Sign in to Peak Brain Path
Sign in to access your courses, books, and progress tracker. New here? Signing in creates your account automatically.
Want to explore courses first?
Browse courses and booksSign in to access your courses, books, and progress tracker. New here? Signing in creates your account automatically.
Want to explore courses first?
Browse courses and booksModule 5
Chapter 5 · 2 h · 8 quiz items · pass at 80%
This module opens IQCB Domain III (Technical), 15% of the exam. The filter-type and ADC content is an exam-critical distinction: the candidate must understand how acquisition hardware and signal processing alter the very metrics a QEEG report depends on. The quiz confirms the learner can reason from amplifier and filter settings to their effect on quantitative output.
Everything you will ever say about a brain map is a claim about voltage, and every one of those voltages passed through a chain of equipment before it reached your screen. A few microvolts of cortical activity travel through skin, gel, metal, wire, an amplifier, and a digitizer before a single number lands in your analysis software. A flaw anywhere in that chain corrupts everything after it, and no database comparison, no source model, no z-score correction recovers a signal that was never cleanly acquired. This is the least glamorous material in the IQCB blueprint and among the most consequential. The Diplomate who understands the equipment produces data worth interpreting. The one who does not produces confident reports built on noise.
The IQCB exam tests this domain at a depth the clinical procedure does not require. You are expected to know not only that impedance matters but why imbalance defeats noise rejection. Not only that you digitize the signal but what sampling rate and bit depth do to the spectral estimate you will later z-score; and, the topic most candidates underprepare, how filter choice changes the numbers a normative database scores. This chapter walks the acquisition chain from the scalp inward: electrodes and their materials, the electrode-skin interface and the impedance that governs it, the amplifier and the handful of specifications that separate a clinical instrument from a toy, the analog-to-digital conversion that turns the waveform into data, the filters that shape it, and the safety standards that make scalp recording on a mains-connected system legitimate. The recording-procedure depth, where to put your hands and how to coach the patient, is Chapter 6 and the companion QEEG Field Guide; this chapter is the electronics underneath the procedure.
An EEG electrode is a transducer. The brain does not speak in the electrons that move through a copper wire. It speaks in ions, charged particles moving through the conductive medium of the body. The electrode converts that ionic current at the scalp into the electronic current the amplifier reads, and it does so at a chemical interface between the metal of the electrode and the electrolyte of the conductive gel. How well it converts, how stable the interface is, how little noise it adds, how faithfully it passes the slowest potentials along with the fast ones, depends heavily on what the electrode is made of.
Silver/silver-chloride (Ag/AgCl). This is the reference standard for EEG and your default for clinical QEEG. A silver electrode coated with a layer of silver chloride forms what physicists call a nearly non-polarizable interface: charge crosses the metal-electrolyte boundary in both directions with minimal buildup, so the electrode does not accumulate a standing voltage that drifts over time. Two consequences matter for QEEG. First, Ag/AgCl passes the lowest frequencies faithfully, down toward DC, which makes it the correct choice for slow cortical potential work and for any recording where slow drift carries information. Second, it is electrically quiet, adding little noise of its own to the small signals you are trying to capture. The drawback is that the chloride layer wears with use and aggressive cleaning, so sintered Ag/AgCl electrodes, where the silver and silver chloride are fused throughout rather than coated on the surface, are preferred for durability in a busy practice.
Gold (gold-plated silver, or gold cup). Gold electrodes are durable, corrosion-resistant, and easy to clean, which has made them popular in practices that run many sessions and need electrodes that survive repeated use. Gold is an excellent conductor and performs well across the standard EEG bands. Its limitation relative to Ag/AgCl is at the low-frequency end: gold is a polarizable interface, less faithful for DC and near-DC recording and prone to more low-frequency drift. For routine spectral analysis in the 1 to 30 Hz range this rarely matters; for slow cortical potential work it does, and you want Ag/AgCl there.
Tin (Sn). Tin electrodes are inexpensive and were common in older clinical EEG systems. They are serviceable for routine band-range work but noisier than gold or Ag/AgCl and poorer at low frequencies. Tin is the budget option, and in a clinical context where signal quality is the whole point, it is rarely the right choice when better materials are affordable.
One rule prevents a category of artifact: do not mix electrode materials within a single recording. Dissimilar metals in contact with the same electrolyte generate small standing voltages of their own, a battery effect, introducing offsets and drift that show up as low-frequency artifact. Match your recording electrodes to each other, and match your reference and ground electrodes to your recording electrodes.
Beyond material, electrodes come in physical forms. Cup electrodes are small concave discs, typically gold or Ag/AgCl, with a hole in the center through which conductive gel or paste is injected once the cup is placed against the scalp. They are the workhorse of clinical EEG and QEEG: applied individually, held with paste or a small adhesive collar, filled through the central hole, they give excellent contact and signal quality at the cost of setup time. Disc electrodes are flat or slightly curved and similar in use. Ring electrodes present an annular contact surface and appear in some montage and reference configurations. An active electrode is a category that cuts across the forms, covered below, where a preamplifier sits at the electrode itself.
Electrode caps embed cup electrodes at the standard 10-20 positions in a stretchable fabric or mesh cap, pre-spacing the sites so setup is faster and placement more reproducible across sessions, with gel injected through each cup's hole once the cap is positioned. The trade is real and worth stating plainly. Caps buy speed and between-session consistency: the geometry is fixed, so you are less likely to drift a site from one recording to the next, which matters for serial QEEG. They cost you per-site tuning, because you cannot individually reposition a single cup that happens to sit over a poorly conducting patch, and the elastic tension that holds a cap can vary with head shape, loosening contact at some sites. Individual cup placement is slower and more operator-dependent but lets you adjust and troubleshoot each site, which a meticulous QEEG lab often prefers for the baseline a treatment plan or a forensic report will rest on. A busy practice frequently runs caps for throughput and accepts the trade. The IQCB-relevant point is that you know what each choice gives up. Chapter 6 covers the placement geometry these forms realize.
Two further distinctions shape what kind of recording you can produce.
Wet versus dry. Wet electrodes use a conductive gel or paste between the metal and the scalp. The gel fills the microscopic gaps, displaces insulating skin oils and dead cells, and creates the stable, low-impedance electrolytic bridge good EEG depends on. Wet Ag/AgCl or gold electrodes give the best signal quality available and remain the clinical standard, at the cost of skin preparation, the mess of gel, and the patient washing their hair afterward. Dry electrodes dispense with gel and rely on direct mechanical contact, often through fingered or pin-shaped tips that push through hair to reach the scalp. They are fast to apply and clean, which suits repeated measurement and field use, but they run at higher impedance, pick up more movement artifact, and generally require amplifiers built to tolerate the higher impedance. For clinical QEEG, where signal quality is paramount and a database comparison is at stake, wet electrodes are the correct default. Dry systems trade quality for convenience and belong to consumer and ambulatory contexts more than to the brain map a treatment plan rests on.
Active versus passive. A passive electrode is a simple conductor connected to the amplifier by a cable. All the amplification happens at the amplifier, some distance away. The problem is that the tiny EEG signal travels down that cable at high impedance, and the cable acts as an antenna, picking up environmental noise and generating artifact whenever it moves. An active electrode places a small preamplifier right at the electrode site, buffering the signal and dropping its impedance before it travels down the cable. The benefit is real: active electrodes are far less susceptible to cable-movement artifact and environmental pickup, they tolerate higher electrode-skin impedance (sometimes reducing or eliminating skin abrasion), and they perform better over long cable runs. The costs are that they are more expensive, they require power delivered to each electrode, and troubleshooting is harder because there is active circuitry at every site. Many modern research and high-density systems use active electrodes for exactly these reasons; many clinical systems use passive electrodes with careful cable management and good impedance technique, which works well when the technique is good.
If one number separates clean data from garbage at the acquisition stage, it is impedance. A practitioner who controls impedance has solved most of the signal-quality problem before the recording begins. A practitioner who ignores it is gambling, and the IQCB exam expects you to know precisely why.
Impedance is opposition to the flow of alternating current, measured in ohms, the AC generalization of plain resistance. At the electrode-skin interface, impedance is dominated by the outermost layer of skin, the stratum corneum, a thin shell of dead, dry, poorly conductive cells, along with skin oils and any hair caught between electrode and scalp. The conductive gel lowers this interface impedance by hydrating and bridging the skin; gentle abrasion lowers it further by thinning the dead-cell layer. A clean, well-prepared, well-gelled site presents low impedance and passes the signal faithfully. A dry, oily, or hair-obstructed site presents high impedance and degrades everything that follows.
Impedance hurts the recording through three distinct mechanisms, and the exam is after all three.
High impedance raises noise. Any impedance generates a small random voltage simply from the thermal motion of charge carriers, Johnson-Nyquist noise, and that noise grows with the impedance. A high-impedance electrode is therefore a noisier electrode independent of anything else, lowering the signal-to-noise ratio of the channel. The brain's signal is only tens of microvolts. You cannot afford to bury it under interface noise.
Imbalanced impedance defeats noise rejection. This is the subtle one and the most important for QEEG. EEG amplifiers reject environmental noise by exploiting the fact that interference, mains hum above all, appears almost identically on both inputs of a differential pair, while the brain signal differs between them. That rejection works cleanly only when the two inputs are electrically matched. If one electrode sits at 3 kilohms and its partner at 12, the common interference no longer divides equally between them, the amplifier can no longer subtract it cleanly, and line noise leaks into the recording. Balance between electrodes can matter more than the absolute value of any one. Two electrodes both at 8 kilohms can outperform one at 2 and one at 15.
Variable impedance creates instability and drift. An electrode whose contact is changing, gel drying over a long session or an electrode slowly loosening, presents drifting impedance, and the signal drifts and wanders with it, sometimes producing slow waves that masquerade as real low-frequency brain activity, the kind of thing that later reads as a spurious delta excess on the map.
The field's working targets, codified in the 2025 IQCB technical requirements for QEEG (Collura et al., 2025), give you concrete numbers to fix:
Caveats sit on top of these targets. Children and sensitive-scalp patients may force you to accept the upper end of the acceptable range to preserve cooperation, a reasonable trade documented in the record. And the targets interact with your amplifier: a high-input-impedance amplifier, discussed below, is far more forgiving of elevated electrode impedance than an older low-input-impedance design, which is part of why active-electrode and modern systems can sometimes run with minimal skin prep.
Modern amplifiers include built-in impedance checking: the system injects a tiny, harmless alternating current and measures the resulting voltage to compute impedance at each site, typically displaying the result on screen with color coding (green below 5 kilohms, yellow in the 5 to 10 range, red above). You run this check after applying electrodes and before recording, and you address every red and ideally every yellow before proceeding. When a site reads high, the corrections are mechanical and reliable: add gel (insufficient gel is the single most common cause), part the hair more thoroughly so the electrode contacts scalp rather than hair, abrade the skin gently to thin the insulating dead-cell layer (a blunt prep, never breaking skin), and confirm the electrode and its cable connection are sound. A site that stays stubbornly high despite all of this may reflect a scalp lesion or anatomical issue. Document it, and consider omitting that channel rather than recording known-bad data through it. Older analog systems without a built-in check require a dedicated impedance meter, testing each electrode against ground and logging the value by hand.
The amplifier is where microvolts become numbers. EEG signals at the scalp run roughly 10 to 100 microvolts, with the larger rhythms like eyes-closed posterior alpha reaching the tens of microvolts and much cortical activity smaller still. These are vanishingly small voltages, smaller than the electrical noise filling any ordinary room, and the amplifier's job is to lift them faithfully into a range a computer can digitize while rejecting the much larger interference around them. A handful of specifications govern how well it does this, and they are core IQCB material.
EEG amplification is differential, and understanding why is the conceptual heart of this domain. A differential amplifier does not measure the voltage at an electrode against some absolute zero. It measures the difference between two inputs and amplifies only that difference. Each EEG channel therefore has two inputs, an active electrode and a reference, and the amplifier outputs and magnifies the voltage between them.
The reason this matters is noise. Environmental electrical interference, above all the 50 or 60 Hz field radiating from mains power and wiring, couples into the head, the leads, and the body more or less equally. It appears on the active input and the reference input at nearly the same amplitude and phase: it is common to both, which is why it is called common-mode signal. Because the differential amplifier outputs only the difference between its inputs, anything common to both is subtracted away and disappears. The brain signal, by contrast, differs between the active site and the reference, so it survives the subtraction and is amplified. In one stroke, differential amplification rejects the dominant source of environmental noise while preserving the signal of interest. This is also exactly why electrode-impedance balance matters so much: the subtraction cancels common-mode noise cleanly only when the two inputs are electrically matched, so an impedance mismatch lets common-mode interference leak through as a difference the amplifier then faithfully amplifies.
How well an amplifier performs that subtraction is quantified by its common-mode rejection ratio, the single most important amplifier specification for EEG. CMRR is the ratio of the amplifier's gain for the wanted differential signal to its gain for the unwanted common-mode signal, expressed in decibels. A high CMRR means the amplifier passes the difference enthusiastically while suppressing the common interference almost completely.
The numbers to fix for the exam:
Translate the decibels into intuition. Each 20 dB is a factor of ten in voltage ratio, so 100 dB means the amplifier suppresses common-mode noise by a factor of 100,000 relative to the differential signal. That is what lets a clean EEG emerge from a room full of mains hum. A cheap amplifier with poor CMRR cannot make that separation, and its recordings carry line noise no post-processing fully removes. When you evaluate a hardware platform, CMRR is the first number to ask for.
For the differential trick to work, the amplifier must not load down the source it is measuring. This is the role of input impedance, the impedance the amplifier presents to the electrodes at its inputs. The principle is a voltage divider: the electrode-skin impedance and the amplifier's input impedance sit in series and split the signal voltage between them in proportion to their sizes. If the amplifier's input impedance is enormous compared to the electrode impedance, essentially all the signal voltage develops across the amplifier input and is measured, and almost none is lost. If the input impedance were comparable to the electrode impedance, a meaningful fraction of the signal would drop across the electrode and vanish before measurement, and worse, the loss would vary with each electrode's impedance, reintroducing the very imbalance differential amplification is trying to escape.
The targets:
This specification is also why high-input-impedance amplifiers tolerate higher electrode impedances, and why active-electrode systems can sometimes skip skin abrasion: when the input impedance is high enough, even an elevated electrode impedance drops a negligible fraction of the signal.
Gain is the factor by which the amplifier multiplies the differential signal. Its purpose follows directly from the size mismatch already named: the brain produces tens of microvolts, and the analog-to-digital converter that digitizes the signal wants something closer to a volt-scale input to use its full resolution. Gain bridges that gap. In practice the amplification is staged, an initial high-impedance buffer or instrumentation amplifier near the input followed by further gain before digitization, which keeps each stage operating in its linear range rather than asking one stage to do all the lifting. The exact figure is less important to memorize than the reason gain exists, which is to scale a tiny biological voltage into the digitizer's working range without clipping it. Modern systems often fold gain together with high-resolution digitization so that you set a recording range rather than a raw gain number, but conceptually gain is the amplification that makes the small signal large enough to measure precisely.
Dynamic range is the span between the smallest signal the amplifier can resolve and the largest it can handle without clipping or distortion, reported in decibels or, equivalently, in the bit depth of the digitizer. The lower bound is set by the noise floor, the amplifier's own internal electronic noise, the small random voltage it generates regardless of input, usually expressed as microvolts root-mean-square referred to the input. Activity below the noise floor is lost in the instrument's own hiss, so for EEG, where the signals are already small, a low noise floor is essential, and a good EEG amplifier's input-referred noise sits well below a microvolt. The upper bound is the largest excursion the amplifier passes without saturating. EEG demands a wide dynamic range because a single recording can contain both small fast rhythms of a few microvolts and large slow excursions or artifacts of hundreds of microvolts (a blink, a movement), and the instrument must capture the small ones faithfully without clipping on the large ones. This is why EEG digitizers run at high resolution, the subject of the next section.
Once the amplifier has lifted and conditioned the signal, the analog-to-digital converter (ADC) turns the continuous waveform into the stream of numbers your software analyzes. Two ADC specifications, sampling rate and bit depth, are routinely tested on the IQCB exam, and both have direct consequences for the QEEG metrics you compute downstream. This is where candidates who learned acquisition only as a procedure get caught: these are not arbitrary settings, they are the resolution limits of every number on the map.
Sampling rate is how many times per second the ADC measures the signal, expressed in hertz or samples per second. The constraint that governs it is the Nyquist theorem: to represent a frequency faithfully, you must sample at more than twice that frequency. Sample too slowly and you do not merely miss the fast activity, you actively corrupt the record, because frequencies above half the sampling rate (the Nyquist frequency) fold back down and masquerade as lower frequencies that were never there. This artifact, aliasing, is insidious for QEEG specifically: an aliased component lands inside a band you are about to quantify and z-score, inflating power at a frequency the brain never produced. Amplifiers guard against it with an anti-aliasing filter, an analog low-pass placed before the ADC that removes content above the Nyquist frequency so it cannot fold back (Nunez & Srinivasan, 2006).
The practical numbers:
Why does the higher headroom exist if QEEG cares about 30 to 50 Hz? Two reasons worth holding. First, a sampling rate comfortably above twice the highest band of interest keeps the anti-aliasing filter's roll-off from intruding on the signal you want. Second, faster sampling improves the temporal precision of the waveform, which matters for the phase and timing measures (coherence, phase lag) you may compute later, where the relationship between two channels is resolved in fractions of a cycle. Sampling faster costs you larger files and no real downside for QEEG; sampling too slowly costs you aliased, corrupted data you cannot repair.
Bit depth is the number of discrete voltage levels the ADC can distinguish, and it sets the amplitude resolution of the recording. An n-bit converter divides its input range into 2 to the n levels: a 16-bit converter resolves 65,536 steps across its range, a 24-bit converter resolves more than 16 million. The size of one step, the smallest voltage change the system can register, is the input range divided by the number of levels.
The consequence for QEEG is dynamic range, the same span discussed at the amplifier but now quantified by the digitizer. A recording must hold a small fast rhythm of a few microvolts and a large slow artifact of hundreds of microvolts in the same data without either clipping the large one or losing the small one in quantization steps. The more bits, the finer the steps and the wider the range the converter spans cleanly:
The QEEG-specific point: insufficient amplitude resolution shows up as quantization noise in the spectrum, a noise floor in your power estimates that can swamp genuinely low-amplitude activity, particularly the small high-frequency content (low-voltage fast activity, gamma) you most want to measure accurately. Bit depth is not a luxury spec. It is the floor under the precision of every power value on the map.
Distinct from sampling and bit depth, but tested alongside them, is the amplifier's frequency response, the band of frequencies the analog hardware passes before digitization. The doctrine for QEEG is to keep the analog front end deliberately broad: a low-frequency cutoff near DC or 0.1 Hz to preserve slow potentials, and a high-frequency cutoff above 100 Hz (consistent with the anti-aliasing requirement). The reason is that you can always apply a more restrictive digital filter during analysis, but you can never recover content the analog hardware discarded at acquisition. Filter narrowly at the front end and you have thrown away information for good; filter broadly and you retain the flexibility to shape the signal digitally, knowingly, after the fact. That hand-off from analog to digital filtering is the subject of the next section.
Filters shape which frequencies survive into your analysis, and the IQCB exam treats them with unusual seriousness because, for QEEG specifically, the filter you choose changes the numbers a normative database scores. Two practitioners can record the same brain, apply different filters, and produce different power values, different coherence, and therefore different z-scores. Understanding why is the difference between trusting a map and being fooled by one.
A filter removes or attenuates unwanted frequencies while passing the wanted ones. An analog filter is built from physical components, resistors, capacitors, op-amps, and acts on the continuous voltage before digitization. The anti-aliasing low-pass discussed above is analog by necessity, because it must remove out-of-band content before the ADC sees it. A digital filter is an algorithm applied to the already-sampled data, implemented in software after acquisition. The practical division of labor is the one named in the previous section: the analog front end stays broad and does only what must happen before digitization (anti-aliasing, gross DC blocking), while the precise band-shaping for analysis is done digitally, where it is repeatable, adjustable, and reversible in the sense that the raw data is preserved and can be refiltered. Digital filtering is where the QEEG-critical distinctions live, because the same nominal filter can be implemented two ways with sharply different effects on the signal's timing.
Digital filters come in two families, and the distinction is squarely an IQCB exam topic because it determines whether your filter distorts the phase of the signal.
A finite impulse response (FIR) filter computes each output sample as a weighted sum of a finite window of input samples, with no feedback. Its defining virtue for QEEG is that an FIR filter can be designed to have linear phase, meaning every frequency component is delayed by the same amount of time as it passes through. A constant time delay shifts the whole waveform later without distorting the relationships among its frequency components, and that delay can be compensated exactly, so the filtered signal preserves the timing and phase structure of the original. The cost is computational: achieving a sharp transition between passband and stopband requires a long FIR filter (many coefficients), which is more work to compute and introduces a longer delay.
An infinite impulse response (IIR) filter feeds its own past outputs back into the computation, which lets it achieve a sharp frequency response with far fewer coefficients than an FIR filter would need. It is computationally cheaper and steeper for its length. The cost is phase: a standard IIR filter has nonlinear phase, delaying different frequencies by different amounts, which distorts the temporal relationships among frequency components, smearing the waveform's shape and shifting the apparent timing of one band relative to another. For ordinary power-by-band analysis at a single site this distortion is often tolerable, because total power in a band is not sensitive to the phase within it. For any measure that depends on timing between sites, coherence, phase lag, the latency of an evoked component, IIR phase distortion is a genuine problem, because it alters the very quantity being measured. (Software can sidestep IIR phase distortion by filtering the data forward and then backward, which cancels the phase shift at the price of doubling the effective filter order, a technique worth knowing exists (Nunez & Srinivasan, 2006).)
The exam-relevant summary: FIR filters can be linear-phase and so preserve timing, at higher computational cost; IIR filters are efficient and steep but distort phase unless specifically corrected. When timing and phase matter, and in QEEG connectivity they always do, the phase behavior of the filter is not a technicality, it is part of the measurement.
Within filter design, the choice of response shape carries the same lesson at the level the exam most often probes. Two classic filter types, Butterworth and Bessel, are each designed for a different priority, and the difference is exactly the amplitude-versus-phase trade that runs through this whole section.
A Butterworth filter is designed for a maximally flat passband: it passes the frequencies you want with as little ripple in amplitude as possible and rolls off steeply into the stopband. This makes it excellent when faithful amplitude across the passband is what you care about, which is why it has long been a standard for clinical EEG, where you want the band power represented with full amplitude fidelity. Its cost is phase: the Butterworth response is not phase-linear, so it delays different frequencies by different amounts and distorts timing.
A Bessel filter is designed for maximally linear phase, a nearly constant group delay across the passband, which means it preserves the temporal relationships among frequency components and keeps the waveform's shape intact as it passes through. The cost is the mirror image: its amplitude roll-off is gentler than a Butterworth's, so it separates passband from stopband less sharply.
Here is the QEEG consequence, and it is the heart of why this section exists. Choose a Butterworth filter and you get faithful amplitude but distorted phase; choose a Bessel filter and you get faithful phase but a softer amplitude transition. For power-by-band analysis, where you are quantifying amplitude in a band, the Butterworth's flat passband is an asset and its phase distortion is largely irrelevant, because total band power does not depend on phase. For coherence and timing measures, where the synchronization and phase relationship between two sites is the quantity of interest, the Bessel's linear phase is what protects the measurement, and a Butterworth's phase distortion can corrupt the coherence and phase-lag values you report. The same EEG, filtered through a Butterworth versus a Bessel, can yield different connectivity numbers, neither filter wrong, but the two not interchangeable, and a coherence z-score computed after the wrong filter choice is a statement partly about the filter and only partly about the brain (Nunez & Srinivasan, 2006).
The operational takeaway for the Diplomate: filter choice is a methodological decision with metric-specific consequences, not a default to be clicked through. When you compare a recording to a normative database, the comparison is valid only if the filtering matches what the database assumed. When you read someone else's coherence finding, you cannot fully interpret it unless you know how the data was filtered. This is the deepest reason the IQCB exam treats filters as more than plumbing.
Beyond the filter's family and shape, its cutoff frequencies directly set which activity reaches your analysis, and careless cutoffs distort band power in ways that propagate into the z-score. A high-pass filter attenuates frequencies below its cutoff (removing slow drift and DC offset); a low-pass filter attenuates frequencies above its cutoff (removing high-frequency content and, before the ADC, preventing aliasing); together they define a band-pass that admits a window of frequencies. The QEEG-relevant hazard is that no real filter has a perfectly vertical edge. A cutoff placed at or near the boundary of a band you intend to measure will attenuate part of that band, because the filter's roll-off begins before the nominal cutoff and continues past it. Set a high-pass cutoff at 1 Hz and you begin attenuating the low delta you are about to quantify; set a low-pass cutoff at 30 Hz and you erode the top of the beta band and most of gamma. The principle that protects you is the one stated at the front end: keep acquisition cutoffs well outside the bands of interest (a broad analog pass, high-pass near 0.1 Hz and low-pass above 100 Hz), and apply analysis band-passes digitally with cutoffs and roll-offs chosen knowingly, so the filter shapes the data the way your method intends rather than silently eating the edges of your bands.
The notch filter is a narrow band-stop centered on the mains frequency, 60 Hz in North America, 50 Hz in much of the rest of the world, designed to remove line noise that survives the differential amplifier's rejection. It is genuinely useful when line contamination is present and cannot be eliminated at the source, and it is genuinely abused, which is why the exam expects you to know its costs.
Three cautions define responsible notch use. First, a notch filter does not fix the underlying problem. It masks one symptom of it. Line noise in the recording signals an environmental or impedance fault (poor grounding, an unbalanced electrode, a nearby device), and the correct response is to find and remove the source, not to paint over the 60 Hz peak and proceed. Second, the notch removes a slice of real EEG along with the noise: genuine brain activity exists at and around 60 Hz (in the gamma range), and the notch deletes it indiscriminately, so any analysis of high-frequency activity is compromised by a notch sitting in the middle of it. Third, the notch addresses only the fundamental and not its harmonics: line noise also appears at 120 Hz, 180 Hz, and beyond, which a single 60 Hz notch leaves untouched, and a filter with a sharp notch can itself introduce phase distortion near the stop frequency that perturbs nearby activity (Nunez & Srinivasan, 2006). The discipline is to prevent line noise at acquisition through grounding, balanced impedance, and environmental control (Chapter 6), to reserve the notch for residual contamination you genuinely cannot eliminate, and to document its use, because a reader of your report needs to know that the gamma band passed through a 60 Hz stop before any conclusion about it was drawn.
You are attaching conductive electrodes to a person's scalp and connecting them, through the amplifier, to equipment that ultimately draws on mains power. That arrangement carries a real, if small, electrical-safety obligation, and the IQCB exam expects you to understand the principles even though modern equipment handles most of the engineering for you.
Every EEG montage includes a ground electrode in addition to the recording electrodes and the reference. The ground establishes a common electrical reference point for the amplifier, the zero against which the differential measurements are stabilized, and it gives common-mode interference a defined path so the differential front end can do its rejection job. The ground's placement is not critical to the brain signal (it is commonly sited on the forehead, a mastoid, or an earlobe, off the regions of interest), but its impedance is critical: a poor ground undermines common-mode rejection and can contaminate the entire recording, which is why the ground, like the reference, is held to a stricter low-impedance standard than the recording electrodes. One ground is used per system. Connecting a patient to two separate grounds on different equipment can create a ground loop, a circulating current that injects noise and, in a worst case, poses a safety hazard, which is one reason battery-powered amplifiers (which float free of mains ground) are favored for clean, safe recording.
A point of frequent confusion worth fixing: the EEG ground is not the building's electrical safety earth, and it must never be connected to it. The patient ground is an isolated reference internal to the amplifier. Tying a person's scalp to mains earth would defeat the patient isolation described next and is exactly what the safety architecture exists to prevent.
The central safety principle in any equipment connected to a person is patient isolation: the part of the system touching the patient must be electrically isolated from the mains-powered part, so that a fault in the equipment or the building wiring cannot drive a dangerous current through the person. EEG amplifiers achieve this with an isolation barrier, an optically or transformer-coupled gap across which the signal passes but mains current cannot, separating the patient-connected front end from the line-powered electronics and computer behind it.
The quantity this controls is leakage current, the tiny stray current that can flow from equipment through the patient connection to ground under normal or single-fault conditions. Medical electrical-safety standards cap patient leakage current at levels far below the threshold of sensation or harm, and equipment built to those standards is designed and tested so that even a component failure cannot push leakage past the safe limit. The governing framework is the international standard for the safety of medical electrical equipment, IEC 60601-1, with particular requirements for EEG equipment addressed in its collateral and particular standards, which specify isolation, leakage limits, and the patient-applied-part protections that make scalp recording safe (Collura et al., 2025). You do not need to memorize the clause numbers, but you should know that clinical EEG equipment is held to a recognized medical-device safety standard, that the standard governs isolation and leakage, and that this is why you use equipment cleared or certified for patient connection rather than improvising with general-purpose electronics.
The principles translate into a few concrete habits. Use equipment certified or cleared for patient connection and intended for EEG. Do not connect a person to consumer hardware never designed or tested as a patient-applied part. Inspect leads and connectors for damage, because a frayed cable compromises both signal and safety. Keep the patient's electrode connections on a single isolated system and avoid connecting them simultaneously to other line-powered equipment that could create a ground loop. Favor battery-powered amplifiers where available, both for the cleaner recording (no mains ground loop, less line noise) and for the inherent isolation of running off a battery rather than the wall. And maintain the equipment on the manufacturer's schedule, since the isolation and leakage protections are guaranteed only in equipment kept in spec.
The acquisition chain in this chapter is what separates clinical-grade equipment from the consumer EEG devices now widely sold, and the IQCB exam expects you to articulate the difference. A clinical amplifier offers high CMRR, high input impedance, a low noise floor, wide dynamic range with adequate bit depth, a sampling rate at or above 256 Hz with anti-aliasing, full 10-20 electrode coverage using wet Ag/AgCl or gold electrodes, built-in impedance checking, and certified patient isolation. Consumer devices typically compromise on most of these to gain convenience and price: dry electrodes at higher impedance, far fewer channels, lower-grade amplification, coarser or limited digitization, little or no impedance verification, and no medical-device safety certification. None of that makes consumer EEG worthless for what it is built for, wellness, gaming, casual self-tracking, but it makes it unsuitable for the clinical brain map a treatment plan or a forensic opinion rests on, and a candidate should be able to say why in the specific terms of this chapter: the consumer device's signal-to-noise, its rejection of line noise, its amplitude and timing resolution, and its impedance control are not adequate to the clinical question, and it is not certified as a patient-applied part.
When you sit down to acquire a recording, the chain you are managing runs in order. Choose electrodes whose material suits the recording (sintered Ag/AgCl for anything touching slow potentials, gold a durable alternative for band-range work, matched throughout including reference and ground), decide between a cap and individual placement for the throughput-versus-per-site-control trade the recording warrants, and prepare each site so impedance comes in below 5 kilohms and balanced across electrodes, holding the reference and ground to the stricter standard because a fault there hits every channel. Run the built-in impedance check and fix every high site before recording. Confirm the amplifier digitizes at 256 Hz or above with adequate bit depth, keep the analog filters broad and do your band-shaping digitally, and reserve the notch for line noise you genuinely cannot remove at the source, documenting it when you use it. Trust the differential amplifier to reject line noise its CMRR is built to suppress, provided you gave it the balanced inputs it needs. Confirm the equipment is certified for patient connection and the leads are sound before anything touches the patient.
For the IQCB exam, fix the specification numbers and the reasons behind them. Electrode impedance: below 5 kilohms preferred, up to 10 acceptable, balanced within 2 to 3, never above 20, reference and ground highest priority. Differential amplification rejects common-mode noise (above all 50/60 Hz line interference) by amplifying only the difference between active and reference, which is why impedance balance is essential and why a mismatch lets line noise through. CMRR quantifies that rejection: at least 90 dB, 100 dB or better in good equipment, where every 20 dB is a tenfold voltage ratio. Input impedance must vastly exceed electrode impedance, at least 10 megohms and often 100-plus, so the signal is measured rather than lost across the electrode. Gain scales the tens-of-microvolts brain signal into the digitizer's range; the noise floor sets the smallest resolvable signal and must sit below a microvolt; dynamic range and bit depth (16-bit adequate, 24-bit ample) let one recording hold both small fast rhythms and large slow excursions. Analog-to-digital conversion samples at 256 Hz or more (Nyquist demands more than twice the highest frequency of interest, and the anti-aliasing filter prevents out-of-band content from folding back as aliasing), with bit depth setting amplitude resolution and therefore the noise floor of your power estimates. And the filters: analog filters act before digitization and stay broad, digital filters do the analysis band-shaping; FIR filters can be linear-phase and preserve timing while IIR filters are efficient and steep but distort phase; Butterworth gives flat amplitude with distorted phase (good for power, fine for band power) while Bessel gives linear phase with a softer roll-off (good for coherence and timing); cutoffs placed at a band edge erode that band; and the notch removes line noise at the cost of real activity around 60 Hz, masks rather than fixes the underlying fault, and leaves the harmonics untouched. Know the chain, and you know why a clean recording, and a defensible z-score, is something you build at the scalp and the settings, not something you rescue in software.