RF phased array testing measures each element and channel in amplitude and phase so the array's correction tables can steer the beam where the design intends. Element-level work means conducted S-parameter measurement on a network analysis bench: return loss per element, gain and phase per channel, and coupling between neighbors.

The same name also covers phased array ultrasonic testing (PAUT), a non-destructive testing method that steers ultrasound through welds and metal to find flaws. The two share the physics of phased excitation, but little else on the bench. This guide stays with RF fundamentals and RF phased arrays.

RF array test runs at 3 levels:

1. Element and channel level: conducted S-parameters on a network analyzer bench, covered in this guide.
2. Subarray or panel level: multiport or switched measurements, also covered here.
3. Full-array radiated level: near-field or far-field range measurements, the step after the bench.

The work moves from bench to range: element data come from a calibrated analyzer, and steer-angle checks follow over the air.

A typical panel has far more elements than the analyzer has ports. Port count determines what one connection measures.

On the E5071C-2K5, every 4-port job becomes sequential 2-port measurements, with one channel or element pair per connection. The 4K5 or 480 can measure a channel and its neighbors in one connection. The vector network analyzer glossary defines the multiport terms, and the entry on what a network analyzer measures covers ports and receivers.

Past 4 ports, the E5092A configurable multiport test set switches more element ports onto a 4-port E5071C, such as the 4K5, from 50 MHz to 20 GHz. Keysight's data sheet lists E5092A Option 020 insertion loss per switch of up to 11.5 dB for SPDT and 12 dB for SP4T at 18 to 20 GHz, with selected-port load match of 4 dB from 10 to 20 GHz. That is why calibration is done through the switch path. Per-switch stability is 0.008 dB/°C from 12 to 20 GHz, listed as supplemental performance data (SPD) at 23 °C ±3 °C. Drift across a long switched run can show up in the results.

From 10 MHz to 4.38 GHz, the 2K5 and 4K5 are specified at 0.035 deg rms. Phase trace noise is the analyzer's random sweep-to-sweep contribution. It does not include drift or post-calibration uncertainty. Keysight's E5071C data sheet lists phase trace noise for Options 2D5, 2K5, 4D5, and 4K5 at 70 kHz IFBW and maximum output power of the sweep range. The value rises by band.

Every trace-noise value in the table is under 3% of a 6-bit phase step. Narrowing the IFBW lowers trace noise further at the cost of sweep time.

On many element benches, the analyzer is a smaller term in the error budget than the setup around it. Cable flexure, connector repeatability, temperature drift, and calibration can contribute more than the analyzer does. The data sheet's corrected performance assumes less than 1 °C deviation from the calibration temperature.

The starting configuration for an element bench is the Keysight E5071C-2K5, with 2 ports, 300 kHz to 20 GHz coverage, bias tees, and 3.5 mm male test ports. Across its options, the E5071C covers 9 kHz to 20 GHz on 2 or 4 ports. The store sells the certified used E5071C ENA, and each unit's installed options are confirmed on its listing.

On the 2K5 and 4K5, bias tees feed DC through the test ports to active evaluation boards. Keysight lists a maximum bias tee input voltage of ±35 VDC and a maximum current of ±200 mA with no degradation in RF specifications. Fit depends on your fixture and supply needs.

Option 010 time domain is required for gating element return loss, and Option 008 frequency offset mode supports channels that convert frequency.

By the band letters the European Space Agency uses, Ku-band spans 12 to 18 GHz and Ka-band 26 to 40 GHz. Ku-band satcom testing therefore sits inside the 2K5 and 4K5 range. The 20 GHz ceiling leaves Ka-band and mmWave phased arrays for a higher-frequency Keysight analyzer, such as an E5080B ENA option that covers the band.

Keysight's page on reasons to buy the E5071C used versus new sets out the cost case. For the configuration, request one of these:

• Development bench, up to Ku-band, one channel at a time: E5071C-2K5.
• 4 channels at once, relative phase without reconnects, or a production station: E5071C-4K5.
• Type-N fixtures, 4 ports, up to 8.5 GHz: E5071C-480, for example, or the E5071C-485 when bias tees are needed.
• Above 20 GHz: an E5080B option that covers the band.

The third decision is the calibration reference plane: where it sits on your fixture. It also includes whether the calibration module you plan to use is listed by Keysight as supported for the E5071C. A plane left at the analyzer port puts cable and fixture variation into the data, then into every correction table built on that data.

Two common paths set the reference plane on a 2K5 or 4K5. A mechanical 3.5 mm SOLT kit connects short, open, load, and thru standards in sequence. Electronic calibration (ECal) replaces that sequence with one module. Keysight's ECal technical overview 5963-3743E (March 2025) lists which ECal modules each analyzer supports and from which firmware revision.

For the E5071C, it lists the N4690D, N4691D, N4692D, and N4696D as supported from firmware B.14.03 over USB. The 3.5 mm N4691D matches the 3.5 mm test ports. Compatibility is model-specific, so check the table before ordering. The 2.4 mm N4693D, which reaches 50 GHz, and the N4694D are not listed for the E5071C, while the same table lists both for E5080B-class analyzers. Keysight states that a full calibration "can be accomplished with a single connection" to the module, which reduces the chance of operator error and the wear on connectors.
 

Analyzer calibration is separate from phased array antenna calibration, which builds the array's own per-element correction tables, often in situ. As of October 2026, IEEE P3582, a recommended practice for calibrating digital phased arrays, is an active project (PAR authorized March 27, 2025) rather than a published standard, so confirm its status before citing it in a test plan. Keysight calibration services cover the analyzer itself, and the calibration certificates guide explains what the certificate documents.

Before you calibrate an element bench, confirm these:

• Connector type and gender at the element fixture (3.5 mm on the 2K5 and 4K5 ports).
• Analyzer firmware revision (B.14.03 or later for N4691D support).
• Temperature stability within 1 °C of the planned calibration temperature.
• Reference plane location: at the element connector, not at the analyzer port.
• A verification standard measured after calibration and before element data is taken.

Check the store for current ECal module listings, and ask on the quote whether a supported 3.5 mm module is available alongside the analyzer.
 

Element S11 that shifts when someone stands near the fixture typically points to reflections from nearby objects. Time-domain gating keeps the antenna's own response and drops reflections from its surroundings. Keysight Application Note 5989-9805 walks through gating on the 20 GHz E5071C. Option 010 time domain is required.The procedure:

1. Set the measurement conditions, run a full 1-port calibration on the port, connect the element, and measure S11.
2. Turn on the time-domain transform, choose low-pass impulse, and run Set Freq Low Pass. It changes the start frequency, so recalibrate.
3. Set a bandpass gate around the element's reflection. The note's example uses a 1.7 ns center and a 1 ns span, which keeps the response within 150 mm of the antenna.
4. Turn the transform off. The gate stays applied to the frequency-domain trace.

In the note's example, ripples caused by an operator's hand near the antenna disappear from the gated trace, and “a 2 ­ 3 dB peak margin can be obtained by removing ripples,” which allows smaller guard bands.

Gate center and span depend on fixture geometry and cable length, so set them from your own trace. The gate also leaves systematic error terms such as directivity to the calibration.

A 256-element panel with 64 states needs 16,384 sweeps. At 12 ms each (11 to 12 GHz, 201 points, 500 kHz IFBW), that's 196.6 s, about 3.3 minutes of sweep time, with about 16 s of data transfer on top. A full 300 kHz to 20 GHz sweep takes 35 ms under the same conditions (Table 40), which makes about 9.6 minutes for 16,384 sweeps. Sweep time is the small part: reconnecting, switching, and settling can take longer, which is where the port-count choice comes back in.

Keysight's supplemental throughput data for the 2K5 and 4K5 lists 12 ms per cycle with 2-port calibration and 24 ms with 4-port calibration (data sheet 5989-5479EN, Table 40). Data transfer times are also supplemental (Table 46). Each row counts one sweep per element and state. Trace-noise values are specified at 70 kHz IFBW, while sweep times are specified at 500 kHz IFBW (Tables 21 and 40).

Automation runs over SCPI on LAN, USB or GPIB, with a handler I/O port for production. Application Note 5989-9805 states that the 20 GHz ENA provides "the same user interface and programming commands" as the 4.5 GHz and 8.5 GHz options, so software written for those options can be leveraged on the 20 GHz unit.

A 2K5 or 4K5 bench covers element and channel S-parameters, state tables, coupling, and active-board checks with bias applied. Beam patterns, gain, and effective isotropic radiated power (EIRP) need a range. EIRP and over-the-air calibration are measured afterward in near-field or far-field setups, such as anechoic chambers and scanners.

A 2020 NIST study of a dual-beam 28 GHz phased-array channel sounder found that over-the-air corrections are steer-angle specific. Angle-specific pre-distortion filters achieved a peak-to-sidelobe ratio of up to 56.5 dB in the impulse response, and applying a filter designed for one steer angle to other angles degraded performance by up to 8.6 dB.

As of October 2026, IEEE lists IEEE Std 149-2021, the recommended practice for antenna measurements, as an active standard. Published in February 2022 as a revision of the 1977 edition, it covers test facility design, instrumentation, and range evaluation and operation. Confirm the applicable revision before writing a new range procedure.

For the step after the bench, see how to measure antenna gain using a network analyzer. The antenna gain calculator and the EIRP calculator help check range results against the link budget.

"The E5071C is end of life. Why build a new array bench on it?" Discontinued changes how you buy and support the analyzer, not how it measures. Keysight's data sheet gives May 31, 2021, as the last order date for the E5071C and recommends the E5080B for new purchases. The ECal technical overview, revised in March 2025, still lists ECal modules supported on the E5071C. Application Note 5989-9805, published December 5, 2017, explains that RF models of vector network analyzers "such as the E5071C ENA (4.5 GHz/8.5 GHz)" are widely used in the design process and on the production line to measure return loss or VSWR for commercial antennas.

A certified used E5071C is sold direct from Keysight and listed with its calibration status, so you know the analyzer's condition before you run your own bench calibration. Keysight Premium Used equipment is refurbished by Keysight and updated to the latest compatible firmware and software, with the same warranty as a new instrument; terms vary by model and region. Each unit goes through a performance verification and a final 101-point quality check. Warranty options run up to 5 years, and KeysightCare adds technical support.

For procurement, the useful document is one quote that states the calibration status, warranty options, and savings, which run across the store up to 90% for Keysight Used and up to 70% for Keysight Premium Used. Keysight reports that 78% of its customers lose more than $25,000 a day when equipment fails, so support terms belong on that quote. The network analyzer buying guide covers the wider category decision.

A complete configuration line saves a round trip. Put these 9 items on the request:

1. Array band and the E5071C option that covers it: 2K5, 4K5, or 480.
2. Ports needed in one connection (2 or 4), or a switched E5092A setup on a 4-port unit.
3. Option 010 time domain for gating element return loss.
4. Calibration path: a 3.5 mm SOLT kit or a supported ECal module, such as the N4691D, if available with the unit.
5. Firmware at B.14.03 or later, if the quote includes an N4691D.
6. Bias needs for active boards, within the ±35 VDC maximum and ±200 mA maximum bias current.
7. Option 008 frequency offset, if the path converts frequency.
8. Calibration status and certificate needed on delivery.
9. Warranty option and KeysightCare coverage.

Send it through the certified used network analyzers page. One decision remains design-specific: define how your array's unmeasured elements are terminated during coupling measurements before choosing between sequential 2-port work and 4 ports at once.