Cochlear Implant Atlas
CI Atlas · Tuning the Electric Ear: Activation and Programming the Implant · Module 06

6Frequency Allocation and the Filter Map

Frequency allocation is the filter-bank table that decides which input frequencies are routed to which electrode, and therefore which cochlear place each sound stimulates. Because most arrays sit shallower than the cochlear region that naturally codes their assigned frequencies, recipients begin life with a place-frequency mismatch they must adapt to, and the audiologist must weigh matching cochlear place against keeping a wide speech bandwidth. This module covers the default filter maps, the mismatch from shallow insertion, frequency importance, and when (and whether) to reassign frequencies.

FThe filter bank: how input frequency becomes electrode place

Frequency allocation determines how the analysed input spectrum is divided into bands and assigned to active channels, recreating an electrical version of the cochlea's normal frequency-to-place mapping. In the Nucleus default map, the full input range across all 22 channels spans about 188 to 7938 Hz. Nucleus channel bandwidths are deliberately uneven: low-frequency channels are narrow at about 125 Hz, mid channels about 250 Hz, and high channels range from roughly 375 up to 1000 Hz, mirroring the narrower auditory filters of the normal low-frequency cochlea. Channel 1 is the most basal contact and channel numbers rise toward the apex, so allocation maps high frequencies basally and low frequencies apically, matching cochlear tonotopy. In MED-EL software the audiologist can move the lower (70 to 350 Hz) and upper (3500 to 8500 Hz) frequency limits in 10 Hz and 500 Hz steps and choose among four psychoacoustically derived tables; in Advanced Bionics allocation is automatic and not user-adjustable.[2020][2009]

Nucleus filter bank: disabling a channel widens its neighbours

1885001k2k4k7.9k12345678910111213141516171819202122frequency (Hz, log) — 22 channels
Channel11Band1030-1222 HzWidth192 Hz

The Nucleus filter bank divides 188-7938 Hz across 22 channels on a quasi-log scale, so low channels are narrow (~125 Hz) and high channels wide (a few hundred Hz to ~1 kHz). Tap a band to read its edges, then disable a channel: its neighbour widens (gold) to span the gap so no frequency region is left uncoded. This is why a deactivated electrode still leaves a working — if coarser — frequency map. Schematic.

TPlace-frequency mismatch from shallow insertion

Tonotopy, the orderly frequency-to-place relationship, is the cochlea's first organising principle; an electrode placed basal to the site that naturally codes its assigned frequency creates a place-frequency mismatch. Default allocations push frequencies onto electrodes that sit more basally than the Greenwood place for those frequencies, because a one-turn lateral-wall insertion only reaches roughly the cochlear region that natively codes around 1000 Hz. Insertion angle and the default allocation jointly determine the spiral-ganglion place pitch each electrode actually stimulates, so the mismatch is systematic, not random. Electrical place pitch saturates and rate-pitch cues are limited to frequencies below roughly 300 Hz, so much of the frequency information must be carried by place, magnifying the cost of a place mismatch. Most recipients adapt to the initial mismatch over months, so default maps are generally retained rather than re-aligned electrode-by-electrode to the Greenwood place.[2020][2009][2015]

Insertion depth and place-pitch mismatch

apex (low f)Apical electrodenatural place:868 Hzassigned band:1000 Hzmismatch:2.4 stplace-pitch region

The filterbank assigns each electrode a fixed band, but a shallow lateral-wall insertion (~one turn) only reaches the ~1000 Hz Greenwood region, so the deepest electrode is told to deliver a frequency far higher than its natural place — a basal shift heard as upward-shifted, “Mickey-Mouse” pitch. Deeper insertion lowers the natural place frequency and shrinks the mismatch, but below roughly 300 Hz rate-pitch cues saturate and added depth buys little extra pitch. Schematic.

CFrequency importance and bandwidth trade-offs

Narrow mid-frequency bands resolve the formant frequencies of vowels and consonants, which is why several studies favour narrow mid-frequency allocation for better speech recognition. Restricting the upper input limit from about 10000 Hz to roughly 8000 Hz narrows the channel bandwidths across the speech range and improved speech recognition. There is a bandwidth trade-off: extending the upper limit captures more high-frequency speech energy but widens each band and worsens spectral resolution, while restricting it sharpens resolution but discards high-frequency cues. Adults are typically allocated out to at least about 6000 Hz (some clinicians cap at 6500 to 7000 Hz), whereas children are given a wider band out to at least 7000 to 8000 Hz because they need high-frequency cues to recognise and produce speech, especially female and child talkers. The Nucleus default frequency table widens the bandwidth for children, while a narrowed mid-frequency table is often chosen for adults.[2020][1995][2001]

Upper input limit: bandwidth vs spectral resolution

adult target8.0 kHz2505001k2k4k8kinput frequency (Hz, log)
Mapped width7.8 kHzPer channel355 HzSpeech cuesspread thin

Pushing the upper input limit from 8000 toward 10000 Hz captures more high-frequency speech energy, but the same fixed channels must spread over a wider band, so each channel covers more hertz and spectral resolution coarsens. Adults are usually mapped to about 6000-7000 Hz and children a little higher at 7000-8000 Hz to preserve the high-frequency cues important for developing speech. Where the marker sits inside the shaded target is the compromise that keeps both bandwidth and resolution useful. Illustrative.

CWhen (and whether) to reassign frequencies

Frequency allocation matters most when an electrode must be disabled; on disabling, the remaining channels widen their bands (the total input range stays the same), which can blur two close frequencies into one channel and degrade spectral resolution. Because disabling reshapes allocation and can cost performance, electrodes should be turned off only when justified, for example abnormal impedance, non-auditory side effects such as facial-nerve stimulation or pain, poor sound quality, or a need for far higher levels than neighbours. When a recipient has few usable channels, double-channel mapping routes two channels into one electrode to restore overall bandwidth. Manufacturers caution against haphazard manual editing of the filter table, but two accepted uses are setting the acoustic-electric crossover frequency for EAS users and recreating a legacy allocation when an upgraded recipient rejects the new defaults. Channel-to-electrode reassignment is rarely warranted, but is reasonable when a sweep reveals a non-tonotopic pitch order on a channel; for EAS, frequency allocation must be coordinated with preserved acoustic low-frequency hearing to avoid overlap.[2020]

Case 17.6 · Frequency Allocation and the Filte
An adult upgrading from a legacy Nucleus processor to a current device complains that, with the default allocation, voices sound unnaturally high-pitched and wrong, even though objective measures and stimulation levels are fine. Speech scores have dropped since the upgrade.

What is the most appropriate frequency-allocation step?

Self-assessment — Module 62 questions
Question 1

Why does a shallow lateral-wall insertion create a place-frequency mismatch with default allocations?

Question 2

What happens to frequency allocation when an electrode is disabled in a contemporary system?

Tracked locally in your browser — see /progress for the dashboard.