Combine observations when one instrument can’t tell the likely causes apart. Keep the firmware, workload, and target state the same across measurements, and note any reset or halt between them.
Pair a log with power
A Joulescope JS220 can decode UART (8N1) on one of its GPI pins, gpi0 to gpi3, while it measures current and voltage. The decoded bytes have no timestamps that line up with the current samples. They show which markers arrived during the capture, but not where each state begins in the power trace.
To measure current in each state, use one of these approaches:
- Hold the target in a known state, let it settle, and capture each state separately.
- Have the firmware drive a GPIO during the state, wire it to a Joulescope GPI, and ask Embedder to capture between the marker’s edges. Edges are polled in software, so the capture boundaries can lag the marker by up to about 50 ms. Keep each state longer than that.
A separate serial log and a power capture don’t share a clock. When transition timing matters, use a marker that the measuring instrument samples itself, with a suitable timebase.
Separate protocol and electrical problems
The logic capture checks the frames, and the scope checks the signal. Use a common marker or trigger if your conclusion depends on matching one exact event across instruments.
Investigate a reset
Check the register’s access rules and the debugger attachment precautions first. Don’t clear a status register that still holds useful evidence.
Move from GDB to trace
Save stack and register observations before you start J-Trace coverage or a bounded instruction capture. Those captures take the probe from any GDB or RTT session, then reset and run the target.
Treat the paused debugger observation and the later trace as separate runs.
Make the test repeatable
Ask Embedder for one hardware script that contains the setup, ordered actions, pass condition, and cleanup. Keep a longer investigation in a hardware case with the raw captures and findings.
Record:
- The firmware artifact and board revision.
- Wiring, supply, channels, trigger, and the sample rate actually achieved.
- Any reset, halt, or other action that changed target behavior.
- The observed result, and what it does or doesn’t establish.
Stop active captures when you finish so the next hardware task can run. Check /hardware-queue if a task is waiting for the board. Last modified on November 13, 2026
