Intermodulation
Distortion (IMD)
Scorpion Option 13 ™
Application Note
Fast, Flexible, and Accurate IMD Measurements Using
a Vector Network Measurement System
Introduction Offset
In this application note methods to successfully measure two-
tone intermodulation distortion products using the IMD option
on the MS462xx Vector Network Measurement System are
reviewed. The application is designed to properly configure the Relative
receiver to perform CW or swept measurements of all IMD Product
necessary signals, set up the sources, coordinate necessary
calibrations, and present the results in a useful form. Since the
IMD application is quite flexible and there are a number of
setup options, a careful analysis of the measurement needs will
lead to a better measurement system. 2f1–f2 2f2–f1
f1 f2
There are many ways to characterize the distortion properties
of a device but two-tone intermodulation distortion (IMD, Figure 1. The frequency relationships in a two-tone IMD
related terms referring to third order intercept point TOI and measurement are illustrated in this diagram. Tones at f1 and f2 are
the term IP3 are also used) is one of the most common. The applied to the DUT and third order products at 2f1–f2 and 2f2–f1 may
two-tone IMD measurement describes a fairly common be generated. The relative magnitude of these tones defines the
communications scenario: if strong tones in neighboring intermodulation distortion.
channels are present, can they introduce spurious signal
energy in a troublesome amount into an adjacent channel? 3rd Order Intercept
This is a very real and legal issue for equipment makers and Fundamental
renders this measurement of significant importance. While
the two-tone measurement is not perfectly realistic in that it Pout
assumes that the signals are CW (or narrowband modulated 3rd Order IM Product
at best), it is a relatively straightforward measurement
conceptually, is well defined, and is quite repeatable.
The general assumption for this measurement is that two Pin
tones are applied to a device under testing (DUT) at
frequencies of f1 and f2 (|f1–f2| small). The difference in
A typical measurement would be made of a lower power level.
frequency between the two tones, termed the offset, is often The fundamental and IMD lines would then be extrapolated
between a few hundred kHz and a few MHz, although other to predict their intersection at 3rd order intercept.
values are possible. The non-linear characteristics of the
DUT generates intermodulation products; among them are Figure 2. An illustration of the power relationship between the
signals at 2f1–f2 and 2f2–f1, which will be very close to the fundamental tones and the third order products in a simple non-linear
original tones in frequency and represent potential adjacent DUT is shown in this diagram. The Third Order Intercept is based on
an extrapolation from measurements at low power and assumes ideal
channel spurious signals. The relationship of these various
slopes for the output tone level dependencies.
spectral components is illustrated in Figure 1. Since these are
third order products, their amplitude dependence on input
power will be about three times that of the fundamental
(prior to saturation). Hence, if one plots Pout (fundamental)
and Pout (third order product) as a function of Pin, the slopes
of the curves will be approximately 1:1 and 3:1, respectively,
at sufficiently low power where the system is linear. With
these disparate slopes, the lines will eventually intersect and
the point of intersection is termed the third order intercept
point (TOI). The amplitude of Pout (third order) relative to Pout
(fundamental) at a given input power level is termed the IMD
level. These concepts are illustrated in Figure 2.
2
Generally, the IP3/TOI measurement is done at a single The Measurement Flow
power level and straight lines constructed (since the slopes The measurement process is illustrated in the flow chart of
are assumed and the one data point will then establish the Figure 3. The first two steps, designing the setup and
equation of a line). The intersection of these extrapolation analyzing potential uncertainties are the most important since
lines determines the IP3/TOI value. The intercept point there are many potential choices.
generated by this measurement has its limitations; foremost
Configure Setup; Frequencies, Power Levels,
among which is that the intermodulation product does not Combining Network (Consider IMD Levels,
always follow the 3:1 slope ratio. Thus the simple analysis Power Levels, Matches, ...)
will lead to an erroneous intercept point. Also, the details of
the intermodulation power may be of interest in itself.
Evaluate Uncertainties,
An initial thought may be to do a measurement set at a few Review Setup if necessary
different power levels and attempt to do a more accurate
extrapolation of the intercept point based on this data. This
would seem inadvisable for several reasons: Desired Measurement Variable:
IMD Product or IP3 Intercept?
• From experiments and manufacturer data, a few added
points will not help accuracy very much and, if properly Product Intercept
positioned, could produce an even worse result than that
obtained with the single point.
Connect Combiner Perform a Receiver Cal.
• Definition problems arise if the tones are of unequal Network Connect a Thru
amplitude. For the second power level, would the between Ports 1 and 2
instrument raise both tones equally? Raise only one?
In the single point scenario, the instrument has the
luxury of assuming the default Pin scale definition: the Connect Combiner
Network
tones are assumed to rise by the same amount at each
succeeding step along the Pin axis If using Internal Source(s),
perform Flat Port Power Cal
For these and other reasons, the two most common at Combiner Output if desired
measurements are the raw IMD product (usually
expressed relative to one of the main tones) and the IP3
extrapolation discussed above. The MS462xx Vector If the results are to be DUT input
Network Measurement System is set up to measure these referred, perform an IMD Cal
(Thru Line between
two quantities directly. Combiner Network and Port 2)
Connect DUT between Combiner
Network and Port 2 to measure
Figure 3. A flow chart describing the IMD measurement process in
the MS4623X. Generally less effort is required for strict product
measurements since absolute power knowledge is not required.
3
• Which sources to use? If the MS462xx was purchased
with the optional second source, all that the user must MS4623B
provide is a combining network (see below). In cases
where extremely low phase noise is important, or for a
number of other reasons, external synthesizers can be
used. In all cases, the sources will be controlled by the
MS462xx for more convenient measurements. If external
Amp LPF
sources are used, evaluate their pulling sensitivity prior Passive
Combiner
to designing the combiner network. Use the source Amp LPF DUT
submenu within IMD to set these up: Source 1 is always
internal, Sources 3 and 4 are always external and source Pads Pads
2 will be internal (port 3 access) if the optional second
source is installed.
Figure 4. An example of a reasonably elaborate combining network.
• CW or Swept measurement? Under the application The added isolation (from the amplifiers) and signal filtering provided
menu, one can select the CW receiver mode for a by this assembly may be needed for measuring extremely small
display similar to that seen on a spectrum analyzer intermodulation products.
(tones fixed, all four tones visible as spectral entities).
This level of complexity may be required if very low IMD
The markers can be positioned for direct IMD readings.
levels are being measured (of order –70 or –80 dBc), if the
This is a quick and familiar measurement but limited.
sources can be easily pulled (external), if the source spur
The swept measurement allows the tones to be swept
levels are very high, if the DUT is very sensitive to match, or
over any band and, at each frequency point, all four
for other reasons. In some cases, this structure can be
tones are measured to perform the required
simplified to that shown in Figure 5.
computations.
• What power levels and frequencies are to be used?
MS4623B
These are generally set by the DUT and the test
requirements. The power levels needed may affect the
combiner network since the maximum internal source
power may be as low as 5 dBm and the combiner
network will usually contribute loss. Amplifiers may be
required. The offset frequency (tone spacing) may not be LPF
Passive
set by the test requirements. When choosing an offset, Combiner
DUT
select one as large as practical for the application LPF
(1 MHz is common) and avoid the image offset
Pads
frequencies discussed in the appendix. Because of
source phase noise, available dynamic range for the
IMD measurement will start to decrease as the offset Figure 5. A simpler combining network. The lower levels of isolation
drops below about 200 kHz. provided by this network may be acceptable for measuring larger
intermodulation products. Additional amplifiers may be needed if the
• Is the DUT a mixer? If so, set up the DUT LO using DUT drive level must be higher than the system can provide (+5 to
the source menus and select the DUT type as mixer at +10 dBm/tone prior to the network losses).
application control. Be aware the mixer spurs and
mixer match may be an issue and filters or pads may It is recommended that filters always be used to reduce
be helpful. source harmonics to well below the expected IMD levels
(can cause additional mix products in the DUT) and that a
• What combining network should be used? This is a
true combiner (resistive, Wilkinson, etc.) instead of a tee
complicated question dependent on DUT, measurement,
connection be used. The pads, which may be small, provide
and source details. A very complete combining network
some additional isolation and impedance buffering. The pad
is shown in Figure 4.
before the receiver is still optional and is mainly there to
ensure that receiver compression is not an issue. This
approach is most likely to be valid if the expected IMD
products are larger than –60 dBc.
4
In both the minimal and maximal combining scenarios, and • Images Because of the receiver structure of the
anything in between, the frequency range selected may need MS462xx, there are certain offset frequencies that will
to be revisited since the filters will reduce the available span cause a main tone to land in the image response of the
to usually less than an octave (obviously not too important if receiver while measuring the IMD product. The issue is
in CW RCVR mode). If it is known that the given source detailed in the appendix but the most troublesome
impurities are particularly small or irrelevant, this condition offsets are 78.125, 125, 156.25 and 250 kHz. Avoiding
can be relaxed somewhat. these frequencies by even 5 kHz can usually remove
any difficulty.
Before accepting a setup, the user should consider a number
of possible sources of uncertainty. • Impedances Ensure that any filters used see their
desired impedance levels and that the DUT is not
• DUT Compression If the DUT is already compressing, presented with an impedance that may lead to
the extrapolated intercept point will be inaccurate altered behavior.
(see Figure 2).
The remaining steps in the measurement flow center around
• Receiver Compression and IMD The receiver IP3 is performing calibrations and selecting measurement types.
typically about +35 dBm (excitation near 0 dBm) and The available calibrations can be described as characterizing
can limit the measurement in some cases. Also the the receiver behavior, flattening the source behavior with
receiver enters compression above about +10 dBm frequency, or characterizing the path between the source and
port power. If the DUT output levels are high, use an the receiver ports.
output pad.
• Dynamic Range and Noise Levels While receiver IMD
Measurement Variables and Receiver Cals
may limit the upper end of dynamic range, the source or These selections are usually dictated by the test requirements
receiver noise levels will usually limit the low end. If the but there are some important effects. If just the IMD product
internal sources are used, source phase noise will start to is required, this is a relative power measurement (product
become important for offsets below a few hundred kHz relative to main tone). As a result, the receiver calibration
(results will vary for external sources). A typical noise may be skipped. This receiver calibration, required for
floor (0 dBm main tone levels) will be about –75 dBc or intercept measurements, establishes an absolute power
lower for higher offsets, degrading to about –70 dBc for calibration at the receiver port and is available under the
a 200 kHz offset (internal sources). If an inadequate power menu and under the IMD cal menu. This very general
combining network is used, in isolation terms, these calibration uses a simple thru line between port 1 and port 2
levels will rise due to pulling. Use a sufficiently small and uses the internal ALC accuracy to determine the
IF bandwidth to minimize noise effects. frequency response of the receiver.
• Input Signal Effects If the signal entering the DUT is Flat Port Power Calibrations
already corrupt with harmonics, IMD products, or spurs, Because the input power to the DUT is often a critical test
the results could be distorted. Filters are suggested for parameter and the combining network often has a frequency
harmonic (and some spur) suppression. To keep self- dependent insertion loss, it may be desired to flatten the
IMD products low, ensure that there is sufficient power at the output of the combining network. This
combiner network isolation to prevent source pulling calibration, available if using internal sources, uses a routine
(frequency and amplitude pull) and that any amplifiers under the power menus and a power meter to adjust the ALC
in the combing network are sufficiently linear. systems as a function of frequency to produce a set level.
• Spurs from DUT Particularly if the DUT is a mixer, This calibration may be done for any number of frequency
spurs into the receiver may present a problem. Use points and at any accessible power level. It must be
adequate filtering. performed separately for source 1 and source 2 although the
• Frequency Accuracy If using external sources, power meter connection does not have to change. This step is
ensure that the MS462X is linked to the same 10 MHz less relevant if in CW RCVR mode.
reference. Even a few PPM difference in references can
affect readings when using small IF bandwidths.
5
Input Referrals and the IMD Calibration The plot in Figure 6 is an example of the spectrum-analyzer-
If the desired result (product or intercept) is to be referred to like display of CW RCVR mode. The amplifier DUT in this
the DUT input, the tone level entering the DUT must be case was measured with a tone offset of 200 kHz and an
known so that the DUT gain can be computed. The IMD output tone amplitude of about 0 dBm. The tones were
calibration performs this task and requires only a simple thru placed at 890 and 890.2 MHz for this example and the
line connection between port 2 and the output of the markers show an IMD product on the order of –50 dBc.
combining network. The main tone levels to be input to the This product amplitude is high enough that a very simple
DUT are stored as a function of frequency for later gain combining network could be used.
computation. Obviously if the power levels or frequencies
are changed, this cal will have to be repeated. This step can
be omitted if the measurements are to be output-referred or if
CW RCVR mode is to be used.
The Measurements
At last, the DUT can be connected between the output of the
combining network and port 2 and the desired variable
measured. Aside from selecting the input/output referral and
the intercept vs. product, the user must also select the
distortion relative to Tone 1 or Tone 2. If the tone amplitudes
are equal, these results will often not be too different but
may be for certain DUTs. When distortion relative to Tone 1
is selected, the upper IMD sideband is measured (and lower
sideband for measurements relative to Tone 2). Some
example measurements are shown below.
Figure 7. An example illustrating an amplifier is IMD product as a
function of frequency is shown in this plot. Tone 1 is swept from
20 MHz to 3 GHz with tone 2 always being 200 kHz higher (in this
case), the system receiver measures both tone and the two IMD
products at each sweep point (201 in this case). The product relative
to tone 1 output amplitude reaches a minimum of about –65 dBc at
higher frequencies.
The plot in Figure 7 is a swept version of the measurement
of a DUT similar to that used in Figure 6. The offset is again
200 kHz but now the tones are swept from 20 MHz to
3 GHz. This measurement methodology is much faster than
any spectrum analyzer equivalent when multiple frequencies
must be measured. In this plot, the amplitude of the IMD
product relative to the tone 1 output amplitude is displayed.
Note that in ALL SWEPT IMD measure-ments, the
frequency displayed along the bottom axis of the graph is
Figure 6. A CW RCVR mode example measurement for a test DUT. that of TONE 1.
The tones in this case are fixed at 890 MHz and 890.2 MHz while the
receiver sweeps a small swath of frequencies around the two IMD
products and around the main tones.
6
Appendix: Offset Frequency
Selection
The MS462xx is a Vector Network Measurement System
designed to make a wide variety of measurements necessary
of RF components and subsystem manufacturers and as such,
its receiver architecture is a bit different from that used in the
typical IMD measurement: the spectrum analyzer. The
MS462xx is much more flexible and much faster
(particularly in swept measurement) than a spectrum
analyzer but does have a few image responses that may affect
an IMD measurement but not the other system
measurements. Because multiple tones are obviously present
during an IMD measurement, it is important that one of the
main tones not land on an image response while the system
is measuring a much smaller IMD product. For this reason,
Figure 8. A swept third-order-intercept plot illustrating the absolute certain offset frequencies are to be avoided. The main
power measurement required by many DUT manufacturers and offenders include 125 kHz, 78.125 kHz, 156.25 kHz, and
customers. The common extrapolation algorithm is discussed in the 250 kHz. Certain mix products of these frequencies may
text. As in Figure 7, the tones are swept and the receiver performs occasionally present a problem if the IMD products to be
four measurements at each point. observed are very small.
The plot in Figure 8 is a swept third order intercept plot A simple test is possible to determine if a desired offset
based on a DUT similar to that used in the other examples. frequency is a problem. Assemble the combining network
As in Figure 7, the tones are swept, but this time the and set the system up for the required IMD product
intercept is computed using the algorithm discussed earlier. measurement. Connect a thru line from the combining
Since a receiver cal was performed for this measurement, the network to port 2 and observe the product. If this level is
vertical scale reference is actually in dBm. At the marker sufficiently below the measurement needs, then the desired
frequency, the output referred IP3 of this DUT is about offset is certainly acceptable. If the organic IMD level is
+24 dBm. higher than acceptable, try changing the offset frequency by
about 5 kHz. If the level drops substantially, then the original
Conclusions offset may present an image problem. If the IMD level does
This application note has described the IMD measurement as not improve substantially, it is likely not an image problem
implemented in the MS462xx Vector Network Measurement and the deficiency may lie in the combining network or other
System including a number of choices that the user must setup details (refer back to the text).
make. These include setup details (source type, frequencies,
swept or CW, power levels, combining network, etc.), the
measurement desired (IMD product or intercept), the
references (measure relative to which tone, relative to
input or output), and the calibrations that can be performed.
Since a great deal of flexibility is afforded, the careful
consideration of these issues should allow the user to
make measurements fitting many manufacturing and
customer demands.
7
All trademarks are registered trademarks
of their respective companies.
Sales Centers: Sales Centers:
US (800) ANRITSU Europe 44 (01582) 433200
Canada (800) ANRITSU Japan 81 (03) 3446-1111
South America 55 (21) 286-9141 Asia-Pacific 65-2822400
Microwave Measurements Division • 490 Jarvis Drive • Morgan Hill, CA 95037-2809
[Link] • FAX (408) 778-0239
September 1998; Rev: B 11410-00213
Data subject to change without notice Intermodulation Distortion Application Note /GIP-E
