Improving Network AnalyzerMeasurements of Frequency-translating DevicesApplication Note 1287-7

LO-RFRF

LO

LO+RF

IFIF

RFLO

IF

LO-RFRF

LO

LO+RF

IFIF

RFLO

IF

LO-RFRF

LO

LO+RF

IFIF

RFLO

IF

LO-RFRF

LO

LO+RF

IFIF

RFLO

IF

2

Table of ContentsPage

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Network Analyzer Mixer Measurement Configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3Scalar Network Analyzer Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4Vector Network Analyzer in Frequency Offset Mode Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5The Upconversion/Downconversion Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Conversion Loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6Definition and Importance of Conversion Loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6Measurement Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Mismatch Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7Considerations Unique to the Scalar Network Analyzer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Importance of Proper Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9Frequency Response Error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Considerations Unique to the Vector Network Analyzer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10Importance of Proper Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10Sampling Architecture and Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10R-Channel Phase-Locking Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12LO Accuracy and Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12Power-Meter Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Accuracy Comparison of the 8757D and a Vector Network Analyzer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16Fixed IF Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Relative Phase Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18Relative Phase and Magnitude Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18Group Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

Important Parameters when Specifying Group Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19Absolute Group Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20Upconversion/downconversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20Modulation delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20Time domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21Measuring delay linearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

Reflection Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24Isolation Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

Feedthrough Measurement of Converters and Tuners . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

Absolute Group Delay — A More Accurate, Lower Ripple Technique . . . . . . . . . . . . . . . . . . . . 27Measurement Configuration Using the Mixer Pair . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

Labeling Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28Proper Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29LO Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29System Calibration and Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

Calibrating the Test System with the Calibration Mixer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29Calibration Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29Calibration Error Terms and Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30Test Procedure for Calibrating the Test System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

First-Order Error Correction: Frequency Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31Second-Order Error Correction: Frequency Response and Input Match . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32Third-Order Error Correction: Frequency Response, Input and Output Match . . . . . . . . . . . . . . . . . . . . . . . . 33

Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Calibration mixer attitudes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

Appendix B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 Program for Fixed IF Measurements with One External LO Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

Appendix C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 Uncertainty in Mixer Group Delay Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

Appendix D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Related Application and Product Notes/Other Suggested Reading/Third Party Companies . . . . . . . . . . . . . . . . . 39

3

Introduction

Frequency-translation devices (FTDs)such as mixers, converters, and tunersare critical components in most RFand microwave communication systems. As communication systemsadopt more advanced types of modulation, FTD designs are increas-ingly complex, tests are more stringent with tighter specifications,and the need to reduce costs is moreimportant than ever.

The measurement trade-offs for frequency-translating devices varywidely among different industries.Measurement accuracy, speed, costand ease of setup are among the considerations for determining thebest test equipment. This applicationnote explores current test equipmentsolutions and techniques that can beused to accurately characterize andtest frequency-translating devices.Frequency-translating devices presentunique measurement challenges sincetheir input and output frequencies differ. These require different measurement techniques than thoseused for a linear device such as a filter. This note covers linear frequecy-translation measurements,such as magnitude, relative phase,reflection and isolation. Correspondingaccuracy issues are also discussed.

To get the most from this note, youshould have a basic under-standing offrequency translation terminology,such as “RF port,” “IF port” and “LOport.” Understanding of fundamentalRF and network analyzer terms suchas S-parameters, VSWR, group delay,match, port, full two-port calibration,and test set is also expected. For abetter understanding of such terms, alist of reference material appear in theAppendix section.

Network analyzer mixer measurement configurations

Network analyzers used for testingfrequency-translation devices includescalar network analyzers, vector network analyzers with frequency offset capability, and vector networkanalyzers using an upconversion/downconversion configuration. Each solution has its own advantages anddisadvantages. This section provides asynopsis of the three configurations soyou can quickly evaluate which is thebest fit for your measurement needs.Detailed information about each solution is discussed in later sections.

4

Scalar network analyzer configuration

The most economical instrument forFTD tests is a scalar network analyzer.A scalar network analyzer uses diodedetectors that can detect a very wideband of frequencies. This capabilityenables a scalar network analyzerto detect signals when the receiver frequency is different from the sourcefrequency. Magnitude-only measure-ments such as conversion loss,absolute output power, return loss andisolation can be made, as well as nonlinear magnitude measurementssuch as gain compression. Group delayinformation is available in some scalarnetwork analyzers using an AM-delaytechnique, which employs amplitudemodulation.

AM-delay measurements are less accurate than group delay measure-ments obtained with a vector networkanalyzer. AM delay typically has anuncertainty of around 10 to 20 ns,whereas group delay with a vectornetwork analyzer has an uncertaintyas good as 150 ps. Advantages of thescalar solution include low cost andgood magnitude accuracy. As shown inFigure 1, fully integrated scalar network analyzers such as the 8711Cor 8713C provide economical RF measurements up to 3 GHz, andinclude AM delay capability. The8757D scalar network analyzer, shownin Figure 2, measures up to 110 GHz,and provides very good absolutepower measurements, particularlywhen installed with an internal powercalibrator and used with precisiondetectors. In certain cases, such asmeasuring FTDs with an internal filter,the 8757D with internal power calibrator and precision detector cantypically make more accurate magni-tude measurements than a vector network analyzer.

8711C RF network analyzer

External LO source Lowpass filter

10 dB

10 dB

External LO source

Directional bridge

Precision detector

Sweeper8757D

scalar network analyzer

Lowpass filter

Figure 2. 8757D scalar network analyzer configuration

Figure 1. 8711C scalar network analyzer configuration

5

Vector network analyzer in frequencyoffset mode configuration

A more versatile solution for FTD testis a vector network analyzer. A vectornetwork analyzer uses a tuned-receiv-er narrowband detector, which allowsmeasurements of both magnitude andrelative phase. The vector networkanalyzer’s frequency offset mode off-sets the analyzer’s receiver from it’ssource by a given LO frequency, andmakes frequency-translation measure-ments possible.

There are two common vector net-work analyzer configurations for FTDmeasurements. The simplest configu-ration is shown in Figure 3, and ispractical for testing upconverters anddownconverters. This configurationallows magnitude-only measurementswith a limited dynamic range. Forexample, if you are interested in themagnitude response of the FTD’s passband, the 8753E vector networkanalyzer has 35 dB of dynamic rangein the R channel and provides a quickand easy solution

To also measure the FTD’s relativephase and out-of-band response,Figure 4 illustrates a high-dynamicrange configuration. An alter-nativehigh-dynamic range configuration canbe achieved by splitting the analyzer’sRF output power between the deviceunder test (DUT) and the referencemixer (This configuration is similar tothe one shown in Figure 24). In bothconfigurations the vector network analyzer has around 100 dB of dynam-ic range. A signal into the reference Rchannel is always necessary for properphase-locking of the vector networkanalyzer. In addition, the R channelprovides a reference for ratioed mea-surements such as relative phase ormagnitude and phase tracking. Vectornetwork analyzers such as the 8720Dseries and the 8753E have frequencyoffset capability to 40 GHz and 6 GHz,respectively.

The upconversion/downconversion technique

A vector network analyzer in normaloperating mode can also be configuredfor frequency-translation measure-ments. This configuration has twomain advantages. First, the instrumentcan be used to measure a FTD’s magnitude and relative phase responsewithout the need for frequency offset.

As shown in Figure 5, two mixers areused to upconvert and downconvertthe signals, ensuring the same frequencies at the network analyzer’ssource and receiver ports. Second, thisconfiguration provides a potentiallymore accurate method for measuringabsolute group delay. You can simplymeasure two mixers and halve theresponse, accepting the resultinguncertainty.

Vector Network Analyzer

Lowpass filter

10 dB10 dB

RF in

Start: 900 MHzStop: 650 MHz

Start: 100 MHzStop: 350 MHz

Fixed LO: 1 GHzLO power: 13 dBm

FREQ OFFSON off

LOMENU

DOWNCONVERTER

UP CONVERTER

RF > LO

RF < LO

VIEW MEASURE

RETURN

Figure 3. Vector network analyzer in frequency offset mode

Vector Network Analyzer

Lowpass filter Reference mixer

CH1 CONV MEAS log MAG 10 dB/ REF 10 dB

START 640.000 000 MHz STOP 660.000 000 MHz

RF out

RF in

10 dB10 dB

Power splitter

External LO source

Figure 4. Vector network analyzer, high dynamic range configuration

External LO source

Vector network analyzer

Bandpass filter

Bandpass filterDUT Mixer

6 dB 6 dB6 dB6 dB6 dB

Powersplitter

RF

LO

IFIF RF

LO

Figure 5. Upconversion/downconversion configuration

6

Conversion loss

Definition and importance of conversion loss

Conversion loss, as shown in Figure 6,measures how efficiently a mixer converts energy from one frequencyto another. It is defined as the ratio ofthe output power to the input powerat a given LO (local oscillator) power.A specified LO power is necessarybecause while the conversion loss of amixer is usually very flat within thefrequency span of its intended operation, the average loss will varywith the level of the LO, as the diodeimpedance changes. As shown inFigure 7, conversion loss is usuallymeasured versus frequency, either theIF frequency (with a fixed LO) or theRF frequency (with a fixed IF). Theconfiguration for a fixed IF measure-ment is different from those describedup to this point. (See the Fixed IFMeasurement section.) Figure 8 illustrates the importance of a flatconversion-loss response. The DUT isa standard television-channel convert-er. The input signal consists of a visualcarrier, audio carrier and a color subcarrier. Since the frequencyresponse of the converter has a notchin the passband, the color subcarrieris suppressed and the resulting outputsignal no longer carries a valid color-information signal.

Frequency

Pow

er le

vel

Conversion loss

Conversion loss =

20*log [ ]mag(f )

mag(f )

RF

LO

IFIF

RF

Figure 6. Conversion loss

RF IF

LO

RF IF

LO

Loss

IF freq

Loss

RF freq

Conv loss vs IF freq(fixed LO freq)

Conv loss vs RF freq(fixed IF freq)

0 0

Converter response

Input Signal DUT Output signal

Audio carrier

Visual carrier Color

sub-carrierColor sub-carrier

attenuated

LO

Figure 7. Two types of conversion loss measurements

Figure 8. TV tuner conversion loss example

7

Measurement considerations

Conversion-loss measurements can bemade with either a scalar networkanalyzer or a vector network analyzer,using the configurations shown inFigures 1 through 5. The measure-ment uncertainties are different foreach type of analyzer. For both typesof analyzers, the two main systematicerrors are port mismatch and frequen-cy response. The scalar network ana-lyzer approach requires additionalcare to minimize errors due to theanalyzer’s broadband detector. Forsome vector network analyzers, aninternal process, called sampling, andphase-lock requirements can also cre-ate errors. Next we will examine eachof these error terms and explore tech-niques to minimize their effects.

Mismatch errorsMismatch errors result when there is a connection between two ports thathave different impedances. Commonly,a device’s behavior is characterizedwithin a Z0 environment, typicallyhaving an impedance of 50 or 75 ohms.Although the test ports of a networkanalyzer are designed to be perfect Z0impedances, they are not. The imper-fect source and receiver ports of thenetwork analyzer create errors in thecalibration stage. Therefore, evenbefore a device under test (DUT) isconnected, some errors have alreadybeen created in the calibration stage(see Figure 9). Once the DUT is connected, the total measurementuncertainty is equal to the sum of thecalibration error plus the measurementerror.

Once the DUT is connected, interac-tion between the DUT’s ports and thenetwork analyzer’s ports cause mis-match errors. As shown in Figure 9,mismatch effects generate three first-order error signals. The first is interaction between the network analyzer’s source port and the DUT’sinput port. The second is between thenetwork analyzer’s receiver port andthe DUT’s output port.

The third is between the network analyzer’s source port and receiverport. For an FTD measurement, thisthird interaction is usually negligiblebecause the conversion loss and isolation of the FTD will attenuate thereflected signals. As frequency transla-tion precludes conventional two-porterror correction, attenuators can beused to improve port match.

Source

Receiver

Measurement:

Calibration:

Source

Receiver

Total Uncertainty = Calibration Error + Measurement Error

ρ ρ

(ρ ρ )

(ρ ρ )

(ρ ρ )

RF IF

LO

source

receiver

DUT input

DUT output

source receiver

Calibration plane

source receiverDUT

Figure 9. Mismatch effects

8

By adding a high-quality attenuator toa port, the effective port match isimproved by up to approximatelytwice the value of the attenuation. A high-quality attenuator has around32 dB of port match. The effectivematch is a function of the quality ofthe attenuator as well as its attenua-tion, as shown in Figure 10.

As shown in Figure 11 and Figure 12, awell-matched attenuator can significantly improve the effective portmatch. For example, a 10-dB attenua-tor, with a port match of 32 dB, cantransform an original port match of 10dB into an effective match of 25 dB.However, as the match of the attenua-tor approaches the match of the original source, the improvementdiminishes. As shown in Figure 12, thelarger the attenuation, the more nearlythe resulting match approaches that ofthe attenuator. However, excessiveattenuation is not desired since thiswill decrease the dynamic range of themeasurement system. The port matchof an FTD can be poor, typicallyaround 14 dB. Therefore, it is recommended that attenuators beplaced at the FTD’s input and outputports. Scalar network analyzers usedifferent detection methods than vector network analyzers that shouldbe considered when testing FTDs.

0 5 10 15 20 25 30 3510

15

20

25

30

35

Original match (dB)

Effe

ctiv

e m

atch

(dB

) 32 dB attenuator match

26 dB attenuator match

21 dB attenuator match

18 dB attenuator match

Region when attenuator no longer results in improved match

Figure 11. Effective match as a function of attenuator’s match (fixed 10 dB attenuator)

0 5 10 15 20 25 30 350

5

10

15

20

25

30

35

Original match (dB)

Effe

ctiv

e m

atch

(dB

) 20 dB attenuation

10 dB attenuation

6 dB attenuation

3 dB attenuation

Region when attenuator no longer results in improved match

Figure 12. Effective match as a function of attenuation (attenuator match = 32 dB)

ρ ρ

Attenuator

(ρ )

(ρ )( ) 2

Source

(ρ ) + = (ρ 2

source E source match

attenuator

source attenuation

attenuator source attenuation)( )

ff

ρE source matchff

Figure 10. Effective match as a function of attenuator’s match.

9

Considerations unique to the scalar network analyzer

Scalar network analyzers use broad-band diode detectors. Although capable of both narrowband andbroadband detection, the 8711 series,which includes the 8712C and 8714Cvector network analyzers, uses broadband detection for FTD measure-ments. Therefore, if you use an 8712or 8714, use the same FTD test considerations as you would for ascalar network analyzer.

Importance of proper filteringA scalar network analyzer’s broadbanddiode detector will detect any signalthat falls within its passband. Althougha broadband diode detector is an economical way to measure FTDs, italso can allow certain detection errors.The diode detector will detect thedesired IF signal, as well as other mixing products or spurious signals. To minimize the detection of undesiredsignals, a filter should be placed at thedetector port to pass the desired IFsignal but reject all other signals.Figure 13 shows an example of theincorrect measurements that mightresult when improper IF filtering isused in a scalar network analyzer configuration.

In Figure 13, the conversion loss measurement without the IF filterappears to be better than it really is.The lack of an IF filter generates erroneous results. The broadbanddiode detector cannot discriminate thefrequency of the received signal(s) —it measures the composite response. If the source is set at 1 GHz, it is“assumed” that this is the frequency of the detected signal. Any signal thatfalls within the passband of the diodedetector will be detected. If the output of a DUT is composed of thedesired IF signal plus the image frequency, LO and RF feedthroughand other spurious signals, the diodedetector will detect the composite ofall the signals within its passband.This composite signal will be incor-rectly displayed as a response thatoccurs at 1 GHz.

Frequency response errorWithout performing any sort of calibration on a scalar or vector network analyzer, the frequencyresponse of the test system cannot beseparated from the FTD’s response.One way to correct these errors is toperform a frequency-response normal-ization or calibration, using a throughconnection in place of the DUT.

For scalar network analyzers such asthe 8757D, which very accuratelymeasures absolute power, the normal-ization calibration can be performed in two steps. See Figure 21. First, theabsolute RF power is measured andstored in memory. Second, the DUT is inserted and the absolute IF poweris measured. Conversion loss is displayed using the Data/Memory format. The conversion loss value isvery accurate since the measurementsof the two absolute power levels, RFand IF, are very accurate. Ratioing twovery accurate absolute power levelsremoves the frequency response error.In some cases, a scalar 8757D with aninternal power calibrator and precisiondetector can make more accurate conversion loss measurements than avector network analyzer. In theAccuracy Comparison of the 8757Dand a Vector Network Analyzersection, error terms are used to illustrate how a scalar analyzer withinternal power calibrator can be moreaccurate than a vector network analyzer.

For analyzers that do not preciselymeasure absolute power, correctionsfor the frequency response error areless accurate. The input and output ofthe DUT are at different frequencies,but the normalization can only be performed over one frequency range.The result is that part of the test system is characterized over a differ-ent frequency range than that which isused during the actual measurement.

There are two choices for the frequen-cy range used for the normalization:either the DUT’s input (RF source)range, or the DUT’s output (receiver)range. The normalization should bedone to correct the portion of the testsystem that contributes the largestuncertainty; for example, this wouldbe the portion with the most loss orfrequency roll-off. Systems and components tend to have poorer performance at the higher frequen-cies, therefore the calibration shouldnormally be performed at the higherfrequencies. In general, high-quality,low-loss cables and connectors shouldbe used to minimize frequency-response errors.

For higher accuracy, combine a nor-malization calibration with externalerror-term correction. During the normalization, only one section of thetest configuration should be connect-ed, either the DUT’s input range orthe DUT’s output range. For highestaccuracy, the removed section can becharacterized separately. An externalcomputer is used to extract theremoved section’s S-parameters fromthe network analyzer. This data isthen used to modify the network ana-lyzer’s error terms to account for theeffects of the removed section.

Start 900.000 MHz Stop 1 000.000 MHz

2:Conv Loss /M Log Mag 1.0 dB/ Ref 0.00 dB

Swept Conversion Loss

1

2

1:Conv Loss /M Log Mag 1.0 dB/ Ref 0.00 dB

-9

-8

-7

-6

-5

-4

-3

-2

-1

Abs

dB

IF filter

No IF filter

2

1

Ch1:Mkr1 1000.000 MHz -6.38 dB

Ch2:Mkr1 1000.000 MHz -4.84 dB

Figure 13. Conversion loss response with and without an IF filter

10

Considerations unique to the vector network analyzer

Now that we have covered the important measurement considera-tions of the scalar network analyzer,let’s continue with a discussion of thevector network analyzer. The impor-tant considerations include: the needfor proper filtering, an accurate andstable LO, and power meter calibra-tion for the most accurate measure-ments.

Importance of proper filteringA vector network analyzer has a narrowband tuned receiver. Since thereceived signal is heavily filtered by an internal narrowband IF filter,broadband detection issues encoun-tered by the scalar network analyzerare not present. However, proper filtering is still very important for vector network analyzers with sampler-based receivers, such as the8753E and the 8720D.

Sampling architecture and issuesA sampler-based receiver consists of a voltage-tunable oscillator (VTO), apulse generator, and a sampler(switch). The VTO drives the pulsegenerator, which in turn drives thesampler. As a result, with proper tuning of the VTO, this combinationreplicates a down-converted input signal at the correct intermediate frequency (IF) for further processing.This combination is similar to a har-monic mixer in which the harmonicsof the LO are generated in the mixer,and the input signal can mix with anyharmonic. With proper tuning of theLO, one of the LO harmonics is offsetfrom the input signal to produce thecorrect IF signal.

Since there are many LO harmonics,any signal (desired or not) that is one IF away from any of the LO harmonics will be downconverted tothe network analyzer’s IF and detect-ed. To illustrate this sampler effect,let’s use the 8753E as an example.

The IF of the 8753E vector networkanalyzer is 1 MHz. Errors might resultbecause the incoming signal is not filtered until after it is downconvertedto the IF. If there is only one signal atthe receiver, this signal will mix withone LO harmonic and is properlydownconvert to 1 MHz. However, ifthere are multiple signals that are 1 MHz away from any of the LO harmonics, these signals will be down-converted to 1 MHz, which createserroneous responses.

Figure 14 illustrates an example ofthis sampler effect where the desiredIF output signal of the mixer is 110 MHz. In order to correctly detectthis signal, the 8753E will use a VTOof 54.5 MHz, where its second harmonic (109 MHz) will properlydownconvert 110 MHz to the desired 1 MHz IF signal. In the illustration, weshow two mixer products (6 LO-2RFand 9 LO-RF) that would also produceIFs at 1 MHz. Notice that these twospurs occur on either side of the LOharmonics (18 VTO and 42 VTO,respectively), but as long as they are 1 MHz away, they will be downcon-verted to 1 MHz. Aside from the signals which downconvert to 1 MHz,signals that will directly pass throughthe finite passband of the 1 MHz band-pass filter can cause problems. In the8753E, IF BWs from 10 Hz up to 6 kHzare available.

(RF – LO) – 1

Spur: 18 VTO – (6LO – 2RF) 981 – 980 = 1 MHz

Spur: 42 VTO – (9LO – RF) 2289 – 2290 = 1 MHz

MixerIF output

LOharmonics

RF – LO

6LO – 2RF 9LO – RF

110 980 2290

2VTO 18VTO 42VTO

109Frequency

Frequency

1 MHz 1 MHz 1 MHz

IF 1 MHz

LO harmonics

2

Desired signal: 2 VTO – (RF – LO) = 109 – (410 – 300) = 1 MHz

Given RF = 410 MHz IF = RF – LO = 110 MHz LO = 300 MHz

VTO = = 54.5 MHz

981 2289

Figure 14. Diagram of spurious measurement responses

11

As shown in Figure 15, wide IF BWallows any signal that falls within thepassband to be detected. For this rea-son narrow IF BWs are recommended.However, narrower IF BWs result inslower measurement speed.

Several techniques can significantlyminimize spurious responses. Onetechnique is to use adequate filteringat the network analyzer’s receiver port.Another technique, as illustrated inFigure 15, is to reduce the instru-ment’s IF bandwidth. Reducing the IFbandwidth will more selectively filtersignals in the instrument’s IF signalpath. A third technique is to avoid frequency spacings equal to the IF. Inthe case of the 8753E, the IF is 1 MHz.Therefore 1 MHz and multiples thereofshould be avoided when choosing frequencies and frequency spacings. A fourth technique is to avoid or eliminate certain frequencies thatmight cause spurious responses. Aspur prediction program, available inthe 8753-2 product note, predicts thefrequencies that might cause spuriousresponses. Modifying the frequencyselection or use the analyzer’s frequency list mode, in which theinternal source will step from one continuous wave (CW) signal to thenext, to avoid these frequencies. Thisprogram, written in BASIC 5.0, is customized for the 8753 series andswept IF mixer measurements,although it can be modified for othermeasurements such as fixed IF. Thisprogram only predicts the possibleoccurrence of a spur — it does notpredict its power levels, and it doesnot consider RF and LO subharmonics.

RF

LORF – LO RF + LO

1 MHz

Mixer RF

Mixer LO

Mixer IF

8753E VTO

harmonics

8753E IF

responses

IF detectors

BPF

LO harmonics

Mixer IF signals1 MHz

Selectible passband

Figure 15. Effects of IF BW on spurious measurement responses

12

R-channel phase-locking considerations For FTD tests, using frequency offsetmode with a vector network analyzer,many times you are required toremove the R-channel jumper andinput a signal into the R Ref-In port.Once an external signal is introducedinto the R channel, proper filtering atthe Ref-In port is essential. Certainrequirements must be met in order forthe R channel to phaselock properly.First, the signal into the R channelmust be within a specified power levelthroughout the entire IF range. Forexample, in the 8753E vector networkanalyzer, the R-channel signal must bebetween –35 dBm and 0 dBm through-out the selected IF range, otherwiseproper phase lock will not occur. If the R-channel filter does not coverthe entire IF range, modifying theselection of the IF range accordinglywill eliminate the phase-lock error. For example, if the IF range is set from10 MHz to 300 MHz, but the R-channelfilter is already rolling off at 300 MHz,then change the IF stop frequency toaround 250 MHz, or use a different filter. The second requirement is thatthe R-channel sampler needs to phase-lock with the correct signal. If thereare spurious signals entering the R-channel sampler, it might try to estab-lish phase lock with an improper frequency. This situation can be avoid-ed with proper R-channel filtering. (If you are using the 8712C or 8714Cvector network analyzers, R-channelconsiderations are irrelevant becausebroadband detetors are used.)

LO accuracy and stabilityFor the vector network analyzer, theuse of an accurate, stable LO isrequired for accurate magnitude andrelative phase measurements. Asshown in Figure 3, a network analyzerin the frequency offset mode will automatically set its internal source tosweep over the corresponding RFrange once you have entered the necessary information (i.e., IF range,LO information, upconversion ordownconversion, RF LO).Therefore, if the LO source is notaccurate, a network analyzer will notreceive the anticipated IF signal. As aresult, the IF signal can fall on theskirts of the IF filter or worse, it canfall entirely outside of the filter pass-band. In the first case, this inaccuracyis transferred to the measurementresults, and in both cases, phase lockcan possibly not even occur.

This LO accuracy and stability require-ment can be explained with the help ofFigure 16, which illustrates the blockdiagram of the 8753E in frequency offset mode. At the start of the sweep,the RF source is pretuned to the IFfrequency plus the LO offset, then themain phase-lock loop (PLL) is lockedup. The receiver will sweep over the IF range, and the source will track itwith the fixed offset. As shown, themixer's IF signal is downconverted inthe R-channel sampler to provide the1MHz phase-lock signal. Therefore, theLO source must be stable and accuratein order to provide an LO signal thatwill properly convert the mixer’s RFinput to its desired IF output. Thismixer’s IF signal, once downconvertedby the sampler, is compared to aninternal 1-MHz reference signal wherea resulting voltage, proportional totheir phase difference, is used to fine-tune the RF source. If the mixer’sIF signal is too far from the expectedreceiver’s frequency, then it might lieoutside the acquisition range of thePLL. The phase-lock algorithm will notwork and a phase-lock error is displayed. For example, the 8753Erequires a LO frequency accuracy ofwithin a few kHz of the nominal frequency.

1 MHzBPF

1 MHzBPF

IF detectors

Pulsegenerator

Pretune DAC

RF source0.03 – 3000 MHz

RF out

DUT

ExternalLO

source A / B

R

FLO

15 – 60 MHz

Reference1 MHz

Main PLL

RF IF

LO

Figure 16. 8753E block diagram of frequency offset mode

13

Power meter calibrationSome vector network analyzers, suchas the 8753E and the 8720D, are capable of using power meters toenhance their source and receiveraccuracy. This power-meter calibrationcan significantly minimize frequencyresponse error, as well as correct forsome mismatch effects. The followingdiscussion relates to a downconvertermeasurement, although the same concepts also apply for upconverters.It is a three-step process:

1. Perform power-meter calibration over the IF range.

2. Perform response calibration over the IF range.

3. Perform power-meter calibration over the RF range.

The power-meter calibration proce-dure needs to be done correctly or itcan lead to unexpected errors. Let’slook at each step for performing apower-meter calibration.

Step 1. Perform power-meter calibration over the IF range.The purpose of this step is to calibratethe network analyzer’s source for avery accurate power level over the IF range. After this, a response calibration completely corrects thereceiver’s frequency response over theIF range. With the power-meter calibration, the accuracy of the powermeter is transferred to the networkanalyzer.

First, the power meter is preset andzeroed. The network analyzer is set tosweep over the IF range.

Before connecting the equipment asshown in Figure 17, you need toensure that the R channel will phaselock properly. For network analyzers,such as the 8753E, that do not havean internal/external switch for the R channel, you must turn on the frequency offset before disconnectingthe R-channel jumper. This step isimportant since it prevents the network analyzer from attempting todo a “pretune calibration” when thereis no valid R-channel signal. If the network analyzer attempts a pretunecalibration without a valid R-channelsignal, the pretune constants would beinvalid.

Once the pretune constants are erroneous, proper phase lock will notoccur even with a valid R-channel sig-nal. If this should happen, reconnectthe R-channel jumper and allow theautomatic pretune calibration to fixitself. This step is not necessary for an8720D with Option 089 which providesan internal/external R-channel switch.

At this point, the LO frequency shouldbe at 0 Hz, so that the source andreceiver are at the same frequency.Connect the equipment as shown inFigure 17. The power at point A and Bare assumed to be the same, but actually are not. Therefore, it is important to make points A and B assymmetrical as possible. For example,if point B is connected to one arm ofthe power splitter, then ideally thepower sensor should also be connect-ed directly to the other arm of thepower splitter (that is, without anyadditional accessories such as anadapter or cable). Figure 17 is config-ured so that the mixer’s IF port willeventually be connected to point B.

A 0 dBm power level is optimal. Power meters are most accurate at 0 dBm because they use a precision50-MHz 0-dBm source as their refer-ence. If the frequency response tableof the power meter is used and themeasured power is close to 0 dBm,very high measurement accuracy canbe achieved (typically better than 0.05 dB). However, point B differsfrom point A by the tracking error ofthe splitter and this adds approximate-ly another 0.2 dB of uncertainty. Selectthe network analyzer’s source powerso that point A and B will be 0 dBm.

Set the power-meter calibration powerlevel. The number of calibration pointsshould be set at this time. Using nomore than 26 or 51 points usuallyyield very good results. The number ofpoints can then be increased beforecalibrating the network analyzer’sreceiver. The power-meter calibrationconstants will be interpolated over thenew points. Another time-saving tech-nique is to use one receiver calibrationfor different frequency ranges. This isaccomplished by calibrating over thelargest frequency span, then turningcalibration interpolation on and changing the span and number ofpoints. Interpolation errors are quitesmall, usually adding less than 0.05 dBto the total uncertainty.

During the power-meter calibration, itis important to connect the equipmentas shown in Figure 17. If point B isunterminated, errors will result due toreflections at this port. Terminatepoint B with the components withinthe IF path, such as the attenuator,filter and cabling. This closely reproduces the same condition duringthe calibration as in the transmission measurement. With the componentsconnected, the power-meter calibra-tion will account for mismatches dueto these components (such mismatch-es contribute to unleveled power levels at both points A and B). Analternative technique would be to terminate point B with a good Z0termination.

During the power meter calibration,via the GPIB interface, the networkanalyzer will automatically read theoutput from the power meter at eachfrequency point and adjust its sourcepower exactly by successively reset-ting and measuring the power. In mostinstances, two readings are sufficientto settle to within 0.02 dB of therequired power at point A.

Vector network analyzer

Lowpass filter

Power meter

GPIB

Power splitter 10 dBPower sensor

Point A Point B

Figure 17. Configuration for power meter calibration over the IF range

14

Step 2. Perform response calibration over IF range.With a very accurate power level atpoint B, now calibrate the networkanalyzer’s receiver for accurateabsolute power measurements overthe IF range.

The equipment is still connected asshown in Figure 17. As mentioned earlier, the power into the R channelmust be within the correct powerrange (–35 to 0 dBm, in the case of an 8753E), throughout the selected IF range, or proper phase lock will notoccur.

Examine the R-channel response. Thevalues displayed for the R channel canappear surprising. Notice that the values displayed are positively offsetby approximately 16 dB. For example,with a signal of 0 dBm, the R channelwill display approximately +16 dBm.Normal operation usually displaysratioed measurements (A/R or B/R). A mathematical offset is used toaccount for the differences betweenthe R-channel path and the A-or B-channel path. The network analyzercannot discern that the signal isdirectly applied to the R sampler, sothe signal level is displayed approxi-mately 16 dB too high. This offset andthe frequency-response error areaccounted for after a responsecalibra-tion is performed on the R channel.

Go to the calibration menu and perform a response calibration. Thethrough standard in Figure 17 is composed of the attenuator, filter, andcabling. After the response calibration,the R channel will indicate 0 dBm. At this point, the accuracy of thepower meter has been transferred tothe network analyzer's receiver atpoint B. In addition, the analyzer hasaccounted for the frequency responseof the through standard. Point B isnow calibrated for a very accurate 0 dBm absolute power reading.

Step 3. Perform power-meter calibration for RF range.This last step, another power-metercalibration over the RF range, provides a very accurate power levelat the RF port of the mixer.

An alternative is to do one power-meter calibration for both the RF andIF ranges. As mentioned earlier,power-meter calibration interpolationis a useful, time-saving tool. To savetime, the first power-meter calibrationcan be performed over a frequencyrange that encompasses both the RFand IF range. Then turn on interpola-tion to preserve the calibration in theseparate RF and IF ranges. This doescompromise accuracy to some extent.

Connect the equipment as shown inFigure 18. Once the appropriate information such as the LO informa-tion, upconversion/downconversion,and RF LO have been set, the network analyzer will automaticallyset the RF range. After selecting thedesired calibration power at port B,perform the power-meter calibration.

Upon completion of the power-metercalibration, a very accurate reading ofthe DUT’s absolute output power isdisplayed on the R channel, which hasbeen calibrated to 0 dBm. Notice thatthe reading is not necessarily the conversion loss of the DUT; it is theabsolute power reading of the DUT’soutput power. For example, if theDUT has a 6 dB conversion loss whenstimulated with –10 dBm of RF inputpower, the R channel will display –16 dBm. If you want the R channel todisplay the conversion loss directly,then in Step 1 select the power atpoint B to be the RF drive power forthe DUT’s RF port. The disadvantagehere is some loss in accuracy; powermeter calibration is most accurate at 0 dBm power level.

Vector network analyzer

Lowpassfilter

External LO source

Power meter

GPIB

Power splitter

10 dB

6 dB

Power sensor

Point APoint B

DUT

Figure 18. Configuration for power-meter calibration over the RF range.

15

Power-meter calibration for a high-dynamic-range measurementFor a high-dynamic-range configura-tion, Figure 19 is analogous to Figure17 above. Notice that the R-channeljumper is connected, since a valid signal is required into the R channelfor phase locking. In this configura-tion, the R channel is used only for thepurpose of phase locking. It is notincluded as a direct path in the actualmeasurement. This is the high-dynam-ic range configuration you should usefor performing Steps 1 and 2.

Figure 20 is analogous to Figure 18.This high-dynamic-range configurationis useful for devices that includes a filter, as shown. A filter is especiallyimportant in the R channel to suppress spurious signals that cancause false phase locking. A filter intothe B channel might not be necessary.This is the configuration you woulduse for performing Step 3.

Vector network analyzerPower meter

GPIB

Power splitter 10 dB

Power sensor

Point A Point B

Figure 19. Power meter calibration over the IF range, high-dynamic-range configuration.

Vector network analyzer

External LO Source

Power meter

GPIB

Power splitter 10 dB

Power sensor

RF in

RF out

Reference mixer

Point APoint B

DUT

Lowpass filter

Figure 20. Power meter calibration over the RF range, high-dynamic-range configuration

16

Accuracy comparison of the 8757D and a vector network analyzer

In some cases, an 8757D with an internal power calibrator and precisiondetectors will make more accurateconversion loss measurements than avector network analyzer with powermeter calibration. To illustrate the difference in accuracy, let’s look at thethe error terms involved in a down-converter measurement in either case.

In Figure 21a, the source is set to theRF range. The precision detector veryaccurately detects the absolute powerat its port over the RF range. Theresulting errors are m1, the mismatcherror, and rdetector RF,the uncertaintyin the detector’s calibration over theRF range. The rdetector RF error is verysmall, approximately ± 0.18 dB.

Next, the RF response is stored intomemory. The purpose of this step is to correct for the frequency responseof the attenuator and the source,rsource RF.

In Figure 21b, the DUT is connected.The errors consist of mismatch errors,m2 and m3 , and the uncertainty of thedetector’s calibration over the IFrange, rdetector IF. The latter error isvery small. Notice that an attenuatoris not used at the output of the DUTsince the precision detector measuresmost accurately at its port. If addition-al devices, such as an attenuator or filter, are added, their effects will addadditional calibration steps and there-fore added uncertainty. This configu-ration is especially advantageous forFTDs with internal filters.

In this 8757D calibration and measurement, the total uncertainty is equal to the sum of three uncorrect-ed mismatch error terms, m1, m2, and m3. The detectors errors are verysmall and are negligible.

With the vector network analyzer withpower-meter calibration, five mis-match errors result. Let’s take a closerlook at the error terms involved in avector network analyzer with powermeter calibration measurement.

In Figure 22a, the source is set to theIF range. A power meter calibration isperformed. The errors consist of themismatch error, m1, and the sourcefrequency response error over the IFrange, rsource IF. The latter error issmall.

In Figure 22b, a normalization calibra-tion is performed. Errors = m2 +rreceiver IF, where m2= m1 + rdetector IFand rreceiver IF is the receiver frequen-cy response error over the IF range.

In Figure 22c, the source is set to theRF range and a power meter calibra-tion is performed. Errors = m3 +rsource RF , where rsource RF is thesource frequency response error overthe RF range.

In Figure 22d, the DUT is connected.Mismatch errors, m4 and m5, are generated. As illustrated, the total uncertainty is due to five uncorrectedmismatch error terms, m1, m2, m3, m4,m5. The other errors are relativelysmall.

In summary, the 8757D with its internal power calibrator and precisiondetectors can be more accurate than avector network analyzer with powermeter calibration since there are lessconnections, resulting in less mismatch effects and therefore lessmeasurement uncertainty. However,the vector network analyzer is capableof further error correction using error-term manipulations with anexternal computer. With additionalerror correction, the vector networkanalyzer might be the more accurateinstrument. In the section, AbsoluteGroup Delay — A More Accurate,Lower Ripple Technique, proceduresfor external error-term manipulationsare discussed.

Sweeper

8757D with internal power calibratorSweeper

8757D with internal power calibrator

Precision detector

Precision detector

m1

m3m2

LO

(a)

(b )

B

B

System test ports

Figure 21. (a) Absolute power measurementover the RF range(b) Absolute power measurement over the IF range.

Vector network analyzer

Vector network analyzerVector network analyzer

Power meter

Vector network analyzer

m1

m3

m2

m5m4

Power meter

Detector

Detector

System test ports

LO

RF IF

(a)

(d)

(b)

(c)

System test ports

System test ports

Figure 22. (a) Power meter calibration over the IF range. (b) Normalization calibration over the IF range. (c) Power meter calibration over the RF range. (d) DUT is measured.

(a)

(a) (b)

(c) (d)

(b)

17

Fixed IF measurements

When a communication systemreceives a signal using a superhetero-dyne receiver, the information isdownconverted to an appropriate frequency called the intermediate frequency. The components in the IFsection of the receiver usually providemost of the gain and frequency selectivity of the communication system. Using a fixed IF allows adesigner to optimize the IF filteringand amplification which yield the highest performance. Typical IF frequencies for FM radios are 10.7 and 21.4 MHz, radar receivers are 30and 70 MHz and satellite receivers are70 and 140 MHz.

The other way in which conversionloss or gain flatness can be measuredis by keeping the IF at a constant frequency. This is known as a fixed-IFmeasurement. In order to accomplishthis, the LO must sweep in conjunc-tion with the RF input signal, keepinga constant frequency offset equal tothat of the IF. In many cases, this measurement more closely matchesthe operation of the DUT in the actualapplication.

The most common way to perform this measurement is to use the vectornetwork analyzer in a tuned-receivermode (without frequency offset).Figure 23 illustrates this configuration.As shown, two external sources areused. One external source providesthe RF signal and the other providesthe LO signal. Both sources sweep in a stepped-frequency mode under thedirection of a measurement controllervia the GPIB bus. This measurement is an ideal candidate for automationusing the built-in test-sequencing orIBASIC capability of many modernnetwork analyzers. Also, notice that anexternal reference must be connectedbetween the VNA and the sources (i.e. a 10 MHz reference). When testing a tuner with a fixed-IF sweep,the controller must be able to tune the internal LO. If an analog voltage is used, then a programmable powersupply or digital-to-analog converteris needed. If the LO is tuned digitally,the proper interface must be used.

Testing a mixer under fixed IF conditions can also be accomplishedwith one external source and the network analyzer under the control ofan external computer. The configura-tion is very similar to Figure 23,except there is only one external LOsource and a computer interface forboth the network analyzer and theexternal LO source. The configurationis shown in Appendix C.

Two cases should be considered for a fixed IF measurement. In both, theLO and RF frequencies are steppedover their respective ranges but ineach case the network analyzer is configured slightly differently.

The first considers the FTD as adownconverter, where the RF outputfrequency of the analyzer is steppedover the RF range and the analyzermeasures a fixed IF frequency at theR-channel input. The second uses theFTD as an upconverter and the outputfrequency of the analyzer (test port 1)is fixed at the IF frequency. The analyzer must measure the steppedRF frequency at the R-channel input.Both measurements are accomplishedusing the frequency offset mode capability available on many modernnetwork analyzers. This tuned receiver mode allows measurement tobe made at frequencies other thanthat of the analyzer’s internal source.

The Appendix provides a program fora fixed IF measurement on a mixerthat operates as a downconverter. This Basic for Windows program usesthe 8753E as the RF source andreceiver and the LO signal is providedby a ESG-D3000A signal generator.The LO source commands are writtenin SCPI, so any compliant signal generator will work. The program setsan absolute power level to the mixer’sRF input port and measures theabsolute power level at the IF port.The analyzer’s receiver is first calibrated at the fixed IF frequency,then the analyzer’s RF source level isset. The difference between the twosignals is the conversion loss of themixer under test. The programassumes the RF drive level from thetest port remains constant over thestepped frequency range and alsodoes not include insertion loss fromthe test port cable. The accuracy ofthe measurement can be enhanced byincluding a power-meter calibration atthe IF and RF ranges as detailed inthe 8753E Operating Manual. Caremust be taken to ensure adequate signal level to the R-channel input orthe system will not phase lock to theIF signal. The R-channel inputrequires a signal level between 0 dBmand –35 dBm. Adequate filtering isalso required to prevent spuriousresponse from entering the receiver.

Vector network analyzer

External RF Source External LO Source

GPIB

10 MHzbandpass

filter

Externalreference

Externalreference

10 dB

3 dB6 dBRF LO

IF

DUT

GPIB

Figure 23. Fixed IF configuration

18

Relative phase measurements

Measuring mixer relative phase parameters such as tracking, relativephase linearity and group delayrequires a vector network analyzer.Relative phase between two mixerscan be measured, but the absolutephase response of a single mixer cannot. In order to make a phase measurement of a FTD, a secondmixer is needed to provide a referencephase signal. This reference mixer isneeded because when the analyzer isin frequency-offset mode, the sourceand receiver function at different frequencies and phase is not definedbetween two different frequencies.This reference mixer needs to be driven by the same RF and LO signalsthat are used to drive the mixer undertest. The IF output from the referencemixer is applied to the referencephase-lock R channel of the vectornetwork analyzer.

Relative phase and magnitudetracking

The capability to measure the amplitude and relative phase matchbetween frequency-translation devicessuch as mixers is increasing in impor-tance as the number of multichannelsignal processing systems increases.These multichannel systems, such as direction-finding radars, requirethat the signal transmission througheach channel be amplitude- andphase-matched. To achieve therequired match between channels,each channel usually is manufacturedwith matched sets of components.

The match between FTDs is definedas the difference in amplitude and relative phase response over a specified frequency range. Also, thetracking between FTDs is measuredby how well the devices are matchedover a specified interval after theremoval of any fixed offset. For exam-ple, this interval can be a frequencyinterval or a temperature interval, ora combination of both.

The configuration shown in Figure 24can be used to make magnitude andrelative phase matching measure-ments. First, one DUT is measuredand its response is stored in memory.

Then the second DUT is measuredand the analyzer’s data/memory feature is used to compare them.Theanalyzer accounts for theresponse ofthe test system since the two DUTsare measured with exactly the sameconditions.Figure 25 shows the magnitude and relative phase matchbetween the two DUTs.

If both DUTs must be measured simultaneously (for tuning), then youcan place one DUT in the B channeland one in the R channel. However, inthis configuration you are limited bythe dynamic range of the R channel.

Therefore, the DUT in the R-channelpath should not have a filter. If it does,you can only view the DUT’s passbandmatch. For two DUTs with internal filters, you cannot simultaneously viewthe response using the configurationin Figure 24.

Figure 26 illustrates a configurationthat takes advantage of the highdynamic range of the A and B channels. The analyzer compareschannel A and B and can be used toactively display A/B responses, or formaking high-dynamic-range measure-ments such as tracking of out-of-bandresponses. In this case, the R channelis used only for phase locking.

Vector network analyzer

Power splitter 10 dB

DUT 1

Power splitter

DUT 2External LO Source

RF in

LO

RF

IF

LO

RF

IF10 dB

Low pass filter10 dB

Step 1: Measure DUT 1, store data in memory Step 2: Measure DUT 2, display data/memory

10 dB

Figure 24. Mixer matching — magnitude and group delay

CH1 S21/M log MAG .02 dB/ REF 0 dB

CH1 START 250.000 000 MHz STOP 400.000 000 MHz

PC

Cor

Ofs

CH2 S21/M phase REF 0 200 m /

CH2 START 250.000 000 MHz STOP 400.000 000 MHz

PCCor

HldOfs

Vector network analyzer with direct receiver access

External LO source

LPFDUT 1

DUT 2

RF

LO

IF

RF

R A B

Powersplitter

Powersplitter

Figure 26. Simultaneous tuning and high-dynamic range configuration

Figure 25. Tracking responses

19

Group delay

Group delay is a measure of a signal’stransition time through a device. It is classically defined as the negativederivative of phase vs. frequency.Relative group delay measurementscan be made with the same setup as that shown for relative phase measurements.

Important parameters when specifyinggroup delay

The result of a group-delay measurement depends upon manymeasurement factors, including aperture, averaging and IF bandwidth.The true negative derivative of phasevs. frequency cannot be determinedbecause any measurement made witha network analyzer is discrete, withone phase point per frequency point,so there is no continuous functionfrom which to take a derivative.Instead we calculate the slope of thephase over a frequency range. We usethe values of two frequency points oneither side of the frequency of interestand calculate the corresponding phaseslope. As illustrated in Figure 27, thismethod can yield a result that is dif-ferent than the true derivative of thephase.

The number of frequency points overwhich the slope is calculated is theaperture. The default aperture is thesmallest difference between any twofrequency points. The aperture can beadjusted via the “smoothing aperture”control to use a wider frequency rangeto calculate the phase slope. The maximum aperture is limited to 20%of the frequency span of the measure-ment.

Aperture = frequency span /(number of points – 1)

The aperture determines the informa-tion that is transferred from oneresponse to the other. Any effects onthe phase response carry over directlyto the group-delay response. Forexample, noise in the phase responsedirectly translates into noise in agroup delay response.

For example, if you set a narrow aperture, the noise in the phaseresponse becomes more significant,making the group-delay response noisier. At the same time, a narrowaperture allows you to see variationsin the group-delay response, whichyou would miss otherwise. But, if youincrease the smoothing aperture toomuch, you begin to lose resolutionsince the peaks start to disappear.Figure 28 illustrates how dramatic the trade-off between resolution andnoise as a function of aperture can be.

If you do not want to trade off resolu-tion and noise, you can achieve a balance by trading off some measure-ment speed. Using averaging or narrowing the IF bandwidth resultsin slower measurement speeds

without degrading resolution or noise.

The smoothing aperture should always be specified for a group delaymeasurement. A group delay measure-ment is only truly valid if its apertureis specified. Figure 28 illustrates howdramatically different the group delayresponses can be, depending on thesmoothing aperture settings. Alwaysspecify the smoothing aperture whenmaking a group delay measurement. Ifyou do not want to trade off resolutionand noise, you can achieve a balanceby trading off some measurementspeed. Using averaging or narrowingthe IF bandwidth results in slowermeasurement speeds, withoutdegrading resolution or noise.

Phase

φ∆φ

∆ω

Frequency (ω)

Tangent

∆ω∆φ

Average delay

Frequency

Group delay ripple

t o

Group delay (t )g =− d φ

d ω= –1

360°d φd f*

fo

Gro

up d

elay

Figure 27. Group delay

20% Aperture

0

0.2

0.4

0.6

0.8

1

1.2

Frequency

Del

ay (

ns)

Minimum Aperture

0

0.2

0.4

0.6

0.8

1

1.2

Frequency

Del

ay (

ns)

2% Aperture

0

0.2

0.4

0.6

0.8

1

1.2

Frequency

Del

ay (

ns)

10% Aperture

0

0.2

0.4

0.6

0.8

1

1.2

Frequency

Del

ay (

ns)

Figure 28. Group delay as a function of aperture

20

Absolute group delay

If you are interested in absolutegroup-delay measurements a calibra-tion mixer is required. When measur-ing the group delay of a linear device,standard vector error correction callsfor a through connection (delay=0) to be used as a calibration standard. The solution to the problem of measuring the delay of a nonlineardevice like a mixer is to use a calibra-tion mixer with very small group delayas the calibration standard. There areseveral techniques that can be used tocreate a calibration mixer with knowngroup delay. Two mixers have beencharacterized by Agilent Technologiesfor this purpose:

Mixer ANZAC MDC-123MCL ZFM-4

Frequency range 30 MHz to 3,000 MHzdc to 1,250 MHz

Group delay 0.5 ns0.6 ns

Upconversion/downconversion

The group-delay values above wereobtained by measuring two mixers,using the upconversion/downconver-sion configuration shown in Figure 5.The measured group delay was veryconsistent, with flat frequencyresponse. For example, a pair of MDC-123 mixers was measured,resulting in 1 ns of group delayexcluding the effects of the IF filterand components placed between themixer pair. A single mixer, then, musthave a group delay between 0 and 1ns. Therefore, one can state that thegroup delay of the MDC-123 is 0.5 ns,with an accuracy of ± 0.5 ns. This is avery conservative uncertainty window.Therefore, the uncertainties of bothmixers are equal to their respectivegroup delay, although these are worse-case estimates. After the through calibration you are ready to test thefrequency-translating DUT, enter avalue of electrical delay to compensatefor the group delay of the calibrationmixer (–0.5 ns for the MDC-123 or–0.6 ns for the ZFM-4). The resultingaccuracy derived from this techniqueis typically the group-delay uncertain-ty of the calibration mixer.

Choosing a calibration mixer witheven smaller absolute group delay willfurther improve the measurementuncertainty because the worst caseerror is half the stated value.

Another important consideration toremember when selecting a calibrationmixer is that the delay of the devicemust have a relatively flat group delayresponse over frequency. The conven-tion is to select a mixer with very widebandwidth compared to the mixer tobe tested — the wider the bandwidth,the better the mixers’ delay linearitywill be. For further informationregarding the desirable characteristicsof a calibration mixer, see theAppendix section, Calibration MixerAttributes. When using the calibrationmixer in the test system, typically onlya reponse calibration is used to calthrough the mixer. The mismatchinteraction between the test systemand the cal mixer and subsequentmeasurements with the DUT willintroduce uncertainty in the form ofripple into the measurement. Addingattenuation before and after eachmixer can greatly improve this ripple.

Another technique for improving themeasurement uncertainty in absolutegroup delay and delay linearity usesan upconversion/downconversion withadditional error correction at themixer's test ports. The technique isdiscussed in the section, AbsoluteGroup Delay — A More Accurate,Lower Ripple Technique.

Modulation delay

Measuring the group delay of convert-ers and tuners with inaccessible internal LOs is very difficult using themethod described above because welack the necessary phase-coherent LO for phase locking. One alternatemethod uses a definition of groupdelay as the time delay through themixer of a modulated signal. Thistechnique can measure group delay ondevices without needing a referencemixer. Using a frequency-responsenormalization, absolute and relativegroup delay can be measured.

As shown in Figure 29, the modula-tion-delay method for measuringgroup delay is accomplished by modulating an RF carrier. AM, FM or pulse modulation can be used. The carrier is swept through the frequency-translating device, the IFoutput is demodulated, and finally the phase of the demodulated signal iscompared with the original basebandmodulation tone (which can bederived by demodulating the carrier atthe input to the DUT). If square wavemodulation is used, zero-crossings canprovide the delay detection mecha-nism; if sine wave modulation is used,the phases of the sine waves must becompared. One main drawback forthe AM-delay technique is that itrelies on the device not having ampli-tude limiting, which would strip offthe amplitude modulation.

AM ModulatorRF

Sweep fmod

DUT

Phase detector

Gd = e (360 * f )– ÷ mod

LO

θ

Measure phasebetween twodemodulated signals

°

Figure 29. Modulation delay method for measuring group delay

21

As the carrier is swept through theDUT, the phase of the modulationvaries proportionately to the phaseresponse of the device itself. Groupdelay is calculated as follows:

Gd = – θe° ÷ (360*fmod)

where θe° is the phase differencebetween the demodulated input andoutput carriers in degrees, and fmodis the modulation frequency. The aperture of this measurement is equalto twice the modulation frequency.The effects of aperture are the sameas the phase-derivative method ofmeasuring group delay previously discussed. As the aperture gets larger,the measurement becomes less noisy,but resolution degrades.

AM-delay measurements are easilyperformed with the 8711C series ofnetwork analyzers (see Figure 30). A response normalization calibration is performed without the DUT connected. Then the DUT is connect-ed for relative or absolute group-delaymeasurement. The aperture of thismeasurement is fixed at 27.8 kHz,with typical accuracy around ± 10 to20 ns. The accuracy of this techniqueis limited by the diode detectors,which have less sensitivity than thenarrowband detectors used in the8653E or 8720D. Group-delay accura-cy with the 8753E or 8720D is better,typically around the picosecond range,depending on the smoothing aperturesettings. As shown in Figure 30, thephase-derived method (top response)is less noisy than the AM delaymethod (bottom response). Vectornetwork analyzers use a phase-derivedmethod of group-delay measurement,making it difficult to measure groupdelay of FTDs with internal inaccessi-ble LOs.

The FM-delay technique is very similar to the AM technique, in that ituses a modulated signal to measurethe group delay. The key difference is that FM modulation is used, whicheliminates errors due to amplitudenoise, and is not sensitive to ampli-tude limiting in the device under test.However, since the frequency at theoutput is dependent on the LocalOscillator (LO) frequency, the FMtechnique has difficulties when the LO is not synthesized.

Time domain

The measurement of group delaythrough a mixer can be easily accom-plished using the time-domain optionavailable on some vector network analyzers. This does not require theextensive hardware and filtering thatother traditional methods demand. Allthat is required to measure absolutegroup delay and delay linearity is asingle mixer, a 50-ohm airline, a short,and a VNA capable of time-domaintransformation. (This capability isavailable on the 8510C and 8720ESanalyzers with Option 010.)

The technique uses the measuredreflection coefficient from the mixerthat is terminated in the 50-ohm airline and the short. The measuredfrequency response is transformedinto the impulse response using thetime-domain option on the VNA.Knowing the absolute delay of the airline, the absolute delay through the mixer can be calculated by examining the two-way reflection from the short in time. In addition, the delay linearity of the mixer as a function of frequency can be mea-sured using the gating function on the VNA. The gating function filtersthe effects of reflections internal to the mixer and isolates only thetransmitted signal through the mixer.This gated signal contains the delaydistortion introduced by the frequencytranslation process. In this way, delaylinearity through the mixer can now be directly measured in the frequency domain.

The impulse response on a VNA simulates a traditional Time DomainReflectometry (TDR) measurement.TDR is a technique to generate animpulse or step waveform in time that is propagated down a transmis-sion line. The reflections from discontinuities can then be detected intime on the TDR display. The measured time delay represents thetwo-way electrical distance to the discontinuity in the transmission line.Individual discontinuities can beexamined in time if there is adequateelectrical separation between them.This same TDR measurement can besimulated using a high-performanceVNA capable of time-domain transfor-mation. In this case, the measured frequency response from the VNA ismathematically transformed into theimpulse response using the InverseFourier Transform algorithm presentwithin the VNA.The measurementaccuracy is improved by applying the standard one-port vector errorcorrection to the initial reflection mea-surement.

Ext detX – input

Ext detY – input

DUT

Detectors

871XC scalar orvector network analyzer

Normalize without DUT first,then measure with DUT

Power splitter

Figure 30. AM delay measurements using the 8711 series

Center 175.000 MHz Span 100.000 MHz

5 10 15 20 25 30 35

Abs

2

1:Memory Delay 5 ns/ Ref 0 sns

35

Center 175.000 MHz Span 100.000 MHz

5 10 15 20 25 30

Abs

ns 2:AM Delay /M 55.6 kHz 5 ns/ Ref 0 s

1

Ch2: Mkr1 25.370 MHz -16.45 ns

M1

1

Ch1: Mkr1 25.480 MHz -16.31 ns

22

As a measurement example, thereturn loss as a function of frequencyof the IF port of a broadband mixer ismeasured on the VNA. The RF port ofthe mixer is terminated with an airlineand a short. The proper LO drive is also applied to the mixer. The hardware configuration is shown inFigure 31. The VNA is used to mea-sure and transform the frequencyresponse of the mixer under test. The resolution of the time-domaintransform on the VNA is directly related to the measured frequencyspan. In order to examine two closelyspaced discontinuities in the timedomain, a large frequency span mustbe measured. However, limitations inthe operating frequency range of the device under test also limit thetime-domain resolution. It may benecessary to electrically separate the discontinuities whenever possible.Because we wish to isolate the reflected signal from the short in the time domain, placing the shortdirectly on the output of the mixermay not provide enough separationbetween the mixer’s own internalreflections and the signal reflectedfrom the short. Placing a 50-ohm airline between the mixer and theshort can improve the measurementresolution by electrically separatingthe reflections from the mixer and theshort, as shown in Figure 31.

For example, the VNA is calibrated foran S11 one-port cal over the IF fre-quency range of 50 MHz to 10.05 GHz.The LO was set to a fixed value of 20GHz. This IF range is chosen for tworeasons; it is within the operatingrange of the mixer’s IF port, and itallows the VNA to be used in the low-pass mode of operation. The low-passmode yields the highest resolution inthe time domain.

The measured return loss of the IFport as a function of frequency isshown in Figure 32. The return lossshows the typical peaks and valleysthat occur when multiple reflectionsare added and subtracted over themeasured frequency range. This fre-quency response is then transformedinto the impulse response using thetime-domain option on the VNA.Figure 33 shows the impulse responseof the mixer terminated in theairline/short.

This figure represents the TDRresponse as a function of time. Itshows the individual reflections from several discontinuities as theimpulse waveform propagates alongthe transmission line. The position of each impulse shows the electricaldistance in time from the calibrationplane.

The amplitude of each impulse showsthe average amount of reflected signalfrom each discontinuity. The threereflections to the left of the figure arecaused by discontinuities presentwithin the mixer: for example, theinput and output connectors andtransformers.

H

PORT 2PORT 2PORT 1PORT 1

to LO Source

Airline

ShortMixer

Figure 31. Hardware configuration for measuring the absolute delay and delay linearity of frequency translation devices

Figure 32. Measured return loss from the IF port of the mixer as a function offrequency, measurement made using a shorted airline placed on the RF port

CH1 S11 LIN 50mU / REF -100mU1

STOP 1.9 nsSTART -100 ps

1: 204.42 mU 888 ps

Lowpass Mode

Airline / Short

Mixer Response

Second Reflection fromAirline / Short

Figure 33. Low-pass impulse response of the mixer using a shorted airline placed on the RF port

23

The largest impulse on the figure iscaused by the reflection from theshort placed at the end of the airline.This is the discontinuity of interest, as it represents the signal that istransmitted through the mixer undertest. A marker is used to measure the position of the short at the end of the airline in time. Because the dis-continuities are electrically separatedin time, each reflection can be individ-ually examined. The measured markervalue of 888 ps represents the two-way reflection from the short as thesignal passes through the mixer andthe airline. By knowing the electricaldelay of the airline, we can subtract it from the measured delay, and thenby dividing the result in half, we can obtain the mixer’s absolute groupdelay of 168 ps. Here we are assumingthat the delay characteristics of the mixer are the same as the signalpasses through the mixer from the IFto RF and RF to IF ports.

Measuring delay linearity

Delay linearity is also an importantcharacteristic of components used inbroadband communication systems. In order to measure the delay linearityof the mixer using this technique, a time filter needs to be applied to the frequency domain data. The timefilter is used to isolate the reflectionfrom the shorted airline. The reflec-tion from the short contains the delaydistortion introduced by the frequencytranslation process as this signal ispassed through the mixer. The timefilter or gating function is positionedon the time-domain response aroundthe reflection from the short, asshown in Figure 34. This gated mea-surement now represents only theresponse of the signal that passedthrough the mixer. Eliminated are allother reflections from the measure-ment: the mixer’s own internal reflec-tions and any secondary reflectionsbetween the mixer and the short. Onlythe reflection from the short isallowed to pass through the time filter.

Once the gate is properly centered,the time-domain function can then be turned off while leaving this gateactivated. This will yield the frequencyresponse of just the first reflectionfrom the short, the signal that passedthrough the mixer.

The delay of this gated frequencyresponse is shown in Figure 35. Thisfigure shows the mixer’s delay lineari-ty as a function of frequency. In thiscase, the measured delay linearity isless than 100 ps peak-to-peak over a10 GHz span, and only 50 ps over themiddle 8 GHz.

Note that the ends of the frequencyrange show a downturn in the measured delay response. This may be the result of the windowingfunction and data extrapolation that is applied to the measured frequencyresponse data. Generally, this window-ing process distorts the measurementsat the very ends of the gated frequen-cy response. The initial frequencyspan should be chosen larger thanrequired to avoid possible errors inthe frequency range of interest.

The measurement of group delaythrough the mixer can also be charac-terized using the reflection from theRF port of the mixer. In this case, theIF port of the mixer is terminated with the airline and short. Here, thetime-domain transformation is typical-ly operated in the band-pass mode,because the frequency range of themixer’s RF port does not typicallyoperate down to DC. The band-passmode allows arbitrary selection of thestart and stop frequencies, which isvery useful for band-limited devices.Keep in mind that the band-pass modewill reduce the time-domain resolutionwhen compared to the low-pass modecovering the same frequency range. In this case, a larger frequency rangeor longer airline may be required inorder to improve the time-domain resolution.

STOP 1.9 nsSTART -100 ps

CH1 S11 LIN 50mU / REF -100mU 1: 204.42 mU 888 ps

Lowpass Mode

CH1 S11 DEL 50 ps / REF 880 ps

1

STOP 10.050 000 000 GHzSTART .050 000 000 GHz

1: 881 psec 5.000 000 GHz

Gating ON

Figure 34. Low-pass impulse response using a time gate to isolate the reflection from the short

Figure 35. Measured delay linearity of the mixer as a function of frequency

24

Reflection measurements

Now that transmission measurementshave been covered, let’s move on toreflection measurements. Reflectionmeasurements are linear, even whentesting frequency-translating devicessince the reflected signal does notundergo a frequency shift. Therefore,these measurements are essentiallythe same as for filters and amplifiers,with a few minor variations. The systematic errors in the test systemand the calibration techniques used to remove them are the same. Narrowband detection should be usedif available.

Mixers are three-port devices, and thereflection from any one port dependson the conditions of the other twoports. When measuring reflection on athree-port device, it is important toterminate the ports that are not beingtested with an impedance typical ofactual operation. While this is oftenthe characteristic impedance Z0(usually 50 or 75 ohms), it could alsobe more complex. For example, if theIF port of the mixer is directly connected to a filter, then this filtershould be used when testing RF- orLO-port reflection. In this case, itmight be necessary to disconnect theIF port of the mixer from the test setand connect the appropriate loadimpedance.

When testing the RF or IF ports, theLO signal should be applied at thepower level that will be used in actualoperation. Since the bias point andimpedance of the mixer diodes is afunction of LO level, the reflection atthe other ports will also be a functionof LO level.

Figure 36 shows the different RF-portVSWR results obtained with a broad-band load versus the IF terminationthat would be used in actual opera-tion. The lower trace was measuredwith the IF port terminated with a 50-ohm load. The upper trace was terminated with a filter and then a 50-ohm load. The effect of mismatchbetween the mixer and filter is readilyapparent. Usually, the mismatch iseven worse in the stopband where thematch is typically quite poor. This is avery common operating environmentfor a mixer.

This measurement was done usingnarrowband detection.The measure-ments of VSWR on the LO and IFports are very similar. For IF-portSWR, the RF port should be terminat-ed with a matched impedance and theLO signal should be applied at its normal operating level. For the LOport VSWR, the RF and IF portsshould both be terminated in conditions similar to what will be present during normal operation.

Testing converters and tuners is easiersince there are only two ports and theLO will already be at the correctpower level. The same requirement forproper termination of either the RF orIF port also applies.

Start 900.000 MHz Stop 1 000.000 MHz

RF Port SWR

1:Reflection &MSWR 0.1 / Ref 1.00 C

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

AbsCh1

1

M1

IF filter

50-ohmload

Figure 36. Reflection measurement

25

Isolation measurements

Isolation is a measure of the leakageor feedthrough from one port toanother. The more isolation a mixerprovides, the lower the feedthroughwill be. Isolation is a transmissionmeasurement since the stimulus isapplied at one port and the responseis measured at another. However, youare measuring a signal at the same frequency as the stimulus, and not thefrequency-shifted signal. As is the casefor reflection measurements, isolation/feedthrough is dependent also on thetermination and LO power on theports not being tested.

Three main isolation terms are ofinterest: RF-to-IF feedthrough, LO-to-IF feedthrough and LO-to-RF feed-through. The latter term is oftenimportant if the mixer is near thefront end of the receiver or tuner thatis connected to a cable or antenna. Inthis case, significant LO leakage couldcause interference in other frequencybands. Both the LO isolation terms aresmall for single- and double-balancedmixers. RF to IF feedthrough is alsolow in double-balanced mixers. RF feedthrough is usually less of aproblem than LO to IF feedthroughbecause in most cases, the LO powerlevel is significantly higher than theRF power level.

RF-to-IF feedthrough is measuredwith the same instruments and setupused to measure conversion loss(except frequency-offset mode is notneeded when using a vector networkanalyzer such as the 8753E). Becausethe source and receiver frequenciesare the same, set up the network analyzer to use narrowband detectionto make the measurement. The onlydifference in the hardware configura-tion is that the IF filter needs to beremoved so the RF feedthrough willnot be filtered out. Depending onwhich network analyzer is used, a frequency-response calibration or fulltwo-port calibration should be performed to improve measurementaccuracy.

The RF feedthrough level is verydependent on the level of applied LOsignal. For this reason, the measure-ment should be made with the LO signal present at its normal operatinglevel.

LO feedthrough measurements aremade the same way as describedabove. Remember to terminate the unused port with the properimpedance. The LO power level usedfor the measurement should be thesame level that is used during actualoperation of the mixer.

LO to IFIsolation

RF Feedthrough

V = 0RF(1)

VLO(2)

LO(3)V

LO to RFIsolation V = 0IF(3)

VLO(2)

LO(1)V

V = 0LO(2)

VRF(1)

RF(3)V

VRF

VIF

VLO

1

VLO

VLO

VRF

2

3

Figure 37. Isolation

Start 900.000 MHz Stop 1 000.000 MHz

RF Feed through

1:Transmission /M Log Mag 2.0 dB/ Ref 0.00 dB

–18

–16

–14

–12

–10

–8

–6

–4

–2

Abs

dB

Ch1

1

Remove IF filterApply proper LO power

Figure 38. RF feedthrough

26

Feedthrough measurement of converters and tuners

The procedure for measuring RF to IF feedthrough of a converter or tuner is identical to that of a mixer.Since an IF filter is often included inthe DUT, the RF leakage is often verysmall. An instrument with highdynamic range should be used tomake the measurement, and averagingand a reduced IF bandwidth can alsobe required to lower the noise floor ofthe test instrument.

Measuring LO leakage requires a different technique and test instru-ment. Since the LO port is typicallyinaccessible, a normal network-analyzer transmission measurementcannot be used. The best instrumentto use for measuring LO leakage is aspectrum analyzer. The spectrum analyzer is connected to either the RFor IF port and tuned to the frequencyof the LO signal. Again, the unusedport should be terminated. If the LOfrequency of the converter or tuner isknown to within a few kHz, then thetuned-receiver mode of a networkanalyzer such as the 8753E could alsobe used to make this measurement.

The plot in Figure 39 shows a fairlyhigh level of LO leakage present at the output of a converter. If a scalarnetwork analyzer were used to measure conversion gain of thisdevice, this signal could cause somemeasurement inaccuracy because theanalyzer’s broadband detectors wouldinclude the LO as well as the desiredIF signal. In cases such as this, additional filtering should be added to reduce the level of the LO feed-through when measuring conversiongain.