Call legs that have a PSTN number don't display the audio level. Plivo SDKs can't detect one-way audio on a PSTN leg if the audio level is too low on one of the streams. We can detect the audio level if the packet count on one of the streams is zero. If both of the streams have a packet count of zero, the call leg is labeled as “no audio.”
Why is the audio level not displayed for a few call UUIDs?
Why are ring start time and answer time displayed as zero seconds if a call has all the other stats?
This happens when a call is auto-answered.
What do the audio quality stats mean?
Audio quality stats display the network-level measures such as jitter, packet loss, audio level, and round trip time (RTT). These are color-coded as green (good), yellow (average), and red (bad) based on industry-standard thresholds.
<= 10 ms
10 – 30 ms
0.5% – 0.9%
-80 to -40 dB
< 200 ms
200 – 300 ms
> 300 ms
The values for each of these metrics are displayed using percentiles.
Reading audio quality stats
Statement 1: 30th percentile of jitter is 8 ms.
- If all the jitter values captured on a call are sorted in ascending order, then 30% of these values are ≤ 8 ms.
- This equates the capture of each value with a unit of time in the call. It can also be read as 30% of overall call duration — not necessarily the first or the last 30% — has jitter ≤ 8 ms.
- Since percentiles indicate values sorted in ascending order, this also indicates that 100 - 30 = 70% of the call has jitter > 8 ms.
Statement 2: 95th percentile of jitter is 32 ms while all the lower percentiles are ≤ 30 ms.
- 95% of the jitter values captured on the call, or 95% duration of the overall call, has jitter ≤ 32 ms.
- This falls in the bad range of jitter and will be color-coded in red.
- 100 - 95 = 5% of call has jitter ≥ 32 ms. This is also in the bad range.
Statement 3: 99th percentile of jitter is 30 ms.
- 99% of the call has jitter ≤ 30 ms; therefore, 99% of the call is not within the bad range.
- 100 - 99 = 1% of the call has jitter > 30 ms, hence 1% of the call is within the bad range.
What is the difference between an external SIP device and an external SIP endpoint?
An external SIP device is any non-Plivo SIP device (such as X-Lite) that's registered with Plivo. An external SIP endpoint is a non-Plivo SIP device that's not registered.
Why is data missing for few calls?
The Call Insights UI displays data for calls over the most recent three months. Suspected issue (displayed at the top) and audio quality stats require call duration to be greater than 10 and five seconds respectively. Plivo displays quality scores only for calls longer than 10 seconds.
You can see suspected issues, insights, and the audio quality stats sections for all client SDK calls (Browser SDK, iOS SDK, and Android SDK). Plivo can display data for PSTN calls when carriers have enabled RTP reports, but they will not show audio level in the audio quality stats section.
What do “ring start time” and “answer time” mean?
Ring start time represents the time it takes for the ring to start after the call has been initiated, represented in + seconds. Answer time represents the duration it takes for the call to be answered after the call starts to ring. Answer time in the call details section is the same as the ring duration in the call stats section. Ring start time is the same as the post-dial delay.
How are audio quality stats categorized as good or bad?
See our post on audio quality stats for the full explanation.
Why is Plivo’s quality score "good" when a call had suspected issues?
Plivo’s quality score is the standard mean opinion score (MOS). The MOS is computed based on factors of jitter, packet loss, and RTT. Unless the suspected issue is broken audio, robotic audio, and audio lag, the quality score computation is not affected.
In addition, the quality score may not match the quality if the suspected issue is related to an associated call_UUID.
Why are suspected issues displayed on top beside the call_UUID, and those displayed under the call relations section different for the call_UUID?
Suspected issues that are listed beside the call_UUID at the top are displayed considering all the related call legs. An issue for one call_UUID may have an impact across all call_UUID legs. The call relations section displays suspected issues for respective call legs.
Please note that
- Suspected issues of “siblings” of a call_UUID are not displayed at the top, but are shown in the call relations section.
- In case of acquaintances, a suspected issue of the specific leg involved in a conference call with the searched-for call_UUID are included at the top, while the suspected issues for all the internal legs are displayed under the call relations section.
What do the percentages mean in the audio quality stats section?
Call values for each of the metrics (jitter, racket loss, round trip time) vary with each voice data packet transmitted; these are represented as percentiles. A 30% value of jitter 45 ms means out of all jitter values that occurred during the course of a call (with each voice data packet), 30% of the values are less than or equal to 45 ms. That means 30% of your call had jitter less than or equal to 45 ms.
Why were some participants disconnected from a conference call?
Use the call_UUID of the conference call to search for call details in the call relations section of the UI. Hangup source and hangup reason are mentioned beside each call_UUID marked as acquaintances.
Why did my call disconnect?
You'll find hangup causes and reasons why a call was ended in the Call Stats section of the Call Debug UI.