My journey to good sound quality

Posts: 3 · New Forum Member Joined: Oct 09, 2008 Posts: 3

Background: I was a network engineer working for a company that owns 2 class A subnets. Today, I work from home and hosts a lot of conference calls. I keep hearing good things from 2 coworkers who use Vonage so I decided to give it a try. I had bad experience with Packet8.

VOIP depends on upstream bandwidth when you're talking. Applications like netmeeting or Webex also depend on upstream bandwidth when you're sharing your desktop/slides. They compete for the same upstream bandwidth.

I first tried the WRT54G Firmware Version: v1.02.0. For traffic shaping, I told my Linksys that my upload bandwidth is 384kbps. My ATA is plugged into switchport 4 so I added these rules: Switchports 1-3 have low priority. Switchport 4 has high priority. The ATA's MAC address have highest priority.

I also artificially suppressed certain classes of packets based on destination ports. Ports 80, 8080, 8088, 443 were all set to low priority. I do this because I have quite a bit of ssl and ssl-tunneling applications that demand bandwidth.

All these still didn't help. I got great sound quality when I was just talking. But once I fired up my Webex equivalent application (my company uses a proprietary system) and shared my desktop, sound quality dropped.

I switched to a Netgear WPN824v3. I felt that its QoS setup HMI (human-machine-interface) was much more ergonomic. There is actually a policy named Vonage_IP_Phone which I set to the highest priority. While it does not say, it would make sense that policy prioritizes traffic form udp ports 10,000 to 20,000. I saw marked performance improvement but still the voice quality was not comparable with POTS.

My final solution: asked Comcast to give me 2mbps upload. Suddenly, VOIP sounds as good as POTS even when I'm sharing my desktop/slidedeck. Rolling Eyes

Posts: 83 · Posted: Thu Oct 09, 2008 9:11 pm Post subject:

My ATA is downstream of an old WRT54G using QOS. The telephone audio is excellent even when the computer is uploading.

I followed Nate Hoy's tips in the last post on this page. I believe that the key is setting the upload speed to just 90% of the lowest speed measured.

After doing that, the only other non-default QOS setting in the WRT54G I have is the port of the ATA set to high, with every other QOS option on the screen set to low.

The WRT54G firmware is the original v2.04.4 Aug 3, 2004 that shipped with the unit.

In my case I already had 1600k upload to play with, so I set 1440k in the QOS screen. However, in Nate's post, he acheived great results with an upload setting of only 190k.

The other eye opener of that thread is that when the upload setting is set higher, even at just 100% of the lowest measured upload speed, it can result in behavior similar to having no QOS at all.

Posts: 3 · New Forum Member Joined: Oct 09, 2008 Posts: 3

Thanks for pointing out the thread. I think it's a good thread and it provides great guidelines. Nate's proposal of 90% of tested average is a good rule-of-thumb. However, the post seems to suggest that the test is conducted between your home-PC and the test server.

The best test is a test between the ATA and Vonage server (called a SIP server? Sorry, I'm not a VOIP expert). This is because the upstream bandwidth should be measured endpoint-to-endpoint.

Since the real time upload speed between your home-router and Vonage's server depends on hoping across several routers, it is not possible to nail down a number for the manual upstream bandwith setting. There are also too many other factors which makes nailing the setting down hard.

E.g. I had an OC3 failure today at work and we had re-route traffic to another path. The usual path had 13 hops, the new path had 15. All latencies changed. This the inherent design of the internet.

Imagine a major road shutdown. What would normally be a 15 minutes drive now takes 20 because you have to take detours. Or road expansion. What would normally be a 15 minutes drive now takes 10 because you now have an additional lane. Such things happen on the internet all the time without any notification to you, the home-user. Smile

Posts: 27 · Vonage Forum Junior Joined: Sep 26, 2008 Posts: 27

Very interesting read, I would imagine though that setting port priorities is a best effort for the cause. Having enough good quality bandwidth that would be suitable for the number of, and demands of the current devices on the network, should almost always be the solution perhaps.

Posts: 3 · New Forum Member Joined: Oct 09, 2008 Posts: 3

I always remember this story told to me. Never verified it myself but the it's still a good story. The US space program spent a lot of money trying to make a pen that would work in space. The Russian cosmonauts simply brought their pencils along . Sometimes the simplest solution is the best solution.

So far, what I've seen is that the switchport on consumer-grade 'routers' can be set to ether LOW or HIGH priority whereas there is a higher priority called HIGHEST. HIGHEST is available to traffic identified by source MAC and destination ports. I do not know how the QoS queues are implemented in the Linksys and Netgear products. I'm pretty sure some google-grease would find the answers. (I'm not an authority on consumer-grade products)

I've found over the past few days that my convoluted QoS rules actually slowed the switching down. Wink

Maybe all I need is 1 rule. Highest priority to the ATA's MAC address.

Slyster wrote:

Very interesting read, I would imagine though that setting port priorities is a best effort for the cause. Having enough good quality bandwidth that would be suitable for the number of, and demands of the current devices on the network, should almost always be the solution perhaps.

Posts: 27 · Vonage Forum Junior Joined: Sep 26, 2008 Posts: 27

Indeed! Consumer end devices are what they are right? Ahh, the marketing of it all. I will admit however that for me, one of the joys in life is the ability to "Tweak". Idea