Baby Monitor Sound Quality: What DECT, FHSS and Wi-Fi Do to a Cry

Three radios, three failure modes

DECT 6.0 in North America runs between 1.92 and 1.93 GHz, in a band reserved for cordless telephones. Nothing else in a normal house uses it, which is why a DECT monitor sounds the same on a Tuesday night as it did on the day you plugged it in. The voice path is compressed to roughly 32 kilobits per second, sampled at 8 kHz, and limited to about 300 to 3,400 Hz. That is the telephone band, and it is enough for a cry.

2.4 GHz FHSS hops across the 83.5 MHz between 2,400 and 2,483.5 MHz, changing frequency many times a second. Hopping dodges interference rather than avoiding it, and it leaves room for a video stream. The cost is that the band also carries your router, your doorbell, your microwave, and the neighbour’s camera.

Wi-Fi is a packet network. Audio is encoded, packetised, sent through a router, and reassembled with a jitter buffer that holds 100 to 500 milliseconds of sound so the playback does not stutter. That buffer is why Wi-Fi monitors feel like a walkie-talkie on talk-back. It is also why the audio can be excellent for hours and then turn to gravel during a router firmware update.

What compression removes

An 8 kHz sample rate can carry frequencies up to 4 kHz, and nothing above it. The filter that enforces this is not a defect; it is the Nyquist limit. A baby’s cry has a fundamental around 300 to 600 Hz, harmonics that reach 3 or 4 kHz, and consonant energy from 4 kHz upward. The link keeps the fundamental and most of the harmonics, and loses the top of the consonants.

That is why a cry through a monitor sounds rounder and less urgent than a cry in the room. The vowel-like wail survives, the sharp edges soften.

Codec artifacts are the second loss. Adaptive differential pulse code modulation packs each sample into 4 bits, which is efficient and introduces a low-level granular noise that rises with the signal. When a cry gets loud, the automatic gain control pulls the level down over a few hundred milliseconds, and the background noise comes up with it. That slow breathing effect is pumping, and it is more tiring to listen to than plain hiss.

Why hiss is not a defect

Every audio path has a noise floor, and in a quiet nursery that floor is the only thing the microphone is picking up. The hiss you hear is a mix of the microphone’s own thermal noise, the radio link’s residual noise, and the parent unit’s amplifier. It is steady, broadband, and quiet. A monitor that produced no hiss in a silent room would be a monitor with its gain turned down too far to hear a whimper.

Automatic gain control makes the hiss audible. In a silent room, the unit turns its gain up by 10 or 20 dB to find something to amplify, and the noise floor comes up with it. When the baby cries, the gain drops and the hiss disappears behind the signal. So the hiss is loudest exactly when the room is quietest, which is when you notice it and when it matters least.

Real faults sound different. A hiss that changes when you move your hand near the parent unit is interference. A click or a chirp at regular intervals is often a Wi-Fi beacon. A 60 Hz hum is the power supply. A crackle when the cable flexes is a broken conductor. Unplug the baby unit: if the hiss stays, the problem is in the handset, and no amount of repositioning the camera will fix it.

DECT against FHSS in a house with Wi-Fi

Wall penetration favours the lower frequency, but only slightly. At 1.9 GHz the wavelength is about 6.2 inches; at 2.4 GHz it is about 4.9 inches. Longer waves bend around the edges of doorways and absorb less in drywall. The bigger difference is occupancy. DECT owns its band. FHSS shares 83.5 MHz with every access point in range, and a video stream needs a steady megabit or two that the router is also trying to use.

Neither standard is better at everything. DECT does speech well and video badly, because 32 kbps has no room for pictures. FHSS does video and acceptable audio, and it degrades in a way you can see: blocky frames, then a freeze, then a catch-up. If your priority is hearing a cry clearly, DECT is the safer pick. If you need eyes, you are in FHSS territory and the audio is a passenger.

The interference cases that actually bite are mundane. A monitor within 6 feet of a microwave leaks at 2.45 GHz and stutters when the microwave runs. A camera next to a mesh node shares its airtime with the node. A handset on a metal shelf loses a few dB to the shelf. None of these are visible in a spec sheet, and all of them show up as audio dropouts.

Wi-Fi audio: good until the network hiccups

A Wi-Fi camera can carry 16 kHz audio at 48 kbps, which is wider than any DECT monitor and close to a good phone call on a landline. In a quiet house on a clean channel, it sounds better than the radio alternatives.

The floor is lower. Adaptive bitrate means the camera will trade audio quality for latency and stability when the network is busy, and it does so without telling you. The result is a stream that drops to 8 kHz mono during a video call upstairs and comes back an hour later. Two-way talk is worse, because the buffer that smooths playback also delays your voice, and a reply that arrives 600 milliseconds late is not a reply.

There is also the question of where the audio goes. A Wi-Fi camera encodes the room and sends it to a server, then back to your phone. A DECT or FHSS monitor sends it to the handset in your hand. If the sound of your nursery leaving the house matters to you, the privacy piece covers what that path actually exposes.

How to judge sound in five minutes

Record a cry. Any recording of a baby crying will do, played at roughly 70 dB from the crib position, which is about the level of a real cry at 6 feet. Listen for three things: whether you can make out the catch in the breath, whether the level pumps up and down, and whether the loud parts crackle. A monitor that passes all three is doing its job.

Then test the silence. Put the handset where you sleep, with the nursery quiet, and listen for a minute. Steady hiss is normal. Pulsing or clicking is not.

Finally, walk. Take the handset to the far corner of the house, close the doors, and listen for dropouts rather than distance. The range number on the box is measured in open air, and a monitor that holds 40 feet through two walls is doing better than one that reaches 200 feet down an empty hallway. The audio-only shortlist is where most of this testing gets done, because those units have nothing to hide behind.