Quality you can hear and measure
Listen to what a direct wideband path can preserve
Use CarrierX's published audio samples, call both live comparison numbers, and inspect the measurement method behind the reported listening-quality ranges.
Audio samples
Same source. Different path
These WAV files are currently published by CarrierX for HD Direct. Use headphones or a quiet room, then test with your own endpoints and models for the comparison that matters most.
Hear the difference from a mobile handset
Call from an AT&T, T-Mobile, or Verizon handset so the HD Direct path can be evaluated as intended, then compare it with CarrierX's published standard-route line.
Same recording. Different route.
Hear both paths in sync
What is in the recording?
Both samples contain the same sequence: music, a narrated CarrierX message about the call path, a customer-service greeting, and a second music excerpt. The comparison is sound quality, not different content.
Review the CarrierX source and methodologyAvailability and audio behavior can vary by handset, carrier, codec negotiation, and route. The displayed numbers match CarrierX's current HD Direct product page.
By the numbers
CarrierX's published listening-quality ranges
MOS-LQO is an objective listening-quality score produced by comparing a degraded recording with a high-quality reference. The scale runs from 1 to 5.
CarrierX reports that HD Direct AMR-WB samples averaged 67% better than the third-party PSTN samples in this test. This is an attributed CarrierX test result, not a universal performance guarantee. Review the original CarrierX results.
How CarrierX measured
A transparent summary of the published method
The comparison checks recorded calls against common source material. CarrierX does not publish the route mix, sample count, or independent validation, so the result remains clearly attributed.
- 01
Three call paths
CarrierX recorded third-party PSTN calls, CarrierX Direct calls using G.711 μ-law, and HD Direct calls using G.722.2 AMR-WB.
- 02
Common source material
Multiple high-quality recordings were played into each call so the captured result could be compared with the original source.
- 03
ViSQOL comparison
Google ViSQOL compared each degraded recording with its high-quality reference and produced an objective MOS-LQO score.
- 04
Published ranges
CarrierX published the score ranges and average improvements shown here.
AMR-WB vs. legacy narrowband
More of the voice reaches the other side
A wider speech range preserves more tone, nuance, and phonetic detail before the conversation reaches a listener or model.
Human speech
Approximately 100 Hz–17,500 HzThe broader source range contains tone, nuance, and detail.
G.711 μ-law
Approximately 300 Hz–3,300 HzLegacy narrowband prioritizes intelligibility while losing vocal detail.
AMR-WB
Approximately 50 Hz–7,000 HzWideband audio preserves more of the speaker's natural sound.
Why signal quality matters to AI
AI can only work with the signal it receives
Audio quality significantly influences AI performance, especially in audio processing, speech recognition, and natural-language understanding. Cleaner wideband input gives downstream systems more useful speech information before analysis begins.
Speech detail Preserve more nuance, tone, and phonetic information before downstream processing begins.
Fewer conversions Direct IP routing avoids repeated TDM handoffs and codec conversions on supported routes.
Test in context Evaluate quality with your own models, endpoints, devices, and production-like traffic pattern.
Listen before you believe us
The published sample is the beginning of the evaluation
Request qualified sandbox access to compare codec behavior, route quality, latency contribution, and call integrity in your environment.