Post on 19-Jun-2018
1
NC STATE UNIVERSITY
Gregory Mazzaro
SIAMES Research GroupNorth Carolina State University
Time-Frequency Effects in Wireless Communication Systems
Ph.D. Dissertation Defense
Tektronix TDS684Bdigitizing oscilloscope
Agilent E8267Cvector signal generator
Agilent E4438Cvector signal generator
Tektronix AWG2021waveform generator
September 28, 20094:00pm, MRC 463
NC STATE UNIVERSITY
2
Time-Frequency Effects: Presentation Overview
• Motivations — The “Big Picture”
• Original Contributions vs. Previous Research
• Completion of Tasks & Conclusions
linear transient distortion in narrowband systems
nonlinear transient distortion in narrowband systems linear metrology of bandpass RF components
nonlinear metrology of bandpass RF systems appendices — the “How-To” section
• New Chapter: Linear Amplification by Time-Multiplexed Spectrum
• Summary & What’s Next?
26 slides
2
NC STATE UNIVERSITY
3
Time-Frequency Effects: Motivations (Chapter 1)
• The limitations of the steady-state frequency domain
• Co-site interference
• multiple radios in close proximity operate with same frequencies• circuit-field coupling, nonlinear mixing, long-tail transients
• Testing of integrated assemblies
• higher levels of integration fewer access points for testing• need non-destructive methods for testing integrated structures
• Improvement of transmitter linearity
• new comm. systems implement spectrally-efficient modulation schemes• need ways to meet stringent specs. for linearity & signal quality
NC STATE UNIVERSITY
4
Why look at the time domain? i.e. When is steady-state analysis inadequate? for narrowband systems
e.g. for filters of higher order
simulated
simulated
measured
• What is the nature of these effects?
• How can they be mitigated?
• How can they be usedto the advantageof a system designer?
Time-Frequency Effects: Motivations (Chapter 1)
3
NC STATE UNIVERSITY
5
Time-Frequency Effects: Nature of Transients (Chapter 2)
• time-frequency effects caused by bandpass filters
coupled resonatorsWhy focus on filters?(a) they are ubiquitous in wireless communications devices
(b) they are most likely to produce unintended transient behaviorwith continued application of traditional steady-state design
narrowband, greatest reactive energy storage capability
NC STATE UNIVERSITY
6
Time-Frequency Effects: Prior Work vs. My Work (Chapter 2)
Studies of bandpass transients were done for 2nd-order filters…
I’ve extended the study to 7th-order, narrowband filters
…to show that transients last a lot longer than the expected s
Little quantitative evaluation of such effects on comm. systems was performed…
I’ve measured ISI and IMD for frequency-hopping scenarios
…to demonstrate that filter transients can produce co-site interference
Resonant-circuit measurement is typically linear, steady-state, 2-port…
I’ve created new 1-port time-domain methods to measure Q, B, S21
…which add to our current collection of non-destructive tests
Most linearization techniques require a priori knowledge of nonlinearities…
I’ve developed a time-multiplexing method for reducing AM distortion
…which does not require predistortion, feedback, or calibration
4
NC STATE UNIVERSITY
7
Time-Frequency Effects: Publications
G. J. Mazzaro, M. B. Steer, K. G. Gard, and A. L. Walker,“Response of RF networks to transient waveforms:Interference in frequency-hopped communications,”
IEEE Trans. Microw. Theory Tech., Vol. 56, No. 12, pp. 2808-2814, Dec. 2008.
G. J. Mazzaro, M. B. Steer, and K. G. Gard,“Filter characterisation using one-port pulsed RF measurements,”IET Micro. Ant. Prop., Vol. 3, No. 2, pp. 303-309, Mar. 2009.
G. J. Mazzaro, K. G. Gard, and M. B. Steer,“Low distortion amplification of multisine signalsusing a time-frequency technique,”
2009 IEEE MTT-S Int. Microw. Symp., Boston, MA, June 2009, pp. 901-904.
(2 others as primary author, 3 others as secondary author)
G. J. Mazzaro, M. B. Steer, and K. G. Gard,“Intermodulation distortion in narrowband amplifier circuits,”accepted to IET Micro. Ant. Prop., Sept. 2009.
G. J. Mazzaro, K. G. Gard, and M. B. Steer,“Linear amplification by time-multiplexed spectrum,”submitted to IET Circuits Devices Syst., Sept. 2009.
publishedjournals
publishedconference
papers
unpublishedjournals
NC STATE UNIVERSITY
8
Time-Frequency Effects: Table of Contents, Preliminary Exam
1 Overview
Motivations, ObjectivesOriginal contributionsPublished & unpublished works
2 Review of Time-Frequency Concepts
Important time-frequency quantitiesReview of analysis techniquesTransients in narrowband systemsExtraction of resonant circuit parameters
3 Linear Transient Distortion in Wireless…
Single-tone pulsed filter responseSwitched-tone filter responseTransient response from lowpass prototypingCase study: frequency hopping systems
4 Linear Metrology of Bandpass RF Components
Equivalence of quality factor-- time & frequencyExtraction of resonator Q from RF pulse decayResonator separation methodDetermination of bandwidth from RF pulse decay
5 Nonlinear Metrology of Bandpass RF Systems
Generating multi-tone signal from switched sourceIMD in frequency-hopping systems
6 Remote Characterization of RF Devices
Linear & nonlinear wireless reflectionsCanonical radios & FRS/GMRS radiosSwitched-tone, one-port filter passband extraction
7 Conclusions & Future Work
same as current versionrearranged
5
NC STATE UNIVERSITY
9
Time-Frequency Effects: Table of Contents, Current Version
1 Overview
Motivations, ObjectivesOriginal contributionsPublished & unpublished works
2 Review of Time-Frequency Concepts
Important time-frequency quantitiesReview of analysis techniquesTransients in narrowband systemsExtraction of resonant circuit parameters
3 Linear Transient Distortion in Wireless…
Single-tone pulsed filter responseResonator separation methodSwitched-tone filter responseLinear wireless switched-tone reflectionsTransient response from lowpass prototypingCase study: frequency hopping systems
same as previous versionrearrangednew material
4 Linear Metrology of Bandpass RF Components
Equivalence of quality factor-- time & frequencyExtraction of resonator Q from RF pulse decayDetermination of bandwidth from RF pulse decayDetermination of S-parameters from short pulses
5 Nonlinear Metrology of Bandpass RF Systems
Generating multi-tone signal from switched sourceNon-linear wireless switched-tone reflectionsSwitched-tone, one-port filter passband extractionIMD in frequency-hopping systems
6 Linear Amplification by Time-Multiplexed Spectrum
Distortion reduction theory: 2 and 4 tonesExperimental validation: narrow- and wide-bandLinear signal recovery: theory & measurement
7 Conclusions & Future Work
NC STATE UNIVERSITY
10
Time-Frequency Effects: Linear Transient Distortion (Chapter 3)
7th-order 900-MHz 4% Chebyshev bandpass filter
higher-order filters respondmore slowly than expected
filters store/release reactive energyin a frequency-dependent cascade
0
900 MHz 34 MHz4.7 ns
2 2 900 MHzQ
f
f1
• How do transients behave inhigher-order bandpass systems?
Vin :
single-tone pulse applied to bandpass filter
MTT, 12/08
Nearest prior work:Tucker/Eaglesfield (1946) -- traces & solutions for 3rd-order filters
6
NC STATE UNIVERSITY
11
f2f1
“switched-tone” signal:
1 s 1 s
stimulus frequency switched at t = 0 ns
it is possible to produce amomentary interference patternwith a switched-tone source
+ a bandpass filter
Nearest prior work:Hatton (1951) & McCoy (1954) --
amplitude & frequency transients for pulses & chirps in RLC circuits
• How do transients behave in higher-order bandpass systems?
MTT, 12/08
(slow) (fast) overlap
Time-Frequency Effects: Linear Transient Distortion (Chapter 3)
NC STATE UNIVERSITY
12
2 102
s S j
2 1
2lp
out1
i
n j tS t
ii
V t Ae Be
bandpass transient responsesmay be computed as time-scaledlowpass transient responses
2 10 0
0 0
bp 2out
1 2
2
in S t j t j ti
i
j t j t j t
AV t e e e
B e e e
bp lp 2 1out 0 outcos
2V t t V t stimulus tone switched on at t = 0 ns, off at t = 250 ns
Nearest prior work: Blinchikoff (2001) -- response (only) at center frequency
MTT, 12/08
• How can we model these transient effects (mathematically)?
Time-Frequency Effects: Linear Transient Distortion (Chapter 3)
7
NC STATE UNIVERSITY
13
• What are the effects of these transients on wireless communications?
Filtered frequency-hopping pulses, 900-MHz 4% filterUser 1 at 10 dBm, User 2 at -20 dBm900 MHz, 100 ns guard interval
measured
sharp filtering can degrade received signal-to-noise ratio
Nearest prior work:Chohan/Fidler (1973) -- impact on FSK
& PSK, no metric
measured
MTT, 12/08
Time-Frequency Effects: Linear Transient Distortion (Chapter 3)
NC STATE UNIVERSITY
14
Time-Frequency Effects: Nonlinear Distortion (Chapter 5)
IET, 09/09
Measured IM3 for single switching eventUser 1 at -11 dBm, User 2 at -11 dBm, no guard interval
filter passband = 885 to 915 MHz, Phillips BGA2748 amplifier
• What other effects impactcommunication systems?
sharp filteringwith amplification
generates intermodulation
Nearest prior work:
Vendik/Samoilova (1997) – IMD by filtering alone,charge carrier density & conductor defects
pulses in a neighboring bandcan generate IMD in your band
8
NC STATE UNIVERSITY
15
Time-Frequency Effects: Linear Metrology (Chapter 4)
900-MHz stimulus tone turned off at t = 0 ns
the Q factor of the outer resonatorsin a chain may be determinedwith time-domain analysis
simulated
2 20ln ln R
tV V t
1 1 21
N N N1 N1 N N1
(0) (0) (0) 2 (0)dV V I Vdt R C C R C
• How do we measure parametersof coupled resonator circuits?
IET, 03/09
Nearest prior work:Pereda (1992) -- Prony analysis,
dielectric resonators, not a ‘coupled’ structure
NC STATE UNIVERSITY
16
Time-Frequency Effects: Linear Metrology (Chapter 4)
the bandwidthof a narrowband circuit
may be estimatedfrom its RF pulse decay,
without its S-parameters
• How else can we use transients to measure resonant circuits?
Pulse decay response and RF envelope7th-order 3% Chebyshev filter0-dBm 900-MHz tone removed at t = 1 us
measured
Nearest prior work:Dunsmore (1999) – tuning filtersby equalizing distances between nulls
2 1
1Bt t
bp lp 2 1env2
d g t h tdt
2t1t
2t1t
9
NC STATE UNIVERSITY
17
Time-Frequency Effects: Linear Metrology (Chapter 4)
2-port S-parameters can be extractedfrom short-pulse time-domain responses
• How else can we exploit transients for metrology?
Nearest prior work:Courtney (1999) –
permittivity measurements, nanosecond impulses
Time- & frequency-domain views, short pulses465-MHz 1% Chebyshev filters
simulated
simulated
measured
NC STATE UNIVERSITY
18
Time-Frequency Effects: Nonlinear Metrology (Chapter 5)
a device’s passbandcan be extracted from 1 port
• Can we exploit filter properties for nonlinear measurements?
IP3 of an amplifier can be measuredusing a filter & switched-tone source
Nearest prior work: Walker (2005) –steady-state two-tone testing
Simulated fast-switching filter response7th-order 465-MHz Chebyshev design(a) input, 1 tone, (b) output, 1 tone, (c) output, 2 tones
Passband extraction for bandpass filter7th-order 900-MHz Chebyshev design
IET, 09/09
measured
10
NC STATE UNIVERSITY
19
Time-Frequency Effects: Wireless Testing (Chapters 3 & 5)
Wireless linear switched-tone reflection capture(a) block diagram, (b) oscilloscope trace
Wireless non-linear switched-tone retransmission capture(a) block diagram, (b) average IM3 power plot
• Can filter transientsbe captured remotely?
Yes, switched-tone phenomenacan be captured wirelessly.
NC STATE UNIVERSITY
20
• How can we improve linearity by applying time-frequency techniques?
the IMD associated with amplitude modulationcan be reduced by trading signal bandwidthfor smaller Peak-to-Amplitude Ratio
Wideband & narrowband spectra for N = 10generated by Agilent N6030A + QM3337A modulator
measured
measured
IET, 09/09
Distortion reduction for N = 20non-multiplexed (red) vs. multiplexed (blue)Ophir 5162 amplifier
Linear Amplification by Time-Multiplexed Spectrum (Chapter 6)
LITMUScircuit architecture
11
NC STATE UNIVERSITY
21
• How can we improvelinearity with time-frequency techniques?
by rapidly time-multiplexing / amplifying / filtering a signal,distortion is reduced for equivalent desired AM signal power
Nearest prior work: Hung et. al. (2002) – optical CATV network
Time-Frequency Effects: LITMUS (Chapter 6)
IMS, 06/09
Distortion reduction for N = 4 with filteringnon-multiplexed (red) vs. multiplexed (blue)Ophir 5162 amplifier
Demonstration of linear signal recovery with low-loss filterwideband unfiltered (top) vs. narrowband filtered (bottom)
time domain (left) vs. frequency domain (right)
NC STATE UNIVERSITY
22
Time-Frequency Effects: Summary (Chapter 7)
Narrowband transients last longer than expected.
(a) identified resonant cascade as a source of long tails(b) developed a differential-equation simplification(c) showed frequency-dependence of the tails causes pulse overlap(d) evaluated ISI and IMD for frequency-hopping scenarios
Used filter transients to develop new measurement techniques:
(a) Q-factor of a single resonator(b) bandwidth, without S-parameters(c) broadband S-parameters from a single time-domain trace(d) device passband from a single input port
Time-multiplexing & filtering — LITMUS:reduces IMD associated with amplitude modulation
co-site interference
non-destructive testing
transmitterlinearity
12
NC STATE UNIVERSITY
23
Time-Frequency Effects: Future Research (Chapter 7)
(1) How can we design systemsto avoid narrowband transients?
(2) Can more information be extractedfrom pulsed & switched-tone responses?
(3) How can LITMUS be grafted onto existing modulation formats?
Garfield Minus Garfield
NC STATE UNIVERSITY
24
Time-Frequency Effects: Appendices
A — Instrument Control & Calibration
how to set up & program the equipmentin our lab to reproduce my results
B — fREEDA Elements
documentation for simulations
C — Matlab Helper Functions (data processing)
D — Cauer 1 vs. Cauer 2 Filter Designs
how transient responses differ forType 1 vs. Type 2 (same S21, different S11)
E — Filters Used in This Study
S-parameters, group delay, manufacturers & pricing
13
NC STATE UNIVERSITY
25
What’s Next?
Army Research Laboratory
2800 Powder Mill RoadAdelphi, MD
NC STATE UNIVERSITY
26
Thank you…
…to my committee members,
Dr. Michael Steer, Dr. Kevin Gard,
Dr. Mohammed Zikry, Dr. Keith Townsend,
Dr. Devereux Palmer
…to everyone who shared MRC 438 with me for three years,
…to Vadum,
…and to everyone else here today.
14
NC STATE UNIVERSITY
Tektronix TDS684Bdigitizing oscilloscope
Agilent E8267Cvector signal generator
Agilent E4438Cvector signal generator
Tektronix AWG2021waveform generator
…and they lived happily ever after.
The End.
NC STATE UNIVERSITY
15
NC STATE UNIVERSITY
29
Time-Frequency Effects: Resonator Separation (Chapter 3)
• What is the natureof filter transients?
Long-tail behavior is attributedto the resonant cascade.
Resonator-separation simulation methodcircuit in ADS (above) and voltage at filter output (left)
NC STATE UNIVERSITY
30
Time-Frequency Effects: Tucker/Eaglesfield (1946)
differential operators (precursor to Laplace Transforms) can be used to solve for analytical forms for pulse responses
oscilloscopes can capture filtered pulse envelopes
• commonly-used6-elementbandpassfilter
Filtered Pulse Responses
while analyzing non-ideal (transmission-characteristic) filters…
“Transient response of filters,” Wireless Engineer, Vol. 23, pp. 36-42 & 84-90, Feb-Mar. 1946
16
NC STATE UNIVERSITY
31
Time-Frequency Effects: Hatton (1951) & McCoy (1954)
“Simplified FM transient response,” MIT, Cambridge, MA, Tech. Rep. 196, Apr. 1951“FM transient response of band-pass circuits,” Proc. IRE, vol. 42, no. 3, pp. 574-579, Mar. 1954
overshoots in amplitude & frequency are possiblefor input frequency transitions within a filter’s passband
• amplitude transients & frequency transientsfor a single resonator
Normalized Amplitude Transients Normalized Frequency Transients
while comparing frequency-modulation to amplitude-modulation…
NC STATE UNIVERSITY
32
Time-Frequency Effects: Blinchikoff (2001)
Filtering in the Time and Frequency Domains, Raleigh, NC: SciTech Publishing, Inc., 2001.
• lowpass vs. bandpass transient responses
transient response at midbandis a time-scaled version of thelowpass turn-on response
0cosb l Nu t u t t
17
NC STATE UNIVERSITY
33
Time-Frequency Effects: Chohan/Fidler (1973)
Normalized Frequency Transients
frequency step as a percentage of B,2nd-order filter
phase step,2nd-order filter
• frequency transients,steps of phase/frequencyat filter input
generalized earliernarrowband Laplace methodsfor any order & any Q value
while investigating filtering effectson FSK- and PSK-type signals…
“Generalised transient response of bandpass transfer functions to FSK and PSK-type signals,”
Electronics Letters, vol. 9, no. 14,pp. 320-321, July 1973.
NC STATE UNIVERSITY
34
Time-Frequency Effects: Vendik/Samoilova (1997)
• resonators: transmission line& microstrip
attributed nonlinearities to(a) crystalline structure(b) charge carrier density(c) Abrikosov vortices
resistance is a function of current
“Nonlinearity of superconducting transmission line and microstrip resonator”
IEEE Trans. Microw. Theory Tech., vol. 45,no. 2, pp. 173-178, Feb. 1997.
2
1 1 20
( , )( , ) 1 I x tR x t RI
18
NC STATE UNIVERSITY
35
Time-Frequency Effects: Pereda (1992)
“Computation of resonant frequencies and quality factors of open dielectric resonatorsby a combination of finite-difference time-domain and prony’s methods,”
IEEE Microwave and Guided Wave Letters,vol. 11, no. 2, pp. 431-433, Nov. 1992.
• estimation of quality factor from resonant decay
reduced time to compute resonant frequencies and quality factor using FDTD and Prony analysis
Prony analysis
while investigating resonancein dielectric resonators…
NC STATE UNIVERSITY
36
Time-Frequency Effects: Dunsmore (1999)
“Tuning band pass filters in the time domain,”IEEE MTT-S Int. Microw. Symp., Anaheim, CA, June 1999, pp. 1351-1354.
showed how to tune individual resonators using time-domain return loss
while working at Hewlett-Packard Microwave Instruments Division…
• time-domain coupled-resonator filter tuning
19
NC STATE UNIVERSITY
37
Time-Frequency Effects: Courtney (1999)
• frequency measurements from time-domain traces
while trying to determine broadband permittivity and permeabilityof a sample material…
found a way to measure and by time-domain-reflectometrywith nanosecond impulses
“One-port time-domain measurement of the approximate permittivity and permeability of materials”IEEE Trans. Microw. Theory Tech., vol. 47, no. 5, pp. 551-555, May 1999.
NC STATE UNIVERSITY
38
Time-Frequency Effects: Walker (2005)
“Behavioral modeling and characterization of nonlinear operation in RF and microwave systems,”
Ph.D. dissertation, North Carolina State University, Raleigh, NC, May 2005.
• remote characterizationof wireless devices
probed devices with two-tone continuous waves, received steady-state third-order intermodulation
identified RF communication bands using nonlinear front-end properties
20
NC STATE UNIVERSITY
39
Time-Frequency Effects: Hung et. al. (2002)
• nonlinear distortion reduction by time-multiplexing
“Optical sampled subcarrier multiplexing scheme for nonlinear distortion reduction in lightwave CATV networks,” Electronics Letters, vol. 38, no. 25, pp. 1702-1704, Dec. 2002.
found a way to reduce IMDby transmittingsubcarrier frequenciesin different time slots
working with optical cable television transmission…