phase noise and jitter measurements
DESCRIPTION
Originally presented at DesignCon 2013. Jitter is a very important topic in signal integrity for high speed serial data links. The jitter performance of clock signals used in generating the serial data signal is critical to the overall performance of these signals. Phase noise is the most sensitive and accurate measurement of the performance of precision clocks. This presentation covers the theory and practice for making phase noise measurements on clock signals as well as the relationship between phase noise and total jitter, random jitter and deterministic jitter. Measurements on a typical clock signal is also included. For more information, visit http://rohde-schwarz-scopes.com or call (888) 837-8772 to speak to a local Rohde & Schwarz expert.TRANSCRIPT
Phase Noise andJitter Measurements
DesignCon 2013 | Phase Noise and Jitter | 2
Phase Noise and Jitter MeasurementsIntroduction
Rick Daniel
Application Engineer
Rohde & Schwarz
Colorado Springs, CO
DesignCon 2013 | Phase Noise and Jitter | 3
Agenda
• Jitter Review
• Time-Domain and Frequency-Domain Jitter Measurements
• Phase Noise Concept and Measurement
• Deriving Random and Deterministic Jitter from Phase Noise
• PLL/Filter Weighting of Jitter Spectrum
• Calculating Peak-to-Peak Jitter from RMS Jitter
DesignCon 2013 | Phase Noise and Jitter | 4
J i t t e rJ i t t e rWhat is J i t t e r ?
• Jitter is the short-term time-domainvariations in clock or data signal timing
• Jitter includes instability in signal period, frequency, phase,duty cycle or some other timing characteristic
• Jitter is of interest from cycle to cycle, over manyconsecutive cycle, or as a longer term variation
• Jitter is equivalent to Phase Noise in the frequency domain
• Variations with frequency components >10Hz are Jitter
• Variations with frequency components <10Hz are Wander
DesignCon 2013 | Phase Noise and Jitter | 5
Types of Jitter• Time Interval Error (TIE)
• Fundamental measurement of jitter• Time difference between measured
signal edge and ideal edge• Instantaneous phase of signal
• Period Jitter• Short-term stability, basic parameter for clocks
• Cycle to Cycle• Important for parallel data transfer
• N-Cycle• Important when clock and data routing differ
DesignCon 2013 | Phase Noise and Jitter | 6
Jitter Measurement Techniques• Time Domain (Oscilloscope)
• Direct method for measuring jitter• Measures TIE, Period Jitter, Cycle-to-Cycle Jitter• Measures RMS or Peak-to-Peak Jitter• Measures data or clock signals• Limited sensitivity (100 – 1000 fs)
• Frequency Domain (Phase Noise Analyzer)• Calculates jitter from phase noise• Measures RMS Jitter• Measures clocks, not random data streams• Easy to separate random and discrete jitter components• Highest sensitivity (<5 fs)
DesignCon 2013 | Phase Noise and Jitter | 7
What is Phase Noise?• Ideal Signal (noiseless)
V(t) = A sin(2t)
whereA = nominal amplitude = nominal frequency
• Real SignalV(t) = [A + E(t)] sin(2t + (t))
whereE(t)= amplitude fluctuations(t) = phase fluctuations
Phase Noise is unintentional phase modulation that spreads thesignal spectrum in the frequency domain.Phase Noise is equivalent to jitter in the time domain.
Level
f
Level
f
t
t
Time Domain Frequency Domain
DesignCon 2013 | Phase Noise and Jitter | 8
Phase Noise – Unit of Measure• Phase Noise is expressed as LL(f)
• LL(f) is defined as single sideband power due tophase fluctuations in a rectangular 1Hz bandwidthat a specified offset, f, from the carrier
• LL(f) has units of dBc/Hz
FREQUENCY
AMPLITUDE
1 Hz
fc fc+f
(f)
LOG A
LOG f
LL
DesignCon 2013 | Phase Noise and Jitter | 9
ClockUnderTest
Phase Noise Measurement Setup
DesignCon 2013 | Phase Noise and Jitter | 10
Example Phase Noise Measurement Plot
Offset from Fundamental Frequency
Pha
seN
oise
(dB
c/H
z)
Discrete Spurs
DesignCon 2013 | Phase Noise and Jitter | 11
Phase Noise Measurement• Shows phase noise over a
range of offset frequencies: LL (f)
• RMS Jitter =
• Phase noise including spursyields total jitter (random plusdeterministic)
• Phase noise without spursyields random jitter
dfffc
)(221
L
DesignCon 2013 | Phase Noise and Jitter | 12
Jitter / Phase Noise Measurement LimitationOscilloscope or
Phase NoiseAnalyzer
• Jitter measured by an oscilloscope or phase noise analyzer isalways the RMS sum of the clock jitter and the measuringinstrument’s internal jitter
• Internal jitter/phase noise limits measurement sensitivity
• Examples:• Clock Jitter: 1ps Instrument Jitter: 1ps Measured Jitter: 1.4ps
• Clock Jitter: 500fs Instrument Jitter: 1ps Measured Jitter: 1.118ps
• Clock Jitter: 500fs Instrument Jitter: 300fs Measured Jitter: 583fs
• Clock Jitter: 200fs Instrument Jitter: 5fs Measured Jitter: 200.06fs
DesignCon 2013 | Phase Noise and Jitter | 13
Phase Noise MeasurementPhase Detector Technique
Ref.Source
ClockUnderTest
PhaseDetector
DDFF=90°°
Low PassFilter LNA Baseband
Analyzer
PLL(tracks DUT freq,maintains 90°°offset)PLL Low Pass Filter
(sets loop BW)
Reference source is tuned to same frequency as clock with 90°phase offset (quadrature)
DesignCon 2013 | Phase Noise and Jitter | 14
Phase Detector with Cross-Correlation
ClockUnderTest
PD
DDFF=90°° Low PassFilter
LNA
ADC
PLL
Correlation
PD
DDFF=90°°Low Pass
Filter LNA
ADC
PLLNoise 2
Noise 2Noise DUT
Noise 1
Noise 1Noise DUT
Ref 2
Ref 1
Cross-correlating both measurements effectively reducesreference source noise – improves measurement sensitivity
DesignCon 2013 | Phase Noise and Jitter | 15
Jitter Measurement InstrumentsReal time (Oscilloscope)
Single-shot or repetitive events (clock or data)Bandwidths typically 60 MHz to >30 GHzLowest sensitivity (highest jitter noise floor)Measures adjacent cycles
Repetitive (Sampling Oscilloscope)Repetitive events only (clock or data)Bandwidths typically 20 GHz to 100 GHzGenerally can not discriminate based on jitter frequencyCannot measure adjacent cycles
Phase noise (Phase noise test set)Clock signals only (50% duty cycle)Integrate phase noise over frequency to measure jitterHighest sensitivity (lowest jitter noise floor)Cannot measure adjacent cycles
HighSensitivity
HighFlexibility
DesignCon 2013 | Phase Noise and Jitter | 16
Phase Noise Measurement (without spurs)• Total RMS Jitter (RJ)
• 42.8 fs
DesignCon 2013 | Phase Noise and Jitter | 17
Phase Noise Measurement (with discrete spurs)• Total RMS Jitter (RJ & DJ)
• 59.9 fs
• Total Jitter (TJ) is RMSsum of RJ and DJ:
• DJ can be calculated as:
• TJ = 59.9 fs, RJ = 42.8 fs• Calculated DJ = 41.9 fs
22 DJRJTJ
22 RJTJDJ
DesignCon 2013 | Phase Noise and Jitter | 18
Measurement of DJ from Individual Contributors• Isolate individual discrete components and evaluate their
contribution to total RMS Jitter
DesignCon 2013 | Phase Noise and Jitter | 19
Measurement of DJ from Individual Contributors• DJ from 1.294kHz spur:
• 1.0 fs
DesignCon 2013 | Phase Noise and Jitter | 20
Measurement of DJ from Individual Contributors• DJ from 1.676kHz spur:
• 3.7 fs
DesignCon 2013 | Phase Noise and Jitter | 21
Measurement of DJ from Individual Contributors• DJ from 20kHz spur:
• 40.9 fs
DesignCon 2013 | Phase Noise and Jitter | 22
Measurement of DJ from Individual Contributors• DJ from 40kHz spur:
• 4.6 fs
DesignCon 2013 | Phase Noise and Jitter | 23
Measurement of DJ from Individual Contributors• DJ from 60kHz spur:
• 6.4 fs
DesignCon 2013 | Phase Noise and Jitter | 24
Measurement of DJ from Individual Contributors• DJ from 80kHz spur:
• 2.8 fs
DesignCon 2013 | Phase Noise and Jitter | 25
Summary of Total Jitter• TJ is RSS of all contributors
226
25
24
23
22
21 RJDJDJDJDJDJDJTJ
1fs
3.7fs
40.9fs
4.6fs6.4fs
2.8fs
42.8fs
fs9.598.428.24.66.49.407.31 2222222
DesignCon 2013 | Phase Noise and Jitter | 26
Measurement of a Very Low Jitter Device• Crystal based 640MHz signal with very low phase noise/jitter
• 4.6fs
DesignCon 2013 | Phase Noise and Jitter | 27
Frequency Offset Range is Settable• Measurements in this presentation have used offset range of
1kHz to 10MHz• Offset range can be as wide as 0.01Hz to 30GHz!
DesignCon 2013 | Phase Noise and Jitter | 28
• By default, jitter is calculated over entire measured offset range• A subset of the offset range may be specified for the jitter
calculation
• Jitter calculated over fullmeasured range of 1kHzto 10MHz is 4.6fs
• For reduced range of 10kHzto 3MHz it is 2.8fs
Jitter Calculation over Subset of Offsets
DesignCon 2013 | Phase Noise and Jitter | 29
• Basic measurement shows raw performance of clock• Real systems use PLLs• FSUP can apply a weighting function to simulate the frequency
response of a PLL
Jitter Calculation with PLL Weighting
Define PLL Freq Response
Select PLL to apply to measurementUnweighted Jitter: 4.6fs Weighted Jitter: 3.3fs
PLL1
DesignCon 2013 | Phase Noise and Jitter | 30
Phase Detector with Cross-Correlation is mostsensitive way to measure phase noise and jitter
Phase Noise/Jitter MeasurementSpectrum Analyzer vs Phase Detector vs PD with Cross-Correlation
SA: 68.7fs
PD: 13.1fs
PD w/CC: 4.6fs
DesignCon 2013 | Phase Noise and Jitter | 31
Calculating Peak-to-Peak Jitter from RMS Jitter• Time-domain histogram of many jitter measurements shows a Gaussian
distribution when jitter is purely random (RJ)• Gaussian distribution mathematics says that 68.3% of all measurements are
within one standard deviation of the mean (+s)• The standard deviation (s) is the RMS jitter (RJ)
RJRMS 2*RJRMS 3*RJRMS 4*RJRMS-RJRMS-3*RJRMS-4*RJRMS -2*RJRMS 0
DesignCon 2013 | Phase Noise and Jitter | 32
Calculating Peak-to-Peak Jitter from RMS Jitter• Phase noise measurement doesn’t provide a histogram, but does provide RMS
jitter value (and therefore standard deviation)• Since RJ has a Gaussian distribution we can calculate peak-to-peak jitter for a
given BER
Most instantaneousjitter measurements
would be in this region
Never reaches 0%probability
Jitter measurements in darkshaded area represent
error-causing jitter values
RJRMS 2*RJRMS 3*RJRMS 4*RJRMS-RJRMS-3*RJRMS-4*RJRMS -2*RJRMS
0 +s +2s +3s +4s-s-3s-4s -2s
DesignCon 2013 | Phase Noise and Jitter | 33
Calculating Peak-to-Peak Jitter from RMS Jitter• Peak-to-peak random jitter can be calculated from RMS jitter by
multiplying by a•
Source: Maxim Application Note AN462
RMSpp RJRJ *a
where is a is derived from: BERerfc
2221 a
a
DesignCon 2013 | Phase Noise and Jitter | 34
a vs. BER• Factors to calculate RJpp from RJRMS based on BER
• Example: RJpp = 9.507*RJRMS for BER = 10-6
9.507*RJRMS (BER=10-6)
11.996*RJRMS (BER=10-9)
14.069*RJRMS (BER=10-12)
6.180*RJRMS(BER=10-3)
DesignCon 2013 | Phase Noise and Jitter | 35
Useful References• “Analysis of Jitter with the R&S FSUP Signal Source Analyzer”
Rohde & Schwarz Application Note 1EF71
• “Converting Between RMS and Peak-to-Peak Jitter at a Specified BER”Maxim Integrated Application Note HFAN-4.0.2
• “Clock Jitter and Measurement”SiTime Application Note SiT-AN10007
• “A Primer on Jitter, Jitter Measurement and Phase-Locked Loops”Silicon Labs Application Note AN687
• “Determining Peak to Peak Frequency Jitter”Pletronics White Paper
DesignCon 2013 | Phase Noise and Jitter | 36