The center frequency on the upper display read: 307 GHz.
Span: 13.017 GHz. Resolution bandwidth: 6.841 MHz.
This is not millimeter wave. This is sub-terahertz — a frequency regime where the boundary between RF engineering and optical physics begins to blur, and where almost every measurement technique developed for microwave systems needs to be reconsidered from scratch.
The booth at the microwave show in the Netherlands was Keysight's. The story on the left screen was bigger than the hardware.
The University of Stuttgart connection
The vertical display on the left showed a slide with a clear statement:
"Keysight enabling University of Stuttgart to advance 6G IC research."
This is not a generic claim. The system block diagram visible on the same screen was a specific research architecture — a complete sub-THz transceiver test chain built around Keysight instruments and VDI frequency extenders, designed for characterizing integrated circuits operating in the H-band (220–330 GHz) and beyond.
The University of Stuttgart's Institute for Robust Power Semiconductor Systems and the affiliated research groups working on millimeter-wave and sub-THz IC design are among the leading European academic teams in this space. The fact that Keysight chose to highlight this collaboration at the microwave show — rather than just showing the hardware — says something about where the company sees its differentiation: not in raw specification numbers, but in enabling research that produces the next generation of devices.
The system architecture — what the block diagram showed
The block diagram on the left screen was detailed enough to reconstruct the signal chain:
Transmit path: → Reference source: J11 Src1, f_oe = 8.10 GHz → CCU (Clock Control Unit) distributing the sub clock lock reference → VCA (Variable Carrier Amplifier) stage → LO frequency: f = 70.95 GHz, Px = −35.0 dBm → VNAX in H-band (WR 3.4) — the transmit upconversion stage → Output frequency: f_IF = 280–320 GHz (H-band) → ×8 frequency multiplier chain visible in the diagram
Receive path: → Second VNAX module — H-band (WR 3.4) receive downconversion → LO: f = 70.76 GHz (slightly offset from TX LO — intentional for heterodyne receive) → W-band (WR 10) interface: f = 70–110 GHz → Output: < −25 dBm receive power level
The W-band / H-band interface: The VDI modules on the bench spanned two waveguide bands: → VDI WR 10 (75–110 GHz, W-band) — the lower frequency interface → VDI f = 110 GHz transition module → H-band waveguide (WR 3.4, 220–330 GHz) — the actual sub-THz measurement interface
The gold waveguide components visible on the bench — the small rectangular blocks with precision-machined waveguide flanges — are the H-band hardware. At 300 GHz, the WR 3.4 waveguide aperture is 0.864 mm × 0.432 mm. These are components you handle with tweezers, not fingers.
The headline claim: less than 1% residual EVM
The left screen stated explicitly:
"Achieve less than 1% residual error vector magnitude (EVM) of the test system with up to 40 GHz of modulation bandwidth."
This number requires careful interpretation.
1% EVM corresponds to approximately −40 dB in EVM power terms. For context: → 5G NR 64-QAM requires EVM ≤ 2.5% (−32 dB) → 5G NR 256-QAM requires EVM ≤ 1% (−40 dB) → Next-generation sub-THz links targeting high spectral efficiency will need EVM floors well below 1%
The claim is specifically about the test system's own residual EVM — the measurement floor of the instrument chain itself, not the EVM of the device under test.
This distinction is critical.
When you characterize a sub-THz IC that achieves, say, 3% EVM, you need your test system's own EVM contribution to be significantly lower — ideally 5–10× lower — so that you can attribute the measured impairment to the DUT and not to the measurement setup.
At 40 GHz modulation bandwidth and 300+ GHz carrier frequency, achieving sub-1% system EVM is an extraordinary engineering challenge. It requires:
→ Phase noise performance at the LO that doesn't corrupt the high-order modulation → I/Q imbalance calibration across the full 40 GHz instantaneous bandwidth → Amplitude flatness of the frequency extender chain across the measurement band → Timing alignment between the transmit and receive paths at sub-picosecond precision
The Keysight claim is that this bench achieves that floor. The PNA-X on the right side of the bench — with its live spectrum display showing a clean signal shape centered in the measurement window — was the live validation.
What the PNA-X is doing in this setup
The Keysight PNA-X (Performance Network Analyzer, X-Series) is primarily known as a vector network analyzer. Its presence in a sub-THz modulation quality test setup is not immediately obvious.
The PNA-X has a unique capability: it combines VNA S-parameter measurement with signal source and receiver hardware that can be used for noise figure, gain compression, and — with appropriate software options — modulation quality analysis.
In this configuration, the PNA-X was functioning as the signal source and receiver backbone, with the VDI frequency extenders handling the up/downconversion to sub-THz frequencies. The PNA-X's built-in phase coherence between its source and receiver ports — the same coherence that makes it a precision VNA — is what enables the low-residual-EVM claim.
The display on the PNA-X showed a clearly shaped spectrum occupying approximately 40 GHz of bandwidth — the yellow/green spectral shape characteristic of a filtered wideband modulated signal — centered within the Sub-THz measurement band.
307 GHz: why this specific frequency
The upper display center frequency of 307 GHz places this squarely in the H-band (220–330 GHz), also sometimes called the sub-THz or THz-adjacent band.
The choice of H-band for 6G IC research is driven by several converging factors:
Available spectrum: H-band contains large contiguous spectrum allocations — tens of gigahertz — that simply don't exist at lower frequencies. A single H-band channel can carry 40+ GHz of modulation bandwidth, enabling multi-hundred-gigabit-per-second data rates.
IC technology readiness: Silicon-germanium (SiGe) BiCMOS processes have reached transit frequencies (fT) above 500 GHz, making H-band transceiver ICs feasible in a production-compatible technology. The University of Stuttgart's research specifically targets SiGe IC design for this band.
Atmospheric window: 300 GHz sits within a relatively low-absorption atmospheric window between the 183 GHz water vapor line and the 325 GHz line — making short-range (< 1 km) terrestrial links viable without excessive path loss from molecular absorption.
Imaging and sensing: 300 GHz provides millimeter-scale spatial resolution with penetration through clothing, packaging materials, and some biological tissue — relevant for security screening, industrial inspection, and medical imaging applications beyond communications.
The 40 GHz bandwidth challenge
The "Characterize to 40 GHz BW" bullet point on the booth sign is the most demanding specification on the list.
40 GHz of modulation bandwidth at a 307 GHz carrier corresponds to a fractional bandwidth of approximately 13% — manageable for a narrowband design, but extremely challenging for the waveguide components, frequency extenders, and calibration chain.
Specifically: → The VDI WR 3.4 extenders must maintain flat amplitude and linear phase response across the full 40 GHz measurement window → Any amplitude ripple in the extender chain appears as in-band distortion that degrades the measured EVM → Group delay variation across 40 GHz causes inter-symbol interference that is indistinguishable from DUT impairments unless carefully characterized and de-embedded
This is why the "Increase EVM confidence" bullet is listed first — before customizing modulation formats or achieving broad bandwidth. You cannot trust any other measurement result until you understand and characterize your instrument chain's own imperfections.
The calibration methodology for a sub-THz modulation quality measurement is itself an active research topic. There is no established standard — IEEE, ETSI, and 3GPP are all in early stages of defining what "EVM measurement" means at these frequencies and bandwidths.
What "6G IC research" means in practice
The University of Stuttgart collaboration visible on the booth display represents a specific type of research infrastructure challenge.
Designing a sub-THz IC is one problem. Verifying that it performs as designed is a different and arguably harder problem.
A 300 GHz SiGe chip comes out of the fab as a few square millimeters of silicon. To measure its RF performance, you need:
→ A probe station capable of making electrical contact to pads spaced at sub-100 μm pitch → On-wafer sub-THz probes — mechanical structures with WR 3.4 waveguide outputs, aligned to micrometer precision under a microscope → A source and analyzer chain with enough dynamic range to characterize the chip's gain, noise figure, and modulation quality in a single measurement session
The Keysight + VDI system in this demo is the analyzer half of that chain. The MWC MR19S visible in the block diagram on the left screen is a specific millimeter-wave controller module in the chain.
Building this infrastructure requires significant capital investment — the instrument stack visible on the bench represents well over €500,000 in equipment, before probes, calibration standards, and the wafer probing station.
This is why the collaboration model matters: the University of Stuttgart gets access to instrument capability that no academic budget could purchase independently, and Keysight gets early access to the IC performance data that will define the requirements for the next generation of test equipment.
The live spectrum at 307 GHz
The upper display showed a live spectrum capture: a flat-topped, filtered signal shape occupying the center of a 13.017 GHz span. The spectral rolloff at both edges was visible — the characteristic shape of a Root Raised Cosine filtered modulated signal.
Below it, the modulation quality analysis panel showed the decoded bit sequence — a dense block of hexadecimal data representing decoded symbols — and an EVM summary table.
The green signal on the spectrum display was clean — no obvious spurious emissions, no visible phase noise skirt at the carrier, no amplitude ripple across the occupied bandwidth.
That is the starting point. The instrument floor. The reference.
The actual SiGe ICs from Stuttgart will show more impairment than this.
The question the research is trying to answer: exactly how much more, at which modulation order, at which power level, across which temperature range.
Those answers don't exist yet in any published literature for H-band ICs at 40 GHz modulation bandwidth.
This bench is how you get them.
All photos: Thomas · @SignalByThomas