VOIP Tech Chat → Re: Best way to test lag on a voip call or phone system?

If you want a controlled test, I recommend using a sound recorder and calling an echo test number (a good PSTN number is 909-390-0003). Then make a snap or clap noise while recording the sending and return sound. Then look at the resulting wave form and measure the exact time delay (for Windows, a nice freeware editor for this is Audacity).

said by »www.voip-info.org/wiki/view/QoS :

---

Callers usually notice roundtrip voice delays of 250ms or more. ITU-T G.114 recommends a maximum of a 150 ms one-way latency. Since this includes the entire voice path, part of which may be on the public Internet, your own network should have transit latencies of considerably less than 150 ms.

---

That's a good idea then. So just to confirm I'm doing it right, as a quick test from my cell using Bria softphone connected to voip.ms, I put it on speakerphone called that number and recorded it with Audacity. I then snapped my finger and had it play that back.

Then in Audacity I zoomed in, and clipped the recording so that the very start of it was the immediate start of the snap, and the end was the immediate start of the playback snap.

1) Using Bria registered to my hosted Asterisk provider which uses voip.ms as a trunk, it showed the delay between the two was about 0.52 of a second. Not sure how that works as that's 520 ms which is really high isn't it?

2) I then called from my cell phone into a voip.ms DISA directly and did the same. This time it was 0.68 seconds.

3) I then called into RingCentral who we have as a virtual PBX, so when someone calls it, it just forwards the calls to our cell, so it's sort of like a hosted PBX provider and used their DISA which they told me has very low latency for this type of application as they said they are connected directly to the CLEC. This one however was about 1.05 seconds.

4) Using my cell phone I called the number directly. This resulted in about 0.71 seconds.

5) Using my hosted Asterisk connected also to voip.ms, I repeated it and this time got about 0.73 seconds.

Now not sure how to interpret that because when I use option 3 it always seems like calls sound totally fine, yet that seems to be the worst of the results when I thought it would be the second best after using the cell directly. If I interpret the above correctly, then Bria wins by far, as it was much shorter where as my RingCentral which was supposed to be really good was twice as long as Bria. However am I correct to assume that basically whether I make a call direct from my cell phone, through voip.ms DISA, or through DISA on my own PBX, the delay is pretty much the same and that all 3 options should provide good results?

Thanks for your suggestions.

Yes, that sounds like some pretty valid tests. Note, with Audacity, you likely don't need to actually crop anything since you should be able to just highlight the start/stop and it should show the time delta.

Your phone may add extra delay due to echo suppression when you turn on the speaker phone, so if you have a microphone for the recorder to listen to the phone earpiece while also being able to hear the originating sound, I'd use that method. Or you could try to conference in a third connection from your phone and record it from there (via recorder or voice-mail).

Those latency number seem a bit high, but when you include all of the paths, they may really be that high and that might be the norm for these connections. The goal should be less than 250-300ms round-trip. The best round-trip number I recorded was when I used a Verizon PSTN connection in MD calling the PSTN echo test number in CA (conferencing in a third number for the recording) and measure 70ms round-trip. Using my Asterisk box recording feature, I typically measured 170-230ms with VoIP.ms and GV (tests conducted a couple years ago). Note, these Asterisk tests missed the extra delay on my ATA and my phone.

In summary, I think your tests are all valid and I'd shoot for the best latency when possible, but the most important thing to test is your actual call experience with real conversations. If you start walking on each other, then you may want to try to change your speaking/listening patterns or keep looking for the ultimate path.