Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Université de Mons Papy NDUNGIDI KIAZAYILA, Fortunato DUALIBE, Carlos VALDERRAMA FPMs - Service d‘Electronique et de Microélectronique (SEMi)
Rapid Top-Down Specification of an RF Receiver using Genetic Multi-Objective
Optimization Method
Introduction
Methodology
Results
Conclusion This work presents a novel approach to refine the parameters of an RF receiver to satisfy a standard
specification by using a genetic algorithm for multi-objective optimization. The interest of this approach is
to come rapidly up with a feasible specification avoiding a hard time-consuming iterative simulation
procedure. The methodology has been applied to the design of an LNA for complying with the IEEE
802.15.4 standard. In particular, it provided feasible design parameters for an inductively-degenerated
common-source cascode LNA with LC tank load, which was implemented on a CMOS 150nm process.
Compared to previously reported works, this approach shows a very low noise figure, acceptable power
consumption and a excellent trade-off between linearity and gain.
The design of RF receivers requires low power consumption, small
area, high performance and, at the same time, the shortest “time-to-
market”. To accelerate the design time, our work proposes:
To automate some calculations and transitions between design
steps;
To gain a significant time-saver by using the optimization instead
of the simulation in the specifications calculation phase;
To automate the circuit sizing of the receiver's blocks.
Standard (e.g. IEEE 802.15.4)
Performances comparison
Architecture (e.g. Homodyne)
Parameters IEEE 802.15.4 (PHY 2.4 GHz)
Frequency range 2400-2483.5 MHz
Channel band width 2 MHz
Channel spacing 5 MHz
Coverage Worldwide
Spreading technique DSSS
Modulation O-QPSQ with MSK
Data rate 250 kbit/s
Symbol rate 62.5 ksymbol/s
Number of channels 16
Sensitivity -85 dBm
Receiver maximum input level -20 dBm
PER 1%
Adjacent channel rejection 0 dB
Alternate channel rejection 30 dB
Parameters IEEE 802.15.4 PHY 2.4 GHz
Eb/No 7.78 dB
SNRout - 1.14 dB
SNRin 25.82 dB
NF 26.96 dB
IM3 - 68 dBm
IIP3 - 44 dBm
DRin 65 dB
Sfs 13 dBm (1V over 50Ω)
Gmin 33 dB
GMax 100.24 dB
Sta
ndard
Sum
mary
Receiver specification
RF Filter LNA Mixer AGC Channel Filter Ampli AA Filter ADC
Gain
NF
Rec. Band
Order
Gain
NF
IIP3
IIP2
Gain
NF
IIP3
NF
IIP3
GainMax
Gainmin
Gain
NF
Band width
Order
Gain
NF
IIP3
IIP1
Gain
NF
fsampling
Dynamic
Order
Resolution
IIP3, …
Antagonist global parameters : F and IIP3 Better noise factor (Fmin) Increase the Gains
Better linearity (IIP3Max) Decrease the Gains
Layout of the LNA (LF 150 nm Technology)
IIP3Total Max NFTotal Min
NFTotal (dB) 7.60 2.54
IIP3Total (dBm) 3.83 -2.78
GLNA (dB) 7.57 8.72
GMIX (dB) 3.49 16.50
GCF (dB) -1.5 -1.5
GAGC (dB) 22.44 25.1
GAAF (dB) -1.5 -1.5
NFLNA (dB) 1.25 1.1
NFMIX (dB) 6.77 5.61
NFCF (dB) 1.5 1.5
NFAGC (dB) 15.42 15.55
NFAAF (dB) 1.5 1.5
IIP3LNA (dBm) 4.98 4.78
IIP3MIX (dBm) 18 12.61
IIP3CF (dBm) 19.93 19.72
IIP3AGC (dBm) 18.22 16.46
IIP3AAF (dBm) 21.94 21.8
[1] [2] [3] This work
F (GHz) 2.4 2.4 2. 4 2.4
CMOS Technology 0.35 µm 0.25 µm 0.09 µm 0.15 µm
G (dB) 19.8 20 20 10.5
NF (dB) 3 2.4 4.65 1.7
IIP3 (dBm) 4.5 -3.4 -15.5 4.36
P1dB (dBm) -4.7 -21 -23.2 -2.7
Pc (mW) 22.4 4.8 0.24 8.7
[*] Christophe VERSÈLE, "Contribution à l'optimisation multiobjectif de convertisseurs électroniques de puissance. Application aux convertisseurs continu–continu isolés", PhD thesis, University of Mons,
Belgium, 2012.
[1] J.C. Huang, Ro-Min Weng, Hsiao Chih-Lung, and Kun-Yi Lin. "A 2 V 2.4 GHz fully integrated CMOS LNA with Q-ehhancement circuit". In Asia-Pacific Microwave Conference, pages 1028-1031, 2001.
[2] Xiaomin Yang, Thomas Wu, and John McMacken. "Design of LNA at 2.4 GHz using 0.25 µm technology". In Topical Meeting on Silicon Monolithic IC in Radiofrequence Systems, pages 12-17, 2001.
[3] Nils Kåre Midtflå, "A 2.4 GHz Ultra-Low-Power Low-Noise- Amplifier", Master thesis, Norwegian University of Science and Technology - Department of Electronics and Telecommunications, 2010.
RF frequency 2.4 GHz
S21 10.5 dB
S11 -17.7 dB
S22 -13.3dB
NF 1.7 dB
IIP3 4.36 dBm
Supply 1.8 V
Power Consumption 8.7 mW
Area (including pads) 1.06 mm x 0.898 mm
Frequency 2.44175 GHz
Voltage 1.8 V
Curant 5 mA
LL 1.3 nH
CL 3.26 pF
LM1,M2 0.15 um
WM1,2 768.5 um
NFMIN 1.24
LS 0.34 nH
LG 4.97 nH
Reuse of existing blocs
User and technology constraints
Example of analog design flow
Functional
Architectural
Structural
“Electrical”
Physical
Vout = A(V+ - V- )
Architectural exploration
Topology choice
Sizing
Layout
Input stage + N gain stages +
ESD protection circuit + …
Schematic: differential pair ,
current mirror
Choice of technology and sizing
Layout and fabrication
Gain ? Bandwidth ? Power ? T° ?
Traditional methodology Proposed methodology
requires a lot of time (in order of hours or days),
because it needs parametric simulation;
leads to hundreds of solutions (blocs
parameters) , not all of them are achievable
is very time-saver because optimization
takes a few seconds;
yields optimal and realistic blocs parameters
by taking into account the constraints
imposed by designers without any simulation
The b
ounds o
f
realis
tic v
alu
es
Blocs Gain (dB) NF (dB) IIP3 (dBm)
RF filter -1.5 1.5 13.42 to 22
LNA 5 to20 1 to 4.4 -22 to 6
Mixer 0 to 17.5 5 to 25 -12 to 24
Channel filter -1.5 1.5 13.42 to 22
AGC 5 to 40 5 to 35 4.38 to 20
AAF -1.5 1.5 13.42 to 22
We have solved a nonlinear multi-objective optimization
problem with constraints, by using the genetic algorithm
NSGA-II. Currently, this algorithm is one of the most powerful
used for optimizing a large number of engineering problems.
Its effectiveness was proven through many scientific
publications [*]
Each solution is composed by different blocs parameters values
(GLNA, GMIX, GCF, GAGC, GAAF, NFLNA, NFMIX, NfCF, NFAGC, NFAAF,
IIP3LNA, IIP3MIX, IIP3CF, IIP3AGC, AIP3AAF.)
An algorithm has been implemented for obtaining feasible design parameters of an inductively-degenerated
common-source cascode Low Noise Amplifier (LNA) from a chosen solution.
For choosing a solution in the Pareto front, the user can be refine or filter the set of solution taking into account
the reality of each design (Design strategies). For example: keep only the solutions that have a minimum noise
figure for LNA and a maximum linearity for the mixer, etc.
The NSGA-II algorithm is based on the biological evolution
theory of natural selection.
From a set of initial solutions, the algorithm randomly selects
some of these solutions to produce another set called new
generation. Through the successive generations, the set of
solutions evolves towards an optimum that minimizes the cost
functions (both noise figure and linearity, in this case).
At each step, the algorithm creates a new generation from the
current one by applying three types of operations:
the selection, for choosing some solutions from the current
set to produce a new set of solutions;
the crossover, for combining two current solutions to create
the next generation;
the mutation, for applying random changes to current
solutions to form new solutions;
Elapsed time is 8.516682 seconds
The Matlab© Optimization
tool (Optimtool) can be
used to optimize both, NF
and IIP3, cost functions.
The Pareto front represents a
set of non-dominated solutions
ZOOM IN on 2 solutions
Post-layout simulations results
LN
A s
pe
cific
atio
n