The label on the booth said "IMPROVED PERFORMANCE."
The method used to achieve that improvement was the opposite of what you'd expect.
At the microwave show in the Netherlands, the Rohde & Schwarz Wideband Modulated Load Pull demo was not trying to make the amplifier work well.
It was trying to make it work badly.
Deliberately. Systematically. Until the failure boundary was mapped.
That map is what optimization actually requires.
What was on the bench
Left side: R&S RTP real-time oscilloscope/spectrum analyzer — the live display showing a dense colorful waterfall spectrum, red and yellow peaks indicating high-power signal activity across a wideband capture. This was the output monitor — watching what the PA under test was actually producing at every moment of the impedance sweep.
Right side: R&S SMW200A Vector Signal Generator — display reading 1.110 000 000 000 GHz, output level 57.00 dB, timing parameter 22.00 ns. The SMW200A is R&S's flagship vector signal generator, capable of generating complex modulated waveforms — 5G NR, LTE, custom — at the exact parameters needed to stress a power amplifier with realistic traffic signals rather than a simple CW tone.
Center: a white rectangular load pull test fixture — the DUT holder that interfaces the power amplifier to the measurement system while allowing the load impedance to be varied systematically.
The system block diagram on the left screen showed the complete signal chain:
SMW200A → RF1 → Coupler 1 → PA (DUT) → Coupler 2 → RF2
With a and b wave notation at each port — a1 (incident wave at port 1), b1 (reflected wave at port 1), a2, b2 — and an R&S RTP Oscilloscope at the top handling the real-time processing of all four wave quantities simultaneously.
The right screen showed the R&S LoadPull software — a Smith chart displaying "Gamma Measured vs. Gamma Wanted" — the actual impedance points achieved by the tuner overlaid on the target impedance grid. The color gradient from red through yellow to green mapped the power-added efficiency (PAE) contours across the Smith chart plane.
The slide on the left screen carried a notation that caught my eye: "COMPANY RESTRICTED" and dated September 2024 — this was internal R&S application engineering material, presented at the show just weeks after its internal release.
What load pull actually does — and why "wideband modulated" changes everything
Standard load pull is a mature technique. You connect a PA between a controllable source impedance and a controllable load impedance, sweep both through a grid of values on the Smith chart, and measure output power, gain, and efficiency at each point. The result is a set of contour plots — circles of constant PAE, constant output power, constant gain — overlaid on the Smith chart. The intersection of these contours defines the optimal impedance for your target tradeoff.
This works well for CW (continuous wave) signals. A single frequency tone, constant amplitude, no modulation.
Real 5G signals are not CW.
A 5G NR signal has a peak-to-average power ratio (PAPR) of 8–12 dB. The instantaneous power varies dramatically — the amplifier spends most of its time at backoff, with occasional peaks that briefly approach the 1 dB compression point. The efficiency at these backoff power levels is dramatically lower than at peak power.
A CW load pull tells you the impedance that maximizes efficiency at peak power. But peak power is not where the amplifier operates most of the time.
Wideband modulated load pull changes the excitation signal.
Instead of a CW tone, the SMW200A generates a full 5G NR modulated waveform — or LTE, or any custom wideband signal — and the load pull measurement is performed with this realistic signal. The efficiency contours on the Smith chart now reflect the average efficiency under real operating conditions, not the peak-power efficiency under an artificially simple stimulus.
The difference is not academic. The optimal impedance for CW efficiency and the optimal impedance for modulated signal average efficiency are not the same point on the Smith chart. A PA optimized using CW load pull and then deployed in a 5G base station is suboptimal — sometimes significantly.
The "IMPROVED PERFORMANCE" claim on the booth sign was specifically about this: using modulated load pull instead of CW load pull to find the impedance that actually maximizes efficiency in the real operating condition.
The a/b wave measurement — why it matters
The system block diagram showed explicit a and b wave labeling at each coupler port.
This is not incidental notation. It defines the measurement approach.
In a load pull system, you need to know the actual impedance being presented to the DUT — not the impedance you commanded, but what the DUT actually sees. At microwave frequencies, cable and connector reflections, coupler directivity limitations, and mechanical resonances mean the impedance delivered to the DUT differs from what the tuner was set to.
a-wave = incident power wave traveling toward the DUT port b-wave = reflected power wave traveling away from the DUT port
The ratio b/a defines the actual reflection coefficient — the actual load impedance — at the DUT reference plane.
By measuring all four waves (a1, b1, a2, b2) simultaneously with the RTP oscilloscope, the R&S system can correct for the difference between commanded and actual impedance in real time. The "Gamma Measured vs. Gamma Wanted" display on the right screen was showing exactly this correction — the dots scattered around each target point on the Smith chart represent the actual measured impedance at each tuner setting, with the color coding showing where efficiency peaked.
This correction is what transforms a load pull system from approximate to rigorous. Without it, you're assuming your tuner is ideal. It never is.
The colorful waterfall: what the RTP was capturing
The left instrument's live display showed a time-frequency waterfall with vivid red, yellow, and green coloring across a wide bandwidth. This was the PA output spectrum under wideband excitation.
The density and color intensity of the waterfall encode instantaneous power at each frequency and time. In a load pull sweep, the impedance is changing — the PA's operating point shifts — and the output spectrum changes in response. Regions of increased spectral regrowth (the spreading of energy outside the intended channel bandwidth, driven by nonlinearity) appear as broadening of the spectral shoulders.
This is the fingerprint of a PA operating near or into nonlinearity. Under CW excitation, you'd see a clean single tone and its harmonics. Under modulated excitation at various impedance points, the spectral regrowth pattern tells you not just efficiency but linearity — whether the PA is producing distortion products that would violate the adjacent channel leakage ratio (ACLR) specification.
A PA that is efficient but nonlinear at a given impedance is not a useful operating point for 5G. The modulated load pull system captures both dimensions simultaneously.
The SMW200A at 1.110 GHz: why that frequency
The SMW200A display showed a carrier frequency of 1.110 000 000 000 GHz — precisely in the L-band, covering part of the sub-2 GHz spectrum used by LTE and early 5G NR deployments.
This frequency range is relevant for a specific reason in the context of load pull research: L-band PAs are often the reference devices used to develop and validate new load pull methodologies before those methods are extended to higher frequencies. The amplifier physics is well understood, the components are readily available, and the measurement calibration is more tractable than at mmWave.
The 57.00 dB level parameter and 22.00 ns timing visible on the SMW200A display suggest a pulsed or time-gated signal configuration — consistent with load pull measurements that use pulsed excitation to avoid thermal drift during the impedance sweep.
The September 2024 COMPANY RESTRICTED slide
One detail worth noting: the left display slide was marked "COMPANY RESTRICTED, Sept 2024."
This means the content shown at the microwave show was R&S internal application engineering material — not a published product brochure, but a technical characterization slide from the product team. Showing it at the show (even with the "restricted" watermark) signals that R&S was using the conference as a platform to demonstrate capability to customers and partners before formal public documentation.
For engineers attending the show, this was a direct window into R&S's internal understanding of their own system's performance — more technically candid than a marketing datasheet.
What load pull tells you that nothing else can
S-parameters tell you how the device responds to small signals in a linear regime.
A gain compression measurement tells you where the device starts to depart from linearity.
Load pull tells you how every performance parameter — gain, output power, efficiency, linearity — changes as a function of the impedance environment.
And wideband modulated load pull extends this to realistic operating conditions.
The contour map on the Smith chart is not an optimization — it is a characterization. It says: here is the full landscape of how this amplifier behaves. The optimal point depends on what you're optimizing for. Maximum PAE. Maximum output power. Maximum linearity at a given output level. These are different points on the map.
The engineering decision — which point to design for — requires knowing the full landscape first.
That landscape only appears when you deliberately vary the conditions away from the ideal operating point.
Which is to say: the optimization begins exactly where the controlled, comfortable operating conditions end.
And that is why load pull exists.
All photos: Thomas · @SignalByThomas