Frequency-specific gain prescription, multi-channel WDRC programming, and real-ear verification — the clinical workflow that turns an audiogram into a matched acoustic fitting
Every hearing aid fitting starts from the same physiological measurement: the air-conduction audiogram. Pure-tone thresholds recorded at 250Hz, 500Hz, 1kHz, 2kHz, 3kHz, 4kHz, 6kHz and 8kHz define, in dB HL, the softest sound the patient can detect at each frequency. This threshold map — combined with bone-conduction thresholds to rule out a conductive component, and uncomfortable loudness levels (UCL) to bound the dynamic range — is the entire physiological input that every downstream prescriptive calculation depends on.
Air-conduction thresholds are measured under headphones or insert earphones in a sound-treated booth, in 5dB steps, using the modified Hughson-Westlake ascending-descending method. Bone-conduction thresholds (via a mastoid oscillator) are measured alongside to separate sensorineural loss (cochlear/neural) from conductive loss (middle-ear pathway) — the air-bone gap. A pure sensorineural loss with air and bone thresholds tracking together is the primary candidacy profile for amplification; a large air-bone gap instead points toward medical/surgical referral.
The audiogram used in this simulator models a moderate, high-frequency-sloping sensorineural loss: 30dB HL at 250Hz descending to 70dB HL at 4kHz — the single most common configuration in age-related (presbycusis) and noise-induced hearing loss, because the basal, high-frequency-coding region of the cochlea is both mechanically and metabolically the most vulnerable part of the organ of Corti.
Beyond thresholds, most probe-microphone fitting systems also require the patient's real-ear-to-coupler difference (RECD) or real-ear unaided response (REAR) — an acoustic transfer function of the individual ear canal — because a 2cc coupler measurement in the fitting software does not equal what actually reaches the eardrum. Ear canal volume varies enormously, especially in children, which is precisely why DSL v5.0 built RECD correction into its core algorithm rather than treating it as an optional step.
A prescriptive formula is a validated algorithm that converts the audiogram into a target insertion gain (or target real-ear output) for every frequency and every input level. The two dominant formulas in modern clinical practice, NAL-NL2 and DSL v5.0, share the same acoustic inputs but optimize for different goals — one for adult speech-intelligibility-per-loudness-unit, the other for a still-developing pediatric auditory system that needs full audibility of soft speech without discomfort.
NAL-NL2 (Keidser, Dillon, Flax, Ching, Brewer 2011) is a nonlinear descendant of the original 1976 NAL linear formula. It computes target gain by maximizing the Speech Intelligibility Index subject to a loudness-normalization constraint: amplified speech should sound neither too loud nor too soft relative to a normal-hearing listener's loudness perception. It adjusts for hearing loss asymmetry, bilateral vs. unilateral fitting, adult vs. child, language (tonal languages get slightly different high-frequency emphasis), and experienced vs. new users (new users are prescribed slightly less gain than the full target, ramped up over weeks — "acclimatization").
DSL v5.0 (Developmental / Desired Sensation Level, Scollie et al., Western University) instead targets a loudness-growth curve: the goal is that amplified sound occupies the same proportion of the child's residual dynamic range as it would in a normal-hearing child, ensuring soft speech (~50dB SPL) stays fully audible above threshold while loud sound stays below UCL. Because pediatric ear canals are small and resonant properties change dramatically with growth, DSL v5.0 mandates individual RECD measurement rather than using age-based averages, and it has a dedicated cochlear-implant variant, DSL v5.0-CI, for mapping electrical stimulation levels.
Both formulas generally prescribe more low-frequency gain restraint than the linear 1976 NAL formula did (to avoid low-frequency upward spread of masking) and both prescribe compression ratios roughly proportional to the slope of the loss per channel — steeper loss, higher compression ratio — which is exactly the channel-by-channel target this simulator's gain-frequency plot renders as the gold dashed curve.
Modern digital hearing aids do not apply one flat volume boost. They split the incoming signal into multiple frequency channels — commonly 6 to 20 depending on the platform — and apply independent wide dynamic range compression (WDRC) to each: its own gain, its own compression kneepoint, and its own compression ratio, so that soft sounds are amplified more than loud sounds within every band, restoring audibility without exceeding loudness comfort.
A hearing aid channel behaves linearly (1:1, output rises 1dB for every 1dB of input) below its kneepoint, then compressively above it: for a channel with a 2.5:1 compression ratio, a 10dB increase in input level produces only a 4dB increase in output. The kneepoint is usually set low — 40 to 55dB SPL — so that compression is already engaged for average conversational speech (~65dB SPL), which is why WDRC is sometimes described as acting almost continuously rather than only "kicking in" for loud sounds.
Channel-by-channel assignment follows the audiogram slope: channels covering frequencies with steep threshold elevation (here, 3–6kHz, where thresholds drop to 65–70dB HL) are given higher compression ratios, because the residual dynamic range between threshold and UCL is narrower there and needs more aggressive squeezing to fit amplified sound inside it — this is the recruitment phenomenon, the abnormally rapid growth of loudness above threshold that is a hallmark of cochlear sensorineural loss. Channels over frequencies with milder loss (250–500Hz) use lower ratios closer to linear, preserving natural loudness contrasts where the ear still has close to full dynamic range.
At this stage the gain curve is only coarsely matched to the prescriptive target — programmed from the formula's calculated values and the manufacturer's fitting software default curves, but not yet verified against what is physically happening in this patient's ear canal. Individual differences in ear canal resonance, venting, and receiver coupling routinely produce 5–10dB deviations from the predicted response at this point, which is exactly the gap real-ear measurement exists to close.
Real-ear measurement (REM) places a thin silicone probe tube, terminating within 4–6mm of the tympanic membrane, alongside the hearing aid in the patient's own ear canal. A calibrated loudspeaker presents a standardized signal — typically the International Speech Test Signal (ISTS) or a speech-shaped composite noise at soft (55dB SPL), average (65dB SPL) and loud (80–90dB SPL) input levels — and the probe microphone measures the real aided output. Gain is then trimmed until the measured curve matches the prescriptive target within tolerance.
Manufacturer first-fit algorithms only approximate the prescriptive target: they rely on population-average real-ear-to-coupler difference (RECD) values rather than this patient's actual ear canal acoustics, and they cannot account for venting, dome/mold style, or slit leaks that change the low-frequency response. Studies auditing first-fit accuracy have repeatedly found average deviations from prescriptive target of 5–10dB at some frequencies before any real-ear adjustment — enough to meaningfully undershoot audibility for high-frequency consonants or overshoot comfort for low-frequency vowels.
During REM, the clinician records the Real-Ear Aided Response (REAR) at each test level and overlays it on the NAL-NL2 or DSL v5.0 target. Where the measured curve falls outside the ±5dB tolerance band, gain in that channel is trimmed up or down in the fitting software and the measurement repeated — an iterative closed loop rather than a one-shot calculation. This step also verifies that the Real-Ear Aided Response at loud input levels does not exceed the Real-Ear Saturation Response ceiling relative to the patient's uncomfortable loudness level (UCL), protecting against acoustic overload.
Despite unambiguous evidence that REM-verified fittings produce better speech recognition and patient satisfaction than fittings accepted at manufacturer first-fit, published survey data (Mueller & Picou, and subsequent AAA practice surveys) has repeatedly found that only a minority of US dispensing audiologists perform REM on every fitting — a persistent best-practice gap despite REM appearing explicitly in the American Academy of Audiology Pediatric and Adult Amplification Guidelines.
Practice surveys spanning more than a decade have found that fewer than half of US hearing aid fittings are routinely verified with real-ear measurement, even though the American Academy of Audiology's Clinical Practice Guidelines list REM as the single recommended method for confirming a fitting matches prescriptive targets. Clinics that skip REM and accept the manufacturer's simulated first-fit curve are, in effect, fitting to a population average ear canal rather than the patient in the chair — a gap directly tied to lower measured speech-in-noise benefit in outcome studies.
A verified gain curve is only usable if it is stable against acoustic feedback and intelligible in noisy environments. Adaptive feedback cancellation algorithms extend how much gain can be applied before whistling; directional microphone arrays and beamforming steer sensitivity toward the target talker; and digital noise reduction attenuates channels dominated by stationary noise — three layers of signal processing tuned on top of the frequency-gain response already verified by REM.
Adaptive feedback cancellation (AFC) models the feedback path — the acoustic loop from receiver back to microphone through vents and slit leaks — using an adaptive filter (commonly an NLMS, normalized least-mean-squares, algorithm) that continuously estimates and subtracts the feedback signal before it re-enters the amplification path. Modern AFC, combined with phase-shift and frequency-shift techniques, typically buys back 10–15dB of usable gain before the onset of feedback compared to a device with static suppression alone — gain that can now safely be delivered at the high-frequency channels the audiogram most needs, e.g. Phonak's and Widex's adaptive AFC implementations, or Signia's.
Directional microphone arrays (typically 2 omnidirectional mics per ear, combined adaptively) create a polar pattern that attenuates sound arriving from the sides and rear relative to the front, improving the signal-to-noise ratio for a talker directly ahead by roughly 4–6dB versus an omnidirectional microphone — one of the most consistently reproduced benefits in hearing aid outcome literature. Modern platforms extend this with full 360-degree scene analysis: Oticon's OpenSound Navigator, for instance, continuously analyzes the acoustic scene at a high update rate and applies noise removal before directionality and gain, rather than after, aiming to preserve spatial awareness of background sound rather than aggressively narrowing focus. Widex SoundRelax and Phonak's SoundRecover2 (frequency lowering, shifting inaudible high-frequency information like /s/ and /f/ down into a residual-hearing region) address a complementary problem: audibility of high-frequency phonemes that even maximum gain cannot restore for a steeply sloping loss.
Digital noise reduction (DNR) estimates the modulation depth of each frequency channel over time — speech is modulated (amplitude rises and falls with syllables), continuous noise (fan hum, engine noise) is not — and applies channel-specific attenuation, commonly 3–8dB of gain reduction in noise-dominated channels, to reduce listening effort even where it does not directly raise measured speech recognition scores.
A fitting that measures correctly on a probe microphone must still be validated against real-world function. The final stage confirms the aided sound-field audiogram, quantifies speech understanding in background noise, and captures the patient's own report of benefit in the specific listening situations that matter to them — three independent lines of evidence that the electroacoustic fitting has translated into usable hearing.
Functional gain testing repeats the audiogram in the sound field with the hearing aid worn and active, comparing aided to unaided thresholds. Because sound-field testing includes head diffraction, real ear canal acoustics, and the full signal chain, functional gain typically runs a few dB below the insertion gain figure measured by REM — both are valid, complementary measures rather than contradictory ones.
Speech-in-noise tests directly quantify the outcome patients care most about: understanding speech against a competing background, which pure-tone thresholds and even REM curves cannot fully predict. QuickSIN presents sentences in four-talker babble at decreasing SNR and reports an SNR-loss score in dB relative to normal-hearing performance; HINT (Hearing in Noise Test) adaptively finds the SNR needed for 50% sentence recognition. Well-fitted, well-processed hearing aids with effective directional/noise-reduction processing can measurably reduce SNR loss compared to an unaided or poorly fitted condition, though no current technology fully restores normal-hearing performance in noise.
Patient-reported outcome measures complete the picture. The APHAB (Abbreviated Profile of Hearing Aid Benefit) scores everyday difficulty across four subscales — Ease of Communication, Background Noise, Reverberation, and Aversiveness of sounds — before and after fitting. The COSI (Client Oriented Scale of Improvement), developed at the National Acoustic Laboratories in 1997, instead starts from the patient's own nominated goals ("hear grandchildren at the dinner table," "hear the TV without disturbing others") and rates degree of change and final ability for each — a individualized outcome measure precisely because "successful fitting" ultimately means the goals the patient walked in with were met, not just that a curve on a screen matched another curve within 5dB.
A landmark finding repeated across decades of hearing aid outcome research: objective electroacoustic accuracy (a REM curve matching target within tolerance) correlates only moderately with patient-reported satisfaction. Two patients can leave the clinic with electroacoustically identical, perfectly verified fittings and report very different real-world benefit — which is precisely why the clinical workflow does not stop at REM, and why COSI and APHAB remain standard final steps rather than optional extras.