5.2 Amplifiers, Filters, Averaging, and Display Settings

Once Ohm’s law explains the voltages at the skin, the instrument must amplify the difference between two electrodes, filter frequencies you did not come to measure, average when the sensory nerve action potential (SNAP) is buried in noise, and display the result on a time-versus-voltage screen. This OpenExamPrep section is independent teaching on those Domain II instrumentation items (II.B–C) from the American Association of Electrodiagnostic Technologists (AAET) R.NCS.T. outline. Display numbers in the table at the end are common lab starting points, not unpublished official exam figures.

Differential amplifier: G1 (active) versus G2 (reference)

A differential amplifier measures the difference between two inputs rather than the voltage of one wire versus earth.

• G1 (grid 1), also called the active electrode, sits over the signal of interest: the muscle motor point for a belly-tendon compound muscle action potential (CMAP), or the designated active site in a SNAP montage.
• G2 (grid 2), the reference, sits over a relatively quieter site: tendon for motor, or the protocol’s reference site for sensory recording.

The displayed voltage is essentially G1 minus G2. A voltage that appears equally on G1 and G2 is common-mode and is rejected. Line-frequency 60 Hz on the body is largely common-mode if impedances are matched and cables are quiet; that is why differential recording, not brute-force gain, is the first defense against wall-power interference.

If G1 and G2 are swapped, the waveform inverts. Belly-tendon motor recording is set so the main CMAP spike is a negative-up (upgoing) deflection when G1 is negative relative to G2 — the electrodiagnostic display convention. An upside-down CMAP is a lead-assignment or invert / non-invert display problem until proven otherwise. Use the invert control to match laboratory convention; do not silently reverse G1 and G2 from the written protocol just to make the picture “look right” without documenting the change. Invert flips the display polarity. It does not repair a reference electrode that is sitting on another active muscle.

Common-mode rejection ratio (CMRR) and input impedance

Common-mode rejection ratio (CMRR) is how well the amplifier ignores the identical voltage on both inputs. It is a ratio, often discussed in decibels (dB). Teaching instruments quote high CMRR, commonly on the order of 90 dB or greater. CMRR collapses in practice when electrode impedances are high or unmatched, because the “common” 60 Hz is no longer equal at G1 and G2. A superb amplifier specification cannot salvage a 2 kΩ G1 paired with a 40 kΩ G2.

Input impedance of the amplifier must be very high (modern designs are often tens to hundreds of megaohms) so the amplifier does not load the signal. Think of a voltage divider: the biologic source and electrode-skin impedance form one part; the amplifier input is the other. If amplifier input impedance is not far larger than electrode impedance, part of the SNAP or CMAP drops across the electrodes and never appears on screen. Pair that rule with Section 5.1: electrode-skin impedance low and matched; amplifier input impedance high. Those two impedances are easy to confuse on a multiple-choice stem. High at the amplifier is good. High at the skin is bad.

Filters: high-pass, low-pass, and 60 Hz notch

Filters remove frequency content. NCS instruments name the same two filters in two vocabularies. Learn both:

A high-pass / LFF setting of 10 Hz means frequencies below 10 Hz are attenuated. A low-pass / HFF setting of 10,000 Hz (10 kHz) means frequencies above 10 kHz are attenuated. Students mix the names because “low-frequency filter” sounds as if it keeps low frequencies. It does the opposite: it is the control that filters out low frequencies (high-pass).

How a high-pass that is too high distorts the tracing

CMAP duration and negative-peak area live in relatively low frequencies. If you raise the high-pass (for example from about 2–10 Hz toward 100–500 Hz), the amplifier throws away that slow content. Effects you should expect:

• Duration shortens and area falls.
• Amplitude may fall.
• Onset can change: a false dip or an earlier-looking takeoff appears because the filter rings and because the slow initial rise was removed.
• The waveform looks peakier and less like the physiologic CMAP you meant to measure.

Sensory potentials are briefer, so they tolerate a somewhat higher high-pass than motor studies, which is why teaching starting points differ in the table below. They are not immune: an overly high high-pass still chews onset and area.

Teaching illustration (directions of change, not a conversion table for reports): a CMAP recorded at high-pass 10 Hz / low-pass 10 kHz might show onset 3.2 ms, peak 4.1 ms, amplitude 8.0 mV, duration 5.5 ms. After the high-pass is raised to 500 Hz, duration and area drop and the onset cursor may no longer sit on the same takeoff. Do not interpret that new onset as faster nerve conduction.

How a low-pass that is too low distorts the tracing

SNAP and CMAP peaks contain faster frequency content. If you lower the low-pass (for example from 10 kHz toward 500–2,000 Hz on a motor study), the amplifier rounds sharp turns:

• Peaks delay (peak latency lengthens).
• Peaks round and amplitude falls.
• Onset latency is often less shifted than peak latency, which is one reason motor onset is preferred over peak for distal motor latency.
• Fine notches that might be noise — or might be a second component — disappear, so you can hide a technical or physiologic detail.

Same illustration continued: drop the low-pass to 500 Hz and the 4.1 ms peak may read later (for example 4.4 ms) with a lower amplitude. That is filter delay and rounding, not new demyelination.

Notch filter

A 60 Hz notch can clean a terrible baseline in a hostile electrical environment, but the SNAP and CMAP themselves have energy near 60 Hz. Notch filtering can ring, add extra phases, and move peaks. Prefer lowering electrode impedance, regrouping cables, turning off nearby motors, and trusting CMRR. Treat notch as a documented last resort, not a default motor-study setting.

Sweep speed (timebase) and gain (sensitivity)

The screen is a graph:

• Horizontal axis = time. Sweep speed or timebase is time per division (ms/div). Faster sweep (smaller ms/div, such as 1–2 ms/div) spreads the waveform so onset and peak cursors are easier to place. Slower sweep (5–10 ms/div) packs more time onto the screen, which you need for F-waves and H-reflexes but which makes a 3 ms distal latency occupy almost no width. Ten divisions at 5 ms/div display a 50 ms window. Ten divisions at 2 ms/div display 20 ms.
• Vertical axis = voltage. Gain or sensitivity is voltage per division (mV/div or µV/div). A more sensitive setting (20 µV/div rather than 5 mV/div) makes a small SNAP visible. If gain is so high that the waveform clips at the screen edge, measured amplitude is invalid. If gain is so low that the CMAP is a nub, onset placement becomes guesswork.

Changing sweep or gain does not change the nerve. It only changes how faithfully you can see and cursor the same voltage-versus-time signal. Sloppy display settings still produce sloppy numbers, and those numbers are what the report sends out. A clipped 15 mV CMAP displayed at 2 mV/div is not “15 mV”; it is “off-scale, repeat at lower sensitivity.”

Signal averaging

A SNAP may be only a few microvolts. Random noise is often larger. Signal averaging adds successive sweeps that are time-locked to the stimulus.

• The time-locked nerve response adds constructively.
• Random noise tends to cancel. Signal-to-noise ratio (SNR) improves roughly by the square root of the number of sweeps (√N). Four sweeps ≈ 2× SNR; 16 sweeps ≈ 4×; 64 sweeps ≈ 8×.

CMAPs are typically millivolts and often need little or no averaging. SNAPs commonly need averaging, especially in thick limbs, edema, or noisy rooms.

Critical limitation: stimulus artifact is also time-locked. Averaging does not remove artifact; it can make a pretty, averaged artifact that masquerades as a SNAP. Fix impedance, polarity, and stimulator orientation first. Do not “average away” a technical problem. Averaging also assumes the response is stable from sweep to sweep. Movement, changing stimulus intensity, or a coached muscle twitch between sweeps smears the average.

Phase cancellation

An NCS waveform is the sum of many axon action potentials (APs). If those APs arrive together, their negative phases add and the recorded SNAP or CMAP is large. If they arrive asynchronously — temporal dispersion from long distance, mixed conduction velocities, or demyelination — the negative phase of a fast axon overlaps the positive phase of a slow axon. Opposite phases cancel. Amplitude falls more than a simple “lost axons” story would predict. SNAPs suffer more than CMAPs because sensory waveforms are briefer and spikier; a small time smear produces large cancellation.

Instrumentation does not create demyelination, but a filter that further smears peaks (low-pass too low) can exaggerate cancellation, and a high-pass that is too high can change the phases you are summing. When two volleys are out of step, the display can look small even though axons are still conducting. That is a physiologic and technical reading skill, not a reason to crank gain until noise looks like a SNAP.

Cursor placement: onset versus peak

Numbers come from cursors, not from the pretty picture.

• Motor distal latency is measured to onset: the first reproducible departure from baseline that belongs to the CMAP, not to residual artifact.
• Amplitude follows laboratory convention, commonly baseline-to-peak for the motor negative spike, and either baseline-to-peak or peak-to-peak for SNAP — but the convention must be consistent across visits.
• Peak latency marks the negative peak. It is often easier to reproduce on a small SNAP than onset, but it is not the same physiologic event as onset. Peak is later than onset. Mixing peak on one visit with onset on the next fabricates a latency change.

Typical errors:

• Placing motor onset on the peak lengthens reported distal latency.
• Placing onset in noise before true takeoff shortens latency.
• Measuring a volume-conducted far-field bump as a SNAP onset (Section 5.3).
• Changing filters between studies and then comparing latencies as if the nerve changed.
• Leaving a peak cursor on a notch created by residual 60 Hz or by an overly aggressive notch filter.

If the takeoff is buried in artifact, fix the setup (Sections 5.1 and 5.3). Do not invent an onset in the middle of the shock spike.

Common lab starting points (not official exam figures)

Use this table as orientation for how motor and sensory displays usually differ. Individual laboratories publish their own protocols. The exam-relevant skill is understanding why sensory traces use more sensitivity, a faster sweep, and somewhat different filters — not memorizing one vendor’s hidden table as if it were an unpublished official number.

If a motor CMAP looks “too peaky” with a short duration, look at the high-pass before you invent a myopathic story. If a SNAP peak latency is longer than last year but the low-pass was dropped from 10 kHz toward 2 kHz, look at the filter before you invent slowing. Display settings are part of the measurement.