We have a fairly complex VOIP setup consisting of three freePBX servers, two on the mesh wan and one on the internet. They are trunked together and phones on any of the freePBXs can dial any other phone.
The issue that we experience occasionally is that phones attached to certain 2.4 GHz nodes (a mix of ubiquity gear) experience a loss of voice quality. When the issue begins voices begin to sound tinny, then voice quality deteriorates rather rapidly, eventually eventually ending up with completely unintelligible voice quality. All you hear are occasional bits of audio that remind me of very long satellite audio paths.
Rebooting the affected node will bring things back to normal until it happens again sometime in the future. The intervals between occurrences of this issue vary. Could be days or could be minutes.
I don't know if this is a firmware issue or equipment issue and honestly don't know where to start troubleshooting. It's not server related, two of the PBXs are on real servers (not rPIs) and one is an (rPI). Personally I think its the node itself although I can't tell if it's the hardware or if it's the firmware. The node that gives us the most trouble is at 3.16.1.1. The signal quality is very good.
Just wondering if someone else has seen this issue or can point me in the right direction.
73,
Thanks!
Cheers!
Chuck...
WB6YOK
You are here
VOIP Bandwidth Issues
Thu, 09/13/2018 - 18:21
#1
VOIP Bandwidth Issues
Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer
Joe AE6XE
Orv W6BI
Ran MTR for about sixteen hours and picked up an instance of a 4500 ms latency. Something is up. I'd bet my hat the radio is fubar. We've been thinking about replacing it anyway so this jsut adds fuel to the fire.
73,
Chuck...
WB6YOK
Chuck,
Some things I'd check.
1. You might check to see if this is a memory issue in the nodes in question. The next time you have to reboot a node, make note of the free space just before you do the reboot. And then make note of the memory after the reboot. After doing this several times and keeping track of what you find, a trend may become apparent.
2. Do any of the nodes in question run meshchat, or anything else that might consume memory or processor?
3. You might keep track of your nodes list size and see if there's any correlation between your issue and if the the nodes list size suddenly increases. Measuring nodes list size is a new feature in the 3.18.9.0 code and is displayed both on the node's home page and the nodes list page.
4. Speaking of 3.18.9.0, you should consider upgrading regardless as there are several enhancements that reduce the memory size of a node. This in itself may give you better performance.
5. I'd check your RF network for packet loss. In particular, if you have a lot of people all on one channel and who don't all hear each other well, you most likely are suffering from "hidden terminal syndrome". When people can't hear each other, they start stepping on each other with the result that packets get corrupted and dropped. With tcp based connections, such as web, the retry mechanism eventually gets packets through. However, voip is udp based and once a packet is dropped, it is lost for good. Get enough of these lost packets and you start to get lost speech.
In general, I don't think the horsepower of your asterisk boxes is insufficient. I've found that a raspberry pi is capable of more performance than what people give it credit for. If you do suspect that processor power is an issue, asterisk is capable of a mode that directly connects end points together (phones) once the call is set up. It then drops out of the streaming audio connection, saving processor. Of course, all bets are off if asterisk has to transcode between different codecs for a call or certain other conditions.
Mark