Techniques for Characterizing

Riadh Said
Product Manager
Microwave and Communications Division

Spurious Signals
Keysight Technologies

October 21, 2014

Our Goals today

Review the sweep time equation to trade-

off dynamic range for sweep speed.

Review the basics of applying good spur

searching strategy.

Introduce new spectrum analysis

technologies to accelerate spur searching
for both R&D and manufacturing.

Page 2
Agenda:
1. What is a spur and why we care about them.
2. Introduce the sweep time equation for spectrum analysis.
3. Establishing your spur searching strategy.
4. Translate sweep time equation to real-life example
5. Techniques and technologies to manage your spur search time more effectively:
• New wideband, high dynamic range ADC & display signal processing
• New fast sweep capabilities vs. traditional sweep
• New stepped FFT approach with spur subtraction
• Noise subtraction & new pre-amps for best sensitivity

Page 3
Assumptions & Definitions

– Spectrum analyzer basics such RBW, TOI, and SHI

– “Noise” = Displayed average Noise level (DANL), sensitivity, noise

figure

– Distortion = Harmonics 2nd and 3rd order and their products

Page 4
What is a spur?
Definition: spu·ri·ous, \ˈspyu̇r-ē-əs\  Adjective
1. not genuine, sincere, or authentic
2. of illegitimate birth
3. outwardly similar or corresponding to something without having its genuine qualities
4. of falsified or erroneously attributed origin
5. of a deceitful nature or quality

Your Design

What you got.

What you wanted.
Spurs & noise

Page 5
 Where do spurs come from?
…Many many places
• Mixers  2nd and 3rd order harmonics + their mixing products with LO and IF inputs…
• Multipliers  ½ rate , ¾ rate harmonics or “sub-harmonics”
• Dividers  Odd Harmonics
• Oscillators, LO, VCO’s, clocks  Leakage
• PLL’s  Frac-N-Loops spurs
• Amplifiers  2nd and 3rd order Harmonics
• DAC’s  repetitive quantization errors…
• Poor filtering & isolation/coupling
• Incidental resonance from parasitic capacitance or inductance in circuit
• Vibration  microphonic noise/spurs
• Power supply: line noise or switching harmonics, 50/60 Hz and their harmonics
• Glitch's or discontinuities in digital IF or baseband FPGA’s, ASIC’s etc…
• The mixing products of any of the above.

When a non-linear device is presented with two or more input frequencies the output will generator both the input frequencies
and the intermodulation distortion products of those input at the same time.

Page 6
Why do we care?
…Link Budgets, Sensitivity, Range, Quality of Service
Radar/EW example: False target /threat detection or inference
Transmitter Spurious Signal
Signal, f0 Return Signal, f1 (-85 dBm)

Satellite example: In channel Rx interference degrades sensitivity & range

Spurious Signal – Jam yourself
Very small received signal (-130 dBm)

Cellular example: Out of channel Interference pollutes neighbors receiver and

degrades range & data rates. Regulatory requirements. (i.e. FCC)
Small received signal (-50 dBm)
Spurious Signal – Jam your neighbor

Page 7
Spectrum Analyzer Dynamic range
Balancing Distortion, Noise & Test time

Block diagram of a classic superheterodyne spectrum analyzer

Spectrum Analyzer Dynamic range is limited by three factors:

1. Distortion performance of the input mixer  2nd & 3rd order products
2. Broadband noise floor of the system.  Sensitivity/Displayed Average Noise level (DANL)
3. Phase noise of the local oscillator  Narrow band measurements.

The above three factors must be optimized in combination with your DUT and measurement
uncertainties requirements.

Page 8
Spectrum Analyzer Dynamic range
Distortion and Noise
• Increasing input attenuation reduces harmonic
distortion from spectrum analyzer. However this
also adds more IF gain which degrades noise
floor.

• To compensate for this you can reduce the RBW

and reduce the noise floor.

For more information see: Page 9

App Note 150: Spectrum analyzer basics
Spectrum Analyzer Dynamic range
Uncertainties Budget for Noise + Distortion
Spectrum Analyzer Dynamic range must be optimized in combination with
your DUT requirements and measurement uncertainties you can tolerate.

Example of uncertainty budget:

Distortion Error budget: +/- 1 dB error = -18 dBc margin relative to DUT input
Noise Error budget: +/- 0.3 dB error = 5 dB margin relative to noise floor
Maximum total error: (+/- 1 dB) + (+/- 0.3 dB) = +/- 1.3 dB (excludes instrument uncertainty)

Distortion Error Noise Error

Uncertainty versus difference in
amplitude between two sinusoids Error in displayed signal
at the same frequency amplitude due to noise

For more information see: App Note 150: Spectrum Page 10

analyzer basics and http://mwrf.com/author/bob-nelson
The sweep time equation for spectrum analysis
Balancing Dynamic Range & Test Time
RBW = Resolution Bandwidth Filter The rise time of a filter (RBW) is inversely
k (Span) ST = Sweep Time proportional to its bandwidth, and if we
ST = k = the constant of proportionality include a constant of proportionality, k,
RBW 2 then: Rise time = k/RBW

100x delta in sweep time

with a 10x delta in RBW!
RBW has a squared
relationship with time.

42 ms (300kHz RBW)

-10 dB
Noise floor change* =
10 log (BW2/BW1)

Where 4.2 sec (30 kHz RBW)

BW1 = starting resolution bandwidth
BW2 = ending resolution bandwidth
* Peak-detectors do not accurately represent the noise floor.

Note: The k value varies based on a number of conditions including filter shape for
RBW, VBW and detector types. Generally a value of 2 or 3 for Gaussian filters. Page 11
 Probability of Intercept for Swept Analysis
Odds of detecting intermittent spurious signals
Perfect POI = 1
Range: zero to one

(R+T)
POI =
(R+R’)
T = duration of the signal of interest
R = listening time at frequency
R’ = time not listening
R+R’ = revisit time
Note: Assumes signal can be discerned

(RBW+SBW)*ST
R (approx) =
Span The Term (R+R’) is the sum of the sweep time and the dead time
between sweeps. And it is also called the revisit time.
RBW = Resolution BW
SBW = spectral width of the signal
ST = spectrum analyzer sweep time Listening time of swept-LO spectrum analyzers can easily be
Span = spectrum analyzer span approximated as the amount of time that some portion of the
resolution–bandwidth filter overlaps some part of the signal energy.

Page 12
Establishing your spur searching strategy
maximize test efficiency
Know where & what to look for:
• Start with DUT block diagram, modes of operation & know design issues.
• Establish required target levels (Balance schedule & search time)
• Establish the type of spurious signals (static, moving, harmonic, random, modulated)

Balance the needs in production vs. R&D:

• R&D and production teams should partner to focus on problem areas.
• Production target test levels ideally more relaxed if designed with margin.
• Focus on known product/component variation problems.

Determine uncertainty budget:

• Just enough margin to yield required results and minimize total test time.
• Remove tests that deliver 100% yield with high margins. Go to sampling.
• Apply appropriate measuring tools for max speed. (Swept, FFT, Real-time)

Unexpected ASIC errors Band Filters In Band Out of Band

1
2
3 Don’t waste time
Baseband
LPF X2 looking outside
/IF/ ASIC
filtered bands?
N
LO Don’t waste time
Focus on known
looking outside
problems first.
filtered bands?
Page 13
 Real-life example
The sweep time equation in action
 Target non-harmonic spurs: >-100 dBm
 Max signal input size = - 10 dBm
 Target attenuation = 16 dB
 Target measurement error budget ~ 2 to 3 dB
 RBW = 3 kHz, (peak detector, with pre-amp)
 Span = 10 MHz to 18 GHz
 Sweep time = 2.4k sec or 40 minutes
 300 different test modes of DUT
 Total test time = 200 hours!
 Repeat every time a new design change is made…

Page 14
Manage your spur search time more effectively
New Techniques and technologies

Large spurs > -75 dBc

Medium spurs > -100 dBc

Small Spurs < - 100 dBc

Page 15
Large Spurs >-75 dBc
New wideband high dynamic range digitization
& Processing

510 MHz BW
500 MHz SDRAM

ADC FPGA FPGA

ASIC
CLK

Page 16
The Follow features available on Keysight X-Series
Analyzer & new N9040B UXA Signal Analyzer

Page
Agilent Confidential
Page 17 July 2014
510 MHz, High Performance IF for spur searching
New N9040B UXA Spectrum Analyzer

– See your spurs clearly with SFDR

of > -75dBc across 510MHz BW

– Monitor and capture highly elusive

spurs across the full analysis
bandwidth with real time signal
analysis

– Maximise dynamic range and

accuracy with excellent IF
frequency response of <0.7dB

For more information see application note: Page

Using Wider, Deeper Views of Elusive Signals to Characterize Complex Page 18
Systems and Environments 5992-0102EN
How to Capture the Intermittent Spurious Signal

– Questions about the signal of

interest
• What Frequency?
• How Often?
• How much Power?
• What is the Bandwidth?
• What Modulation?
• Where is the Noise Floor?
• What is the Phase Noise?
• Is there more than one signal?

Page 19
The Swept Analysis Mode

• A swept LO w/ an assigned RBW.

Swept LO

• Covers much wider span.

• Good for events that are stable in Freq

Lost Information
the freq domain.

• Magnitude ONLY, no phase

Lost Information
information (scalar info).

• Captures only events that occur at

Lost Information
right time and right frequency
point.

• Data (info) loss when LO is “not

there”.
Time

Page 20
IQ Analyzer (Basic) Mode – Complex Spectrum and Waveform
Measurements

• A parked LO w/ a given IF BW
• Collects IQ data over an interval Parked LO
of time.
• Performs FFT for time- freq-
domain conversion Freq

• Captures both magnitude and Meas Time

or
phase information (vector info). FFT
Window
Length
• Data is collected in bursts with Lost Information
data loss between acquisitions.
Meas Time
or
FFT
Window
Length
Analysis BW

Time

Page 21
Real–Time Spectrum Analysis

• A parked LO w/ a given IF BW
• Collects IQ data over an interval
Parked LO
of time.
• Data is corrected and FFT’d in
parallel Freq

• Vector information is lost

Acquisition or
• Advanced displays for large slice time

amounts of FFT’s
Acquisition or
slice time

Real-time BW

Time

Page 22
Real-Time Spectrum Analysis Hardware
ADC
(400 MSA/s, 14-bit)

Real-time corrections and decimation

Time Domain Overlap

Processor Memory

FFT Engine
(292,968 FFT’s/s)

Power vs Time Spectrum Density trace Frequency

trace memory trace memory memory Mask Trigger

Display processor

Page 23
Using Real-Time Spectrum Analysis

Benefits
– Gap free capture
– Supports wide bandwidths
– Real-Time capture of signals that are
present for only 2 ns with large S/N
ratio’s.
– For full amplitude accuracy UXA POI
expressed in time = 3.57 us

– Best at measuring the

shortest duration signals that
are infrequent or occur only
one time within 510 MHz

POI express in time duration T is the minimum length of the

signal of interest if it is to be detected with 100-percent
probability and measured with the same amplitude accuracy as
Window size = FFT windowing in points
Time record length = FFT bin size in points that of a CW signal.
P = Overlap FFT processing points
fs= sample rate
For more information see application note:
Understanding and Applying Probability of Intercept in Real-time Spectrum
Analysis 5991-4317EN Page 24
Using Wider, Deeper Views of Elusive Signals to Characterize Complex
Systems and Environments 5992-0102EN
Real-Time Spectrum Analysis – Density Display
Color coded for fast visualization & triggering

Page 25
Wider, cleaner analysis BW
Quickly Analyze large spans for spurs with confidence

Maximize the dynamic range for

>-75 dBc optimum headroom.

510 MHz

Page
Minimize Measurement Uncertainty
IF Frequency Response UXA N9040B

<0.7

Minimize measurement
uncertainties across wide
instantaneous bandwidths

Page
Page 27
Stepped Density Method
Using Real-time Dwell
Capture of repetitive non CW signals over large bandwidths Freq

Time

For more information see application note:

Using Wider, Deeper Views of Elusive Signals to Characterize Complex
Systems and Environments 5992-0102EN Page
28
Stepped Density Method
Using Real-time Dwell over full span

Full span of spectrum analyzer

Page
29
Medium to low Spurs (<-75 dBc)
New fast sweep vs. traditional sweep

500 MHz SDRAM

ADC FPGA FPGA

ASIC
CLK

Page 30
The Follow features available on Keysight X-Series
Analyzer & new N9040B UXA Signal Analyzer

Page 31
Agilent Confidential
July 2014
Traditional Sweep – non-continuous signals
Limited by analog RBW rise time vs. accuracy needs
Standard Analog Sweep
Freq

k (Span)
Sweep
ST =
Signals RBW 2
RBW
(R+T)
POI =
(R+R’)

(RBW+SBW)*ST
R =
Span

Time

Page 32
 Fast Sweep – non-continuous signals
Compensated RBW rise time with improved accuracy
UXA Fast Sweep Freq

Modified
k value

k (Span)
ST =
RBW 2
(R+T)
POI =
(R+R’)

(RBW+SBW)*ST
R =
Span
Time

For more information see :

5991-3739EN-Using Fast-Sweep Techniques to Accelerate Spur Searches Page 33
 Up to 50x faster vs. Traditional Sweep

Fast Sweep Traditional Sweep

Full 26.5 GHz span Full 26.5 GHz span

For more information see :

5991-3739EN-Using Fast-Sweep Techniques to Accelerate Spur Searches Page 34
34
 The Effects of Over-sweeping
RBW filter limits the rise time of signal

Oversweeping produces errors in frequency, amplitude and bandwidth

• Low amplitude: The displayed amplitude of the spurious

and other signals is lower than the true value, and by an
amount larger than the analyzer specifications would
indicate.

• Bandwidth spreading: The effective RBW of the

measurement is significantly wider than the selected value.

• Frequency shift: The apparent center frequency of

spurious and other signals is higher than the true value, and
by an amount larger than the analyzer specifica- tions would
indicate.

Leveraging real-time DSP during fast sweep, the phase response of the RBW filter is
adjusted based on the sweep rate to compensate for oversweeping effects. This maintains
the correct amplitude and bandwidth of the detected signal, even at very high sweep rates.

For more information see :

5991-3739EN-Using Fast-Sweep Techniques to Accelerate Spur Searches Page 35
FFT S WEEP SPEED IMPROVEMENT
Sweep Speed Improvement with IF CHIRP Processing

> 10 x Faster!
Area where Chirp IF
Without Chirp processing improves
speed

With Chirp

FFT Swept

Page
Page 36
Fast Sweep Repeatability

Holding sweep time constant while

using a narrower RBW to measure
CW signals reduces measurement
variance because the narrower
filter blocks more of the broadband
noise.

Comparing fast sweep to traditional sweep, the lower values and shallower slope of the blue data points
(fast sweep) show that repeatability is improved and varies less with sweep time.
For more information see :
5991-3739EN-Using Fast-Sweep Techniques to Accelerate Spur Searches Page 37
37
 Fast Sweep Benefits Summary

Fast-sweep technology provides at least four important benefits:

1. Dramatically reduced sweep times for CW spur searches over wide spans and
narrow RBWs. (> 10x faster)

2. Improved measurement throughput while maintaining accuracy, frequency

selectivity and consistent bandwidth

3. Improved measurement repeatability at faster sweep rates

4. Simplified selection of RBW to get a desired combination of dynamic range and

repeatability, because repeatability depends almost entirely on dynamic range
rather than both dynamic range and sweep time.

Note: Installed base Keysight (Agilent) X-Series spectrum analyzers can be upgraded with fast sweep.

Page 38
38
Medium to Low Spurs (< -75 dBc)
Stepped FFT

FFT- Based

Page 39
The following features available on the M9393A
PXIe Performance Vector Signal Analyzer

Page 40
Stepped spectrum analysis
High-speed stepped FFTs to analyze wide spans
 Benefits: Single FFT up to max
analysis BW (160 MHz)
 Fast “sweeps” and excellent dynamic
range with narrow RBW
 Search for spurs, measure harmonics
or analyze multiple signals at once.
 Trade-off
 Solid state front end limits max
sensitivity, but can be balance with
narrower RBW’s and speed.

 Method:
Span up to frequency range of analyzer (27 GHz)
 Multiple FFTs are concatenated to
create a span >the IF bandwidth
 High speed LO and digitizer list mode  Fastest stepped spectrum analysis requires wide
analysis bandwidth and fast frequency tuning
enables fast stepping across span options, and powerful computer.
 Software stitches together FFTs and
displays single trace result up to 27GHz

Page 41
Digital Image Rejection
Fast, accurate measurements without hardware pre-selector

High & Low Side Mixing Smart Image rejection processing

Minimum
detect
 Method:
1. Adjust LO making 2 acquisitions: high-side
mix then low-side mix
2. Use minimum detection algorithm to
determine real signals
3. Use additional techniques including max-
hold, IF dithering and using narrow RBW to
accurately measure even challenging signals
For more information see application notes:
Page 42
5991-4039EN - Achieving Excellent Spectrum Analysis Results Using
Innovative Noise, Image and Spur-Suppression Techniques
Digital image rejection
Fast, accurate measurements without hardware pre-selector
 Benefits:
 Fast tuning speed
 Excellent amplitude accuracy
 Compact physical size

2 GHz LFM
signal

AWG 12
GHz clock

AWG alias
product

Image-protect
Challenging test scenario: Measuring 2 GHz wide linear FM chirp
from arbitrary waveform generator centered at 2 GHz

For more information see application notes:

Page 43
5991-4039EN - Achieving Excellent Spectrum Analysis Results Using
Innovative Noise, Image and Spur-Suppression Techniques
Power Spectrum mode for high-speed stepped FFT analysis
Results

– Measure spurs & harmonics across 27 GHz @ 10 kHz RBW in < 1

second

– Achieve > 300 GHz/sec sweep speeds

Page 44
Stitched FFT with Digital Image Rejection Summary
Optimized for speed & compact form factor

– Fastest tuning speed enabled by:

Frequency tuning speed (nominal)
• Very fast LO & all solid-state design
< 3.6 GHz 175 us
3.6 to 8.4 GHz 135 us
• List mode executes predefined set
of acquisitions from digitizer FPGA
8.4 to 13.6 GHz 135 us
13.6 to 17.1 GHz 155 us • Takes advantage of latest processor
17.1 to 27 GHz 145 us technology – M9037A or high-power
PC

– Outstanding speed to dynamic range to

Meas. time: see low-level signals quickly
105 dB 1 second
dynamic range
– Leading to shorter test times & higher
1 GHz span throughput in design validation &
production

Page 45
Low Spurs (< -100 dBc)
How to improve the noise floor

Pre-Amps Noise subtraction

Page 46
 Noise Floor Extensions/Corrections
Subtracting the noise floor of the spectrum analyzer
Benefit:
 Accurately measure signals
close to the noise floor

Trade off:
 Increases variability typically
requires more averaging when
near the noise floor and
therefore more time.

Method:
1. Measures internally generated noise (N)
2. Measures input signal and noise (S + N) Feature available on both X-Series & M9393A
3. Subtract the two measurements for
corrected result: (S + N) – N = S
4. Use averaging/VBW filter to see effects of
noise correction Real-time ASIC
Software driven

For more information see application notes:

5990-5340EN - Using Noise Floor Extension in the PXA Signal Analyzer
5991-4039EN - Achieving Excellent Spectrum Analysis Results Using
Page 47
Innovative Noise, Image and Spur-Suppression Techniques
USB Pre-Amps
Pre-calibrated plug & play operation

Page 48
U7227x USB Preamplifiers
– Pre-calibrated and ready to use with X-Series
– Noise Fig ~ 5dB

– Gain >17 dB (makes NF of SA

negligible).
U7227x USB Preamps
– USB provides power to Preamp,
and reads gain, noise fig, and S-
parameter data from flash.

– UXA SA app can use preamp as

“remote front end”; correct absolute
amplitude vs frequency displayed.

Model Frequency Range

X-series Analyzers + USB Preamplifiers provide:
U7227A 10 MHz to 4 GHz
 2X improved noise figure beyond 10 GHz up to 50 GHz
U7227C 100 MHz to 26.5 GHz
 Improved measurement uncertainty up to 1/3
U7227F 2 GHz to 50 GHz  Lower DANL/noise floor improving to -171 dBm/Hz

Keysight USB Pre-amplifiers: 5991-4246EN:

Page 49
 Summary

1. Start with a smart spur search strategy

2. Balance it with the Sweep time considerations
3. Leverage modern digital IF processing:
• Real-Time
• Stepped Density Capture of repetitive non CW signals over large bandwidths
• Real-time Density Best at measuring the shortest duration signals that are infrequent or
occur only one time within 510 MHz
• Pre-Selected Fast sweep  Best at measuring both small and large signals
with or without modulation at high speed.
• Stitched FFT with digital image & spur rejection  Ideal for continuous
CW like spurs. Ideal for the fastest speeds when absolute noise floor can be
traded off.
• Noise Floor corrections – When you can trade-off speed for dynamic range.
4. USB Pre-Amps – Simplified calibrated setup to extend noise floor, improve
uncertainties, or increase sweep speed with wider RBW’s.

Page 50
 Where to learn more/References

• Spectrum Analyzer Basics, App Note 150, 5952-0292:

http://literature.cdn.keysight.com/litweb/pdf/5952-0292.pdf
• Using Wider, Deeper Views of Elusive Signals to Characterize Complex Systems and Environments:
http://literature.cdn.keysight.com/litweb/pdf/5992-0102EN.pdf
• Using Fast-Sweep Techniques to Accelerate Spur Searches, 5991-3739EN:
http://literature.cdn.keysight.com/litweb/pdf/5991-3739EN.pdf
• Achieving Excellent Spectrum Analysis Results Using Innovative Noise, Image and Spur-Suppression Techniques
http://cp.literature.agilent.com/litweb/pdf/5991-4039EN.pdf
• Understanding and Applying Probability of Intercept in Real-time Spectrum Analysis :
http://cp.literature.agilent.com/litweb/pdf/5991-4317EN.pdf
• Measuring Agile Signals and Dynamic Signal Environments: 5991-2119EN:
http://cp.literature.agilent.com/litweb/pdf/5991-2119EN.pdf
• Using Noise Floor Extension in the PXA Signal Analyzer:
http://cp.literature.agilent.com/litweb/pdf/5990-5340EN.pdf
• N9040B UXA X-Series Signal Analyzer: www.keysight.com/find/UXA
• M9393A PXIe Performance Vector Signal Analyzer: www.keysight.com/find/M9393A
• Keysight USB Pre-amplifiers: 5991-4246EN: http://literature.cdn.keysight.com/litweb/pdf/5991-4246EN.pdf
• Spur Calculator from Marki Microwave: http://www.markimicrowave.com/WepApps/Spur_Calculator.aspx
• Technical Expert Bob Nelson: http://mwrf.com/author/bob-nelson

Page 51
Thank You!

Page 52