Audio Signal Level

Understand audio signal level measurement in VoIP calls — how CallMeter detects speech volume using VAD-gated RMS levels, and what abnormal levels mean for call quality.

Audio Signal Level measures the loudness of speech in a VoIP audio stream, expressed in dBov (decibels relative to digital overload). Unlike a simple volume meter, this measurement is gated by Voice Activity Detection (VAD) — it only measures level during moments of active speech, ignoring silence periods entirely.

Think of it as measuring how loud someone is speaking, but only when they are actually talking. This gives you a true picture of speech volume without silence pulling the average down.

How It Works

CallMeter analyzes the decoded audio signal and uses Voice Activity Detection to identify speech segments. During those segments, it calculates the RMS (Root Mean Square) power of the audio waveform and converts it to dBov.

The dBov scale has 0 dBov as the absolute maximum (digital full-scale). All real speech falls below this:

-10 to -15 dBov — Loud speech, approaching clipping
-20 to -30 dBov — Normal conversational level
-40 to -50 dBov — Quiet speech, may be difficult to hear
Below -50 dBov — Near-silent, likely inaudible

VAD-gated measurement

Because this metric ignores silence, it reflects the actual speech level regardless of how much of the call is silence. For overall level including silence, see Audio Level RMS.

Why It Matters

Speech volume directly impacts intelligibility. If the signal level is too low, listeners strain to hear or miss words entirely. If it is too high, the audio clips (distorts) at the digital ceiling, producing harsh, unintelligible sound.

For enterprise VoIP testing, signal level reveals:

Endpoint configuration issues — Microphone gain set too low or too high
Codec gain problems — Encoder or decoder applying unexpected gain adjustments
Media file issues — Pre-recorded audio used in tests normalized to the wrong level
Network path transparency — Whether SBCs or transcoders are altering audio levels

Common Causes of Abnormal Signal Levels

Cause	Explanation
Microphone gain too low	Hardware or OS-level input gain producing quiet audio
Microphone gain too high	Input clipping before the signal reaches the codec
AGC (Automatic Gain Control)	Software AGC overcorrecting, causing pumping or inconsistent levels
Transcoding artifacts	SBCs re-encoding audio at different levels
Echo cancellation side effects	Aggressive echo suppression attenuating speech
Test media file levels	Pre-recorded audio files with incorrect normalization

How to Fix It

Check media file levels — If using pre-recorded audio for testing, ensure files are normalized to around -20 dBov RMS for natural speech levels.
Review endpoint gain settings — Verify that microphone and speaker gain are configured to standard levels on test devices.
Disable or tune AGC — If Automatic Gain Control is causing level fluctuations, consider disabling it or adjusting its target level.
Test through the media path — If levels change between send and receive, an intermediary (SBC, transcoder, or conference bridge) may be modifying the audio.
Compare send and receive levels — Large differences between directions indicate a gain mismatch at one endpoint.

Audio Level RMS — Overall RMS level without VAD gating, includes silence periods
Audio Noise Level — Background noise during non-speech moments
Speech Activity (VAD) — Percentage of time with active speech
Opus Decoder Gain — Volume adjustment applied by the Opus decoder

RFC Reference

Audio level measurement methodology follows ITU-T P.56 (Objective Measurement of Active Speech Level). The dBov unit is defined relative to the digital full-scale point of the audio signal.