PNA-X Training

Created: Sep 19, 2014

[email protected]
[email protected]
[email protected]

Abstract: These notes are from the Keysight N5242A PNA-X (aka Vector Network Analyzer)
training[10]. The notes assume basic familiarity with a network analyzer and the material in
recommended reading below. The notes supplement the comprehensive User Reference
Manual.

Recommended reading:
●​ Network Analyzer Basics
●​ High-Accuracy Noise Figure Measurements Using the PNA-X Series Network Analyzer

Table of Contents
Table of Contents
Connectors and cables
PNA hardware
Receiver
Biasing
Pre-Calibration setup
Normal Calibration (not noise figure cal)
Save
Noise Figure Measurement and Calibration
In-fixture calibration
Line-Reflect-Line (LRL) calibration [3]
GGB Picoprobe calibration
Power Calibration Method
Power Calibration Method [continued]
Intermodulation measurement
IM Spectrum
Power sweep (Pout vs Pin single tone, single frequency)
Harmonic Distortion
Gain compression
Markers
Display
Memory
Balanced (differential) measurements [11]
Load pull (Phase Control)
Troubleshooting
References
Connectors and cables
Avoid unnecessary adapters, connectors, cables as much as possible by having the correct
gender cables. Extra components introduce impairments that degrade the fundamental
accuracy and repeatability of measurement. Calibration helps but only to the degree everything
is constant.

SMA and 3.5 mm connectors are compatible but have very different performance. SMA
connectors have very poor high-frequency return loss and are lost cost (~$3). 3.5 mm
connectors are high performance and high cost (~$150). Figure 1 shows the physical
difference. Notice the SMA dielectric has a white plastic dielectric, but the 3.5 mm connector
has an air dielectric and finely machined contacts.

Figure 1: SMA connector (left), 3.5 mm connector (right)

Take care not to damage delicate 3.5 mm connectors and cables by keeping the cable or DUT
stationary when dis/connecting and only rotate the outer nut. If the male pin rotates against the
female sleeve, you will damage the connector. Never spin the cable or DUT to loosen or
tighten a connection. This requires one wrench on the body (stationary) and one wrench on
the outer nut (turning). Always use a torque wrench. Stop tightening as soon as the torque
wrench flexes.
Flexible cables will have varying phase and loss as they flex, but high quality, phase-stable
cables will have much less variation (~$3000). You can check this variation by flexing the cable
while looking at S-parameters (mag and phase) in real-time.

PNA hardware
Figure 2 shows a simplified block diagram of two of the four PNA ports and one of the two
internal sources. Understanding this basic diagram is useful for setting the correct power levels
for your measurement.

Figure 2: Simplified block diagram (only shows 2 ports and basic signal path)

R1 reference receiver to sense port 1 outgoing power

A receiver to sense port 1 incoming power
R2 reference receiver to sense port 2 outgoing power
B receiver to sense port 2 incoming power

All S-parameter measurements are ratios of these powers: S11, S21, S12, S22

Note: 4-port PNA also includes R3, C, R4, D and second source not shown.
Your power sweep should always keep the VNA receivers below compression (and max power
level) and above the noise floor. Your power sweep should also be appropriate for your DUT.

Test point Description Approx value

Source 1 Raw source power -25 to +15 dBm

A, B Raw receiver 0.1dB input compression point -10 dBm

A, B Max receiver input power (damage will occur) +30 dBm

A, B Raw receiver noise floor (NF= ~35dB) -140 dBm/Hz

Directional coupler forward loss (figure 3) 0dB

Directional coupler reverse loss (figure 3) -15 dB

Source 2 to Port 1 coupler (not pictured) -15 dB

Figure 3: Directional coupler loss

Attempts to source more power than can be delivered at the given source attenuator settings
will give “unleveled power” warning, which is not harmful, but actual power level is less than
displayed. You will also get a warning if the receiver is at max input power. Never exceed the
raw max receiver input power (note this is at points A, B not the front panel), which will
damage the instrument. You, however, will not get a warning that the receiver is going into
compression, which occurs at a much lower level. Limiting receiver input power to -15 dBm
should be sufficient for most measurements. Small-signal S-parameter measurements can use
much lower power levels as long as there is sufficient SNR on the traces.
Receiver

Figure 3: Receiver block diagram

Similar to a spectrum analyzer, the IF bandwidth (resolution bandwidth) determines the receiver
noise floor and sweep speed. Lower IF bandwidth gives a lower noise floor but slower sweep
speed. For best accuracy, IF bandwidth of 1kHz (AVERAGING menu) or lower is
recommended, which automatically uses “stepped sweep” mode. During a frequency sweep,
this mode waits for the source and receive LOs to settle before measurement. This mode can
be engaged at higher IF bandwidths through a SWEEP sub-menu. For IMD or swept power
measurements, the IF bandwidth should be set in the SWEPT IMD SETUP sub-menu
(Frequency tab), as the setting the AVERAGING menu has no effect in this mode.

Recommended setting: AVERAGING -> IF bandwidth <= 1kHz

Biasing
The front panel ports are AC coupled and will tolerate DC bias. As shown in figure 2, bias tees
are built into the VNA. These can be accessed on the back panel and do not require any
software switching.

Tip: The VNA takes about 45 minutes to warm up from a cold start to be temperature stable. It
is recommended not to turn power off.
Pre-Calibration setup
Set up your measurement and get it working first before doing any calibration.

MEASURE -> Class -> New Channel

A “channel” is a measurement, such as S-parameters, IMD, compression point. Each channel

has its own independent calibration data and instrument configuration. Part of your the channel
configuration are the key parameters below:

1.​ Source and receive attenuators (POWER)

2.​ Frequency sweep (SWEEP)
3.​ Number of points in sweep (SWEEP)
4.​ IF bandwidth (AVERAGING)

These parameters should be determined and set before calibration. Changing any of these
parameters post-cal will invalidate the calibration.

Set attenuators to manual to avoid auto-switching during your sweep.

Recommended setting: POWER -> Power & Attenuators -> uncheck auto

In manual mode, you will get out of range warnings if the manual attenuator setting does not
cover the power you are trying to output, which is better than quietly going out of cal in auto
mode. Manually set the attenuators to cover the power ranges appropriate for the DUT (source
attenuators) and VNA receivers (receiver attenuators). Receive attenuators are on a sub-menu:

POWER -> Power & Attenuators -> Receive attenuators

If you have an e-cal module inserted into the jumper location on the front panel (see Noise
Figure Calibration below), then you must switch it out using the Path Configuration:

Required setting: POWER -> Power & Attenuators -> Path Configuration
Click “Noise Path Config” tab. Set port 1 noise tune: internal

According to the AE, this is a bug. PRESET erroneously defaults to noise figure path
configuration, which routes the signal through the source-out → cplr-thru connectors at the front
panel. These ports are only used for noise figure measurements with e-cal source pull (see
noise figure measurement below). In all other measurements these ports should be bypassed
internally. You won’t notice if the jumper is in place, but you will get another ~2dB loss if the
e-cal module is inserted, or worse you could damage the e-cal module with high power levels if
you do not bypass it with INTERNAL setting.
TIP: Enable on screen virtual keys (mouse clickable) to avoid using front panel buttons
Right Click on bottom of screen -> Keys

TIP: Each trace has its own Y-axis scaling which can be confusing. To make the Y-axis the
same for all traces in the same window SCALE -> More -> Scale Coupling -> Select Window
Normal Calibration (not noise figure cal)
Once the measurement is set up, you can refine accuracy with a calibration. Out of the box, the
PNA is roughly calibrated to the front panel ports at PRESET settings. An uncalibrated
measurement won’t account for cable loss and will have telltale ripples in the traces due to
reflections at the various connector discontinuities and signal leakage through the couplers (i.e.
directivity), such as shown in figure 4. You may need to increase the frequency sweep range to
see the repeated ripples.

Figure 4: (1) Uncalibrated S11 vs (2) calibrated S11

A setup using a connectorized DUT can use an e-cal module. The e-cal module contains
reference impedance elements that can be switched automatically under USB control to do a
full 2-port vector calibration with a single insertion. Post calibration, the measurement
reference plane will be the points where you connected to the e-cal module, such as the ends of
the cables. Do not exceed the max power level below. The E-cal module is not intended for
power amplifier measurements. A manual cal kit should be used for high power levels.

N4690 E-cal module specifications

Preferred input power - 5 dBm

Max input power +10 dBm

CAL -> Start Cal -> Cal Wizard

Select e-cal and follow the prompts.

Select APC 3.5 mm female connectors

Confirm the unique serial number on the e-cal module matches the selected module in the
dialog box. This step is important when you have two e-cal modules attached, such as with the
Noise Figure Calibration below.

S-parameters do not require absolute power calibration, but IMD and compression
measurements require power calibration. Check “Calibrate source and receiver power” when
calibrating for these measurements. You will need a USB controlled power meter. If the power
meter connector is not the same as the DUT, you will need to insert a high-quality 3.5 mm
adapter, which will degrade the measurement slightly.

You can choose to save the calibration as a separate cal file. This is not necessary as the cal
data is part of the channel state, but this can be useful if you want to apply the same cal data to
another measurement (assumes same atten, power, frequency, IF bandwidth).

Tip: The orientation (port A vs port B) does not matter. The PNA is smart enough to figure out
how the e-cal module is connected.

Tip: If you have many different settings (attn, power, frequency, IF bandwidth), you can copy
your channel and create different versions. These can all be calibrated with one e-cal insertion
using “Cal All”.

Tip: If you are calibrated, you will see a message in the status bar at the bottom of the screen
such as ‘C 2-port’ or ‘C 4-port’. If not, it will say ‘No corr’

Tip: At power up, or first power cal, you need to tell the PNA-X to look for the power meter via
USB (default interface is GPIB). This can be found under Source Calibration->Power meter
settings->USB. Try this if you cannot find the power meter model number listed in the pull-down
menus.
Save
You can save all the measurement channels, their calibration data, and trace data, and
complete instrument state:

SAVE -> Data & State

RECALL -> Restore a previous state

Tip: Put your channels in SINGLE mode before saving to save the trace data. If you save state
in Continuous mode, your data trace will be clobbered as soon as you load the state. Single will
make one sweep and hold the trace data.

TRIGGER -> Single

Tip: When writing instrument automation scripts, you can load the save.csa files to do the bulk
of instrument setup and avoid most of the tedious GPIB or SCPI commands.

The .csa file is binary and not human readable. If you want to save the data in a human
readable format, you can save either Touchstone .s?p files or .csv. For non-s-parameter data,
you need to save as .csv. For example, to save all displayed traces into one big .csv file:

SAVE -> Save data as

Scope: Displayed data

Format: Displayed format
Type: .csv
Noise Figure Measurement and Calibration

Create a noise figure measurement channel:

MEASURE -> Class -> New Channel -> Noise figure cold source

This is the one time the noise path configuration should select port 1 noise tune external.

POWER -> Power & Attenuators -> Path Configuration

Select the Noise Path Config tab -> Port 1 noise tune: external

The noise figure calibration is similar to the normal calibration, but requires two e-cal modules
and one power meter.

N4690 E-cal module specifications

Preferred input power - 5 dBm

Max input power +10 dBm

CAL -> Start Cal -> Cal Wizard

Select e-cal and follow the prompts.

Select APC 3.5 mm female connectors

The second e-cal module will replace one of the port 1 front panel jumpers between the
SOURCE-OUT → CPLR-THRU connectors at the front panel. It’s purpose is to do a source
impedance pull and measure DUT noise parameters (Fmin, Gamma_opt, Rn) so the VNA can
correct for the actual source impedance seen by the DUT [2]. It does not need to be connected
with special high-quality cables, but it is recommended to use a barrel and semi-rigid cables
(e.g. .141”) to keep the characteristics stable. Because we are correcting for source pull and
2-port mismatch, VNA cold source NF can be much more accurate than Y-factor spectrum
analyzer NF.

Confirm the unique serial numbers on each of the e-cal module matches the selected module in
the dialog boxes. This step is important to distinguish the source pull and signal path modules.

Check “Calibrate source and receiver power”. You will need a USB controlled power meter. If
the power meter connector is not the same as the DUT, you will need to insert a high-quality 3.5
mm adapter, which will degrade the measurement slightly.

You can choose to save the calibration as a separate cal file. This is not necessary as the cal
data is part of the channel state, but can be useful if you want to apply the same cal data to
another measurement (assumes same atten, power, frequency, IF bandwidth).

Tip: The orientation (port A vs port B) does not matter. The PNA is smart enough to figure out
how the e-cal module is connected.

In-fixture calibration
Non-connectorized DUTs, such as wafer probe or waveguide cannot be calibrated using an
e-cal module. These require manual calibration in-fixture [4]. RF probes come with a calibration
substrate kit that can be used to move the reference plane to the probe tips [7,8]. To move the
reference plane onto the silicon, a custom TRL/LRL calibration structure needs to be fabricated
on your wafer [9].

Line-Reflect-Line (LRL) calibration [3]

LRL is also known as TRL (Thru-Reflect-Line) with non-zero thru, reference plane set by reflect.
If there is not sufficient separation between the launchers, and between the launch and the DUT,
coupling of higher order modes will produce unwanted variations during the error-corrected
measurements. A minimum of two wavelengths is recommended. Three standards are
required:

1.​ 2-port non-zero length THRU (known Z0)

2.​ 1-port SHORT which defines reference plane
3.​ 2-port LINE (known Z0)

The LINE insertion phase requirement is between 20 and 160 degrees with respect to the
THRU.

phase_per_meter = 360deg * Freq_Hz * sqrt(epsilon_eff) / c

c = 3E8 m/s

Ideally, the lines are identical in all other respects. Differences contribute to calibration errors.

GGB Picoprobe calibration

This section describes how to do a calibration to put the measurement reference plane at the
probe tips.

For source power calibration referenced to the probe tip, you need to first characterize the loss
of the probe. If you are only doing relative power S-parameters, you can skip this step.

1.​ Perform a 2-port ecal at the ends of the cables where it will connect to the probe
connector
2.​ Measure the S21 of with both probes landed on a short low-loss thru
3.​ Record the S21 versus frequency over the range of interest. This data will become your
loss table. If you know the thru loss you can optionally account for that.

For example, I measured this data for the GGB model 40A GSG Picoprobe (with SMA
adapters):

Frequency (GHz) GSG-THRU-GSG S21 (dB) Probe loss table (dB)

9 -0.72 -0.36

11 -0.78 -0.39
 13 -0.82 -0.41

15 -0.93 -0.465

17 -0.96 -0.48

If you have all your measurement channels set up, then you can calibrate all of them with one
SOLT cal. Perform the cal as below. Only non-default settings are shown.

CAL -> Start Cal -> Cal All Wizard

Select DUT Connectors and Cal kits

Port 1: Probe, CS-5 100 (GSG) else CS-3 100 (GSSG)
Port 2: Probe, CS-5 100 (GSG) else CS-3 100 (GSSG)

This assumes the cal kit file from GGB has already been imported.
Check Modify Cal (You want SOLT which puts the reference plane at the probe tips)
-> NEXT
Select Thru cal method: Defined thru
Click “Cal type/Stds…”
Select Calibration type: SOLT (not TRL)
-> NEXT

If you have not already connected power meter, connect it to PNA via USB.
Click Power Meter Settings
Select USB Power Meter. It should have identified the USB power meter connected.
Check Sensor Loss Compensation
Edit table and enter half the loss you measured for the probe-thru-probe measurement above.
e.g. -1dB becomes -0.5dB. Note the numbers in this table should be NEGATIVE. Negative
means that the power measured when the meter is connected will be larger than when the DUT
is inserted. The power meter doesn’t see the probe loss, but the DUT will.
Hit down arrow to add more rows to the loss table.
-> NEXT

Follow the prompts through the cal. Insert the power meter at the indicated port directly to the
cable (you won’t see this step if you don’t have any absolute power measurements set up).
The remainder of the SOLT measurements should be done at the probe tips using the GGB
calibration substrate.

Tip: You can use NEXT and BACK buttons to do the cal measurements out of order to reduce
the number of probe landings
Power Calibration Method
[Updated 2016/04/19 by swirhun@ for bismarck/revB1 iP1dB power calibration].

For Bismarck/revB1, revB boards (S/N ...B0009) Paul measured the on-board TRL thru
structure (board B0009) and got the following loss table. This was smoothed (1%) and averaged
(10x) since the source power was at -50dBm for receiver measurements. We could consider
re-doing it with higher power levels to get more accurate loss data.

Since our USB power meter doesn’t have an SMA connector, we actually need an adapter in
front of it, so these numbers should probably be adjusted up by 0.1dB to account for the
adapter. No attempt was made to do this as it seems like a negligible offset in iP1dB
measurements. For example, instead of -1.52dB seen by the DUT relative to the power meter, it
may only be -1.42dB, since the power meter also has some adapter loss in front of it like the
DUT does.

f[ 0] = 1.00 GHz: S21 = -0.80 dB -> S21/2 = -0.40 dB​

f[ 1775] = 1.50 GHz: S21 = -0.85 dB -> S21/2 = -0.43 dB​
f[ 3553] = 2.00 GHz: S21 = -1.06 dB -> S21/2 = -0.53 dB​
f[ 5330] = 2.50 GHz: S21 = -1.27 dB -> S21/2 = -0.64 dB​
f[ 7108] = 3.00 GHz: S21 = -1.50 dB -> S21/2 = -0.75 dB​
f[ 8886] = 3.50 GHz: S21 = -1.78 dB -> S21/2 = -0.89 dB​
f[10664] = 4.00 GHz: S21 = -1.99 dB -> S21/2 = -0.99 dB​
f[12441] = 4.50 GHz: S21 = -2.41 dB -> S21/2 = -1.21 dB​
f[14219] = 5.00 GHz: S21 = -2.86 dB -> S21/2 = -1.43 dB​
f[15997] = 5.50 GHz: S21 = -3.04 dB -> S21/2 = -1.52 dB​
f[17775] = 6.00 GHz: S21 = -3.32 dB -> S21/2 = -1.66 dB​
f[19553] = 6.50 GHz: S21 = -3.56 dB -> S21/2 = -1.78 dB​
f[21330] = 7.00 GHz: S21 = -3.91 dB -> S21/2 = -1.96 dB​
f[23108] = 7.50 GHz: S21 = -4.57 dB -> S21/2 = -2.29 dB​
f[24886] = 8.00 GHz: S21 = -5.99 dB -> S21/2 = -2.99 dB​
f[26664] = 8.50 GHz: S21 = -7.94 dB -> S21/2 = -3.97 dB​
f[28441] = 9.00 GHz: S21 = -10.05 dB -> S21/2 = -5.03 dB​
f[30219] = 9.50 GHz: S21 = -11.16 dB -> S21/2 = -5.58 dB​
f[31997] = 10.00 GHz: S21 = -10.20 dB -> S21/2 = -5.10 dB
Power Calibration Method [continued]
[Updated 2016/04/19 by swirhun@ for bismarck/revB1 iP1dB power calibration].

1.​ There is a “Gain Compression” measurement class available. These instructions do not
apply to that class, and use a “Standard” measurement class instead.
2.​ Measure your THRU to generate the table of thru-loss as shown above
a.​ Use an e-Cal to calibrate to the connectors
b.​ Measure the connector / THRU / connector combination and divide the S21 (in
dB) by two, as shown in the table above.
i.​ This represents the loss from the connector to the DUT, since our TRL
THRU is designed to be exactly twice that length.
3.​ Plug in the USB power meter and get your RF adapters in order. Minimize any adapters
inserted in front of the power meter as they degrade accuracy if loss is unaccounted-for.
4.​ Create two measurement channels and right click on the x-axis -> Setup Sweep to set
the sweep (or use Sweep and Freq menus)
a.​ CH1 sweeps -30dBm to +5dBm with 36 points at 5.6GHz; plot S21
b.​ CH2 sweeps 4GHz to 7GHz with 31 points at -30dBm; plot all S-parameters
5.​ Start with power cal on the swept-power channel:
a.​ Select a trace window that corresponds to that chan, as shown in lower left of
screen
b.​ Cal -> Power Cal -> Source Cal and choose Channel 1; Port 1; tolerance
0.05dBm; 50 max readings
c.​ Under Options:
i.​ Choose “Use a power meter”
ii.​ Check the box for “Calibrate the source at multiple power levels” and set
them to be -10dBm to -30dBm with 0.5dBm step size.
iii.​ Check the box “Calibrate the PNA reference receiver
d.​ Under Power Meter Config:
i.​ Choose your USB power meter
ii.​ Add some settling settings or use the defaults (0.05dBm, 10 readings)
iii.​ Check Use Loss Table and enter in the table [measured above for rev B
PCBs], which is half of the THRU loss after a connectorized e-Cal
measurement of the TRL THRU. The entry method is not intuitive. Use a
combination of Enter, Tab, and Down-Arrow to enter your table.
e.​ Click “Take Cal Sweep”, and since it doesn’t take much time, you may as well do
it for both ports.
6.​ Do a typical TRL calibration at -30dBm for CH1
7.​ Do a typical TRL calibration at -30dBm for CH2
a.​ Make sure to click a trace window corresponding to CH2 first!
If all this worked, you’ll get a gain compression curve and some [small-signal] s-parameters as
shown below. You can put some markers on manually to extract the iP1dB, which is -14.4dBm
in this case, at about 11.9dB of small-signal gain.
Intermodulation measurement
Two-tone intermodulation distortion can be measured vs power, vs frequency, or vs power and
frequency. The two-tone power coupler is only available on port 1, so make sure you are set up
to drive the DUT from port 1 (figure 5).

Figure 5: IMD configuration

MEAS -> Class -> New Channel -> Swept IMD

Select a relatively wide IF bandwidth (~100kHz) for the fundamental tones which are well above
the noise floor. Select a sufficiently narrow IF bandwidth (<1kHz) for the small distortion tones,
such that they are above the noise floor. You can tune the latter by doing a sweep and
observing how noisy the distortion tone power appears. Reduce the distortion tone IF
bandwidth until it is not noise-limited (which slows the sweep down).

Note: There seems to be a bug in IF bandwidth selection. The IF BW setting only works
through the measurement-->IMD setup but not through the Averaging dialog.
Tip: Use a smaller number of points when tuning the sweep to speed it up, and increase
number points after IF bandwidth is finalized.

With the TRACES menu you can plot absolute tone power (fundamental or nth order intermod in
dBm), IIP, relative power (dBc), and others.

To create two tones on port 1, the second source is routed to a combining coupler on port 1
(figure 5). This coupling combiner has an additional 15 dB loss, which reduces the maximum
tone 2 output power. This additional loss reduces the max output power of your sweep.

IM Spectrum
IM Spectrum measurement is a kind of poor man’s spectrum analyzer. It can be used to spot
check things, but is slow and not intended to make precision measurements.

Power sweep (Pout vs Pin single tone, single frequency)

MEAS -> Measurement Class -> Standard
New Channel

(If balanced stimulus, MEAS -> Balanced parameters -> Change topology -> True balance)

SWEEP -> Sweep type -> Power Sweep

POWER -> Power and attenuators

Set up source and receive attenuators appropriately as above

POWER -> Start power, Stop power

FREQ -> CW
Set desired tone frequency

TRACE -> New trace -> Receivers

Activate receiver power to plot (A = port1, B = port2, C = port3, D = port4)
Source power = set to DUT input port
“Source power” is the x-axis of the power sweep plot.

(If balanced stimulus, source power = port 1--differential power out of BAL1 logical port 1)
(If balanced stimulus, source power = port 2--differential power out of BAL2 logical port 2)
Harmonic Distortion
You need to create separate channels for fundamental and each harmonic. For example

Channel 1: FREQ -> fstart to 1*fstop

Channel 2: FREQ -> fstart to 2*fstop
Channel 3: FREQ -> fstart to 3*fstop

Calibrate

If you want to plot the harmonics on the same frequency axis as the fundamental, then you
Frequency Offset firmware option 080, which will allow you to make the receiver frequency a
multiple of the source frequency.

FREQ -> Frequency Offset

Edit receiver and set coupled multipliers as below for each channel

Channel 2: FREQ -> fstart to fstop, frequency offset mode, multiplier = 2

Channel 3: FREQ -> fstart to fstop, frequency offset mode, multiplier = 3

Gain compression
Requires Gain compression firmware option 086. It appears you can only measure S21 gain
compression.

FREQ -> Gain Compression Setup

You can do single frequency gain compression or gain compression vs frequency. For single
frequency gain compression

ANALYSIS -> Compression Analysis

Analysis Frequency: CW = <set single frequency>

Markers
The marker menu has many convenient functions for finding salient trace features. See help
guide for more details [5]. Some examples:
MARKER -> Marker Function -> Marker-> Ref Level
Assign reference level to the point on the marker

MARKER -> Search -> Compression

Find N-dB compression point on a curve

Display
DISPLAY -> Quad
Split screen into four quadrants for different traces, such as S-parameters

Memory
Traces can be saved into a temporary memory buffer for doing math or comparison.

MEMORY -> Data->Memory

MEMORY -> Normalize

Save current trace to memory and display current trace relative to that memory. Normalize is
useful to compare different conditions. Once enabled, all subsequent data captured on that
channel will also be normalized to (i.e. divided by) the memory trace. Be sure to use use
MATH->DATA (or OFF) when you want to go back to absolute measurements.

Balanced (differential) measurements [11]

With a 4-port VNA, any combination of single-ended and differential S-parameters can be
measured. To create logic ports out of one or more physical ports, select Balanced Parameters:

MEAS -> Balanced Parameters -> Change -> Configure

Example: Sds21
S = S-parameter
d = differential logical port 2
s = single-ended logical port 1
2 = logical port 2 (differential in this case)
1 = logical port 1 (single-ended in this case)

The VNA will compute the differential measurements by doing two single-ended measurements
then post processing the data. If you have a device with different common mode and differential
mode responses, such as an amplifier, you need to use Integrtated True Mode Stimulus
Application (ITMSA). ITMSA can be enabled post calibration. The VNA will use data from the
calibration to ensure balanced amplitude and phase at the DUT in ITMSA mode. The DUT will
be stimulated with true differential or common-mode signals.

MEAS -> Balanced parameters -> Change topology -> True balance

For a passive device with identical differential and common-mode responses, you should get the
same S-parameters whether or not True Balance is enabled.

The ITMSA firmware option requires a one-time calibration when it is first used after install. The
VNA must then be rebooted for this feature to work properly.

Load pull (Phase Control)

The VNA won’t generate a load-pull plot of constant power contours. You can, however, sweep
over a range of load impedances and plot the resulting load power.

A power amplifier load pull can be simulated by using Phase Control option 088 [5]. The
second source can emulate a load reflection with a phase and amplitude relative to source 1.
Only certain combinations of ports are supported, such as port 1 / port 3, port 2 / port 4 (not
port 1 / port 2 or port 3 / port 4). The controlled source generates the simulated reflection. The
reference source generates the input stimulus for the DUT. For example, to drive the DUT with
port 1 and load the DUT with port 4 and do a load pull on port 4:

SWEEP -> More -> Phase Control -> Phase control setup

Controlled port: port 4

Referenced to: port 1
Control parameter: R4 / D
Max iterations: 25

The reflection coefficient (effective impedance) looking into the PNA will be R4/D, which is the
simulated reflection generated by controlled source outputting on port 4 divided by the signal
generated by the DUT. Setting the relative phase is an iterative process since it changes the
impedance seen by the DUT. Increasing max iterations increases the chances the VNA can
successfully achieve the desired phase relationship (but may not be possible).

SWEEP -> Sweep Type -> Phase

Set phase sweep

Start phase, stop phase (as desired)
Controlled port: port 4, phase control: R4/D
Reference port: port 1, phase control: reference

Plot effective PNA reflection coefficient (impedance) looking into VNA port 4
TRACE -> New trace
Receivers tab -> Numerator: R4, Denominator: D, source port: port 4
Format: Smith

Adjust the magnitude of reflection coefficient by changing the port 4 power

POWER -> Power and attenuators
Port4 power: -10dBc (set to desired load reflection coefficient magnitude)
Port4 source atten: 10dB (as desired)
Port4 leveling mode: R4/D, Port1
Note the port4 source atten will need to be set such that the source an output the required
“reflection” power. For small reflection coefficients, such as -29dBc, you need more attenuation
to get the source power down. For large reflection coefficients, such as 0dB, you will need less
attenuation.

Plot Pout vs phase

TRACE -> New trace
Receivers tab -> Numerator: R4, Denominator: 1, source port 1

There is no automatic way to get a power contour plot. Instead, you need to manually change
the reflection coefficient magnitude (port4 power in dBc).

Troubleshooting
A simple way to see if the set up is operating correctly is to do a PRESET, then put a thru
between two ports and measure S-parameters. In this uncalibrated state, you expect to see
about 1dB loss per cable for high-quality phase stable 1m cables at 15GHz. For example,
this is how we found that the front panel is configured for noise figure on PRESET, which is an
apparent bug. Another useful tool is the TDR firmware option 010, which can help pinpoint
which reflections in the setup are causing the most error.

References
[1] Network Analyzer Basics, Agilent Technologies.
[2] High-Accuracy Noise Figure Measurements Using the PNA-X Series Network Analyzer,
Agilent Technologies.
[3] Network Analysis Applying the 8510 TRL Calibration for Non-Coaxial Measurements,
Keysight Technologies.
[4] In-Fixture Measurements Using Vector Network Analyzers, Keysight Technologies.
[5] PNA-X User’s Guide [PDF] [HTML], Keysight Technologies. Same as instrument help dialog.
See also Tutorials > App Notes on HTML version.
[6] Joel P. Dunsmore, Handbook of Microwave Component Measurements: with Advanced VNA
Techniques, John Wiley & Sons, Ltd., 2012. Dunsmore is the Agilent architect and chief guru
for VNAs. This book has all the gory details.
[7] CS-5 SOLT Cal kit, GGB Industries.
[8] CS-3 SOLT Cal kit ENA 5071B, GGB Industries.
[9] Network Analysis Applying the 8510 TRL Calibration for Non-Coaxial Measurements,
Keysight Technologies.
[10] Keysight applications engineer: [email protected]
[11] PNA-X Balanced Measurements