question to mberlant

[nowshow]

hi mberlant , you are in Japan, aren't you? I'd like to ask the question about latency between Japan and other Asian country. When you make calls from Japan to Hongkon or China , do you feel longer latency? Last time I called from beijing to shanghai, the lantency is unacceptable.but calls from beijing to US city had no problem. what is your experience?

[ethernet]

There is almost zero latency between my location in central Russia and the Russian Far East that borders China and Japan (Vladivostok, Magadan or the Central Asia republic of Kyrgyzstan), if that helps.

[mberlant]

Latency is dependent upon the entire route the call must take. In my case, if I am using BroadVoice to call China, for example, the call goes via the internet from my location in Japan to the BroadVoice Los Angeles proxy server I am homed to. From there it goes on to BroadVoice's outbound gateway, also in Los Angeles (I believe) and BV's termination carrier for China brings it to an in inbound gateway city in China (Guanzhou, Beijing and Shanghai are popular), from where it joins the Chinese PSTN.

I can only measure the piece of the puzzle between my own location and proxy.lax.broadvoice.com, which averages about 160ms ping time, down very recently from a 210ms average. In any case, voice quality (which I believe counts the most), generally ranges from "good cellular" to "sounds like next door". This is a very subjective evaluation, but I have no quantitative metrics to measure.

If you are experiencing good quality calling some places (like the US or Western Europe) and poor quality calling other places (like Taiwan or Shanghai), the problem is almost certainly not under your control. As HansJG has complained, and I have echoed, termination to Taiwan is horrible right now.

Your poor calls to certain places are probably being corrupted at the location (wherever that is) where the termination provider hands the call off to the Chinese PSTN. You probably should document the time of day and the phone number you are calling and report the trouble to BroadVoice Customer Service for them to pursue with their termination provider.

By the way, if you are experiencing clear voice quality with poor latency you may wish to reduce the size of your jitter buffer by adjusting the "Network Jitter Level" on your ATA from "very high" to "high", etc., as your ATA permits. This will reduce the amount of time your ATA holds incoming packets waiting for quiet interludes and waylaid packets to show up before presenting the audio to your ear. If your voice quality is so clear that you could tolerate a little bit of "static", as my voice quality often is, this may help your case.

[cayman]

Originally Posted by mberlant

By the way, if you are experiencing clear voice quality with poor latency you may wish to reduce the size of your jitter buffer by adjusting the "Network Jitter Level" on your ATA from "very high" to "high", etc., as your ATA permits. This will reduce the amount of time your ATA holds incoming packets waiting for quiet interludes and waylaid packets to show up before presenting the audio to your ear. If your voice quality is so clear that you could tolerate a little bit of "static", as my voice quality often is, this may help your case.

I've wondered how/why "Jitter Level" worked. Can you explain a bit more about that and other things to consider for its adjustment etc.

TIA,
- Don

[craig]

I'm sure you've directed the question to Michael, but I'll give it a go.

"Latency" is the delay between when you say something and it is heard. The Sipura ATA-3000, which is what I have, measures network latency as well as codec latency. The first is the time it takes the network to send the packets over the network and the second is the time it takes to turn your voice to data and visa versa.

"Jitter" is a measurement in the variability in receipt of the packets. Each packet is expected at a certain time and deviation is considered jitter.

So, after sending a number of packets you have a statistical average of how long it takes them to arrive (network latency). You also have a calculation as to how variable the arrival times are (jitter).

As for the jitter buffer, it is an attempt to smooth out the variability in arrival by holding the packets long enough for subsequent ones to arrive. This results in smooth audio with few gaps/breaks. However, you are holding the packet up and so you increase the audio latency to the user.

There is a trade to be made between loss of audio and delay of audio. By decreasing the buffer, as Michael suggested, you reduce the latency at the expense of possible lost packets.

Craig

[mberlant]

Don,

What he said. ;^)

Michael