Field notes from European Microwave Week 2026, London.
"Validate End-to-End NTN" read the panel on one of Keysight's EuMW 2026 stations. NTN, non-terrestrial networks, means 5G links that run through satellites instead of, or alongside, cell towers. The demo showed a low-Earth-orbit satellite passing over a globe, while a side panel plotted the two quantities that make NTN different from every terrestrial test: how the link delay and the Doppler shift change during the pass.
The setup
Under the display sat a Keysight PROPSIM FS16 radio channel emulator (F8820B) with a Keysight MXG signal generator on top. The emulation software's tabs included cell, timers, NTN, handover and resource allocation, and listed several base stations and two satellites ("SAT1" and "SAT2"). An Android device was mirrored in a window on the right of the screen, which suggests a real handset as the user equipment.
What the runtime plots showed
The values are small in my photo, so I read the shapes rather than exact numbers:
- Downlink delay fell smoothly over the displayed interval, from roughly 3.4 to roughly 1.2 on the plot's millisecond axis, and flattened towards the end.
- Downlink Doppler shift started at about +40 kHz, stayed nearly flat, then fell increasingly fast through zero to about −15 kHz at the end of the window. The live readout showed SAT1 at about −14.5 kHz.
- Throughput bars for downlink and uplink, and an RTT latency gauge.
That is the signature of a satellite approaching and passing closest approach. Delay is shortest near closest approach, and Doppler crosses zero there, changing fastest at exactly that moment.
Why NTN needs an emulator, not just a signal generator
A terrestrial channel model deals with multipath and modest Doppler from a moving handset. A LEO link adds three things on top:
- Large Doppler: a satellite at around 7.5 km/s produces Doppler shifts of tens of kHz at S-band carriers, about the order of the ±40 kHz on this screen. That is far beyond the few hundred Hz a terrestrial receiver normally has to track.
- Doppler rate: near closest approach the Doppler changes fastest, so the receiver's frequency tracking has to follow a moving target, not just a fixed offset.
- Time-varying delay: the propagation delay changes by milliseconds during a pass. Uplink timing advance must follow it, otherwise the base station receives uplink symbols at the wrong time.
Then come handovers. With two satellites in the scenario, the emulator can reproduce the moment when the UE has to move from one beam or satellite to the next, while both Doppler and delay are changing. Testing that over the air needs a real constellation and a lot of patience. In the lab it needs a channel emulator that applies the right delay and Doppler as functions of time, consistent with the orbital geometry.
What to check in an NTN emulation setup
- Trajectory realism: are delay and Doppler derived from a real orbit and ground position, or are they simple ramps? The plots should look like this one, with the steepest Doppler change at closest approach.
- Doppler range and rate at your actual carrier frequency. Doppler scales with frequency, so Ka-band values are an order of magnitude higher than S-band.
- Delay range: LEO, MEO and GEO differ by orders of magnitude in delay. Make sure the emulator covers the orbit you need.
- Pre-compensation: NTN UEs use their own position and the satellite ephemeris to pre-compensate Doppler and timing. The test should verify that the UE's pre-compensation works, not only the network's tracking.
- Handover timing: measure interruption time and throughput during satellite changes, not only steady-state throughput.
Takeaway
The globe was the eye-catcher, but the engineering content was in the two small plots beside it. Delay that changes by milliseconds and Doppler that swings through tens of kHz in a single pass are what an NTN device has to handle, and what a terrestrial test setup cannot reproduce. An end-to-end NTN test is credible when those curves come from real orbital geometry and the device under test has to track them while staying connected.
Notes: Observations and interpretations are my own, based on what was running at the booth during the show. Plot values are approximate readings of curve shapes from my photo. The carrier frequency, orbit parameters and UE type were not confirmed with the exhibitor.