Audio Level RMS
Understand overall audio RMS level measurement — how CallMeter tracks the volume of the complete audio signal, thresholds for detecting silence and clipping, and why it matters for call quality.
Audio Level RMS measures the overall loudness of the entire audio signal, including both speech and silence periods, expressed in dBov (decibels relative to digital overload). Unlike Audio Signal Level, which only measures during active speech, this metric captures the full audio picture — making it ideal for detecting silence, clipping, and overall level problems.
Think of it as a VU meter on a mixing console. It shows the total energy in the audio at every moment, regardless of whether someone is speaking or not. This makes it the go-to metric for answering "is there any audio at all?" and "is the audio too loud or too quiet?"
How It Works
CallMeter computes the Root Mean Square (RMS) of the decoded audio waveform within each measurement interval and converts it to dBov. The dBov scale is anchored at 0 dBov (digital full-scale):
- 0 dBov — Maximum possible level, audio is clipping
- -10 to -15 dBov — Very loud, risk of clipping on peaks
- -20 to -30 dBov — Normal conversational audio
- -40 to -50 dBov — Quiet, but still audible
- Below -50 dBov — Near-silent, likely no meaningful audio
Because this metric includes silence periods, the reading is typically lower than the speech-gated signal level. During a natural conversation with 50% speech activity, the overall RMS will be several dB lower than the speech-only RMS.
Why It Matters
Audio Level RMS is the first metric to check when diagnosing fundamental audio problems:
- Is there audio at all? — Readings below -50 dBov indicate silence or near-silence, suggesting a broken media path
- Is the audio clipping? — Readings approaching 0 dBov indicate digital clipping, which destroys audio quality
- Is the level consistent? — Large fluctuations suggest AGC problems or intermittent audio
Thresholds
| Level | Value | Interpretation |
|---|---|---|
| Good | -20 dBov or higher | Audio present at comfortable listening level |
| Warning | -40 dBov | Audio is quiet, may be difficult to hear clearly |
| Critical | Below -50 dBov | Audio is effectively silent, likely a media path problem |
Clipping detection
Values above -3 dBov strongly suggest clipping. At 0 dBov, the audio signal is hitting the digital ceiling and peaks are being flattened, producing harsh distortion. This cannot be fixed by the receiver — the damage happens at the source.
Common Causes of Abnormal Audio Levels
| Cause | Explanation |
|---|---|
| No media flowing | Broken audio path, muted endpoint, or SDP negotiation failure |
| Microphone not connected | Hardware issue producing digital silence |
| Input gain too high | Audio clipping at the source before encoding |
| Input gain too low | Audio barely above the noise floor |
| AGC malfunction | Automatic gain control oscillating or stuck at wrong level |
| Transcoding level shift | SBC or bridge changing audio levels during re-encoding |
How to Fix It
- Check for silence first — If the level is below -50 dBov, focus on the media path: is RTP flowing? Is the audio file playing? Is the endpoint muted?
- Check for clipping — If the level is above -3 dBov, reduce input gain at the source. Clipping cannot be repaired downstream.
- Compare send vs receive — If send level is normal but receive level is abnormal, the issue is in the network path or at the remote endpoint.
- Review AGC behavior — If levels fluctuate wildly, Automatic Gain Control may be overcorrecting. Consider disabling AGC during tests.
- Validate test media files — Ensure pre-recorded audio files are normalized to appropriate levels (around -20 dBov RMS).
Related Metrics
- Audio Signal Level — Speech-only RMS level, gated by VAD
- Audio Noise Level — Noise floor during non-speech periods
- Speech Activity (VAD) — How much of the audio is speech, explaining the gap between overall and speech-gated levels
- Opus Decoder Gain — Decoder gain adjustments that affect the output level
RFC Reference
Audio level measurement follows the methodology defined in ITU-T P.56 (Objective Measurement of Active Speech Level). The dBov unit is specified in relation to the digital full-scale point, with 0 dBov representing the maximum representable amplitude.
Opus Last Packet Duration
Understand Opus last packet duration measurement — how CallMeter tracks the duration of decoded Opus frames, and what deviations from 20ms reveal about encoder configuration.
Keyframe Requests Sent
Understand outbound keyframe requests in video streams — what they measure, why a high count signals video corruption, and how to reduce keyframe request frequency.