The automotive industry has been talking about Automotive Ethernet for years. What I saw at Embedded World 2026 in Nuremberg was the moment it stopped being a future technology and became an urgent engineering challenge.
At the Rohde & Schwarz booth, a live Automotive Ethernet and ASA testing setup was running. Not a slide. Not a render. A real physical-layer measurement — live signals, live eye diagrams, live compliance results. And what it revealed is something that most ADAS and SDV teams are not yet treating as a first-class problem.
The Architecture Has Already Changed
Traditional vehicles used CAN, LIN, FlexRay, and MOST — bus systems designed for low-bandwidth control signals. Those work fine for activating a window motor or reading a temperature sensor. They do not work for feeding a central compute platform with simultaneous streams from eight ADAS cameras.
The solution the industry settled on is Single Pair Ethernet (SPE): 100BASE-T1 and 1000BASE-T1, transmitting at 100 Mbps to 1 Gbps over a single twisted pair instead of the four pairs in standard office Ethernet. The benefit is clear — lighter harness, lower EMI, lower cost per connection.
But a single pair is also far more sensitive to degradation. And that is where the testing challenge starts.
R&S RTP164B (16 GHz / 40 GSa/s): Live 1000BASE-T1 waveform (top) and eye diagram (bottom). Photo: Thomas (x.com) @SignalByThomas
What the Eye Diagram Shows That Software Cannot
On the oscilloscope screen — the R&S RTP164B, a 16 GHz bandwidth instrument running at 40 GSa/s — you could watch the eye diagram of a live automotive Ethernet signal in real time.
An eye diagram is formed by overlaying thousands of consecutive bit transitions on top of each other. When the "eye" is wide open and symmetric, the receiver can distinguish ones from zeros reliably with plenty of timing margin. When it starts closing — because of jitter, noise, impedance discontinuities, or thermal drift — bit errors follow.
The critical point: no amount of sophisticated ADAS software can compensate for physical-layer errors that are discarded before the data stack even sees them. A corrupted packet dropped at Layer 1 does not generate a warning log. The system simply operates on incomplete sensor data — silently.
In a production vehicle at highway speed, incomplete sensor data is not a software bug. It is a safety risk.
ASA: The Next Wave Has Already Started
The second focus of the demo was ASA — the Automotive SerDes Alliance standard. ASA targets the highest-bandwidth links inside a vehicle: camera-to-ECU, radar-to-compute-platform, display-to-SoC. These run at multi-gigabit speeds over short but electromagnetically noisy harness segments.
ASA is designed to replace proprietary SerDes technologies like FPD-Link and GMSL. As the standard matures, every camera and radar module supplier will face a re-certification cycle. The physical-layer testing discipline that teams are building now for Automotive Ethernet will need to scale to handle ASA as well.
Automotive Ethernet evaluation board (DUT) with differential probe connections. Photo: Thomas @SignalByThomas
From Compliance to Real Validation
The compliance test software running alongside the oscilloscope was doing more than checking boxes. It was running automated amplitude verification, jitter decomposition, rise and fall time measurement, and eye mask validation — and presenting the results as PASS / FAIL against the relevant standard.
That matters because there is a real gap between compliance and operational robustness. A PHY that passes static compliance at room temperature may develop marginal eye openings at −40°C after 150,000 km of vibration. Jitter that looks within tolerance in a clean lab environment may grow deterministically when routed adjacent to a high-frequency switching power supply.
Understanding that gap requires tools that decompose jitter into its components: random jitter (RJ), deterministic jitter (DJ), periodic jitter (PJ). The right tool tells you not just that a signal is degraded, but why — which is the information you need to fix it.
Protocol Triggering and Decoding: 10BASE-T1S / 100BASE-T1 / 1000BASE-T1 automotive Ethernet bus decode. Photo: Thomas @SignalByThomas
Protocol Visibility: The Layer That Connects Physical and Digital
The final capability shown in the demo — protocol triggering and decoding — is where physical-layer and software-layer debugging finally converge. The oscilloscope was decoding 10BASE-T1S, 100BASE-T1, and 1000BASE-T1 frames live: displaying frame structure, MAC addresses, CRC status, and payload content alongside the analog waveform.
This matters enormously for integration testing. A board-level PHY compliance test tells you whether the transmitter meets the standard. Protocol decoding tells you whether the actual data being exchanged makes sense in context — and whether physical-layer marginal conditions correlate with protocol-level errors.
Three Observations for Teams Building Automotive Systems
- Signal integrity is becoming a first-class validation discipline in automotive development — not an afterthought. The teams treating it that way are shipping fewer field surprises.
- The ASA transition will create a re-validation cycle across the entire automotive camera and radar supply chain. Building the testing infrastructure and institutional knowledge now is a competitive advantage.
- Compliance is not the same as margin. Passing a standard at nominal conditions tells you the minimum. Understanding your actual signal margin — across temperature, aging, and EMI environment — tells you what you actually have.
The car of the future runs on data. That data runs on signal. If the signal is not right, nothing downstream saves you.
Observed live at Embedded World 2026, Nuremberg. Equipment: Rohde & Schwarz RTP164B (16 GHz / 40 GSa/s) with Automotive Ethernet compliance and ASA testing setup. All photos: Thomas @SignalByThomas