Voice Quality Choppy/Delayed

[[email protected]]

Hello,
I am getting choppy communication intermittently, sometimes delays. I need someone to look over my configuration and tell me what I am missing. I am not getting dropped packets when talking on the phone – I confirmed this with wireshark.
Hardware/Software/Configuration:
3CX Server – HP DX2400 – Processor – Intel Core 2 Duo @ 2.93GHz Memory – 2GB
• Version 10 with latest service packs
• Abyss Web Server
• NexVortex SIP Trunk
• STUN Turned Off
• Specified IP for the SIP Trunk instead of using STUN – NAT Policies are configured on the firewall
• PX Delivers audio on all extensions and SIP trunk – Supports Re-invite on all extensions
Firewall – Sonicwall NX220 Total Secure
• X1 Interface – DSL – 6MB Down/ 1MB Up
• X0 Interface – IP Address - 192.168.1.251 - unchecked 802.1p tags
• X0:V2 – Vlan-Sub Interface to X0 interface – checked 802.1p tags CoS – Priority 6
• BWM (global) settings for 95% realtime on (LAN) X0:V2 to (WAN) NexVortex Public for 3CX services
• BWM (global) settings for 95% realtime on (WAN) NexVortex Public to (LAN) X0:V2 for 3CX services
• BWM (global) settings for 95% realtime on (LAN) X0:V2 to (LAN) X0:V2 for 3CX services
• ALL VoIP settings disabled
• NAT configuration per 3CX Sample working sonicwall configuration

Distribution Switch – HP V1910-16G
• Trunk1 (int1-4)Untagged vid1 Tagged vid2 – LACP Trunk to Access Switch (Hybrid Mode)
• 3CX Server on int13 Tagged vid2 - No Untagged – (Hybrid mode)
• Uplink to sonicwall – int16 – untagged vid1 – tagged vid2 (Hybrid Mode)
• All other ports are untagged vid1 – access mode
Access Switch – HP 1810-24G
• Trunk1 (int1-4) Untagged vid1 Tagged vid2 – LACP Trunk to Distribution Switch
• All other ports are Untagged vid1 and Tagged vid2
• Yealink T32G and T38G are provisioned for (WAN port – Vlan enabled - vid2 - priority 6) (PC Port – Vlan enabled – vid1 – priority 1
• Company PCs plug into the phones PC port and get vid1 IP. Switch ports plug into WAN port and dish out vid2 IP.
Any help is appreciated. Ask me any questions that I didn’t cover. I can post anything you need.
Thanks,
Jon

 

[[email protected]]

The sonicwall is an NSA220...Sorry

 

[lneblett]

A few things -

1. Do you experience the voice issue on internal calls?
2. You don't really give an indication of scale of your Internet needs. How many calls at one time, how much needed for other than voice, etc.

You have dsl with a 1mbs upstream limit (maybe). I say maybe as have you tested this? Keep in mind that while you have QOS initiated, it (qos) is only applicable while the data is within your network. I am assuming that you are using a G711 codec given nexvortex, so this implies ~100-110kbs in each direction for each call. It has also been my experience that dsl has more latency than cable. If you do not have issue with internal calls, then look at your bandwidth usage versus availability.

 

[jpillow]

I'd say a bandwidth issue just as a guess, with only 1mbps up plus other applications are in use as well that may very well be a key indicator. Plus as stated earlier by sarcosys, latency with MOST DSL providers is a huge issue

 

[[email protected]]

Thank you for the replys. when i first posted i meant to state that it doesnt occur on internal calls and occurs with only one call active. To update, i have been able to get 4-5 simultaneous calls to work by removing the BWM settings on the sonicwall. Today, i am going to be upgrading our 3CX hardware to support 24 simultenous calls. We can only get T1's where we are located. To support the 24 simultaneous calls (16 will be used for remote sites). Do you all believe 4 bonded T1's will be enough bandwidth at our corporate site? It is my knowledge that remote sites will take up double the stream for the corporate site. An RTP stream up/down to the remote site and an RTP stream up/down to the voip provider. All at 128kbps. So 256kbps up/down for each remote site. Thanks guys.

 

[[email protected]]

Forgot to mention. We would be keeping the DSL for all other internet data traffic and the T1's would be 100% 3CX voip traffic. I can shape that on the NSA220 firewall. Thanks. Looking forward to your replies.

 

[lneblett]

Hard to say. When I read your post you give me the impression that the "bonded" t1s will be terminated at the corporate site. You don't mention how the other remote sites will be connected nor how many sites nor how many phones at each. I only know 8 lines at one site and 16 others spread around. Also, I am not certain that 3cx has a 24 call edition out yet. I think it currently goes 16, 32, 64, etc., but I have seen some mention of a 24 call system possibly coming (and I guess it could already be here).

The t1 will handle 24 64kbs channels. the remote sites will depend on the codec chosen, but I am not certain why you think a doubling of bandwidth will be required. Can you provide more detail? Also look at the blog and search for bandwidth or codec or the two together.this will give you a better idea. Finally, do yourself a big favor and use a supported/preferred interface. You may want to solicit opinions on the forums using a new thread.

 

[jpillow]

I'd say the bonded t1s are the correct route to go for voice as they are only ring used for 3 x traffic. Low latency and and more than enough upload.