How do you get quality VOIP calls on a mesh Wi-Fi network?

As more Wi-Fi mesh networks are deployed around the world and as more people use Skype, Gizmo Project and other VOIP services on their Wi-Fi devices, people are wondering what the effect will be on the quality of voice calls over a citywide or countywide Wi-Fi network. This article by Robert Poe on VOIP News addresses some of the issues As more Wi-Fi mesh networks are deployed around the world and as more people use Skype, Gizmo Project and other VOIP services on their Wi-Fi devices, people are wondering what the effect will be on the quality of voice calls over a citywide or countywide Wi-Fi network. This article by Robert Poe on VOIP News addresses some of the issues (quote):

“An emerging trend in wide-area Wifi gives wireless IP telephony fans one more technical issue to worry about. According to a new In-Stat report, there will be some 99,000 mesh Wifi access points in operation in 2010, up from less than 20,000 last year. And while the quality of wireless VoIP calls may be spotty to begin with, having Wifi APs linked via wireless mesh backhaul can degrade it further. All users can do is hope that Wifi operators build out their mesh networks with voice in mind.”

Click here to read the entire article.

Comments

  1. Great post. I think this is spot on. I am glad to see that some other people see where this is heading. Bravo!

  2. Here is the description of my experiences with testing the Google WiFi mesh network with VoIP (Skype):
    I drove around Mtn View earlier today to test how well the Google WiFi network worked for mobile users.

    For mobile users, the WiFi network is very impressive, imho. Even with a very low-power WiFi card on my laptop, the connection was impressive. Trees seem to attenuate the signals considerably. If I could see a Google node, I could connect to it. If I didn’t see the node, the connection was weak.

    Google should not use the MAC address to authenticate users – that sounds a little bit extreme.

    Test setup
    * I hooked up my laptop to the Google WiFi network using an old Cisco Aironet 350 802.11b card.
    * I used SkypeOut to call my home phone #.
    * The laptop and I both stayed in the car.

    The drive
    * I drove a little bit on the West side of El Camino Real (around Castro Street), but mostly on El Camino Real and on the East side of El Camino on Castro Street, Shoreline Blvd, Rengstorff Avenues and streets perpendicular to those. I also drove on Central Expressway between Rengstorff Ave. and Castro St.
    * I stayed close to areas where the mesh nodes have been installed and did not venture into areas where the Google network is not yet established.

    Observations
    * Good connection: The connection was very good in the vicinity of the network nodes/access points and also fairly continuous on streets where there were sufficient street lights and network nodes and few trees. I could talk uninterrupted via Skype in these areas.
    * Trees problematic: Even when the node density was high, if the street was tree-lined, the voice packets would not go through, though the connection would not drop.
    * Works at 35mph: I drove at up to 35 mph speed and the connection wasn’t significantly different at that speed compared to slower speeds.
    * MAC address authentication?: I authenticated myself via the browser the first time I got on to the network (gave my gmail id). Thereafter, it would let me on to the network even if I removed all the cookies and restarted my computer. It seems that Google is using the MAC address of the WiFi card to authenticate me on to the network.
    * Central Expressway: The connection was poor along the Central Expressway since it has few light poles and many trees. The node density is not very high along Central Expressway (http://wifi.google.com/city/mv/apmap.html).

    Comparison with other municipal WiFi networks
    * Corpus Christi, TX: I have used the WiFi network in Corpus Christi, TX (also set up by Tropos Networks, the same company that provided equipment for Mtn View). I tested the Corpus Chrsti network last year and found it to be much spottier than the Mtn View network. Looks like Tropos is using newer equipment and has tuned the Mtn View network much better.
    * Palo Alto, CA: Firetide is setting up a WiFi mesh network for public safety purposes in 2 sq miles of Palo Alto around California Ave. The Firetide network is still being installed and tuned. So far, the Mtn View network performs slightly better than the Firetide network.

    Conclusions
    * WiFi networks are tough to set up and maintain.
    * You need a little extra power and better antennas, but you can do lots of cool mobile applications in a properly tuned WiFi network.
    * The Mtn View Google WiFi network is well-tuned and one of the best when it comes to municipal WiFi networks.

  3. I design our networks with voip and video in mind. We make it part of the code to never let a voip or video packet hop more than 4 times across our mesh. i.e. QoS and then some…

  4. We (Novarum) have rigorously tested over 41 wireless broadband networks in North America in last six months (including 14 Metri Wi-Fi networks). We tested Google’s Mountain View network back in November and the complete ranking of the networks we tested is on our website. We will be reporting on these results at the next Muni Wireless conference.

    The best network in terms of coverage was in St. Cloud, FL (using basically the same equipment as the Google network – Tropos with Motorola backhaul) at roughly the same depoyment density (and thus cost). At the time we measure the Mt. View network, it appears that additional tuning was required since complete coverage of Mt. View from a mobile client was not achieved.

    The best network in terms of performance was the Toronto Hydro network in Toronto, ON. Substantially higher throughput and capacity than Mountain View in a more challenging environment – urban canyons in downtown Toronto with much higher density of other Wi-Fi equipment.

    Our analysis concludes that Wi-Fi networks today, when properly deployed, have coverage similar to 3G at substantially highe performance. We believe this validates the technology – but the business models necessary to deploy Metro Wi-Fi in densities sufficient for adequate performance are still in flux.

  5. Kevin Kapich says

    Thanks for the case study. Here is what needs to change for VoWIFI to truly take off: (1) power problems; (2)QoS (“bandwidth”, “latency”, “jitter”); and (3) propagation characteristics need to be tunned up. In general, Wi-Fi was not envisioned or intended to be used with robust data centric applications. The solution and innovation needs to occur in inner – access point to access point communication. From a business perspective, the promulgation of WIFI standards has evolved the use of WIFI in different ways to meet consumer demand. If the consumer demands VoWIFI(“VOIP WIFI”) and a viable business model persists, then the engineering community and industry mobile pros will make it happen.