How an Electrode Records Neural Activity

An extracellular electrode records the small voltage that a neuron's own currents produce in the tissue around it. A site on a thin-film probe is a patch of metal a few tens of micrometres across; it measures that voltage as a difference from a reference, and it sees only neurons within about a hundred micrometres. How an electrode is chosen and used follows from three facts — what it senses, what it senses against, and how far it can see.

This is the background a reader needs before choosing an electrode. It is written from the electrode's side of the interface: each concept is stated once, in a sentence a reader can keep, and then followed only as far as it changes a decision about a probe or a grid. The neuroscience behind each concept is one link away and is not repeated here.

Research use only. NeuroNexus products and the procedures described here are for non-clinical research use, and must not be used in human or veterinary medical procedures.

What does an extracellular electrode actually measure?#

A neuron firing an action potential drives a brief current across its membrane, and that current has to flow back through the extracellular space to complete the circuit. Because the extracellular space has resistance, the current produces a voltage in the tissue: a few tens to a few hundred microvolts at the cell body, falling away with distance. An extracellular electrode measures that voltage. It never touches the neuron and never measures what happens inside it; it measures the field the neuron leaves in its surroundings.

Three things follow for the electrode. First, the signal is small — microvolts, against a background of thermal noise, other neurons, and the animal's own movement — so the interface that picks it up and the electronics that amplify it decide whether it is recoverable at all. Second, the measurement is always of a difference: an electrode site reports the voltage at its own position relative to the voltage at a reference somewhere else, so the reference is half of every number in the file. Third, the voltage at a site is the sum of every source near enough to contribute. A site does not record a neuron; it records a place, and which neurons are visible from that place depends on where the site sits and how far it can see.

On a thin-film probe the site is a pad of iridium or platinum, from 92 to 1,250 µm² across the NeuroNexus catalog, connected by a thin-film conductor along the shank to the connector at the base. Its position relative to every other site was fixed by the mask set that patterned it, which is why a laminar profile recorded on one probe means the same thing on the next.

Why is a spike only visible within about a hundred micrometres of a site?#

An extracellular spike falls off steeply with distance from the cell body. Within a few tens of micrometres it is large and clearly shaped; at around a hundred micrometres it is at the level of the background; beyond that it is gone. The exact distance depends on the cell type, its orientation, and the tissue, but the order of magnitude is the same in every preparation. This is the detection radius, and it is the single most consequential number in electrode design.

It decides the site pitch. Sites spaced farther apart than the detection radius each see a different set of neurons; sites spaced closer than it see the same neuron from several positions at once, at different amplitudes. That shared view is what a spike sorter uses to tell one unit from another: the ratio of amplitudes across neighbouring sites is specific to the neuron's position. A tetrode is four sites inside one detection radius for exactly this reason; a polytrode extends the idea along the shank; a linear array with sites spaced beyond the radius trades unit isolation for a continuous profile across depth. The question is it good at isolating single neurons is, at the electrode, the question is the pitch inside the detection radius.

It also bounds what any probe can see. A single shank samples a cylinder of tissue roughly two hundred micrometres across along its length, however many sites it has. Adding sites within that cylinder adds views of the same neurons; reaching more neurons means more shanks, or a longer span. That is the reason multi-shank and three-dimensional arrays exist, and the reason a probe with a thousand sites on one shank is a very well-sampled cylinder rather than a large volume.

Are spikes and field potentials different signals?#

They are the same voltage on the same site, separated by frequency. A spike is the extracellular signature of one neuron's action potential: a waveform about a millisecond long, whose energy lies above roughly 300 Hz. The local field potential is what remains below that band — the slower, summed synaptic and population currents of many neurons over a larger volume. The recording chain captures both on every channel; a filter divides them afterwards.

For the electrode, the distinction matters in two places. Site area weighs the two differently: a small site is dominated by the nearest few neurons and resolves spikes sharply, while a larger site averages more tissue and reports the population signal with less unit detail. And the field potential does not obey the detection radius — it is generated by many cells over hundreds of micrometres, so a site sees field potentials from far outside the volume in which it can see spikes. A probe can therefore be positioned to record excellent field potentials from a structure whose single units it cannot isolate at all, which is worth knowing before concluding that a recording has failed.

On the surface of the brain the same logic gives a thin-film grid its character: the sites are separated from the neurons by the pia or the dura, so individual spikes do not reach them, and what a grid records is the field — at site diameters from 25 µm to 1 mm on the NeuroNexus grids, chosen for the scale of the population being mapped.

What does site impedance do to a recording?#

Impedance is the opposition a site presents to current at a given frequency, measured at the electrode–tissue interface and conventionally quoted at 1 kHz. The interface between a metal site and the electrolyte around it is not a simple wire: charge crosses it through a double layer that behaves largely as a capacitor, and the impedance of that layer is set by the site's area, its metal, its surface texture, and any coating on it.

Two consequences reach the data directly. The interface generates thermal noise in proportion to the square root of its resistance, so a higher-impedance site starts with a higher noise floor before the amplifier adds anything. And the amplifier's input impedance forms a divider with the site's: if the two are comparable, part of the signal is lost across the site before it is measured. The working rule is that the amplifier should present an input impedance many times — at least ten times — the site's, and the headstages designed for these probes are built to that rule.

The practical trade is against site area. A smaller site has higher impedance and more thermal noise, but sees a more local signal and discriminates units more sharply; a larger site has lower impedance and less noise, but averages more tissue. The typical range therefore depends on site size: a 177 µm² site on a NeuroNexus probe reads 200 kΩ to 1 MΩ at 1 kHz, larger sites read lower, and smaller sites read higher. Where impedance must come down without giving up a small site, the lever is the surface rather than the area — a coating that increases the effective surface without increasing the footprint — which is what a site coating such as Z-Coat is for. Impedance also sets the limit on stimulation: it caps the charge a site can deliver before the voltage across the interface reaches the point where water electrolyses. A site that records well and a site that stimulates safely are constrained by the same number from two directions.

Impedance is also the first thing that changes when something is wrong. A site that reads far above its typical value has a broken trace, a contaminated surface, or a dried residue; one that reads far below has a short. Measuring impedance before every session is the cheapest check there is, and it is the reason probe setup begins with it.

Why is the reference half of every measurement?#

An electrode never records a voltage; it records the difference between two points. The site is one. The reference electrode is the other, and whatever is common to both — line noise, the animal's muscle activity, the slow drift of the whole preparation — cancels in the subtraction, while whatever differs between them remains. Choosing where the reference sits is therefore a scientific decision about what to keep, not a wiring detail.

A reference placed on the probe near the sites, in quiet tissue, cancels the most and leaves the local signal; a reference on the skull or in a distant structure cancels less but preserves slow, shared activity such as the field potential across a region. A reference inside the structure being recorded will subtract the very signal of interest. Ground is a different thing again: it is the zero the whole recording system is built on, tied to the animal so that the amplifier and the tissue share a voltage, and it should not be confused with the reference even when the two are connected together.

On a NeuroNexus probe the reference and ground are separate wires on the package, and the surgical guides and the EEG referencing guide treat their placement as the decision it is. Most recordings that look like a noisy electrode are a reference problem.

What happens between the site and the file?#

The voltage at the site travels along a thin-film trace on the shank to a bond pad at the base of the probe. The package — the interconnect, connector and covering assembled around the array — routes it from the bond pad to a connector a headstage can mate to; a design is not orderable until it is packaged, and the package decides how the probe mounts, how it survives an animal, and which instruments it can talk to. The headstage amplifies and, in current designs, digitises at the head, a few millimetres from the site, so that the microvolt signal is converted before a cable can add noise to it. The acquisition system receives the digital stream from every headstage, stamps it on one clock with stimulation, video and the rig's own events, and writes it to a file. Software then reads that file — and, with the right workflow, reads the stream while it is being recorded, so that a bad site, a bad reference or a bad session is seen while there is still time to fix it rather than days later. Radiens is NeuroNexus's software for that workflow.

The chain is chosen from the electrode outward. The probe design sets the package options; the package sets the connector; the connector sets the headstage; the headstage sets the acquisition system. A probe is never chosen in isolation from the rest, which is why the selection guide runs in that order.

What does the tissue do to an electrode over days and weeks?#

In an acute preparation the electrode is in tissue for hours and the interface is as it was made. In a chronic implant the tissue responds: over the first weeks, glial cells encapsulate the shank, the tissue around it changes, and the signal at the sites changes with it — some sites lose units, some gain them, and the population that is visible drifts. The recording is not worse in any simple sense, but it is different, and it keeps changing.

For the electrode, this makes chronic work a materials and package question as much as a geometry question. The substrate and its thickness set how much the shank moves against the tissue; the package sets whether the probe stays anchored to the skull or floats with the brain; the connector sets whether it survives months of an animal grooming against it. The acute-versus-chronic guide sets out the choice, and the chronic surgical guide the practice. A probe that records beautifully on day one and is lost by week three has usually failed at the package, not at the sites.

What to read next#

With this much, the rest of the electrode content can be read in the order it was written. Silicon Probes and Thin-Film Grids: An Introduction sets out the three questions to ask of any electrode — is it good, is it right for this experiment, and where does it sit in the space of what can be made. What makes a thin-film electrode good answers the first, Choosing an electrode for your experiment the second as a checklist, and The electrode design space the third as a reference; Choosing a probe package and Thin-film grids for surface recording complete the set. Each term used above is defined in the glossary, one sentence each, with its expansion from the electrode.