
Freeze
Local_update
Error_status
Check_Err
VLSI Algorithmic Design Automation Lab.
35
Simulation – Output SNRSimulation – Output SNR
VLSI Algorithmic Design Automation Lab.
36Simulation – Result Simulation – Result AnalysisAnalysis
• Step Constants(Step Constants() are power-of-2) are power-of-2– 0.5 -> ½ (20.5 -> ½ (2-1-1))– 0.125 -> 1/8 (20.125 -> 1/8 (2-3-3))– 0.03125 -> 1/32 (20.03125 -> 1/32 (2-5-5))– 0.007125 -> 1/128 (20.007125 -> 1/128 (2-7-7))
• The smaller Constant, the longer The smaller Constant, the longer convergence timeconvergence time
• Replace Complex Multiplier to Just Replace Complex Multiplier to Just Exponent AdderExponent Adder
• Low Power Consumption is acquiredLow Power Consumption is acquired
VLSI Algorithmic Design Automation Lab.
37
Lower Power 시스템 설계 설계 자동화 알고리즘 설계 자동화 알고리즘
HW/SW Co-designHW/SW Co-design
저전력 IP’s저전력 IP’s
핵심단위기술핵심구성기술
연구실 역량연구실 역량
연구개발실적
저전력 VLSI Layout 저전력 VLSI Layout
통신모뎀신호처리 통신모뎀신호처리
음성및영상신호처리알고리즘
음성및영상신호처리알고리즘
저전력 Encoding 저전력 Encoding
S/W H/W 통합 설계S/W H/W 통합 설계
HDTV 수신기의 SDR HDTV 수신기의 SDR
암호화 재구성구조 암호화 재구성구조
전력 예측 전력 예측
Data encoding버스반전부호화Data encoding버스반전부호화
저전력 Global/Detailed RouterArea Router for Memory Pheriphral저전력 Global/Detailed RouterArea Router for Memory Pheriphral
IS-95 동기부IS-95 동기부
DSP 및 FPGA 혼합 설계 DSP 및 FPGA 혼합 설계
내용
Gate-level 전력 예측 Gate-level 전력 예측
ECC, RSA, SEED, Rijndael ECC, RSA, SEED, Rijndael
저전력 CDMA2000 비터비 복호기저전력 CDMA 탐색자저전력 WCDMA 상관기Cable Modem/OFDM 등화기
저전력 CDMA2000 비터비 복호기저전력 CDMA 탐색자저전력 WCDMA 상관기Cable Modem/OFDM 등화기
저전력 HMM Scorer저전력 Motion Estimator저전력 HMM Scorer저전력 Motion Estimator
VLSI Algorithmic Design Automation Lab.
38
저전력 IP’s저전력 IP’s
연구실적
DSPDSP
ASICASIC
GIRemoval
GIRemoval FFTFFT Phase
RotatorPhase
Rotator
CRCR
FineSTRFineSTR
ChannelEstimator/Equalizer
ChannelEstimator/Equalizer
ViterbiFEC
ViterbiFEC
Coarse STR
Coarse STR
GI/FFTDetectorGI/FFTDetector
ADCADC
CPE CSI
TimingProcessorTiming
Processor
IF
RFSERSER
DemodDemod
NCONCO
DPAGCDP
AGC
GIRemoval
GIRemoval FFTFFT Phase
RotatorPhase
Rotator
CRCR
FineSTRFineSTR
ChannelEstimator/Equalizer
ChannelEstimator/Equalizer
ViterbiFEC
ViterbiFEC
Coarse STR
Coarse STR
GI/FFTDetectorGI/FFTDetector
ADCADC
CPE CSI
TimingProcessorTiming
Processor
IF
RFSERSER
DemodDemod
NCONCO
DPAGCDP
AGC
Key_add
Mux_1
Mux_2
Mix_Column
Byte_Sub
Shift_Low
Key_add
DIN_Reg
DOUT_Reg
Control
KeyGeneration
clksel_1
enb
sel_2
clk
enb
rst
Key
subKey
clk
rst
start sel_2
enb
sel_1
HOSTCPU
ADDRESS BUS(8BIT)
RESET
CS
RD
WR
CLK
DW
CryptoProcessor
DATA BUS(32BIT)
DATA BUS(32BIT)
C o e ff ic ie n tU p d a te
C o n ju g a to r
E rro rC o n tro l
L e a rn in gC o n s ta n tC o n tro l
x
x *
y z
c
-5
0
5
10
15
20
25
30
35
40
Conventional FEQ Low-Power FEQ
Conventional FEQ
Low-Power FEQ
buffer
PE PE PE PE
comparator comparator comparator comparator
Control Generator
MemoryPDF
( )
Transition( )
1( )j tb w
ija
1( )i tw
( )i tw
search data buffer reference data buffer
addressgenerator
externalmemorysearch
data
clock generator
contorl signalgenerator
comparator
Motion Vector
comparator
c3_sum
c4_sum
comparator
comparator
modified PE
modified PE
modified PE
modified PE
modified PE
modified PE
modified PE
modified PE
modified PE
modified PE
modified PE
modified PE
shift registors
c2_sum
c1_sum
shift register
externalmemorycurrent
data
modified PE
modified PE
modified PE
modified PE
Low-Power Equalizer for xDSL21% 전력 감소 , SNR=40dBLow-Power Equalizer for xDSL21% 전력 감소 , SNR=40dB
Fast and Low Power Viterbi Search Engine using Inverse Hidden Markov Model68% 전력 감소 , 71% 속도개선 , 1.9 배면적증가삼성 휴먼 테크 우수논문상 , ‘02
Fast and Low Power Viterbi Search Engine using Inverse Hidden Markov Model68% 전력 감소 , 71% 속도개선 , 1.9 배면적증가삼성 휴먼 테크 우수논문상 , ‘02
Maximizing Memory Data Reuse for Lower Power Motion Estimation33% 전력 감소 , 52Mhz 2.1 배 면적증가(SCI 논문 )
Maximizing Memory Data Reuse for Lower Power Motion Estimation33% 전력 감소 , 52Mhz 2.1 배 면적증가(SCI 논문 )
IS-95 기반 CDMA 의 Double Dwell Searcher 저전력 및 co-design 설계 67% 전력 감소 , 41% 면적감소
IS-95 기반 CDMA 의 Double Dwell Searcher 저전력 및 co-design 설계 67% 전력 감소 , 41% 면적감소
OFDM-based high-speed wireless LAN platform20.7Mhz, 237000 gates
OFDM-based high-speed wireless LAN platform20.7Mhz, 237000 gates
스마트 카드용 차세대 저전력 보안 프로세서 칩 설계ECC, Rijndael, DES, SHA
스마트 카드용 차세대 저전력 보안 프로세서 칩 설계ECC, Rijndael, DES, SHA
High-Flexible Design of OFDM Tranceiver for DVB-T ( 개발 중 )
High-Flexible Design of OFDM Tranceiver for DVB-T ( 개발 중 )
The “Lower” Power TechnologiesThe “Lower” Power Technologies
VLSI Algorithmic Design Automation Lab.
39
논문 및 특허 실적 (5 년간 )논문 및 특허 실적 (5 년간 )
연구실적연구실적
번호번호 논문논문 // 특허등록특허등록 국명국명 게재일 게재일 기타기타
11 Design Automation of Deep-Submicron VLSI, Design Automation of Deep-Submicron VLSI, A first Patent Application InveA first Patent Application Invention Achievement Awardntion Achievement Award, IBM, IBM 미국미국 2001.42001.4 특허 특허
출원출원
22 Viterbi Decoder ADS Module, No. 10-2000-0000331 Viterbi Decoder ADS Module, No. 10-2000-0000331 w/ SECw/ SEC 미국미국 2000.1 2000.1 특허 출원특허 출원
33 Low-power design and architecture, IEEE Low-power design and architecture, IEEE PotentialsPotentials 미국미국 2001.82001.8 초청논문초청논문
44 Low Power Viterbi Decoder, No. 10-2000-0002479, Low Power Viterbi Decoder, No. 10-2000-0002479, w/ SECw/ SEC 미국미국 2000.12000.1 특허 출원특허 출원
55 Lower Power Bus Encoding, No. 98-27313 wLower Power Bus Encoding, No. 98-27313 w// SEC SEC 미국 미국 1998.71998.7 특허 특허 출원출원
Maximizing Memory Data Reuse for Low Power Motion Estimation,Maximizing Memory Data Reuse for Low Power Motion Estimation,VLSI Design, Vol 14, No 2 VLSI Design, Vol 14, No 2
미국 미국 2002.62002.6 SCISCI
66 Low Power Digital Multimedia Low Power Digital Multimedia Telecommunication DesignsTelecommunication Designs, , VLSI DESIGN VLSI DESIGN ,Vol. 12,Vol. 12, No. 3, No. 3 미국미국 2001.32001.3 SCISCI
Wiring Space and Length Estimation in Two-Dimensional ArraysWiring Space and Length Estimation in Two-Dimensional Arrays,,IEEE Trans. on CAD, Vol. 19, No. 5IEEE Trans. on CAD, Vol. 19, No. 5 미국미국 2002000.50.5 SCISCI
77 A High-Speed, Low-Power Phase Frequency Detector andA High-Speed, Low-Power Phase Frequency Detector andCharge PumpCharge Pump Circuits Circuits, IEICE Trans, IEICE Trans. Fundamentals, Vol. E82, No. 11. Fundamentals, Vol. E82, No. 11 미국미국 1999.111999.11 SCISCI
88Lower-Power and Min-Crosstalk Channel Routing for DeepLower-Power and Min-Crosstalk Channel Routing for DeepSubmicron Layout design, VLSI DESIGN, Vol.10, No.1Submicron Layout design, VLSI DESIGN, Vol.10, No.1
미국미국 1999.11999.100 SCISCI
99A Lower Power Booth Multiplier based on OperandA Lower Power Booth Multiplier based on OperandSwapping inSwapping in Instruction Level, Journal of the Korean Instruction Level, Journal of the KoreanPhysical SocietyPhysical Society
국내국내 1998.111998.11 SCISCI
SCI SCI 논문 논문 (1998 – (1998 – 현재현재 : 13: 13 편편 ) ) 국제 학술 대회 국제 학술 대회 1717 편 편
VLSI Algorithmic Design Automation Lab.
40
연구과제명연구과제명 연구기간연구기간 책임책임여부여부 재원재원
11Low Power ASIC for Personal Network AccessLow Power ASIC for Personal Network AccessModuleModule
2001.3 - 2003.22001.3 - 2003.2 OO 교내 교내 연구비연구비
22High-Flexible Signal Processing Chip Design ofHigh-Flexible Signal Processing Chip Design ofOFDM TranceiverOFDM Tranceiver for DVB-T for DVB-T
2002.5. 2002.5. - - 2003.52003.5 OO 삼성전자삼성전자
33 Software Defined Radio Software Defined Radio 교육 프로그램 개발 교육 프로그램 개발 2002.72002.7-- 2003.2 2003.2 OO전파교육전파교육연구센터연구센터
44Lower Power Physical Design considering logicLower Power Physical Design considering logicdesigns in Deep-Submicrondesigns in Deep-Submicron
1998.2 -2000.21998.2 -2000.2 OO 과학재단과학재단
55Embedded OS-based Wireless Multimedia AccessEmbedded OS-based Wireless Multimedia AccessModule in Personal NetworksModule in Personal Networks
2000.9 - 2003.82000.9 - 2003.8 XX 학술진흥재단학술진흥재단
66Low Power and High Performance ReconfigurableLow Power and High Performance ReconfigurableEqualizer for Cable MODEMEqualizer for Cable MODEM
2000.12000.1 - - 2000.122000.12 OO 삼성전자삼성전자
77Fast and Low Power Search Engine for SpeechFast and Low Power Search Engine for SpeechRecognitionRecognition
2000.5 - 2000.12000.5 - 2000.1 OO 삼성전자삼성전자
88 Lower Power Decoder For Convolutional EncoderLower Power Decoder For Convolutional Encoder 1992.2 - 1999.11992.2 - 1999.1 OO 삼성전자삼성전자
99 HW/SW Codesign on CDMA systemHW/SW Codesign on CDMA system 1998.121998.12 - - 1999.11999.111 OO 학술진흥재단학술진흥재단
1010 MODEM (Viterbi Decoder) ASIC Design forMODEM (Viterbi Decoder) ASIC Design for Lower PowerLower Power
1998.121998.12 - - 1999.11 1999.11 OO 교내 연구비교내 연구비
대표적 연구 프로젝트 실적 (5 년간 ) 대표적 연구 프로젝트 실적 (5 년간 )
연구실적연구실적
VLSI Algorithmic Design Automation Lab.
41
산학협력 및 연구수행 산학협력 및 연구수행
수상 경력 수상 경력
-SDR (ICU,IT-SOC)- 저전력설계 (IT-SOC,SAIT, SEC,IDEC, 서두인칩 )-배움닷컴 (1996- 현재 ) 저전력 강의 콘텐츠 서비스 중
- 삼성전자 연구팀장- 삼성전자 저전력 설계연구회 자문-IBM 연구원- 산학과제 10 건 - 국내논문 (16,8), 국제 (13,18)
-T.C: ICCAD ’03, ISQED ’03,ISLPED’03, ASP-DAC’03,SLIP’00,AP-ASIC’99-Guest Editor: VLSI Design, 2000 특별호 :Low Power Architecture and CAD- IEEE Senior Member ‘96,
-ACM/SIGDA Member - Review: IEEE T. CAD, VLSI System, Journal of VLSI Signal Processing
연구 실적
NobleArchitecture
High Flexibility
HighPerformance
저서 저서
-High-Performance Physical
Design for MCM and Packages
1998-Wiley Encyclopedia of Electrical and Electronics,VLSI Circuit Layout, 1999-Low Power Digital Core
DesignFor Multimedia and Telecommunications,2002외 4 권
연구 및 학술 활동 (5 년간 )
초청 강연 및 콘텐츠 개발 초청 강연 및 콘텐츠 개발
국내외학술활동국내외학술활동
- 최고논문상 : DAC ’93- 최고우수연구교수상 ’ 0
0- 공훈상 : 삼성전자 ’ 93- 삼성 휴먼테크 상 ‘ 02
- 최고논문상 : DAC ’93- 최고우수연구교수상 ’ 0
0- 공훈상 : 삼성전자 ’ 93- 삼성 휴먼테크 상 ‘ 02
Lower PowerLower PowerLower PowerLower Power