Weep moat

The weep moat is a bare copper trace on the interposer of the MR1 Nanopore Reader, laid out to notice salt water in a place where none should be. Dry, the gap between the trace and the neighboring ground plane measures more than 10 GΩ. Bridged by a drop of saline, it measures less than 1 kΩ. In the design that change switches off the power to the converter that digitizes the pore current, after which the firmware stops the pumps and sets the bias across the pore to zero.^[1]

A weep hole is the small opening that lets water out of a masonry wall, and a moat is the ditch around a castle. Neither word is explained in the design documents, which use "weep" for the trace, the pad, the net and the firmware pin alike.^[2] No circuit has been built. The repository's own status line for the whole device reads "Designed. Simulated. Built and measured: not yet."^[3]

Two channels of 1 M potassium chloride, each 1.4 mm wide, run on either side of the membrane that holds the pore. Beside them sit an amplifier meant to see picoamperes and a bias stage that can drive the pore electrodes to ±2.5 V in either direction.^[1] The engineering change order that specifies the weep circuit states the hazard as electrolyte that breaches the gasket and either bridges the moat to ground or touches the path that excites the pore. A comment in the firmware gives the reason for zeroing the bias at once: "to prevent electrode damage/electrolysis".^[3]

The interposer is a hexagonal circuit board that carries the electrodes between the cartridge and the amplifier. Working outward from the sensing pad, the specification lists a driven guard ring, the rib of a cast silicone gasket, the weep trace and a copper pour at analog ground. A uniform gap of 0.25 mm separates the weep trace from the pour. On the front layer the trace follows both sides of the cis channel at an offset of 2.8 mm, and on the back layer it does the same for the trans channel; stitching vias join the two into one continuous fence.^[2]

The specification keeps the guard ring and the weep trace on separate nets and gives the reason. A guard ring is driven by a low-impedance amplifier to the same voltage as the sensing pad, so that no leakage current flows toward the pad. If a leak shorted that ring, the buffer would pour current into the liquid, overload, and hide the collapse in impedance that a detector exists to notice. The weep trace is therefore passive and high in impedance, biased through a weak pull-up of 1 MΩ to 3.3 V.^[2] The document calls the 0.25 mm gap a "detection spark gap". Behind it, the circuit limits the fault current to about 22 µA.^[1]

In the hardware path, the weep pad feeds a 100 kΩ limiting resistor and then the gate of a small N-channel transistor (a 2N7002 or BSS138), which a 1 MΩ resistor holds at 0 V. The transistor's drain is the enable pin of the Texas Instruments LP5907 regulator that makes the 1.8 V rail for the converter. A 100 kΩ pull-up to 3.3 V keeps the regulator on while the moat is dry.^[1]

The third row is the reason the transistor exists. A regulator's enable pin tolerates no more than −0.3 V, and the pore amplifier reaches −2.5 V during negative bias and electrode clearing. Wired straight to the pin, a negative swing would forward-bias the chip's internal protection diodes, draw heavy substrate current and end, in the order's words, in "permanently destroying" the regulator. A Schottky diode from gate to ground clamps the gate near −0.3 V, and the 100 kΩ resistor holds the current to 22 µA.^[1][a] The bench test in the order asks for the enable pin to fall below 0.1 V in under 2 µs and for the pump driver to be de-energized within 10 ms.^[1]

Firmware has its own path. The leak line goes to pin 1 of the ESP32 with the internal pull-up enabled, so liquid in the moat pulls it low. Once the pin has stayed low for 10 ms, the supervisor latches the alarm, zeroes the bias converter and calls a routine that tells the pump module to stop; only a reset clears the latch.^[3] The firmware README describes the cutoff as "<1 ms", though the debounce alone lasts 10 ms.^[3] A note on the schematic of the pump board adds a third route. When the emergency line goes low, a hardware gate puts both motor drivers to sleep "without waiting for software".^[3]

At release the company's own review found that the circuit in the schematic differs from the order. Its 1 MΩ resistor sits in series between the net and the gate, so nothing pulls the gate down while the moat is dry. The clamp diode is a different part, an MBR0540. That resistor is marked "exclude from board", so the layout has no footprint for it, and one connection to the diode is unrouted. Written as a directive, the order has no test of the interlock on record.^[4] The review recommended adding the pull-down and running a saline bench test, which the order itself describes as a 2 µL drop of 1 M potassium chloride laid across the gap.^[1]

Rework instructions in the order cover "existing hardware revisions (MR1 Rev A)" and begin with severing a copper trace using a scalpel.^[1] The list of open items at release ended with building one board and measuring it.^[4]

• MR1 Nanopore Reader
• Microfluidics
• Flow cell
• Four-electrode cell