Application Note 97

January 2006
Accurate Measurement of LT5514 Third Order
Intermodulation Products
Dorin Seremeta

INTRODUCTION
Accurate measurement of the third order intercept point
for low distortion IC products such as the LT5514 requires
certain precautions to be observed in the test setup and
testing procedure. The LT5514 linearity performance is
high enough to push the test equipment and test set-up to
their limits. A method for accurate measurement of third
order intermodulation products, IM3, with standard test
equipment is outlined below.
It is also important to correctly interpret the LT5514
specification with respect to ROUT, and the impact of
demo-board transmission-line termination loss when
evaluating the linearity performance, as explained in the
LT5514 Datasheet and in Note 1 of this document.
EVALUATION OF LT5514 LINEARITY PERFORMANCE.
The difficulty in providing an adequate low distortion, twotone test signal (Note 2) and failure to adjust the spectrum
analyzer for maximum resolution (Note 3) are the most
common issues when evaluating the LT5514. Below is a
description of a test set-up and a method for evaluating
high linearity components using standard test equipment.
This method makes it easy to identify which section of the
set-up limits the measurement, and to optimize the set-up

SIGNAL
GENERATOR #1

10dB PAD

for the maximum resolution achievable with the available

test equipment.
The test equipment required for LT5514 demo board
evaluation consists of two signal generators, a 2-way
power combiner, a spectrum analyzer, and fixed attenuator pads, as shown in Figures 1 and 2.
Two steps are required for the evaluation of LT5514
intermodulation products.
1. Evaluate and Optimize Test-Setup Performance
Begin by combining the signal outputs of two signal
generators (more information is given in Note 2), and
connect the resulting two-tone output directly to a spectrum analyzer as depicted in Figure 1. An example initial
setting for the spectrum analyzer is: ATTEN = 20dB, SPAN
= 100Hz, Resolution Bandwidth = 3Hz to 10Hz, REFERENCE LEVEL = 20dBm, ATTENUATION AUTO (OR
MANUAL, 10dB). Each frequency of interest should be
inspected, one at a time, with the spectrum analyzer at
minimum span. (Spectrum analyzer settings for maximum resolution are discussed in Note 3). The two-tone
test signal should be set to a power level corresponding to
the desired input level for the LT5514 (23dBm for optimum maximum output). The signal displayed on the
, LTC and LT are registered trademarks of Linear Technology Corporation.
All other trademarks are the property of their respective owners.

SPECTRUM
ANALYZER
2-WAY
COMBINER
AN97 F01

SIGNAL
GENERATOR #2

10dB PAD

Figure 1. Basic Set-Up for Two-Tone Signal Source IM3 Performance Verification Measurements
an97f

AN97-1

Application Note 97
spectrum analyzer should be inspected for:
a) Intermodulation products. Adjust the spectrum analyzer to obtain the maximum possible resolution
(intermodulation product amplitude in dBc relative to
tone power).
b) Second and third harmonic content (which will determine if a lowpass filter is required at the combined
signal generator output).
If the observed intermodulation products from the test
setup are of comparable size to those expected from the
LT5514, then an accurate characterization of the LT5514
is not possible. Likewise, if the observed harmonic content
is significant, then it must be reduced in order to avoid
generating spurious responses during the LT5514 characterization. Depending on the available test equipment,
an adjustment of the spectrum analyzer settings and the
signal combiner may be required in order to obtain satisfactory results. The intermodulation product level may be
limited by the two-tone signal source or by the spectrum
analyzer resolution. The insertion of a small attenuation
pad (as described in Note 4) can be used to find which
section of the setup limits the intermodulation product
level. No results better than this initial measurement can
be expected from the LT5514 demo board.
2. Evaluate the LT5514 Demo Board Performance
Remove matching resistor R7 (255) from the LT5514
demo board when no transmission lines are expected in
the LT5514 application (Note 1)
Program the LT5514 for maximum gain
Measure the LT5514 demo board gain

Insert at the LT5514 demo board output (without

cables) a combination of attenuator pads, so that the
net attenuation cancels the LT5514 board gain. The net
gain (LT5514 + attenuator) should be adjusted close to
0dB, so that the spectrum analyzer input level (and
settings) from Step 1 are preserved
Insert the LT5514 demo board with output attenuator
between the two-tone source and the spectrum analyzer (Figure 2)
Evaluate the magnitude of the LT5514 intermodulation
products
Even when the input signal quality is marginally adequate,
the LT5514 performance can still be evaluated by inspecting the relative degradation of intermodulation products
between Step 1 and Step 2.
When testing the LT5514s gain steps, the two-tone test
signal level should be increased by an amount equal to the
LT5514s attenuation relative to its maximum gain. For
example, if the LT5514 gain is decreased 10dB, then the
two-tone test signal should be increased by 10dB, correspondingly. No change is required to the Figure 2 test
setup. However, in order to maintain the same spectrum
analyzer settings, the Figure 1 test setup must be modified: An attenuator pad equal to the increase in the twotone signal amplitude (10dB in this example) should be
inserted in series with the spectrum analyzer input. This
pad is not required in the Figure 2 test setup.
Linearity testing at or near the LT5514s minimum gain
setting (when the LT5514s IIP3 is very high) is particularly demanding of two-tone signal source quality (see
Note 2). Repeating step 1 may help to explain seemingly
unsatisfactory results.
An OIP3 measurement example is given in Note 5.

SIGNAL
GENERATOR #1

10dB PAD

SPECTRUM
ANALYZER

LT5514
DEMO-BOARD
2-WAY
COMBINER

ATTEN
DUT
AN97 F02

SIGNAL
GENERATOR #2

10dB PAD

0dB NET GAIN

Figure 2. Basic Set-Up for LT5514 Two-Tone IM3 Measurements

an97f

AN97-2

Application Note 97
NOTE 1. The Demo-Board Output Linearity and the
LT5514 Device Linearity

50, and is matched to a 50 transmission line (R7 =

255 in Figure 4 legend). Resistor R7 provides a transmission line back termination, but dissipates almost half
of the LT5514s output power. This situation is appropriate for applications where transmission lines must be
driven.

The LT5514 demo board shown in Figure 3, as shipped,

will drive 50 transmission lines, while providing proper
line termination. Figure 4 (reproduced from the LT5514
Datasheet Figure 12) plots the LT5514 demo board OIP3
vs frequency for different measurement conditions. As
shown in Figure 4, the LT5514 demo board output linearity
as shipped (bottom thick line) is about 6dB lower than the
LT5514 device performance specified in the datasheet for
ROUT = 200 (top thin line). The transmission line termination loss (due to R7) accounts for 3dB. The lower
impedance seen by the LT5514 differential outputs (ROUT
= 100 instead of 200) accounts for another 3dB. This
explanation is elaborated in the following paragraphs.

Case 2: The LT5514 demo board output is connected to the

load without a transmission line, and the board output
impedance is high (R7 = open in legend). This situation is
appropriate for an application where there is a local 50
load, and no transmission line back termination is required. No power is lost in the transmission lines back
termination in this case.
It is important to note that the differential impedance
presented to the LT5514 outputs (ROUT) is different for
these two conditions. For R7 = open, ROUT = 200 (50
board impedance multiplied by the 4:1 impedance transformation ratio of the on-board transformer). For R7 =
255, this matching resistor in parallel with the 200

LT5514 OIP3 performance is plotted in Figure 4 for two

different conditions:
Case 1: The LT5514 demo board output impedance is

VCC
C2
0.1F
IF
IN

ENA
1
2
3
4
5
6
7
8
9
10

T1
1:2

TC2-1T

TRANSFORMER DEMO BOARD

C1
0.47F

ENA
ENB
VCC1
VCC2
GND
GND
LT5514
GND
GND
IN+
OUT
IN
OUT+
GND
GND
GND
GND
PGA0
PGA3
PGA1
PGA2

PGA0 PGA1

VOSUP
C4
0.1F

C3
4.7F

IF
OUT

T2
4:1
ROUT
100

R7
255

TC4-1W

RLOAD
50
J2
0

PGA2 PGA3
AN97 F03

Figure 3. Output Transformer Application Board (Simplified Schematic)

58
55
52
OIP3 (dBm)

J1
0

ENB
20
19
18
17
16
15
14
13
12
11

49
46
43
40
37
34

25

50

75 100 125 150 175 200

FREQUENCY (MHz)

DUT ROUT = 100, R7 = 255

BOARD ROUT = 100, R7 = 255
DUT ROUT = 200, R7 = OPEN
BOARD ROUT = 200, R7 = OPEN
AN97 F04

Figure 4. Typical OIP3 for Demo Transformer Board

an97f

AN97-3

Application Note 97
transformed impedance gives an approximate ROUT =
100.

NOTE 2. A Clean Two-Tone Signal Must be Generated

to Drive the LT5514

Thick lines (BOARD in legend) are the OIP3 figures for the
board output power as measured with a spectrum analyzer
for both conditions.

The two most common problems for the two-tone signal

source generation are:

Thin lines (DUT in legend) are the OIP3 figures calculated

for the actual power the LT5514 part delivers at its differential outputs. These curves depict the LT5514 performance for local load applications.
When comparing the BOARD and DUT OIP3 curves for
R7 = open, the difference (about 1dB) is simply the
transformer insertion loss. The OIP3 is highest at ROUT =
200. When comparing the BOARD and DUT OIP3 curves
for R7 = 255, the difference (about 3dB) increases due
to the additional dissipated power in the matching resistor. The OIP3 is reduced when ROUT = 100.
As shipped, the LT5514 demo transformer board is not
optimized for the best linearity performance (bottom OIP3
curve in Figure 4, and is only the starting point for LT5514
evaluation. The LT5514 demo board may be customized
depending on the application requirements for the best
linearity performance.
There are several factors to consider when optimizing the
LT5514 for the highest linearity in a particular application
that may differ from the default LT5514 demo board setup:
The transmission line termination resistor, R7, (and its
associated 3dB power loss) is unnecessary when the
LT5514 drives a local load in the application.
RLOAD (next stage input impedance) may be increased
above 50, thereby effectively increasing ROUT. This
will increase OIP3.
A bandpass LC network (instead of the transformer
between the LT5514 outputs and the following stage)
has the potential to further improve the application
linearity above 100MHz, when compared to a wideband transformer demo board. The bandpass LC network can provide any impedance transformation ratio
(thereby increasing ROUT), and can also absorb the
LT5514s output capacitance (responsible in part for
linearity degradation at high frequencies).

a. Failure to provide adequate isolation between the two

signal generators and proper wideband impedance
matching.
b. Insufficient filtering of the two input test signals.
The presence of the signal from one generator at the
output of the second generator will modulate the second
generators output. If the unwanted signal is within the
signal generators automatic level control (ALC) bandwidth, the signal generator will try to cancel the interfering signal by modulating the amplitude of the desired
signal. This results in the signal generator no longer
producing a single signal at the required frequency, but
also producing side bands at an offset from this signal.
These side bands are indistinguishable from
intermodulation products generated by the device under
test.
There are a number of different ways in which the two
generator signals may be combined, in order to minimize
these distortion problems:
1. A resistive power combiner is the preferred method for
combining the two signal generator outputs. A resistive
combiner is well matched, and it presents a constant,
power-independent impedance at all three ports. In addition, a resistive combiner is broadband and linear. However, the resistive combiner offers only 6dB isolation
between the two signal generators.
2. Reactive combiners give better isolation (>40dB typically) and lower insertion loss (<4dB typically) but over a
limited frequency range. In theory, this offers >34dB
improvement in isolation over simple resistive combiners.
However, the load VSWR must be very good. Otherwise,
the benefits of the isolation are lost. Therefore, attenuation
pads at the combiner ports are recommended.
When testing at or near maximum gain settings for the
LT5514, these simple methods of combining the signal
generator tones are adequate. However, when testing the
LT5514 at minimum gain, the signal generator levels must

an97f

AN97-4

Application Note 97
increase to about 10dBm. In this case, methods (1) and (2)
fail to provide an adequately clean (low distortion) signal.
3. The best isolation between the two signal generators
can be achieved by buffering the individual signal generator outputs with high linearity amplifiers, prior to driving
a resistive combiner. This method was used to generate
the two-tone test signals for the LT5514 during datasheet
characterization.
A lowpass filter may be inserted between the combiner
output and the LT5514 board input in order to remove
harmonic content from the signal generator output. This
precaution is particularly recommended when signal generators with a high second harmonic output are used.
When the signal generators harmonic level is below
40dBc, the LT5514 intermodulation products can be
tested without this lowpass filter.
NOTE 3. Dynamic Range of the Spectrum Analyzer
Spectrum analyzers produce significant intermodulation
products if operated incorrectly.
As a rule, spectrum analyzers are designed to operate at
their best with about 30dBm to 40dBm at the analyzer
input. Normally the only device in front of the analyzer is
the instruments input attenuator. Therefore, with tone
levels of 10dBm, the (total) input attenuation should be set
to about 40dB to 50dB. As the spectrum analyzer attenuation level is increased, the instruments noise floor rises
and potentially hides any intermodulation products.
Attempting to display both the high side and low side
intermodulation products simultaneously is the most common problem in making use of the full resolution potential
of the spectrum analyzer. For example, a spectrum analyzer span setting of 4MHz is required to observe the
intermodulation products with 1MHz tone spacing. This
results in unnecessary elevation of the displayed noise
floor and reduced resolution.
The spectrum analyzer measurement span and resolution
bandwidth must be reduced in order to lower the displayed
noise floor, and to increase resolution. The distortion
products of the LT5514 should be measured by setting the
spectrum analyzer center frequency to that of each individual intermodulation product and tone frequency, one at
a time. The spectrum analyzer span is set to a minimum
value (100Hz) and resolution bandwidth is set to 10Hz. For

high quality spectrum analyzers, this will be sufficient to

exclude the main tones during measurement of the
intermodulation products. For some spectrum analyzers,
however, the main tones may not be displayed, but their
high level may still degrade the performance of the spectrum analyzer. Be careful.
The dynamic range of a typical spectrum analyzer is 70dB.
The best spectrum analyzers can accurately measure
intermodulation products at about 80dBc. This sets a
limit on the LT5514s minimum output power for accurate
intermodulation product measurements. For example, the
LT5514 specification at 140MHz is OIP3 = 45dBm. When
a 70dB dynamic range spectrum analyzer is used, the
minimum LT5514 output power can be calculated as
Ptone(dBm) = OIP3(dBm) IMD3(dBc)/2 = 45 70/2 =
10dBm. The performance at lower LT5514 power levels
may be extrapolated from high power level measurements
(IMD3 improves with twice the power difference, assuming constant OIP3).
At 50MHz, accurate measurement of the LT5514s OIP3
(54dBm) with this direct method is simply not practical
with a 70dB dynamic range spectrum analyzer.
For datasheet characterization and for production testing
of the LT5514, two different methods are used to measure
the third order nonlinearity. Both methods are based on
reduction of the main tone amplitude at the spectrum
analyzer input, thereby rendering the intermodulation
products more easily detectable by the spectrum analyzer.
1. The first method uses two bandpass filters centered on
the frequencies of the 3rd order intermodulation products. These filters provide at least 20dB attenuation of
the main tones. The principle limitation of this method
is that the tone spacing is limited by the filter out-ofband roll-off, and cannot be lower than 20% to 30% of
the center frequency. Availability and insertion loss
calibration of these filters are additional difficulties of
this method. A lowpass filter for the IMD3 product and
a highpass filter for the IMD3+ product may allow
reduced tone spacing if the resulting out-of-band rolloff is sharper than that of a bandpass filter.
2. The LT5514 datasheet characterization is based on
wide band cancellation of the main signal tones at the
spectrum analyzer input. This has the advantage that
very low tone spacing can be used, limited only by the
an97f

AN97-5

Application Note 97
spectrum analyzer filter. This method uses a second,
low distortion reference signal path that is placed in
parallel with the tested LT5514 device path. The outputs
of these two paths are subtracted, one from the other,
at the input to the spectrum analyzer. The net gain and
electrical delay of the two paths are tuned such that the
main tone level presented to the spectrum analyzer is
reduced by about 40dB typically, while the tested LT5514
output signal intermodulation products are not altered.
This method is most accurate, but requires a large
amount of hardware. The calibration procedure for this
method is also cumbersome, particularly when all gain
settings are to be tested.

NOTE 5. OIP3 Measurement Example

The example test set-up from Figures 1 and 2 consists of:
Signal Sources: HP8644B
Combiner: Mini-Circuits ZFSC2-2 (25dB isolation,
3.5dB Insertion loss)
Combiner input attenuation pads: 10 dB
Spectrum Analyzer: HP8594E (ATTEN = 10dB, Span =
1kHz, RBW = 300Hz, VBW = 300Hz)
This example describes the steps for an actual two tone (at
100MHz) LT5514 Demo Board OIP3 measurement at
MaxGain and at 21dB Attenuation Step.

NOTE 4. Check the Dynamic Range of the Spectrum

Analyzer

The Max-Gain measurement will start with signal source

qualification (step 1) as follows:

As a quick check, if the intermodulation products displayed on the spectrum analyzer are real, and not artifacts
of the test setup, then their amplitude relative to the main
tone level will not be affected by varying the input attenuation on the analyzer (typically in 10dB steps). External
attenuator pads (3dB or 6dB) can be used for smaller level
variation for this test.

Connect instruments for the test setup depicted in

Figure 1

If the IMD product levels relative to the main tone level

(dBc) do change when the input attenuation is changed,
then the spectrum analyzer is generating internal distortion products. That means that the spectrum analyzer
input level is too high to make a valid measurement and
more input attenuation is required.
If the IMD (dBc) readings are constant, then the spectrum
analyzer resolution is adequate. Adjustments of the twotone combining method (Note 2) may be required when
the two-tone source is applied directly to the spectrum
analyzer input (setup calibration step 1).
Whenever the spectrum analyzers input attenuator is
based on pin-diode attenuators, the spectrum analyzer
attenuation should be set to 0dB, and all required input
attenuation should be implemented with external attenuator pads. This gives the best dynamic range capability for
the spectrum analyzer. There is a simple test to find out if
external attenuators are advantageous. Apply a low distortion, two-tone signal to the spectrum analyzer input, and
compare the intermodulation product levels observed
using only internal attenuation versus those obtained
using only external attenuation.

Set two tone frequencies to 99.9MHz and 100.1MHz

Adjust Signal Source levels for Pwr_tone_src = 23dBm
at combiner output
The combined signal measured 45dBc 2nd harmonic,
64dBc 3rd harmonic. These levels are considered
satisfactory, and no lowpass filter is used
Record the Spectrum Analyzer readings for tone and
intermodulation products in the SOURCE column of the
LT5514 Max-Gain Measurement Table 1
The Max-Gain measurement will continue according to the
Figure 2 test setup (step 2) as follows:
The LT5514 demo board has R7 = 255 resistor
removed and the measured demo board Gain is 29.8dB
The LT5514 demo board drives an Ext_Atten = 30.1dB
attenuator pad (at the LT5514 demo board output
connector)
Record the Spectrum Analyzer levels in the LT5514
column of the LT5514 Max-Gain Measurement Table 1
(bold face fields)
The LT5514 demo board OIP3 calculation formula is:
OIP3 = Pwr_tone_board + (IMD3)/2 = Pwr_tone_sa +
Ext_Atten + |(IMD3)/2|
an97f

AN97-6

Application Note 97
For the previous example, the OIP3 at demo board output
when LT5514 is set at Max-Gain is:
OIP3 = 23.3 + 30.1 + (77.4)/2 = 45.5dBm
Note that the higher (worst) sidetone measurement is
used for this calculation.
The OIP3 measurement for the LT5514 Demo Board at
21dB Attenuation Step follows a similar sequence.
The set-up changes for step 1 only (source qualification,
Figure1) are:
Insert about 21dB pad at the combiner output
Increase by 21dB the signal generator output levels
Fill in the SOURCE column in Table 2
The set-up changes for step 2 only (Demo Board measurement, Figure 2) are:
Set LT5514 PGA0 high and PGA1 = PGA2 = PGA3
inputs low
Remove the 21dB pad from the combiner output in
step 1

Insert the LT5514 Demo Board driving the same

Ext_Atten = 30.1dB attenuator pad used in the MaxGain measurement previously
Fill in the LT5514 column in Table 2
The Spectrum Analyzer measured power levels are given
in Table 2 (in bold face).
Then, for the demo board output when the LT5514 is set
at the 21dB Attenuation Step:
OIP3 = 23.1 + 30.1 + (74.7)/2 44.4 dBm
This calculation is accurate provided the relative difference between the IMD3 levels for step 1 and step 2
(IMD3LT5514 IMD3SOURCE Difference) is high enough.
A cascaded OIP3 system calculation (valid only for this
0dB net gain method) was used to generate the OIP3 error
curve as a function of IMD3 relative difference, shown in
Figure 5. The result of this calculation is plotted there as
the thick top curve, and applies whenever the IMD3SOURCE
(from step 1) is well above the spectrum analyzer noise
floor. Since this may not be the case, Figure 5 presents a
family of OIP3 error curves, where the legend dB value
represents the IMD3SOURCE elevation above the spectrum
analyzer noise floor.

Table 1. LT5514 Max-Gain Measurement

FREQUENCY
(MHz)

PARAMETER

UNIT

Column Suffix:

SOURCE
(STEP1)

LT5514
(STEP2)

_src

_sa

99.7

Pwr_IMD3

dBm

116.5

101.3

99.9

Pwr_tone

dBm

23

23.3

100.1

Pwr_tone

dBm

23

23.3

100.3

Pwr_IMD3+

dBm

116.0

100.7

IMD3

dBc

93.5

78

IMD3+

dBc

93

77.4

UNIT

SOURCE
(STEP1)

LT5514
(STEP2)

IMD3LT5514 IMDSOURCE
DIFFERENCE
15.2dB

15.3dB

Table 2. LT5514 21dB Attenuation Step Measurement

FREQUENCY
(MHz)

PARAMETER

_src

_sa

99.7

Column Suffix:
Pwr_IMD3

dBm

116.5

98.5

99.9

Pwr_tone

dBm

23

23.1

100.1

Pwr_tone

dBm

23

23.1

100.3

Pwr_IMD3+

dBm

114.5

97.8

IMD3

dBc

93.5

75.4

IMD3+

dBc

91.5

74.7

IMD3LT5514 IMD3SOURCE
DIFFERENCE
18dB

16.7dB

an97f

Information furnished by Linear Technology Corporation is believed to be accurate and reliable.

However, no responsibility is assumed for its use. Linear Technology Corporation makes no representation that the interconnection of its circuits as described herein will not infringe on existing patent rights.

AN97-7

Application Note 97
12

40dB
20dB
10dB
6dB
3dB

10

OIP3 ERROR (dBm)

One final word of caution should be noted when using the

Figure 5 OIP3 correction factor. This correction is valid
only when the two-tone source quality is limiting. When
the Spectrum Analyzer is limiting the step 1 intermodulation
product reading, the Figure 5 OIP3 correction factor does
not apply. In this case, we have found that the step 2
intermodulation product readings may be optimistic. The
reason is that comparable levels of LT5514 generated
intermodulation products and Spectrum Analyzer
intermodulation products may have an arbitrary phase
relationship. If these intermodulation products are in
opposite phase, the Spectrum Analyzer readings may be
lower than the actual values (due to intermodulation
product cancellation).

IMD3 ELEVATION ABOVE

SPECTRUM ANALYZER
NOISE FLOOR

8
6
4
2
0

6
8
4
10 12
IMD3 DIFFERENCE (dBc)

14

16

AN97 F05

Figure 5. OIP3 Correction Factor vs Relative Difference Between

IMD3 Levels for Step 1 (Source Only) and Step 2 (Device Under
Test)

an97f

AN97-8

Linear Technology Corporation

LT 0206 PRINTED IN USA

1630 McCarthy Blvd., Milpitas, CA 95035-7417

(408) 432-1900

FAX: (408) 434-0507 www.linear.com

LINEAR TECHNOLOGY CORPORATION 2006