İşitmeAtölyesiDevice Technology Workshop
From compression to noise reduction
Brochures list features, but rarely say what they actually do, under which conditions they make a difference, and where their limits begin. This workshop takes the processing inside the device one step at a time: compressing sound, suppressing noise, steering the microphone and moving high frequencies down. Every module comes with a demonstration you can listen to or try for yourself.
Educational · no brands, no models, no recommendations
- 15 modules · audio demos
The workshop runs in three stages
Fifteen modules across three stages: first how sound is shaped inside the device, then which solution actually makes a difference in noise, and finally the limits of the feature package plus a short self-test. Click any heading to jump straight to that module.
How is sound shaped?
The logic and parameters of compression, channel structure and the stability limit.
What actually works in noise?
Solutions that genuinely raise the signal-to-noise ratio, and moving high frequencies down.
Limits and expectations
What a feature tier is really worth, automation, technical limits, daily care and a short self-test.
What does the device do? — core features
—
—
—
—
Active: WDRC · DSP processing stage
Steady background noise (grey) is suppressed; speech (green) is preserved. NR mostly improves comfort and usually does not markedly change intelligibility.
Pick a sound type, then listen uncompressed and with WDRC and compare the output waveform: soft sounds become audible, loud ones are held back.
⚠ Lower your device volume before listening; these are short, low-level simulations. Approximated with the WebAudio DynamicsCompressor.
Source: NIDCD, ASHA and related signal-processing research. The benefit of features varies by environment and person.
Threshold, ratio and time constants
Compression is what makes the whole range of everyday sounds audible in an ear with a narrow dynamic range — but how it is configured matters as much as whether it is used at all. The level at which compression starts (TK), its ratio and its time constants together define the character of the sound. Fast compression lifts soft sounds quickly but flattens the natural peaks and valleys of speech; slow compression preserves that contrast but is late to catch sudden soft sounds.
—
Sources: reviews of WDRC and time constants (Moore; Souza). The fast–slow choice depends on the person, the setting and the shape of the loss; there is no single right answer.
Channel or band? They are not the same thing
Brochures print “16 channels” and “20 bands” side by side, and they are usually read as the same thing. They describe two different functions: band is the frequency region in which gain can be adjusted independently; channel is the region in which compression (and, in most devices, noise-reduction decisions) operates independently. Try it below: you can drag each gain handle (a band) on its own, but you choose the compression ratio at channel level — one channel's ratio moves every band inside it at once.
- 1 At 65 dB, drag the handles onto the target
- 2 Switch the input level to 50 / 80 dB
- 3 Click a channel and change its ratio — every handle in that region moves together
—
Gain handles
Drag the points up and down: each one is the gain of a band. The more bands, the finer the match to target; in a real-ear measurement, these are the handles you use to correct a deviation.
Compression regions
The wide blocks below are channels. Click one and change its ratio: the curve at 65 dB stays where it is, but at soft (50) and loud (80) inputs every bands move together.
This is where the difference shows
Hitting the target at 65 dB with bands is easy. The real test is at 50 and 80 dB, where the curve is governed by channel compression. That is why real-ear measurement is done at three levels; measuring at one level hides what the channels are doing.
Are more channels better?
The number of channels tells you in how many separate frequency regions gain can be set independently. More channels allow a closer fit to the prescription target — but the return shrinks beyond a certain point. Change the channel count and watch how the target-match error responds.
—
Feedback management and the gain limit
When amplified sound escapes from the ear canal and returns to the microphone, the loop closes — this is feedback — and the system starts to whistle. The highest gain that can be delivered before whistling begins is called the maximum stable gain (MSG) . Change the coupling and the gain to see where the limit sits and how much headroom the feedback manager buys you.
Almost no leakage: the widest gain headroom, but the strongest occlusion effect.
Once you know how the sound is shaped, the next question is how much of that processing survives in noise.
What really helps in noise?
In quiet, the processing inside a hearing aid has an easy job; the real exam starts in noise. Digital noise reduction (DNR) improves listening comfort and reduces listening effort, but its contribution to speech understanding has been found limited in the literature. Only two approaches genuinely raise the signal-to-noise ratio: the directional microphone, and the remote microphone — the most extreme form of simply getting closer to the talker. Change the distance and the solutions and see the difference for yourself.
—
These are typical magnitudes: doubling the distance drops the direct sound by about 6 dB (free-field assumption); directionality buys roughly 3–5 dB in real life; a remote microphone buys 10–15 dB of SNR. DNR does not meaningfully improve the signal-to-noise ratio; it improves comfort and listening effort.
Directional microphone: where is the noise coming from?
A directional microphone varies its sensitivity with the angle of arrival: speech from the front is preserved while noise from the sides and behind is attenuated. Choose a pattern below and rotate the noise source to see the theoretical SNR benefit at that angle.
Cardioid: a full null to the rear (180°); noise from the sides is only partly attenuated.
—
Compare omni and directional by listening
A talker in front of you reads the same sentence while café babble comes from behind. This is the most favourable condition for directionality: speech in front, noise behind. Switch to the directional mode and the speech stays at the same level while the noise behind you drops by about 5 dB — that is what raises the SNR.
“I've set the table in the garden; if you like we can have our tea outside, it's quite warm today.”
⚠ Turn your volume down before listening. The sentence is read by your browser's speech synthesiser (the voice varies by browser); the babble is generated with WebAudio. This is a simplified simulation.
Which solution actually buys you SNR?
Understanding speech in noise is governed by the signal-to-noise ratio. Processing inside the device improves it only to a limited degree; moving the microphone next to the talker moves it to another level entirely. Pick a setting and compare what each solution is worth.
—
Improves comfort and listening effort; usually does not meaningfully change intelligibility.
Works when speech is in front and noise behind; the benefit fades in reverberation and when the talker is at your side.
The microphone sits on the talker's lapel: distance and reverberation drop out of the equation.
Frequency lowering and dead regions
Another reason speech is hard to follow in noise is that the high frequencies are only partly heard: sounds such as /s/ and /ʃ/ concentrate their energy between 4 and 8 kHz. Where the loss in that region is profound, adding gain often buys nothing — the cochlea may simply not be responding there (adead region). Frequency lowering moves that energy down into a region the person can still hear. Move the sliders and watch whether the /s/ energy rises above threshold.
Audibility margin: how far the /s/ energy sits above (or below) threshold.
—
Where the high-frequency loss is profound and understanding does not improve with gain, assess with the TEN test.
Live speech mapping with /s/ and /ʃ/, followed by a discrimination test (e.g. /s/–/ʃ/).
Overly aggressive lowering damages naturalness and music perception; that is the criterion.
Sources: Moore (dead regions, TEN test); Alexander et al. (verification of frequency lowering). Schematic representation.
Can you hear the /s/?
The three buttons below play the same /s/-like fricative: first in a normal-hearing ear, then in an ear with a high-frequency loss (the sound disappears), and finally with frequency lowering switched on (the energy is moved down and becomes audible again).
⚠ Turn your volume down before listening. This is a simulation: the impaired ear is represented by a filter that strongly attenuates everything above 3 kHz.
Bilateral fitting, binaural processing and CROS
Hearing does not end at one ear: localisation, separating speech from noise and the sheer ease of listening all rest on the two ears working together. This section gathers what a bilateral fitting buys you, how the two devices talk to each other, and where CROS/BiCROS comes in for a single-sided loss.
—
—
Head shadow, ITD and ILD
At high frequencies the head attenuates the sound reaching the far ear (the head shadow). The brain uses the resulting time (ITD) and level (ILD) differences between the ears to locate a sound and pull it out of noise. With one ear only, those cues are lost; a bilateral fitting improves localisation, separation in noise and listening effort.
Binaural synchronisation
Modern pairs use a wireless link to align level, program and noise decisions; some systems stream sound from one ear to the other (ear-to-ear streaming) and run both microphone arrays as a single beamformer. This can add SNR in noise — but aggressive synchronisation that flattens the natural cues (ILD) can weaken localisation.
CROS and BiCROS
Sound picked up at the unusable ear is transmitted wirelessly to the better ear (CROS). If the better ear also has a loss, that transmission is combined with amplification (BiCROS). The head shadow disappears — but because all the sound arrives at one ear, true localisation does not come back, and expectations in noise must be set carefully.
Once the limits of the processing inside the device are clear, what remains is the real value of the feature package.
Does a premium device let you hear better?
Manufacturers sell features in tiers, and the premium tiers promise more channels, more advanced directionality and smarter automatic programs. Independent research has found the everyday value of that promise to be limited.
Switch the items below on and off, and see how much the factors that really drive everyday outcomes outweigh the feature tier.
—
The weights are schematic: they reflect the relative ordering of effect sizes in the literature, not precise percentages. The point is to show how much the feature tier weighs next to everything else.
The everyday difference is small
In the double-blind, crossover series by Cox, Johnson and Xu, premium and basic devices produced similar everyday outcomes, and participants did not systematically prefer the premium tier. In the laboratory, a small advantage appeared only for localisation of high-frequency stimuli in quiet.
Accuracy of the fitting, and coupling
A basic device placed on target with real-ear measurement usually outperforms a premium device fitted without measurement. Accuracy of gain, appropriate coupling and hours of use decide more than the feature tier does.
Aligning expectations
What a premium tier does add sits on the comfort-and-convenience side, mainly in noisy and changing environments. Recommending a tier before the person's listening needs (COSI goals) are established is a common source of dissatisfaction.
Source: Cox, R. M., Johnson, J. A., & Xu, J. — Impact of Hearing Aid Technology on Outcomes in Daily Life (I–III) and the related laboratory studies.
Automatic programs and datalogging
Modern devices classify the environment continuously (quiet, speech, speech in noise, music, wind) and adjust gain, directionality and noise reduction accordingly. The same system also logs how the device is used — and that log is the most concrete counselling tool you have in the clinic.
—
—
—The average change the user makes with the volume control.
—
Next stepThe example logs are representative data reflecting patterns commonly seen in the clinic. Datalogging is a counselling tool, not a surveillance tool.
Environment classification
The device estimates the environment from the level, modulation and spectrum of the signal, and moves between programs gradually so the user does not hear an abrupt switch. Misclassification — treating music as noise, for instance — can degrade quality, which is why a dedicated music program still makes sense.
What does datalogging tell you?
Average daily hours, the distribution of environments, program changes and volume adjustments are all recorded. Low usage rarely means "the device is bad" — it usually points to comfort, occlusion or expectations.
Using it in the clinic
The log must be read alongside what the patient says: for someone who reports struggling in noise but wears the device only two hours a day, the priority is not a noise program but wearing time. Datalogging is a counselling tool, not a surveillance tool.
Wind, impulse sounds and processing delay
Every process has a price. This section gathers the three technical limits behind the most common everyday complaints.
—
—
Wind noise
Wind creates turbulence directly on the microphone membrane; the source is not out in the room but on the device itself. Signal processing can only do so much: when wind is detected, low-frequency gain is reduced and, where possible, the device reverts to a single (omni) microphone. Microphone position and behind-the-ear placement matter too.
Impulse sounds
Very short, loud sounds — a spoon against a cup, a door slamming — reach peaks that fast compression cannot catch in time. Impulse-noise softening suppresses those peaks within milliseconds; it improves comfort, but set too aggressively it also blunts the plosives of speech (/p/, /t/, /k/).
Processing delay
Digital processing typically introduces 5–10 ms of delay. In closed fittings this goes unnoticed; in open fittings the delayed sound from the device mixes with the sound entering the ear directly and produces comb filtering — the person may hear their own voice as "hollow" or "echoey". In open fittings, low-delay processing and restrained low-frequency gain matter.
Tinnitus generator, charging, moisture and care
How the device performs in the clinic matters — but so does whether it survives daily life. These are the topics most often skipped in counselling.
—
⚠ Turn your volume down before listening. This is a simplified simulation of tinnitus; it is not a diagnostic or therapeutic tool.
—
Tinnitus sound generator
Most devices include an adjustable sound generator (broadband noise, shaped noise or nature sounds). The aim is not to mask the tinnitus completely, but to reduce its perceptual dominance and support habituation. A well-fitted hearing aid already pushes tinnitus into the background by restoring ambient sound; the generator is added alongside that, not instead of it.
Rechargeable or battery?
Rechargeable devices are a clear convenience for users with limited dexterity, and dropping the battery door improves moisture resistance. In return, heavy all-day streaming can run into the limits of a charge; long journeys and power cuts are worth discussing. Zinc-air batteries can be carried as spares, but they are small parts and carry a swallowing risk.
Moisture, cerumen and care
Sweat, moisture and cerumen cause the majority of failures. Wiping the device daily, changing the wax guard regularly, using a drying box overnight and checking the microphone grille all extend its life. An IP rating (e.g. IP68) describes resistance to dust and water — but no hearing aid is "swimmable". A sudden drop in level or a muffled sound is far more often a blocked filter than a broken device.
Which feature solves which complaint?
Eight short cases: read what the patient says and choose the most appropriate solution. After each answer you get the reasoning and a link to the relevant section.
—
—
—
Frequently asked questions
No. Beyond a certain point, extra channels bring no measurable benefit. What decides the outcome is whether the gain sits on the prescription target and whether the time constants suit the person. The big number in the brochure does not fix the curve measured in the ear.
Its effect is limited. Noise reduction improves listening comfort and reduces listening effort; because it does not change the signal-to-noise ratio, no meaningful gain in understanding should be expected. In noise, the real benefit comes from the directional microphone and the remote microphone.
There is no single right answer. Fast compression makes soft sounds audible quickly but reduces the natural contrast of speech; slow compression preserves that contrast but is late to catch sudden soft sounds. The choice follows the person's dynamic range, the shape of the loss and the listening environment.
No. Where gain alone provides adequate audibility in the high frequencies, it is unnecessary. The indication is a profound high-frequency loss or a suspected dead region; when it is used, verify with /s/ and /ʃ/ and avoid excessive lowering.
What is this page based on?
- Moore, B. C. J. Cochlear Hearing Loss — compression, dead regions and the TEN test.
- Souza, P. Effects of compression on speech acoustics and intelligibility — time constants.
- Ricketts, T. A. Directional hearing aids — directionality and real-world SNR benefit.
- Bentler, R. Digital noise reduction: outcomes — the effect of DNR on comfort and intelligibility.
- Alexander, J. M. Individual variability in recognition of frequency-lowered speech — verification of frequency lowering.
- Kates, J. M. Digital Hearing Aids — band (gain adjustment) and channel (compression) architecture.
- Dillon, H. Hearing Aids — multichannel compression, channel count and target matching.
- Moore, B. C. J. et al. — studies on the benefit of multichannel compression as a function of channel count.
- American Speech-Language-Hearing Association (ASHA) and American Academy of Audiology (AAA) — best-practice guidelines.