Whisper in a thunderstorm problem

The whisper in a thunderstorm problem is Molecular Reality Corporation's name for the difficulty that the difference in current between two DNA bases is smaller than the electrical noise of a measurement fast enough to see it. In the members' help file it appears, with the speeding bullet problem, as a reason the company's devices cannot sequence DNA: "the 'whisper in a thunderstorm' problem (signals are too faint)."^[1] Speed sets the bandwidth a measurement needs, and the noise of a current amplifier rises with bandwidth.

Every resistor in the circuit, the amplifier's feedback resistor and the pore itself among them, produces Johnson-Nyquist noise from the thermal motion of its charge. For a resistance $R$ over a bandwidth $B$ at temperature $T$ its root-mean-square current is $\sqrt{4k_{B}TB/R}$, which grows as the square root of the bandwidth.^[2] The pore adds flicker noise, which dominates below roughly 100 Hz and whose power follows the inverse of the number of charge carriers, a relation named for Hooge.^[2] At the other end of the spectrum the input voltage noise of the amplifier acts on the capacitance at its input, which includes the chip, the wiring and the membrane, and produces a current that grows as $B^{3/2}$. Losses in the insulators around the pore contribute dielectric noise. See nanopore noise.

The table gives numbers for three bandwidths. The values are computed here from the formulas at 300 K, with an assumed amplifier noise of 5 nV/$\sqrt{\text{Hz}}$ and an assumed input capacitance of 10 pF; none is a measurement.

Independent sources add in quadrature, so the largest row sets the floor. Between 10 kHz and 1 MHz the thermal rows grow tenfold and the capacitive row a thousandfold.

An amplifier must be fast enough to follow the event. A rectangular blockade lasting 1 microsecond that passes through a first-order filter set at 10 kHz reaches about 6 percent of its depth, and the charge it carries is spread over some 16 microseconds. The company's guide states the trade in one line: "you can filter the noise, but filtering smears the fast events you came to see."^[3] Opening the filter to 1 MHz restores the depth and, in the table, brings the noise to the scale of tens or hundreds of picoamperes.

In the 2012 experiment that read single nucleotides with an MspA pore and a polymerase motor, each current level lasted a median of about 28 milliseconds, and levels differed by up to 40 pA.^[4] A level that long can be read through a filter of a few hundred hertz, where the thermal noise of a 1 MΩ resistor is about 2 pA. An unmotored strand gives the amplifier a microsecond, which calls for megahertz and brings about 130 pA from the same resistor.

Slowing the molecule, the subject of the speeding bullet problem, allows a narrower filter and so less noise. Averaging $N$ passes of one molecule divides the noise by $\sqrt{N}$, which is the method of ping-pong sensing. The capacitive term, which dominates at megahertz, falls with the capacitance at the amplifier input. Large parasitic capacitance in the membrane is one of the limits named in a 2016 paper on recording at 100 nanoseconds,^[5] and insulating chips, small membranes and amplifiers built into the chip all reduce it. An integrated CMOS amplifier resolved signals of 1 microsecond at 1 MHz, and a later one reached 5 MHz with a signal-to-noise ratio above 10.^[6][5] A resistor of 1 GΩ with 0.1 pF of stray capacitance has a time constant of 100 microseconds, so amplifiers that must be both quiet and fast use other feedback schemes, such as a capacitor in place of the resistor.

A draft of the company's Disclosure Crawl, dated May 2025, titles the problem "Signal-to-Noise for Base Discrimination". It observes that the four DNA bases are structurally very similar and that telling them apart from "the minute, fleeting changes they induce in ionic current" calls for extraordinarily low-noise measurement and pore geometries often at the sub-nanometer scale. The draft ends with a critique in a skeptic's voice, which lists both problems among "fundamental physics and engineering limitations".^[7]

The 2022 technical overview planned a first-batch amplifier that would run at "up to 125,000 samples of current per second", and it estimated that "Maybe half" of the people working on solid-state sequencing do not use ionic current modulation, describing readout by transverse current and by light as alternatives.^[8] The open design of the MR1 Nanopore Reader documents a 1 GΩ feedback resistor, a full scale of 2.048 nA, and a noise floor under 1.2 pA RMS at 10 kHz "on PTFE interposer with driven guard ring", a low-loss insulator and a shield held at the input voltage.^[9] The resistor alone contributes about 0.4 pA at that bandwidth.

• Signal-to-noise ratio
• Bandwidth (signal processing)
• Bessel filter
• Dumb problems