Dashboards¶
Edge Insights embeds three Grafana dashboards in the portal. Open them from Dashboards in the sidebar.
- Bus Overview: how does my system work? Frame counts, rates, CAN IDs, and timing classification.
- Bus Health: what is wrong? Anomalies, errors, foreign IDs, and congestion, with drill-down.
- Signal Explorer: decoded signal values, statistics, and a signal catalog. Needs a DBC file loaded on the interface, see CAN Settings.
Each dashboard opens embedded in the portal, with a link to open it directly in Grafana. The time-range picker and refresh interval at the top apply within the embed, same as in Grafana itself. Every panel has a (?) icon next to its title. Hover over it for a description of what the panel shows.
Bus Overview¶
- At a glance: frames in range, active CAN IDs in range, latest bus utilization, and how many DBC messages are loaded.
- Bus activity: bus utilization over time, and CAN frame rate broken down by ID.
- Bus behavior: a timing summary per CAN ID, covering classification, confidence, sample count, and average interval, plus the overall classification distribution across periodic, aperiodic, and burst categories.
Bus Health¶
- At a glance: total anomalies, timing anomalies, congestion windows, and bus errors in range.
- What's wrong, when: an anomaly timeline, split by type: timing, congestion, foreign ID, bus error.
- Anomaly list: every anomaly in range, with timestamp, type, CAN ID, interface, a summary, and a count. Filter by CAN interface or anomaly type using the controls at the top.
Anomaly drill-down¶
Click a row in the Anomaly list on Bus Health to open a deep-dive dashboard for that anomaly. Which dashboard opens depends on the type in the row.
| Type | Opens | Filtered to |
|---|---|---|
| Timing | Timing Anomaly Deep Dive | The CAN ID in the row |
| Congestion | Congestion Deep Dive | The interface |
| Foreign ID | Foreign ID Deep Dive | The interface |
| Bus Error | CAN Errors Deep Dive | The interface |
Only a timing anomaly belongs to a single CAN ID. The other three describe the state of the bus, so they open filtered to the interface. Every deep dive opens on a time range around the event, with about 30 seconds of context on each side. Change it with the time picker.
Timing¶
The Interval timeline panel plots two lines.
- Measured (window avg): the average time between frames for that CAN ID, over one aggregation window.
- Baseline (expected): the reference cadence the analyzer scores against. Edge Insights learns this value from the bus. It is not the cycle time declared in the DBC file.
Anomalies are scored against the baseline, not against the measured average. Red points mark the individual intervals that fell outside the threshold.
The two lines answer different questions. If they sit on top of each other, the message is running at its normal cadence. If the measured line drifts away from the baseline while few frames are flagged, the cadence has shifted but each frame is still inside the allowed band. This is an early sign of a timing problem. If the measured line stays on the baseline while red points appear, the average cadence is normal and individual frames are jittering.
The Anomaly details table lists every flagged frame with its measured interval, the baseline it was scored against, and the deviation between them.
Congestion¶
A microburst is a short window, well under a second, where bus utilization crossed the congestion threshold. Set the threshold in Analyzer Settings.
The dashboard shows the same traffic at two time scales, and the difference between them is the point.
- Sub-second (why it triggered): the Micro-burst timeline plots one point per event, at the utilization measured during that short window. These are individual events, not a continuous signal.
- Per-second (context): the Bus utilization panel plots the per-second average over the same range.
A burst that saturates the bus for 200 ms stands out in the first panel and is almost invisible in the second, because the per-second average spreads it over five times its duration. Read the two together to tell an isolated burst from sustained pressure. The Avg bus utilization and Peak utilization numbers at the top make the same comparison in a single figure each.
The Recent events table lists every microburst with its utilization, frame count, payload bytes, and the threshold that applied at the time. The threshold is recorded per event, so events from before a settings change still show the value they were judged against.
Foreign ID¶
A foreign ID is a CAN ID that is not defined in the loaded DBC file, not learned into the baseline, and not listed under Known CAN IDs. It can mean a new or misconfigured ECU, a device left connected from an earlier test, or traffic that does not belong on the bus.
The Frame count timeline plots one series per foreign ID, so a steady talker and a one-time appearance look different at a glance. The Event details table lists every event with its frame count and the timestamp the ID was first seen.
The drill-down opens with every foreign ID on the interface selected. Narrow it with the CAN ID filter at the top when one ID is of interest.
Before treating a foreign ID as a fault, check that it is not simply missing from your configuration. Adding it under Known CAN IDs stops the anomalies.
Bus error¶
These are errors reported by the CAN controller itself, not anomalies derived from traffic patterns. A Bus Error anomaly fires when the number of error frames in a window exceeds the error rate threshold, both set in Analyzer Settings.
Error rate over time breaks the errors down by class: TX timeout, lost arbitration, controller, protocol, transceiver, no ACK, bus off, bus error, restarted, and error counter. The class points at the layer to investigate. No ACK usually means nobody is listening, for example a single node on the bus or a wiring fault. Lost arbitration is normal in small numbers on a busy bus. Bus off means the controller took itself off the bus after too many errors.
Two details when reading the stacked chart. A single error frame that carries several class bits is counted once in the totals but appears in each matching series, so the stacked total can be higher than the raw frame count. Bus off and restarted are state transitions rather than running counters, so they appear as isolated marks rather than a rate.
The Recent threshold events table lists the windows that crossed the threshold, each with its own class breakdown.
Signal Explorer¶
- Selection: how many signals are selected versus available, and a summary table with one row per selected signal: CAN ID, message, signal, unit, sample count, min, max, average, and range over the selected time range. Pick signals from the Signal filter at the top.
- Signal time series: a plot per selected signal.
