Cutting Skew at 224G: URF’s 110 GHz Phase-Matched Cables, Matched to ±0.5 ps

The bottleneck has moved from loss to timing

Optical modules have moved from 400G to 800G, and 1.6T products built on 224 Gbps lanes are now in validation. As data rates climb, the timing budget shrinks, and a tiny difference in electrical length between two paths becomes a real measurement error.

At 224 Gbps PAM4 (112 GBd), one unit interval lasts about 8.9 ps. Unmatched cables of the same nominal length can differ by a picosecond or more, and 1 ps of skew takes up more than 11% of that interval, or about 40° of phase at 110 GHz.

This is why URF built a 110 GHz cable set that holds every pair within ±0.5 ps, verified pair by pair.

Why loss and VSWR specs alone are not enough

Insertion loss, VSWR, and phase stability still matter. But none of them guarantees that two cables have the same electrical length. Two assemblies can each pass every spec and still arrive at the DUT a picosecond apart.

In differential and multi-channel setups, that gap shows up as skew, eye closure, and inflated jitter, and engineers end up chasing errors that belong to the cables, not the device. Matching the paths removes the problem at the source.

What URF built

URF's DC to 110 GHz, 2-channel cable sets use 1.0mm connectors and are delay matched to within ±0.5 ps per pair, roughly a tenth of a millimeter of electrical length. They come in 6 in (ST110-201-201-6-PM/2-0.5ps) and 12 in (ST110-201-201-12-PM/2-0.5ps) lengths.

The match holds when the cables move, too. Phase change under flexure is ±8° max at 110 GHz, about ±0.2 ps, well inside the matching window.

How the matching is achieved and verified

Three steps get each pair inside ±0.5 ps. Consistent materials (LD-PTFE dielectric, solid silver-plated copper center conductor) keep propagation velocity uniform from cable to cable. A controlled termination process keeps the 1.0mm interface from adding stray delay. Each pair is then measured, trimmed to match, and 100% tested for delay match, insertion loss, and VSWR before it ships.

Measured performance of a production pair

The measurements below come from a production 12 in pair (ST110-201-201-12-PM/2-0.5ps), tested on a 110 GHz VNA from 10 MHz to 110 GHz.

VSWR at both ports of Cable 1 (left) and Cable 2 (right). Both cables stay well below the 1.35:1 max spec across the full band.

Insertion loss of Cable 1 (left) and Cable 2 (right). Both measure about 4.7 to 4.8 dB at 110 GHz, within the 5.2 dB max for the 12 in set.

Time-domain delay of Cable 1. The peak marks when the signal arrives through the cable.

Time-domain delay of Cable 2, on the same time axis. Its peak lines up with Cable 1.

Where engineers and researchers use it

• Differential S-parameter characterization. Connector and cable designers running 4-port differential measurements get mode conversion results that reflect the DUT, not the test cables.
• SerDes, jitter, and BER testing. Aligned P and N paths keep setup skew out of eye diagrams and jitter numbers.
• Optical module and modulator validation. Teams testing 800G and 1.6T modules, or driving electro-optic modulators differentially, keep every picosecond of margin.
• Phased array, Massive MIMO, and OTA testing. Matched paths keep channel-to-channel phase comparisons valid up to 110 GHz.
• On-wafer probing and mmWave calibration. Researchers and test engineers get a known, matched reference path for differential probing and system verification.

More matching options for larger setups

Standard sets are matched within each pair. For setups that use several pairs, such as multiple test stations or multi-channel arrays, URF also offers absolute delay matching: every cable is matched to a common reference delay, so any cable can be swapped with any other. For 112G work up to 67 GHz, 1.85mm phase-matched sets are also available.

See the 110 GHz ±0.5 ps phase-matched cable sets, or contact our team for absolute matching, 67 GHz sets, and volume pricing.