Field notes from European Microwave Week 2026, London.
"Elevate Signal Purity" was the headline over one of Keysight's demo stations at EuMW 2026. The product was a two-channel signal generator in a 2U chassis, built on what the booth display called a hybrid YIG/DDS architecture. The screen above it showed the evidence: phase noise and AM noise of three 10 GHz carriers, all measured simultaneously. The display rewards careful reading, because two of its marker readouts look comparable and are not.
What the screen showed
- Three channels, each measuring a carrier near 10 GHz. Ch1 and Ch2 read 10.0000001 GHz at 8.754 dBm and 8.704 dBm. Ch3 read 10.0000006 GHz at 8.628 dBm.
- Offset range 100 Hz to 100 MHz on all channels, at 10 dB/div.
- Top window: phase noise, reference level −120 dBc/Hz.
- Bottom window: AM noise, reference level −131 dBc/Hz.
- Marker readouts, phase noise: 312.830 kHz, −141.39 dBc/Hz (yellow trace), and 278.346 kHz, −126.13 dBc/Hz (blue trace).
- Marker readouts, AM noise: 312.830 kHz, −156.08 dBc/Hz, and 278.346 kHz, −142.91 dBc/Hz.
The display was partly blocked when I took the photo, so I cannot read the curves in the 10 kHz to 1 MHz region. Everything below is based on the visible parts and the marker table.
Reading 1: two markers, two different offsets
At first glance the marker table seems to say that one trace is about 15 dB better than the other: −141.39 versus −126.13 dBc/Hz. But the two markers are not at the same offset. One is at 312.830 kHz, the other at 278.346 kHz.
On a phase noise plot that slopes at roughly −20 dB/decade in this region, a 12% difference in offset frequency accounts for only about 1 dB. So most of the 15 dB difference is real. Still, when comparing sources, the comparison only holds when the markers sit at exactly the same offset. With markers placed by search (peak, or a specific spur), as these odd values suggest, it is easy to compare two different things.
The general rule: for source comparison, put all markers at the same round offsets, such as 10 kHz, 100 kHz, 1 MHz and 10 MHz, and read them in a table. Search markers are useful for finding spurs, not for comparing noise floors.
Reading 2: AM noise sits well below phase noise
At the same marker positions, AM noise read about 15–17 dB below phase noise: −156.08 against −141.39 dBc/Hz, and −142.91 against −126.13 dBc/Hz. That is typical of a good synthesiser. The oscillator's amplitude is stabilised by its own limiting mechanism, while its phase has no such restoring force.
The reason to measure AM noise anyway is that some systems care about both. In a radar receiver, close-in AM noise on the LO adds to the noise floor just as phase noise does. In an ADC clock, amplitude noise can be converted into timing jitter by the comparator threshold. A phase noise analyser that reports AM noise from the same acquisition makes it easy to check that AM noise really is negligible, rather than assuming it.
Reading 3: why "hybrid YIG/DDS"
The architecture named on the booth panel reflects a classic trade-off in microwave synthesis:
- A YIG oscillator has a very high-Q resonator. It offers excellent phase noise at larger offsets and a low broadband noise floor, but it tunes slowly.
- A DDS (direct digital synthesiser) offers very fine frequency resolution and fast, phase-continuous switching, but its spurious content and its noise at microwave frequencies depend on the clock and the multiplication.
Combining them, typically with the DDS providing fine steps and the YIG providing the low-noise microwave carrier, aims to get both. The fine frequency offsets visible on screen (10.0000001 GHz versus 10.0000006 GHz) are a reminder that the carriers were set with sub-Hz resolution. The design challenge in any such hybrid is the spurious content of the DDS path, which is why the full offset range out to 100 MHz matters as much as the close-in curve.
Checklist for reading a phase noise comparison
- Same offsets: compare traces only at identical offset frequencies.
- Same carrier frequency and power: phase noise scales with 20·log(N) under frequency multiplication, and the analyser's sensitivity depends on input power.
- Analyser floor: make sure the measurement is not limited by the analyser itself. Cross-correlation instruments show their floor; ask for it.
- Spurs versus noise: discrete spurs appear as sharp lines and are specified in dBc, not dBc/Hz.
- AM noise: check it separately when the application is amplitude-sensitive.
Takeaway
A phase noise screen packs a lot of information into a small space, and marker readouts are the part people quote. On this one, the two phase noise markers sat at different offsets, a detail that is easy to miss and that changes what a comparison means. AM noise read about 15 dB below phase noise, as expected from a good source, and it was measured rather than assumed. That is the right way to show signal purity.
Notes: Observations and interpretations are my own, based on what was running at the booth during the show. Screen values are as read from my photo; part of the display was obscured. The product model, the analyser used and the source-to-channel mapping were not confirmed with the exhibitor. Architecture features are paraphrased from the booth display.