Impedance, Noise & Electrode Materials

Reading the interface: impedance spectroscopy

Electrode impedance spectroscopy (EIS) sweeps a small AC current across frequency and records the complex impedance. Plotted as magnitude and phase versus frequency (a Bode plot) or as a Nyquist arc, the spectrum reveals the Randles parameters: the high-frequency plateau is R_s, the mid-band capacitive slope is C_{dl}/CPE, and the low-frequency rise toward R_s + R_{ct} is the faradaic leak.

# Randles impedance with a constant-phase element (CPE)
import numpy as np

def randles_cpe(f, Rs, Rct, Q, n):
    w = 2*np.pi*f
    Z_cpe = 1.0 / (Q * (1j*w)**n)          # CPE replaces the ideal C_dl
    Z_par = 1.0 / (1.0/Rct + 1.0/Z_cpe)    # C_dl || Rct
    return Rs + Z_par                       # complex impedance vs frequency

The Randles+CPE impedance model you fit to a measured EIS sweep to recover Rs, Rct, Q and n.

The single number people quote — 'impedance at 1 kHz' — is just one point on this curve, chosen because it sits in the spike band. It is a handy health check, but it hides the shape that actually matters.

Impedance and electrode size

To first order the interface impedance scales inversely with geometric surface area: halve the area and you roughly double the impedance. Small tips give spatial selectivity (single units) but high impedance; large contacts give low impedance but average over more tissue. This is a fundamental selectivity-versus-impedance tradeoff.

|Z_{if}| \;\propto\; \dfrac{1}{A}\quad(\text{interface}),\qquad R_s \;\propto\; \dfrac{1}{\sqrt{A}}\quad(\text{spreading})

Interface (double-layer) impedance falls roughly as 1/A; the geometric spreading resistance falls more slowly (~1/√A for a disc). Both approximate.

Shrinking an electrode raises its impedance, but the two contributions grow at different rates. The interface part scales as 1/A while the geometric spreading resistance grows more slowly (1/\sqrt{A}). So tiny electrodes are dominated by interface impedance — which is why coatings that boost effective surface area help so much.

|Z_{if}|: The interface (double-layer) impedance magnitude.
A: The electrode's surface area.
R_s: The geometric spreading resistance.
1/A \text{ vs } 1/\sqrt{A}: The two different rates at which the terms fall with area.

Cutting the area to a quarter roughly quadruples the interface impedance but only doubles the spreading resistance.

The noise floor: Johnson–Nyquist

Even a perfect amplifier cannot beat the thermal noise of the electrode's real impedance. The Johnson–Nyquist noise voltage is:

v_{n,\mathrm{rms}} = \sqrt{4 k_B T\,\mathrm{Re}\{Z\}\,\Delta f}

Thermal noise voltage; k_B is Boltzmann's constant, T absolute temperature, Re{Z} the resistive part of the electrode impedance, Δf the bandwidth.

Every resistor hisses with thermal noise, just from heat jiggling its electrons — an unavoidable voltage floor. This says how loud that hiss is: more resistance, higher temperature, or wider bandwidth all make it worse. It sets the quietest signal your electrode can ever resolve.

v_{n,\mathrm{rms}}: The RMS thermal noise voltage.
k_B\,T: Boltzmann's constant times absolute temperature — the thermal energy scale.
\mathrm{Re}\{Z\}: The resistive part of the electrode impedance (only this makes noise).
\Delta f: The measurement bandwidth — a wider band lets in more noise.

Plug in numbers: a ~100 kΩ resistive part over a ~10 kHz spike band at body temperature gives on the order of a few µV rms — right at the scale of small extracellular spikes. Lowering impedance directly lowers this floor, which is the practical argument for coatings.

The electrode's thermal noise competes with amplifier input-referred noise and the biological background; a well-designed system is roughly balanced so no single term dominates (see the low-noise amplifier and noise-efficiency discussion in the hardware track).

Materials: buying surface area without buying size

The trick behind modern microelectrodes is to increase the effective (microscopic) surface area while keeping the geometric footprint — and hence the spatial selectivity — small. That cuts impedance and raises charge capacity at the same time.

Smooth noble metals (Pt, PtIr): robust, but modest capacitance and limited charge injection.
Iridium oxide (IrOx / AIROF): reversible faradaic Ir³⁺/Ir⁴⁺ valence changes give a high charge injection capacity — a stimulation workhorse.
PEDOT:PSS conducting polymer: a porous polymer sponge with huge effective area — very low impedance and high charge capacity, excellent for recording; the tradeoff is long-term mechanical and adhesion stability.

Across these coatings, 1 kHz impedance can drop by one to two orders of magnitude versus bare metal of the same footprint, with charge capacity rising correspondingly — the enabling step for both high-channel-count recording and safe microstimulation.

The engineer's summary of the front end

Put it together: the Randles/CPE interface sets impedance and offset; impedance sets the thermal noise floor; materials reshape the interface to lower both; and everything must be read out by a low-noise, AC-coupled front end. A decision at any one stage constrains all the others.