Two booths. Different companies. Different frequency ranges. Different technologies.
But standing in front of both of them at the microwave show in the Netherlands, one observation kept coming back:
Nobody owns the whole system.
MACOM: the component layer
The MACOM booth ran a long wall of live demos. Blue and white. Dense with hardware.
The headline numbers were hard to miss.
Left side: Multi-stage GaN amplifier, 2.7–3.1 GHz, 250W peak. The center monitor showed "In Circuit Data showing 250W from 2.7–3.1 GHz" — a live power sweep confirming the spec on a real device. The output power reading on the adjacent display: 47.38 dBm — exactly 55W continuous, consistent with the pulsed 250W peak claim at low duty cycle.
The MAPC-P1010 pallet on the bench was the physical device behind that number. A multi-stage GaN die mounted on a copper-tungsten carrier, pre-matched at the input and output for the S-band radar frequency range.
Moving right along the wall:
C-Band 100W Power Amplifier with 57% PAE. This efficiency number is worth pausing on. 57% power-added efficiency at 100W output means the device is converting more than half the DC input power into RF — the remaining 43% becomes heat. For a 100W PA, that's still 43W of dissipation in a small package. The thermal management implications are non-trivial, but 57% PAE at C-band in a GaN MMIC is a strong result.
MMIC Band Pass Switched Filter Bank, 21 GHz. A compact switchable filter solution for frequency-agile front ends — the kind of component that goes between a wideband PA and an antenna in a system that needs to cover multiple bands without retuning the amplifier.
High Peak Power Surface Mount Limiter, 2–18 GHz, MADL-011121. The datasheet visible on the right monitor showed the key specs: → Operating frequency: 2–18 GHz → Peak power operation: 60 dBm, 1 μs, 1% duty cycle → Flat leakage power: 33 dBm @ 60 dBm input, 4 GHz → Insertion loss: 2.2 dB @ 18 GHz → Package: 1.8 × 1.8 mm surface mount → Passive device — no DC bias required → Applications: receiver protection, ship and airborne radar
A limiter that handles 60 dBm peak input in a 1.8mm package with no power supply. This is a component designed for radar front ends where the receiver must survive being hit by the transmitter's own leakage — and recover fast enough to see the return pulse.
mHEMT E-Band Low Noise Amplifier. E-band spans 60–90 GHz. An mHEMT (metamorphic high-electron-mobility transistor) process enables sub-1 dB noise figure at these frequencies — relevant for E-band point-to-point links and satellite receive chains.
GaN Si TRM with Fast Recovery LNA. A transmit-receive module combining a GaN power stage for transmit and a fast-recovery low-noise amplifier for receive, with the switching speed needed to recover the LNA quickly after the transmit pulse. The "Si" in "GaN Si" indicates a GaN-on-silicon process rather than GaN-on-SiC — relevant for cost-sensitive volume applications.
In the center of the booth: a glass display case containing multiple bare MMIC dies and packaged devices — the actual silicon and compound semiconductor chips behind all of the above.
NI + Emerson: the system layer, 40 centimeters away
The adjacent booth — green background, clean geometric design — was a different world.
"Real-Time Data Streaming for Ultra-Wide Bandwidths."
The hardware on the optical breadboard was minimal by comparison: a NI PXIe-1095 chassis with SDR modules, and two VDI WR 6.5 (110–170 GHz) frequency extenders mounted on opposing ends of the breadboard — facing each other across approximately 40 cm of free space.
TX on the left. RX on the right. Gold horn apertures pointing at each other.
The screen above showed: "Real time SISO Modulation/Demodulation" — a live interface running NI's LabVIEW-based SDR software, displaying the transmitted and received waveform, constellation diagram, and BER metrics in real time.
The placard on the bench said: "Evolve 5G and 6G Research with SDR Technology" — a joint NI and Emerson branding, with three QR codes pointing to their Sub-THz Reference Architecture documentation.
This demo was not characterizing a component. It was demonstrating a link.
The 40 cm of air between those two WR 6.5 horns was the point.
What that 40 cm actually contains
When you look at the NI demo, the 40 cm gap between the TX and RX modules looks like nothing.
It is not nothing.
That gap contains: → Free-space path loss at 140 GHz — approximately 68 dB over 40 cm, far more than at microwave frequencies over the same distance → Antenna gain from the WR 6.5 horn — the narrow beam pattern that makes alignment critical; a 1° misalignment at 140 GHz degrades received power by several dB → Atmospheric effects — even over 40 cm, water vapor absorption at these frequencies is measurable in a high-humidity environment → Multipath from reflections off the optical table surface and surrounding equipment
The SDR on the PXIe chassis is processing all of this in real time — compensating for path loss, correcting phase errors, decoding the modulation — and displaying the result as a working link.
The system works. But it works because every interface in that chain has been addressed: the VDI extender output impedance matches the horn antenna, the horn gain is characterized, the SDR baseband processing is calibrated for the expected path loss.
Remove any one of those interface conditions, and the link breaks.
The observation that connects both booths
MACOM was showing components. NI was showing a system.
But neither was showing the complete picture.
The MACOM GaN PA at 47.38 dBm cannot connect directly to the VDI WR 6.5 extender — different frequency bands, different impedances, different power levels. To get from a 250W S-band GaN pallet to a 140 GHz SDR link, you need a chain of components that were not all on either booth.
The limiter that survives 60 dBm needs a receiver behind it that can process the returned signal. The NI SDR is that receiver — but at a completely different frequency.
The mHEMT E-band LNA from MACOM could feed into a VDI WR 10 extender to push the NI SDR link up to W-band. But that integration doesn't exist as a product. It exists as an engineering project.
This is what the show floor reveals that no catalog can: the distance between components and systems is not a specification gap. It is an integration gap.
Why this matters
Every demo at the microwave show is a controlled environment.
The MACOM PA hits 250W because it's biased at exactly the right quiescent point, driven at exactly the right input power, temperature-stabilized on a test fixture.
The NI link closes at 140 GHz because the TX and RX are aligned on a precision optical breadboard, 40 cm apart, in a controlled lab environment on the show floor.
Neither of these conditions describes your application.
Your application has a PCB with traces that add insertion loss. An enclosure that creates thermal gradients. An antenna with return loss that varies with mounting. Adjacent channels that create interference. A power supply that has ripple.
The components work. The system is what you build from them.
And what you build from them is not specified anywhere.
It emerges from every interface — every connection, every transition, every boundary between one block and the next.
That's where the real engineering happens.
Not inside the blocks.
Between them.
All photos: Thomas · @SignalByThomas