Cochlear Implant Atlas
CI Atlas · Speech-Coding Strategies: The Complete Lineage · Module 11

11Chasing Fine Structure: FSP, FS4 and the Temporal-Pitch Ceiling

Envelope strategies throw away the fast temporal fine structure within each band — the very cues that carry pitch, melody and tonal-language information. MED-EL's fine-structure strategies phase-lock the apical channels to that fine structure, but the electric temporal-pitch ceiling near 300 Hz caps how much benefit is possible.

TWhat fine structure is and why it matters

CIS, SPEAK and ACE deliver only band envelopes and discard the temporal fine structure within each band Lacking fine structure, implant users do poorly with speech in noise, speaker identification, melody and tonal languages, and are limited in bilateral benefit Fine-structure timing carries cues important for pitch, music and tonal languages such as Cantonese and Mandarin Encoding frequency modulation derived from fine structure has been proposed to improve performance in noise.[2004][2005]

33.1%FS4 high-rate monosyllables, best of four settings (Riss 2016) [2016]
13.2dB long-term speech-in-noise SRT gain on FSP over 24 months (Kleine Punte 2014) [2014]

CFS4: phase-locking the apex

FS4 extends envelope coding by transmitting the temporal fine structure on the most apical (low-frequency) channels It uses a 12-band FIR filter bank for 12 electrodes, with the four most apical channels delivering pulses synchronised to the zero-crossings/phase of the band waveform The higher-frequency electrodes continue with conventional CIS-like envelope coding Default MED-EL stimulation rates are set to the highest possible within voltage compliance, roughly 1500-2000 pps on average.[2014][2012]

Envelope only, or envelope + fine structure?

band envelopeapical pulses timed to zero-crossings

Inside every band sits a fast fine structure (the carrier) under a slow envelope. CIS keeps only the envelope. MED-EL's fine-structure strategies (FSP, FS4) add, on the apical channels that carry the voice pitch, pulses timed to the waveform's zero-crossings — delivering temporal pitch cues that pure envelope coding throws away. The catch comes next: the percept those timing cues evoke stops improving above a few hundred hertz. Schematic.

CMixed clinical evidence

Some studies show benefit for music appreciation and tonal languages with fine-structure processing Other studies found no significant difference between fine-structure and CIS-envelope strategies Crossover studies compared FS4, FS4-p and FSP fine-structure variants over a 4-month period Stimulation-rate effects were studied directly between FS4 and HDCIS.[2012][2011]

CThe temporal-pitch ceiling

Electric rate-pitch saturates near 300 Hz for most patients, with some listeners perceiving up to about 800-1000 Hz Rate-pitch resolution is roughly an order of magnitude poorer than acoustic: normal listeners discriminate ~1-2 Hz at 100 Hz, while CI listeners need ~10-20 Hz Because of this ceiling, fine-structure benefits plateau even when the coding faithfully delivers the timing These fundamental temporal-processing limits of electric hearing motivate current steering and future AI/optical approaches.[2004][2008]

Perceived pitch saturates near 300 Hz

~300 Hz ceilingpitchrate (pps)
Rate150 ppsPerceived pitch148 Hz

Fine-structure strategies bet that faster, waveform-timed pulses would convey finer pitch. They do — but only up to a point. Across implant recipients, the pitch evoked by pulse rate rises steeply at low rates and then flattens near 300 Hz: beyond it, adding rate adds little pitch. This temporal-pitch ceiling is why fine-structure coding helps low-frequency voicing and music yet cannot, on its own, restore full pitch range. Schematic.

TFine structure: real benefit, or extended low-frequency range?

FSP/FS4 place explicit temporal fine-structure timing on the apical (low-frequency) channels; FS4 codes fine structure on the 4 most apical channels, FS4-p adds parallel stimulation (Riss 2014). Among the MED-EL fine-structure strategies, FSP scored highest on Freiburger monosyllables (54.3%) vs FS4-p 51.8% and FS4 49.7% (omnibus p=0.03), yet 20/33 users chose an FS4 variant over default FSP (Riss 2014). When CIS and FSP were matched at an IDENTICAL frequency spectrum, there was no significant difference on sentences-in-noise, monosyllables, or melody, suggesting the apparent FSP benefit came from extended low-frequency range rather than fine structure itself (Riss 2011a). FSP vs CIS at 12/8/5/3/2 channels showed no significant difference at any channel count (Riss 2008), reinforcing the extended-low-frequency interpretation. Independent non-manufacturer data found HDCIS and FSP statistically equivalent on CNC, HINT, AzBio, and BKB-SIN (Dillon 2016); fine structure is favored for telephone (+10-11.5 pp, Galindo 2013) and music sound quality is often rated higher with HDCIS (Magnusson 2011).[2014][2011]

TBy the numbers

Fine-Structure (FS4) vs Envelope (HDCIS) Monosyllables by Stimulation Rate

010203040Monosyllables % correct (65 dB, quiet)HDCISFS4
Strategy and envelope rateFS4Low rate (750 pps/ch)25.2High rate (~1376 pps/ch)33.1

Fine-structure strategies (FSP/FS4) deliver explicit timing on the most apical channels rather than envelope-only pulses. In Riss 2016 (n=26) FS4 only pulled ahead of envelope HDCIS at high rate (33.1% vs 27.2%); at low rate the two were equivalent (25.2% vs 25.5%). Stimulation rate mattered more than fine-structure-vs-envelope. Verified means from Riss 2016.

FHear it

Two ways to hear pitch — place and rate

13002479376641225519576312875000apex · low pitchbase · high pitch

Tap the electrodes from apex to base. Each plays a band of noise at a different cochlear place, and the pitch rises left to right — this is the tonotopic place code an array reconstructs by lighting different positions. It is the dominant pitch cue a cochlear implant has, and the reason spreading sound across electrodes matters. Synthesised in your browser.

Hear what the envelope keeps — and what it loses

The original vowel phrase carries fast temporal fine structure — the harmonics that give it a clear pitch and identity. Reduce it to a single broadband envelope and you hear what one channel transmits: rhythm and loudness survive, but the vowels and pitch collapse into a wash. An 8-channel vocoder restores the place pattern across bands and the words re-emerge — yet the within-band fine structure is still discarded, which is exactly the gap fine-structure strategies (FSP, FS4) set out to fill. Synthesised in your browser.

Case 14.11 · Chasing Fine Structure
A bilingual Mandarin-speaking recipient is switched from an envelope-only strategy to FS4 and reports that lexical tones and music sound somewhat clearer, but tone discrimination still falls short of a normal-hearing peer.

Which factor best explains why FS4 helps but does not fully restore tone perception?

Self-assessment — Module 112 questions
Question 1

To which channels does FS4 apply temporal fine-structure timing?

Question 2

Approximately where does electric rate-pitch saturate for most CI users?

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