Measuring and Debugging

Two questions dominate tag development, and they need different instruments. How much current does this cost? is answered by a Joulescope. What is the driver actually doing on the bus? is answered by a Saleae logic analyzer.

The hardware is arranged so that the first can be asked of anything, and the second only of a prototype. That asymmetry is the reason prototype boards exist.

Current: the Joulescope reaches everything

A Joulescope connects at the 4-pin JST-XH, J1, on every base board — the net is EXT_PWR on the coin-cell boards and PWR_IN on the LiPo ones. Because every tag is powered through a base board, whether it is a finished tag on a pogo connector or a prototype on a header pair, the same measurement works in both cases.

The requirement this imposes is the one described in the base board overview: when the instrument is connected, the board's own supply to the tag must be out of the circuit. Two sources on one rail would measure the wrong thing. How that is done follows the battery chemistry — a physical changeover for LiPo, a low-leakage analog switch for coin cells.

The bases also carry a current mirror, TAG_MIRROR on the tag bases and PWR_IN_MIRROR on the breakout bases, for observing the supply without interrupting it.

Signals: the Saleae reaches only a prototype

A finished tag has six pogo pads, carrying power and SWD. Everything else is buried between two packages on a board that has to weigh under a gram. There is nowhere to attach a probe, and no room to add one.

A prototype board is the opposite: the signals between the processor and the sensors cross a pair of 1×17 headers on 2.54 mm pitch. The breakout base brings out essentially the entire processor — PA0 through PA15, PB0 through PB15, plus NRST, boot0, clkout and the RTC's rtc_scl and rtc_sda — and the prototype side names them for what they do:

Prototype	What a logic analyzer can see
UIUCBreakout	Three SPI buses — ACCEL_*, AT25_*, LPS_* — plus LPS_PWR, LPS_RDY and WKUP1
IMUTagNandBMP581-breakout	SPI to the flash and pressure sensor, I²C on SCL/SDA, interrupts BMM_INT and LPS_DRDY, the trigger LSM_TRG, and FLASH_PWR

That is the full working surface of a device driver: the transactions, the interrupt and data-ready timing, and the chip selects.

It also catches board defects, not just firmware ones

tag-breakout-l432v2 has its RTC's SDA and SCL crossed. On the bus that is unmistakable — clock and data doing each other's job — while from inside the firmware it looks like a clock that does not answer. A capture distinguishes a driver bug from a wiring bug in seconds, which is worth remembering before rewriting a driver that was already correct.

Using them together

The combination is more useful than either alone, because the power-gating lines are visible to both instruments at once. LPS_PWR and FLASH_PWR appear on the headers as logic signals and as steps in the current trace, so a question like how long does the pressure sensor take to settle after its rail comes up, and what does the inrush look like can be answered in one capture rather than inferred from two.

That is the measurement that settled two decisions already documented elsewhere: the 10 Ω inrush-limiting resistor on the UIUC tag's pressure sensor, and the removal of the NAND load switch from IMUTag once the saving turned out to be negligible.

Why this argues for prototyping drivers first

A driver written against a finished tag can only be debugged by what the firmware reports about itself. The same driver written against a prototype can be watched directly, on the real sensor, at the real clock rate, with the real current cost visible beside it.

By the time a tag-shaped board is fabricated, the drivers should already work — and the only genuinely new questions should be mechanical.