dslreports logo
Search similar:


uniqs
1071
Stewart
join:2005-07-13

1 edit

Stewart

Member

What might cause this choppy voice?

Click for full size
While waiting to pick up his kid, a friend in suburban New York with an iPhone 6 on VZW called my Reno DID (AnveoDirect, CLEC is Bandwidth.com). Inbound audio was terribly choppy, though he could hear me fine.

The audio sounded like bad VoIP in the old days, when missing samples from lost or excessively delayed packets were simply replaced with zeros, causing audible clicks at the start and end of dropouts.

Modern systems don't do that. The engineers of cellular codecs included clever "error concealment" algorithms that replace the missing data with an interpolation of the previous and following sounds. Even with G.711 and VoIP, decent IP phones fade down the last few good samples and fade up after the gap, eliminating the clicks.

When he got home, he called from his VZ VoIP line and sounded perfect. As a test, I asked him to call from his cell and it was bad again, even though he's now on another tower. I called him back and there was no problem, so it's probably not his phone.

I called another friend with VZW (Las Vegas, cheap Android) and asked him to call back; there was no trouble.

Looking at the recorded audio (see screenshot), the missing intervals are 20 ms long, but are not aligned with the incoming packets. The alignment was different for the two calls. How can that even happen? I would think that it would require cellular -> (flaky) VoIP -> TDM -> VoIP. But VZW and Bandwidth are both huge companies and I presume that there is a well defined and direct route between them, either TDM or VoIP.

Any ideas from the experts what may be wrong? Or where the fault may lie? (It can't be Anveo; audio came directly from Bandwidth.) Just a guess: VZW experimenting with VoLTE?
zamarac
join:2008-11-29
Canada

2 edits

zamarac

Member

said by Stewart:

VZW experimenting with VoLTE?

One doesn't have to be a VoLTE expert to learn over the years that most issues in this world, including seemingly technical in nature, are likely to have economic background. Large companies like VZW tend to structure business by divisions and local branches, and those may have different economic targets and the ways to meet them depending on region specifics and management personality.

"Even though revenue from voice calls and SMS is falling, a format for voice over LTE and messaging, it was as necessary to have a viable and standardized scheme to provide the voice and SMS services to protect this revenue." »www.radio-electronics.co ··· olte.php

It might be that choppy VoLTE in certain regions aims to encourage customers to call over cell networks and pay roaming charges, or use their "daytime cell minutes", or such.
engineerdan
join:2006-12-07
Washington, DC

engineerdan to Stewart

Member

to Stewart
said by Stewart:

Looking at the recorded audio (see screenshot), the missing intervals are 20 ms long, but are not aligned with the incoming packets. The alignment was different for the two calls. How can that even happen?

20 ms does smell quite a bit like one or more missing ulaw packets. This is probably an important clue.

I haven't looked at enough VoIP audio waveforms to be able to see what you mean, but it's not apparent to me that the missing intervals aren't aligned with the incoming packets.* It is clear that they're not aligned with the audio zero-crossing points (which is what is causing the clicking), but that's to be expected.

To the contrary, the image provided shows two gaps of the same duration (20 ms?) with a period of good audio between them that's exactly twice the duration of either gap (40 ms?). To my untrained eye, that looks like one missing packet, followed by two good packets, one more missing packet, followed by one or more good packets.
said by Stewart:

Any ideas from the experts what may be wrong? Or where the fault may lie? (It can't be Anveo; audio came directly from Bandwidth.)

Clearly I'm no expert. However, I think you're on the right track by eliminating things unlikely to be the cause. I also think your initial instinct that this is just packet loss may be correct.

My guess is that Verizon Wireless is delivering this call from their switch to Bandwidth by ulaw over an IP network that was suffering some packet loss issues. When your friend moved to a different cell, his call was still originating from the same Verizon Wireless switch, resulting in the same route and symptom. However, his Verizon land line call was likely through a different switch and therefore took a different and unaffected route to Bandwidth.

Could it be as simple as that? Just a hunch.

Incidentally, I've been monitoring a number of Internet routes (to VoIP providers that I depend upon) for a three years now. It seems that in the weeks following the Christmas holidays, there are always indications of congestion (increased ping times, occasional packet loss, etc.) to some but not all carriers. This seemed particularly bad this year, even to previously unaffected providers.

* - My apologies to non-native English speakers for this triple-negative sentence.