At EuMW 2025 in the Netherlands, this demo had no robotic arm.
No giant curved display. No constellation visualization. No AI slogans.
Just a black box labeled Mini-Circuits. A VDI frequency multiplier. A gold waveguide sensor. And a small power meter with a green LED display reading 15.306.
Most people walked past.
The engineers who stopped were the ones who understood what that number represents — and how hard it is to trust it at 115 GHz.
What was on the bench
Mini-Circuits SSG-15G-PC synthesizer — a compact, USB-controlled signal generator covering up to 15 GHz. The software on the laptop showed: Synth Freq 9583.333333 MHz, Synth Power 0.00 dBm, Output Freq 115.000000 GHz.
The arithmetic is direct: 9583.333333 MHz × 12 = 115.000000 GHz.
The ×12 frequency multiplication chain — handled by the VDI frequency multiplier module — converts the synthesizer's 9.583 GHz output to 115 GHz D-band. This is the upconversion architecture used throughout sub-THz research: a low-phase-noise synthesizer at a manageable frequency, multiplied up to the target band.
The laptop software interface — "Telemetry Board Control Panel" — showed two simultaneous control windows:
Attenuator Control: Attenuation 07.50 dB, Frequency 115 GHz, Power Out +11.85 dBm, NDAC 0688 Counts.
Frequency Synthesizer Control: Connected, Source State IDLE, Single Sweep mode selected. The synthesizer output frequency confirmed at 115.000000 GHz.
VDI Erickson PM5B mm-submm power meter — the instrument measuring the output. The green LED display in Image 1 read 15.306 — on the 20 mW range, this corresponds to 15.306 mW, which is 11.85 dBm. Consistent with the software readout.
The PM5B's calibration label on the rear panel: → Unit ID: PM5B Sn871V → Calibrated by: FEB → Calibration date: 7/18/2025 → Date due: 7/18/2026 → Virginia Diodes, Inc.
The calibration was 2 months old at the time of the show. Still valid. Expiring in July 2026.
What the Erickson PM5B actually does
Standard microwave power meters — diode-based sensors, thermocouple sensors — lose sensitivity and accuracy above approximately 50 GHz. The diode junction capacitance limits bandwidth. The calibration transfer becomes unreliable. The measurement uncertainty grows.
The Erickson PM5B is a different class of instrument. It uses a calorimetric measurement principle — the power absorbed by the sensor raises its temperature, and that temperature rise is measured with high precision. This approach is frequency-independent above the thermal time constant of the sensor, making it accurate from millimeter wave through sub-millimeter wave frequencies.
The power ranges visible on the PM5B front panel: 200 μW, 2 mW, 20 mW, 200 mW — a four-decade dynamic range spanning from research-level weak signals to moderate output power levels. At 115 GHz, 15 mW represents a reasonably high power output — one that required the Mini-Circuits synthesizer at 0 dBm to be multiplied up through the ×12 chain and amplified.
The CAL FACTOR dB dial on the front panel allows manual entry of the calibration correction for the specific waveguide sensor being used at the specific frequency. Each waveguide frequency band has a different insertion loss characteristic that must be accounted for to get an accurate absolute power reading. This is not automatic — it requires the operator to know the calibration factor for their specific setup.
At 115 GHz, this factor is non-negligible. A 0.5 dB error in the CAL FACTOR translates directly into a 0.5 dB error in every power measurement — affecting gain calculations, efficiency measurements, and link budget predictions throughout any system that uses this as a reference.
The ×12 multiplication chain — what it costs
The Mini-Circuits SSG-15G-PC synthesizer is a well-characterized source at 9.583 GHz. Its phase noise, spurious level, and output power flatness are known and documented.
When this signal passes through a ×12 frequency multiplier to reach 115 GHz, several things happen that are not free:
Phase noise degrades by 20·log(12) = 21.6 dB. Every multiplication step adds phase noise in proportion to the multiplication factor squared (in power terms). A synthesizer with −100 dBc/Hz phase noise at 10 kHz offset at 9.583 GHz will produce approximately −78 dBc/Hz at 115 GHz — a significant degradation that affects every modulation quality measurement made with this source.
Harmonic content increases. A frequency multiplier generates not only the desired 12th harmonic but also other harmonics (6th, 18th, etc.) that must be filtered. At 115 GHz, the filter structures are waveguide bandpass filters with relatively high insertion loss and finite rejection. Residual harmonics appear as spurious signals in the output spectrum.
Output power varies with input level and temperature. The multiplier's conversion efficiency is not flat — it depends on the drive level from the synthesizer and on the operating temperature of the multiplier stage. A 1 dB change in synthesizer output power does not translate to exactly 1 dB change in 115 GHz output power.
All of these effects are present in the 15.306 mW reading on the PM5B.
The power meter is telling you what is actually coming out of the waveguide port. The software is telling you what the synthesizer was set to generate. The difference between these two views — insertion loss, multiplier efficiency, waveguide transition losses — is what calibration is supposed to account for.
Why the calibration date matters
The sticker on the back of the PM5B said: Calibration date: 7/18/2025. Date due: 7/18/2026.
This is a one-year calibration cycle — standard for a precision measurement instrument used in metrological applications.
What does calibration expiry actually mean?
It means the instrument's response has been verified traceable to a national measurement standard within the stated time period. VDI calibrates their PM5B units against standards maintained at a national metrology institute — NIST in the US, PTB in Germany — which are themselves traceable to the international definition of the watt.
After the calibration expires, the instrument still works. The display still shows a number. But the uncertainty of that number is no longer formally bounded. If the calorimetric sensor has drifted — from thermal cycling, from mechanical stress, from connector wear — the reading may be wrong by an unknown amount.
At 115 GHz, an absolute power uncertainty of ±0.5 dB is considered acceptable for most research measurements. An uncertainty of ±1.5 dB makes gain and efficiency calculations unreliable. An uncertainty of ±3 dB makes the measurement nearly useless as a reference.
The calibration sticker on the back of that power meter is not a bureaucratic formality.
It is the last link in the traceability chain that gives the number on the display any meaning.
The infrastructure argument
This demo represented something that appears in none of the headline presentations at a microwave show.
It represented the measurement infrastructure layer — the tools that exist not to demonstrate a capability, but to verify that other measurements are trustworthy.
The 300 GHz OTA link in Article 049 showed 63 Gbps throughput and EVM numbers. The D-band active load pull in Article 050 showed gain compression curves and EVM vs. Pout. The W-band VNA extender in Article 051 showed antenna radiation patterns.
All of those measurements depend on source stability and power calibration.
If the LO signal feeding the 300 GHz upconverter is drifting because nobody has verified the multiplier chain output power recently — the EVM numbers are unreliable.
If the DUT input power during active load pull is 1 dB higher than intended because the synthesizer output has aged — the gain compression curves are shifted.
If the antenna pattern measurement reference level is wrong because the power meter was out of calibration — every gain number in the pattern is offset by a systematic error.
None of these failures are dramatic. None produce an obvious error message. The measurement appears to work. The software displays numbers. The curves look reasonable.
The error is silent.
Why this becomes harder above 100 GHz
At 10 GHz, there are multiple independent ways to verify source power and frequency. Spectrum analyzers, power meters, frequency counters, and scalar network analyzers all provide cross-checks that help catch drift and uncertainty.
At 115 GHz, the options narrow dramatically.
There is no low-cost spectrum analyzer that covers 115 GHz. The waveguide frequency counter is expensive and rare. The power meter selection is limited to instruments like the Erickson PM5B — precision, expensive, slow to measure.
The redundancy that engineers take for granted at microwave frequencies disappears at millimeter wave.
If the PM5B is the only power reference in the lab, and if it is wrong, there is no independent verification available without sending the instrument back to VDI for recalibration.
This is the infrastructure problem at sub-THz frequencies.
Not the frequency generation. Not the signal processing. Not the antenna.
The ability to know — with bounded uncertainty — what signal actually exists at the measurement reference plane.
The quiet demo that was the foundation
The 15.306 mW reading on the PM5B was not impressive on its own.
It was the ground truth that made every other measurement on the VDI booth credible.
The OTA link at 300 GHz. The D-band active load pull. The W-band antenna characterization.
All of them ultimately traced back to a calibrated power meter, a known synthesizer, and a verified multiplication chain.
The demo that nobody stopped for was the one that made the others meaningful.
All photos: Thomas · @SignalByThomas