agilent psa high-performance spectrum analyzer series amplitude...

Agilent PSA High-PerformanceSpectrum Analyzer Series

Amplitude Accuracy

Product Note

Accurate measurement of signalpower level is critical in moderncommunications systems.Specifications and margins arevery important, whether one isanalyzing circuits, subsystems orcomplete communications links.Better amplitude accuracy cantranslate into faster design time,more decisive troubleshooting,higher yield, better powerefficiency, tighter specification of the customer’s device, etc.

For a single frequency signal,power meters achieve the bestamplitude accuracy. Whenmultiple signals are involved,spectrum analyzers are veryuseful because of their frequency

selectivity. This frequencyselectivity enables the user toisolate a particular signal andexclude other signals from a measurement.

This product note identifiespossible sources of amplitudeuncertainty in traditionalspectrum analyzers. It thencompares traditional spectrumanalyzer performance with thatof the Agilent PSA Serieshigh-performance spectrumanalyzer (model E4440A).

The PSA Series offers severaltechnical innovations—precisionflatness calibration, an all digitalintermediate frequency (IF)

section, and internal calibrators—that make measurements moreaccurate, faster, and easier. Thisproduct note describes theseinnovations and provides someexample measurements.

The PSA Series is not intended tocompletely replace power meters,although it can replace powermeters in some applicationswhere spectrum analysis is alsoneeded. In addition, the PSASeries is frequency-selective and has the ability to isolate one signal from others and tomeasure that signal with highaccuracy, similar to a powermeter. This makes it an ideal toolfor communications applications.

2

Table of Contents

Sources of Amplitude Uncertainties in Traditional Spectrum Analyzers . . . . . . . . . . . . . 3Sources of relative amplitude uncertainty in a traditional spectrum analyzer . . . . . . . . . . . . . . . . . 3

Sources of absolute amplitude uncertainty in a traditional spectrum analyzer . . . . . . . . . . . . . . . . 5

Repeatability uncertainty in a traditional spectrum analyzer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Temperature drift in a traditional spectrum analyzer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Post-tuning drift in a traditional spectrum analyzer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

PSA Series Improvements in Amplitude Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6Flatness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Digital IF section . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

IF effects on amplitude accuracy in a traditional spectrum analyzer . . . . . . . . . . . . . . . . . . . . . . . . . 7

1. IF gain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2. Log amplifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3. Bringing the measured signal to the reference line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

IF effects on amplitude accuracy in the PSA Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1. IF gain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2. Log–linear fidelity uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3. Dithering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

4. Power bandwidth uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Internal calibrator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Testing and specifying calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Comparing Amplitude Accuracy Specifications of the PSA Series and the Agilent Technologies 8563E . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

Examples of the Measurement Uncertainties in Typical Amplitude Measurements . . 15Example 1. Signal power measurement (absolute measurement) . . . . . . . . . . . . . . . . . . . . . . . . . . 15

Example 2. Relative measurement (using the delta marker) of signals in different bands . . . . . . 16

Example 3. Third-order distortion measurement (relative measurement) . . . . . . . . . . . . . . . . . . . . 17

Example 4. Mismatch measurement error and measurement uncertainty . . . . . . . . . . . . . . . . . . . 18

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

Glossary of Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

Specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

Support, Services, and Assistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

About the Agilent PSA Serieshigh-performance spectrumanalyzer

The Agilent PSA Series is a high-performance radiofrequency (RF) and microwaveline of spectrum analyzers thatoffers an exceptional combinationof dynamic range, accuracy, andmeasurement speed. The PSASeries delivers the highest level of measurement performanceavailable in the AgilentTechnologies spectrum analyzerportfolio. An all-digitalintermediate frequency (IF)section includes fast Fouriertransform (FFT) analysis and adigital implementation of a sweptIF. The digital IF and innovativeanalog design provide muchhigher measurement accuracyand improved dynamic rangecompared to traditional spectrumanalyzers. This performance iscombined with measurementspeed typically 2 to 50 timesfaster than spectrum analyzersusing analog IF filters.

The PSA Series complementsAgilent’s other spectrumanalyzers such as the MXA Series,a family of midrange analyzersthat covers a variety of RF andmicrowave frequency rangeswhile offering a greatcombination of performance,speed, and applications.

3

Sources of Amplitude Uncertainties in Traditional Spectrum Analyzers

The amplitude accuracy ofspectrum analyzers is specified interms of both absolute accuracyand relative accuracy. Absoluteamplitude is the power level of asignal in absolute units such asdBm. Relative amplitude is thedifference between two signallevels, using one signal level as a reference. For example, in a two tone intermodulationmeasurement, we use thefundamental signal as a referenceand measure the third orderintermodulation products usingdecibels relative to the carrierlevel in dBc units. Figure 1, atraditional spectrum analyzerblock diagram, identifies possiblesources of amplitudeuncertainties.

Sources of relative amplitudeuncertainty in a traditionalspectrum analyzer

Because other literaturediscusses spectrum analyzeramplitude accuracy in detail, thisproduct note will just mentioneach uncertainty and give a briefexplanation. Please refer toReferences 1 and 2 on page 19 fordetailed explanations.

Frequency response (flatness)Frequency response, or flatness,is the relative amplitudeuncertainty versus frequencyover a specified frequency range. It is a function of inputattenuator flatness and mixerconversion loss; both of these arefrequency range dependent, andcalibrated with respect to theanalyzer’s calibration frequency.

In the newest generation ofAgilent spectrum analyzers (MXAand PSA Series), this calibrationfrequency has been converged to50 MHz, the same frequency asthe calibrator signal of powermeters.

Flatness in preselected bands(usually above 3 GHz) is alsodependent on the sweep rate (SR) due to errors in keeping the Yttrium Iron Garnet-tuned(YIG-tuned) filter (also known as the YTF) preselector aligned to the tuned frequency. Some errors in tuning the filter arecompensated by modeling themas a delay between the frequencycontrol and the center frequency.YTF delay compensation is notperfect, so the sweep rate (SR =span/sweep time) should not belarger than the YTF delay

compensation allowed. In the PSA Series, the maximum sweeprate is limited by local oscillator(LO) and YTF capabilities to600 MHz/ms for band 0 to band 2 (up to 13.2 GHz); for band 3, SR = 500 MHz/ms; and for band 4,SR = 400 MHz/ms.

Band switching errorA microwave spectrum analyzeruses several frequency bands.These bands use different mixer paths and different LOharmonics. When signals indifferent bands are measured,uncertainties arise when theanalyzer switches from one bandto another. In the PSA E4440A,there are five internal mixingbands: 3 Hz to 3.0 GHz; 2.85 GHzto 6.6 GHz; 6.2 GHz to 13.2 GHz;12.8 GHz to 19.2 GHz; and18.7 GHz to 26.5 GHz.

InputAttenuator

Preselectorof InputFilter Mixer

Resolution BandwidthFilter

LogAmp

EnvelopDetector

VideoFilter

Display

RampGenerator

A/D

LocalOscillator

IF Gain

Figure 1. Traditional spectrum analyzer block diagram

4

Sources of Amplitude Uncertainties in Traditional Spectrum Analyzers (continued)

Scale fidelity (log fidelity or linear fidelity)Scale fidelity, the uncertainty ofthe observed signal level withrespect to a reference level,depends on the linearity of theenvelope detector and linearity(log fidelity) of logarithmicamplifiers.

Reference level accuracyWhen a spectrum analyzer iscalibrated, the reference level is traditionally defined by theamplitude represented at the top line of the graticule on thedisplay. The amplitude of the topgraticule is a function of the input(RF) attenuation level and the IF gain, which are determined by the reference level control.Uncertainty in the amount of IFgain at a particular referencelevel control setting affects theaccuracy of the reference levelamplitude. When a known signalstandard is used to calibrate the reference level, calibratoruncertainty is substituted for reference level controluncertainty. Any subsequentchange in the reference levelcontrol introduces uncertaintyinto the measurement.

RF input attenuator switching errorAttenuator step accuracy, likefrequency response, is a functionof frequency. If the attenuator ischanged between the referenceand measurement positions, itwill introduce uncertainty in themeasurement. For the PSA Series,like many spectrum analyzers, theinput attenuator reference settingis 10 dB.

Resolution bandwidth (RBW)switchingThe available resolutionbandwidth filters of a spectrumanalyzer have uncertaintyassociated with their relativeinsertion loss. As a result, if asignal is measured using differentresolution bandwidths, themeasured amplitudes may differ. Whenever the resolutionbandwidth setting is changedbetween calibration andmeasurement, the accuracy isdegraded and measurementaccuracy is compromised. Thereference resolution bandwidthfor the PSA Series is 30 kHz.

Noise effect on signal amplitudeAt any point in a measurement,the spectrum analyzer ismeasuring the sum of all signalenergy present in the IFpassband. Therefore, themeasured amplitude of the signalis actually signal plus noise.Depending upon the signal levelrelative to the noise level, theinaccuracy in assuming that the“measured” amplitude equals the“signal” amplitude may be smallor large. Please refer to Reference3 on page 19 for a more detailedexplanation.

Impedance mismatchSpectrum analyzers do not haveperfect input impedance, nor domost signal sources have idealoutput impedance. Impedancemismatches produce reflections,which reduce the signal powertransferred to the analyzer and introduce a measurementuncertainty. The generalexpression used to calculatemismatch error limits in dB is:

–20 × log10 ( 1 ± |ρanalyzer × ρsource| )

Where ρ is the reflection coefficient.

As described in the equation, thismismatch error depends on boththe spectrum analyzer inputimpedance and the outputimpedance of the signal source.The use of attenuation at theinput of the analyzer improvesthe input impedance match. Thisis why the reference setting of the spectrum analyzer inputattenuator is 10 dB. For bestamplitude accuracy, use an inputattenuator ≥ 10 dB.

This mismatch error can be quitelarge. See Example 4 on page 18,for an illustration of this error.

5

Sources of Amplitude Uncertainties in Traditional Spectrum Analyzers (continued)

Sources of absolute amplitudeuncertainty in a traditionalspectrum analyzer

Calibrator accuracy gives aspectrum analyzer its absoluteamplitude reference. Forconvenience, calibrators aretypically built into today’sspectrum analyzers and provide a signal with specified amplitudeat a convenient frequency. Therelative accuracy of the analyzeris used to translate the absolutecalibration to other frequenciesand amplitudes. In the PSASeries, the internal calibrator isset at a frequency of 50 MHz anda level of –25 dBm.

Repeatability uncertainty in atraditional spectrum analyzer

Mechanical switches can cause a lack of repeatability. In many spectrum analyzers, thecalibrator switch and inputattenuator use mechanical relays.(Some degree of amplituderepeatability uncertainty is anunavoidable consequence ofmechanical switches.)

Temperature drift in a traditionalspectrum analyzer

Temperature drift is caused bychanges in IF amplifier gain with temperature. Traditionally,amplitude accuracy specificationsmust be relaxed over a widetemperature range to account forthis drift.

Post-tuning drift in a traditionalspectrum analyzer

Self-heating of the YIG-basedhighband preselector filter causesa frequency shift of the filter,which leads to changes in thedisplayed signal amplitude.

6

PSA Series Improvements in Amplitude Accuracy

In the PSA Series, the design andproduction processes have beenfundamentally changed in waysthat minimize most of themeasurement uncertaintiesdescribed previously.

The major changes include: aprecision flatness calibrationprocess, an internal calibrator,and an all digital IF.

Flatness

In a spectrum analyzer, thefrequency response of the inputsection components, such asinput attenuators and the firstmixer, creates amplitudevariations with frequency.

In the PSA Series design, aprecision flatness calibrationmethod corrects frequencyresponse errors.

0 500 1000 1500 2000 2500 3000

0.15

0.1

0.05

0

-0.05

-0.1

-0.15

Am

plitu

de E

rror

(dB

)

Frequency (MHz)

Figure 2. PSA E4440A flatness cal data typical frequency response

This method greatly improves the“absolute flatness” accuracy, thatis, the flatness relative to thefrequency of the absoluteaccuracy reference—in this case,the 50 MHz calibrator signal.

With some spectrum analyzersthe flatness of some bandsrelative to the calibrator is notparticularly well controlled. Inthese analyzers, the relative gainuncertainty between signals intwo bands can be specified as thesum of the relative flatness ineach band plus the band-switchinguncertainty. Such analyzers maybe specified more tightly this waythan as the sum of the flatnesserrors relative to the calibrator ofthe signals for both bands. In thePSA Series, the flatness relativeto the calibrator is very wellcontrolled in all bands. As aconsequence, no specification

improvement is possible using the band switching uncertaintyconcept and there is no suchspecification.

Figure 2 shows a typicalfrequency response of the PSASeries E4440A for frequenciesbelow 3 GHz. The specificationlimits of flatness for frequenciesbelow 3 GHz are ±0.40 dB. Notehow well controlled the frequencyresponse errors are, and comparethe typical response to thespecification limits.

Digital IF section

A comparison of a traditionalanalog IF section with the alldigital IF section in the PSASeries highlights the advantagesof the technology used in the PSA Series.

7

PSA Series Improvements in Amplitude Accuracy (continued)

IF GainMixer

LocalOscillator

Resolution BandwidthFilter

21.4 MHz

Pout = Log (Pin)

Pin

Pout

Figure 3. Block diagram of a traditional spectrum analyzer analogIF section

IF effects on amplitude accuracyin a traditional spectrum analyzer

It is instructive to understandwhy moving the signal under testto the reference level is necessaryin a traditional spectrum analyzerto achieve better amplitudeaccuracy. This is illustrated in theblock diagram in Figure 3.

In Figure 3, we can see that thetraditional spectrum analyzeranalog IF section is composed ofan analog IF amplifier followed bya log amplifier.

As explained in the “Referencelevel accuracy” section on page 4,the top line of the graticule on thedisplay is defined as the reference

level. A known calibration signalcan be used to calibrate thisreference line. Thus the calibratordetermines the absolute accuracyfor the top line of the graticule ata particular analog IF gain settingand input signal power level tothe log amplifier. The referencelevel is the most accuratelyindicated level because it is thecalibrated level.

A traditional spectrum analyzerhas only one calibrator level andone reference input attenuatorsetting, typically 10 dB.Therefore, only one referencelevel can be optimally calibrated,unless additional externalcalibration levels are used.

1. IF gainThe reference level controldetermines the IF gain. When the analyzer is calibrated, thereference level is set to thecalibrator level, and theuncertainty of the IF gain iscalibrated out. However, anychanges to the reference levelfrom that used for calibration will affect the IF gain setting and consequently introduceuncertainty to the signalamplitude measurement.

8

-90 -80 -70 -60 -50 -40 -30 -20 -10 0

0.25

0.2

0.15

0.1

0.05

0

-0.05

-0.1

-0.15

-0.2

-0.25

Line

arity

(dB

)

Input Power re:RL (dB)

Figure 4. Typical log fidelity error in a traditional spectrum analyzer


2. Log amplifierIn order to display a widedynamic range signal, spectrumanalyzers use a logarithmicvertical scale in the display. Asshown in Figure 1 on page 3,there is a log amplifier in thetraditional specification analyzerblock diagram. The log amplifieris used to display the signalpower level on the vertical axis of the display.

Most log amps are realized by physical components in atraditional spectrum analyzer.These components haveperformance limitations anduncertainty, resulting in logfidelity error such as that shownin Figure 4.

3. Bringing the measured signal to thereference lineBecause the calibrator operatesat a single frequency and fixedsignal power level, the referencelevel setting is calibrated for justone particular setting of the IFgain. For most signal amplitudemeasurements, the signal under test will be at a differentfrequency as well as a differentamplitude.

To measure a signal with a powerlevel different from the calibrator,we can either change the IF gainto a value different from thecalibrated gain, or we can readthe signal level at a drive level tothe log amplifier that is not thereference level. For some

spectrum analyzers, the logfidelity has much largeruncertainty than IF gainuncertainty. When thesespectrum analyzers are used, thelowest uncertainty is achieved ifthe IF gain is changed in order tobring the unknown signal level tothe reference level, thus driving aconstant input signal level to thelog amplifier. For other spectrumanalyzers, where the IF gainuncertainty is a largeruncertainty than the log fidelity,the lowest uncertainty is achievedby keeping the IF gain (and thusthe reference level) fixed at thecalibrated setting, and allowingthe drive level to the log amplifierto change.

9


IF effects on amplitude accuracyin the PSA Series

In the PSA Series, an all digital IFsection is used, which eliminatesor minimizes many traditionalamplitude uncertainties such as log fidelity error, IF gainuncertainty, and resolutionbandwidth (RBW) switchingerror. The block diagram in Figure 5 shows theimplementation of digital IFtechnology in the PSA Series.

In Figure 5, we can see that the IF section of the PSA Series iscomposed of an analog to digitalconverter (ADC) and digitalsignal processing (DSP). The DSP

is implemented in an applicationspecific integrated circuit(ASIC) chip, which includesdigital RBW filtering, a digital log amplifier, a digital videobandwidth (VBW) filter, anddigital display detectors (such aspeak, normal, sample, etc).

1. IF gainNone of the stages shown inFigure 5 have gain that changeswith the reference level. All of the operations are performeddigitally. Therefore, there is no IFgain uncertainty. A major benefitto the user is the ability to changethe reference level for the bestdisplay without affecting themeasurement accuracy.

Prefilter Mixer

Dither

ADC

LocalOscillator

7.5 MHz

All Digital IF ASIC

Digital RBWDigital Log AmpDigital VBWDigital Detector

Σ

Figure 5. Block diagram of the PSA Series IF section

10


2. Log–linear fidelity uncertaintyThe specifications of the PSASeries for amplitude accuracyinclude display scale fidelity. Thespecification indicates log-linearfidelity relative to –35 dBm at the input mixer. However, asmentioned previously, there is nolog amplifier in the PSA Series,and so the source and magnitudeof this error require explanation.

In the PSA Series, the logfunction is done mathematicallyand traditional log fidelityuncertainty does not exist. Butthere are other sources of errorsin the PSA Series that causelog-linear fidelity uncertainty.This uncertainty comes from two sources: RF compression

-100 -90 -80 -70 -60 -50 -40 -30 -20 -10 0

0.25 0.2

0.150.1

0.05 0

-0.05 -0.1

-0.15-0.20 -0.25

Line

arity

(dB

)

Input Mixer Level (dB from reference level)

Figure 6. Input power versus typical log fidelity error with dither

(especially for input signal levelsabove –20 dBm) and ADC rangegain alignment limitations. The log fidelity is specified at±0.07 dB for any signal level up to–20 dBm at the input mixer of theanalyzer, and ±0.13 dB for signallevels up to –10 dBm at the inputmixer. Figure 6 shows typical logfidelity error in the PSA Series.

In Figure 6 we see that for signallevels at the input mixer above–20 dBm, there are some logfidelity uncertainties well withinthe specified ±0.13 dB. For signallevels at the input mixer in therange of –80 dBm to –20 dBm, logfidelity error is inside the rangeof ±0.07 dB. For signal levels at

the input mixer below –80 dBm(signal levels close to thespectrum analyzer noise floor) the log fidelity erroruncertainties are mostly due tonoise. Thus these are not logfidelity errors; there is noapparent limitation to low-levellog fidelity except system noise.

Traditional spectrum analyzershave similar uncertainty sources,but the log amplifier errordominates the total log fidelityuncertainty. Log fidelity error caused by first mixercompression is too small to besignificant in the log fidelityuncertainty of traditionalspectrum analyzers.

11


3. DitheringAll ADCs have errors in linearity,and even an ADC with perfectlinearity would have quantizationerrors. When the signal level issmall, the quantization errors can dominate the amplitudeuncertainty. Figure 7 shows idealADC transfer function (input vs.output relationship) with andwithout noise added at the input.

Figure 7 also shows that if aninput sine wave is applied to this ADC with less than 1 leastsignificant bit (LSB) peak-to-peaklevel, the output is a constantzero, giving an amplitude error of an infinite number of decibels.Input signals from 1 to 2 LSBspeak-to-peak produce outputlevels that vary by much less than

the 6 dB of input range. Theseexamples demonstrate whyquantization errors can result invery large linearity errors. In aspectrum analyzer, these errorsare part of “log or linear fidelity”error terms, the relative accuracyof the detection of different signal levels.

However, if the input signal hasnoise added, the average of thetransfer function is smoothed, as shown in Figure 7, becausesignal levels near bit transitionsoccasionally yield results acrossthe bit transition. Analytically,the average transfer function isthe convolution of the noiselesstransfer function with theprobability density function(PDF) of the noise. If the added

ADC Output

ADC Input

Ideal ADC

Average for ADCwith Small Dither

Average for ADCwith Large Dither

Figure 7. Ideal ADC: Input-output relationship

noise is much larger than a most significant bit (MSB), thelinearity can be excellent.

This added noise is called dither.In the PSA Series, the dither isapplied at frequencies outside therange of IF analysis so it is notnormally visible on the screen.Used in this way, it acts toconvert the quantizationdistortion (errors that would becorrelated with the signal) intouncorrelated errors, which wouldthen act as quantization noise.But the dither used is so largethat the maximum input signalmust be reduced, compromisingthe S/N ratio. The displayedaverage noise level gets worse byabout 2.5 dB.

12


In the PSA Series, the user maychoose from the Mode Setup (frontpanel key) for dither on, ditheroff, or automatic selection ofdither. When the dither is off, thelow-level linearity is poorer, butstill nominally within about±0.2 dB. When the dither is on,the increased noise floor of theanalyzer can cause measurementerrors. But even with a noisefloor increase as large as 2.5 dB,versus a detection linearityimprovement from only ±0.2 dB to±0.07 dB, the total measurementerror, on average, is lower withdither on, unless the signal isactually lower than the noise floor.Figure 6 on page 10 shows the logfidelity error with dithering onand Figure 8 shows the logfidelity error with dithering off.

A comparison of Figure 6 andFigure 8 shows that, at low inputsignal levels, log fidelity errorimproves substantially whendithering is used.

In the PSA Series, it is generallyrecommended that dither be usedwhen the measured signal has aS/N ≥ 10 dB. When the S/N isunder 10 dB, the degradations to accuracy of any singlemeasurement (in other words,without averaging) that comefrom a higher noise floor areworse than the linearity problemssolved by adding dither, so ditheris best turned off.

4. Power bandwidth uncertaintyIn the measurement of noisedensity or the measurement of thetotal power of a noise-like signal,the accuracy of the bandwidth ofthe RBW filters has significanteffects on amplitude accuracy.

In the PSA Series, the analogprefilter is set to about 2.5 timesthe RBW width. It will have someuncertainty in bandwidth, gain,and center frequency withdifferent RBW settings. Figure 5on page 9 shows that there is nofurther analog RBW filtering andlog amplification in the IF sectionof the PSA Series.

Although the analog prefilteraffects the shape of the total RBWfiltering, most of the filtering isdone digitally in an ASIC in thedigital IF section. The digitalfiltering is very repeatable, but it cannot be set with perfectresolution. Typical RBWs in thePSA Series have a bandwidthwithin 1.2 percent of the selectedbandwidth.

In any case, the resolution of the digital filtering is accuratelyknown in the PSA Series.Therefore measurements of noise-like signals can becompensated for the digital filterbandwidth, leaving only theuncertainty of the analogprefilter as a bandwidth error.This prefilter is well characterizedand has only a small effect on thetotal bandwidth because itsbandwidth is significantly widerthan the selected RBW (nominally2.5 times wider).

As a result, the accuracy ofmeasurements that depend on the bandwidth of the RBWfilter are as accurate as thosemade with a filter with only±1.0 percent bandwidthuncertainty, or an equivalenterror of ±0.044 dB compared withthose made using an ideal filter.Previous generations of spectrumanalyzers had RBW widthaccuracies of 10 to 20 percent.

-100 -90 -80 -70 -60 -50 -40 -30 -20 -10 -0

0.25 0.2

0.15 0.1

0.05 0

-0.05 -0.10 -0.15

-0.2-0.25

Line

arity

(dB

)

Input Mixer Level (dBm)

Figure 8. Typical log fidelity error with dithering off

13


Internal calibrator

The calibrator of a spectrumanalyzer provides a convenientabsolute power level. Theabsolute accuracy of anyspectrum analyzer measurementmay be determined by adding the calibrator error to any otheruncertainties that arise fromsettings that deviate from thecalibration setting.

The calibrator in the PSA Seriesis different from previousgenerations in three ways: level,frequency, and cabling.

The calibrator level is –25.0 dBm.This level is low enough(–35.0 dBm at the input mixer)that it can serve as the referencepoint for scale fidelityspecifications without significanterrors due to gain compression inthe analyzer circuits.

The calibrator frequency is50 MHz. This calibrationfrequency is now common in thelatest generation of analyzers.The 50 MHz frequency helpstraceability because most

production facilities, includingthose of Agilent Technologies,trace their amplitude accuracy to the National Institute ofStandards and Technology (NIST)through power meters, and powermeters use 50 MHz for theircalibrators. If the referencefrequency of the power meter is the same as the analyzerreference frequency, thefrequency response flatnessaccuracy term in the power metererror budget does not apply.Therefore traceability has a lowererror at 50 MHz than at any otherfrequency.

The PSA Series has an internalcalibrator, so no external cable isinvolved. Calibration is thereforeas convenient as pressing keys.

Testing and specifying calibration

Previous generation analyzers,with externally cabled amplitudecalibration signals, specifiedseparately the absolute accuracyof the calibrator. The totalabsolute accuracy was the sum ofthe errors of the calibrator and

the errors due to differences inanalyzer settings between thecalibration and measurementsettings.

The PSA Series specifies many amplitude uncertaintiesseparately, but, for absoluteamplitude accuracy, it specifiesthe sum of many errors. Over abroad range of RBWs and signallevels, for any reference level anddisplay scale, the PSA Seriesspecifies a single parameter:absolute amplitude uncertainty at 50 MHz. This specification of±0.24 dB is verified in the factoryby testing a set of 43 conditionsthat should exercise the worsterrors.

This specification substitutes for the following parameters inolder analyzers: calibrator error,log-linear fidelity uncertainty, IF gain uncertainty, and RBWswitching error. In a typicalmeasurement, the user needs toonly add flatness and, possibly,attenuator switching uncertaintyto get the total absolute accuracy.

14

Comparing Amplitude Accuracy Specifications of the PSA Series and theAgilent Technologies 8563E

This section compares some of themajor differences in amplitudeaccuracy specifications betweenthe PSA E4440A and the Agilent8563E. For detailed specificationsfor the PSA E4440A, please referto Reference 4 on page 19 orwww.agilent.com/find/psa.

From this comparison, we canconclude that the PSA Series hasmuch less amplitude uncertaintythan the 8563E spectrumanalyzer. This is also true whencomparing the PSA Series withother spectrum analyzers in themarket. Some of the uncertaintiesin traditional spectrum analyzerscan be limited or minimized, suchas by measuring the signal levelwith respect to the calibratedreference level and CRT scalewithout changing the referencelevel or attenuator controls. Suchprocedures can eliminate some of the uncertainties, but theremaining uncertainties are still much larger than the totaluncertainty in the PSA Series.Because the PSA Series does nothave many traditional amplitudeuncertainties, it provides greateruser flexibility in accurateamplitude measurements.

Table 2.

Specifications PSA E4440A 8563E

Absolute frequency response 0 to 3 GHz < ±0.4 dB < ±1.8 dB

Log fidelity–20 dBm or lower < ±0.07 dB < ±0.85 dBRange of log fidelity unlimited 100 dB from

reference level reference level

IF gain uncertainty none < ±1.0 dB

RBW switching (all but the widest) < ±0.03 dB < ±0.5 dB

Calibrator Not specified < ±0.3 dB

Calibrator + log + IF gain + RBW switching < ±0.24 dB Not specified,adds to 2.65 dB

15

Examples of the Measurement Uncertainties in Typical Amplitude Measurements

This section examines sometypical amplitude measurementsusing both the PSA Series and the8563E, with some suggestions onhow to use the PSA Series toobtain the best amplitudeaccurate measurements.

We will list which uncertaintiesapply and calculate the totalerror and root sum of squares(RSS) error in Examples 1-4.

Example 1.

Signal power measurement (absolute measurement)

Measurement Measure a 900 MHz CW signal at –5 dBm, using the 10 kHz RBW.

Important tip In the PSA Series, the error caused by switching the attenuator from the reference setting of 10 dB is smaller than the log fidelity errors due to having an input signal to the first mixer above –20 dBm.

On the PSA Series it is not necessary to move the signal to the reference line to improve accuracy. As long as the input attenuator is set so that the mixer level is at or below –10 dBm, the reference level can be set anywhere within the instrument’s allowed range.

PSA E4440A Agilent 8563E

Factor Specified uncertainty Applicable uncertainty Specified uncertainty Applicable uncertainty (±dB) (±dB) (±dB) (±dB)

Calibrator 0.3 0.3

Reference level 0.27 0.27 1.0 0.0 (if calibrated/used with 0 dBm reference level)

Scale fidelity 0.5 0.5

RBW switching 0.0 (if calibrated in10 kHz RBW)

Frequency response1 0.38 0.38 1.5 1.5

Total uncertainty 0.67 2.3

Total uncertainty in RSS2 0.48 1.61

1 With 10 dB input attenuation.2 Determining the total uncertainty for a measurement by adding all error contributors is a very conservative approach. A more realistic method of combining uncertainties is the

root-sum-of-the-squares (RSS) method. In an amplitude measurement, the RSS uncertainty is based on the fact that most of the errors are independent of each other. Becausethey are independent, it is reasonable to combine the individual uncertainties in an RSS manner.

Finding the RSS uncertainty requires that each individual uncertainty is expressed in the same units. Typically uncertainties are expressed in terms of decibels.

RSS = [ e12 + e2

2 + e32 + …]1/2

Where RSS, e1 , e2 , e3… in dB.

16

Examples of the Measurement Uncertainties in Typical Amplitude Measurements (continued)

Example 2.

Relative measurement (using the delta marker) of signals in different bands

Measurement 10 GHz fundamental signal with its second harmonic at 20 GHz.

Important tip In general, it is best to use the same input attenuator setting and resolution bandwidth setting for both signals. In this way, the uncertainties associated with input attenuation switching uncertainty and resolution bandwidth switching uncertainty will not affect the measurement.



Frequency response 1.0 dB (absolute) 1.0 dB 2.5 dB (relative)1 2.5 dB(at 10 GHz)

Frequency response 2.3 dB (absolute) 2.3 dB 3.0 dB (relative)1 3.0 dB(at 20 GHz)

Band switching N/A 0 1.0 dB1 1.0 dB

Log fidelity 0.072 0.07 0.85 dB max 0.85 dB max(over 90 dB range)

Total uncertainty 3.37 6.55 dB

Total uncertainty in RSS 2.51 4.12 dB

1 The frequency responses relative to the absolute calibrator could be used, in which case band switching would not apply.2 Set the attenuator so that the signal level is below –20 dBm at the mixer.

17


Example 3.

Third-order distortion measurement (relative measurement)

Measurement Measure two 2 GHz signals separated by 50 kHz with third-order distortion at –80 dBc.

Important tip For best distortion products from the PSA Series, the input signal level to the first mixer should be: 1/3 x (2 x TOI + DANL). Please refer to PSA Series technical specifications.

Figure 2 on page 6 shows that, for signals that are very close to the same frequency, frequency response can be ignored.



Log fidelity 0.071 0.07 0.85 dB max 0.85 dB max(over 90 dB range)

Total uncertainty 0.07 0.85 dB

Total uncertainty in RSS 0.07 0.85 dB

1 Set the attenuator so that the signal level is below –20 dBm at the mixer.

18


Example 4.

Mismatch measurement error and measurement uncertainty

Measurement Measure a signal using the PSA E4440A at different input attenuation settings.

Important tip For all spectrum analyzers, some input attenuation is needed (to optimize impedance match) for the most accurate amplitude measurements.

Mismatch errors are due to the impedance difference in source output and load input impedances. In a spectrum analyzer measurement, imperfections in the output impedance of the device under test (DUT) and the input impedance of the spectrum analyzer can both cause a mismatch error.

The complete power transfer equation is:

PL/PS = (1 -ρS2)(1 -ρL

2) / (1 ±ρLρS)2 <1>

10 log (PL/PS) = 10 log (1 -ρS^2) + 10 log (1 -ρL^2) – 20 log (1 ±ρSρL) <2>

In equation <2>, the first two terms determine the expected loss in power due to the reflection coefficient. The third term is the mismatch uncertainty. The actual power transfer can fall anywhere in between the two extremes represented by the + and – cases in the third term.

The following two tables list the calculated result for a DUT with a voltage standing wave ratio (VSWR) 1.4:1, using the PSA Series to measure this device under different input attenuator settings.

VSWR ρ VSWR ρ< 10 dB input attenuation ≥ 10 dB input attenuation

DUT output port 1.4:1 0.167 1.4:1 0.167

PSA 50 MHz – 3 GHz 2.3:1 0.394 1.2:1 0.091(from spec.)

Input attenuator setting < 10 dB input attenuation ≥ 10 dB input attenuation

Measurement error (dB) –0.855 –0.159

Measurement uncertainty –0.591/+0.553 –0.133/+0.131(±dB)

Total error range (dB) –0.264/–1.408 –0.026/–0.29

Mismatch error and uncertainty apply to all instruments used for amplitude measurement, such as power meters and spectrum analyzers.

19

Summary

New technology in the PSA Serieshigh-performance spectrumanalyzer has dramaticallyimproved the amplitude accuracyof many signal measurements.Most traditional amplitude

References

[1] Spectrum Analysis Basics, Application Note 150, literaturenumber 5952-0292

[2] Optimizing Spectrum Analyzer Amplitude Accuracy,Application Note 1316, literature number 5968-3659E

[3] Spectrum Analyzer Measurements and Noise, Application Note1303, literature number 5966-4008E

[4] PSA E4440A Specifications. See page 21.

accuracy errors have beeneliminated or reduced, whileease-of-use and measurementflexibility have both beenimproved.

20

Glossary of Terms

ACP Adjacent Channel Power

ADC Analog to Digital Converter

ASIC Application Specific Integrated Circuit

CW Center Frequency

DANL Displayed Average Noise Level

DSP Digital Signal Process

dBc Decibels Relative to the Carrier Level

DUT Device Under Test

FFT Fast Fourier Transform

IF Intermediate Frequency

LO Local Oscillator

LSB Least Significant Bit

MSB Most Significant Bit

NIST National Institute of Standards and Technology

P1dB One dB Compression Point

RBW Resolution Bandwidth

RF Radio Frequency

RSS Root Sum of Squares

SR Sweep Rate

TOI Third Order Intercept

VBW Video Bandwidth

VSWR Voltage Standing Wave Ratio

YIG Yttrium Iron Garnet

YTF YIG Tuned Filter

21

Product Web site

For the most up-to-date andcomplete application and productinformation, please visit ourproduct Web site at: www.agilent.com/find/psa

Related Literature

Publication Title Publication Type Publication Number

Agilent PSA Series High-Performance Brochure 5980-1283ESpectrum Analyzers

Agilent PSA Series Spectrum Analyzers Data Sheet 5980-1284E

Optimizing Dynamic Range for Product Note 5980-3079ENDistortion Measurements

Measurement Innovations and Benefits Product Note 5980-3082EN

Select the Right PSA Spectrum Analyzer for Selection Guide 5968-3413EYour Needs

Self-Guided Demonstration Product Note 5988-0735EN

Swept and FFT Analysis Product Note 5980-3081EN

Specifications

PSA E4440A Specifications

Frequency coverage 3 Hz to 50 GHz

DANL –165 dBm (10 MHz to 3 GHz)

Absolute accuracy ±0.24 dB (50 MHz)

Frequency response ±0.38 dB (3 Hz to 3 GHz)

Display scale fidelity ±0.07 dB total (below –20 dBm)

TOI (mixer level –30 dBm) +16 dBm (400 MHz to 2 GHz)+17 dBm (2–2.7 GHz)+16 dBm (2.7–3 GHz)

Noise sidebands (10 kHz offset) –116 dBc/Hz (CF = 1 GHz)

1 dB gain compression +3 dBm (200 MHz to 6.6 GHz)

Attenuator 0–70 dB in 2 dB steps

Warranty

The E4440A is supplied with a one year warranty.

Agilent Technologies’ Test and Measurement Support, Services, and AssistanceAgilent Technologies aims to maximize the value you receive, while minimizing your risk andproblems. We strive to ensure that you get the test and measurement capabilities you paidfor and obtain the support you need. Our extensive support resources and services can helpyou choose the right Agilent products for your applications and apply them successfully.Every instrument and system we sell has a global warranty. Two concepts underlie Agilent’soverall support policy: “Our Promise” and “Your Advantage.”

Our PromiseOur Promise means your Agilent test and measurement equipment will meet its advertisedperformance and functionality. When you are choosing new equipment, we will help youwith product information, including realistic performance specifications and practicalrecommendations from experienced test engineers. When you receive your new Agilentequipment, we can help verify that it works properly and help with initial product operation.

Your AdvantageYour Advantage means that Agilent offers a wide range of additional expert test andmeasurement services, which you can purchase according to your unique technical andbusiness needs. Solve problems efficiently and gain a competitive edge by contracting withus for calibration, extra-cost upgrades, out-of-warranty repairs, and on-site education andtraining, as well as design, system integration, project management, and other professionalengineering services. Experienced Agilent engineers and technicians worldwide can help you maximize your productivity, optimize the return on investment of your Agilentinstruments and systems, and obtain dependable measurement accuracy for the life ofthose products.

For more information on Agilent Technologies’products, applications or services, pleasecontact your local Agilent office. The completelist is available at:

www.agilent.com/find/contactus

Phone or FaxUnited States:(tel) 800 829 4444(fax) 800 829 4433

Canada:(tel) 877 894 4414(fax) 800 746 4866

China:(tel) 800 810 0189(fax) 800 820 2816

Europe:(tel) 31 20 547 2111

Japan:(tel) (81) 426 56 7832(fax) (81) 426 56 7840

Korea:(tel) (080) 769 0800(fax) (080) 769 0900

Latin America:(tel) (305) 269 7500

Taiwan:(tel) 0800 047 866(fax) 0800 286 331

Other Asia Pacific Countries:(tel) (65) 6375 8100(fax) (65) 6755 0042Email: [email protected] revised: 05/27/05

Product specifications and descriptions in thisdocument subject to change without notice.

© Agilent Technologies, Inc. 2006Printed in USA, August 31, 20065980-3080EN

Agilent Email Updates

www.agilent.com/find/emailupdatesGet the latest information on the products and applications you select.

www.agilent.com

Agilent Direct

www.agilent.com/find/agilentdirectQuickly choose and use your test equipment solutions with confidence.

Agilent Open

www.agilent.com/find/openAgilent Open simplifies the process of connecting and programming test systems to help engineers design, validate and manufacture electronic products. Agilent offers openconnectivity for a broad range of system-ready instruments, open industry software,PC-standard I/O and global support, which are combined to more easily integrate testsystem development.

agilent psa high-performance spectrum analyzer series amplitude...

Documents