Give me wireless VoIP (but hold the spectrum)
At first glance, running voice over IP on a Wi-Fi network should not be a significant engineering challenge. After all, VoIP applications generally require less than 10 kb/s of throughput, which represents just a sliver of the 11 Mb/s supported by the slowest 802.11 standard.
But ask most equipment providers about running VoIP over their 802.11-based networks and the response usually is tepid. “Well, you could use it for voice, but I wouldn’t recommend it,” is one representative’s stock answer. “It wasn’t designed for voice.”
Indeed, although the 802.11x portfolio has proved to be extremely useful for data applications, the real-time nature of voice applications creates a host of difficulties: packet contention, packet overhead, routing latency, coverage range, interference, multipath conflicts and — in the case of mobile VoIP — handoffs between base stations.
Furthermore, Cohda Wireless CEO Martin Suter said the root of the problem can be traced to the fact that 802.11 was designed to work indoors instead of as a standard for wide-area networking across entire cities.
“I contend that the movement of 802.11 outdoors is requiring the radio to perform unnatural acts,” Suter said. “You are putting 802.11 in an environment for which it was not originally intended or designed.”
However, market opportunities dictate that 802.11 not only venture outdoors but also support voice. And several companies claim to have solutions that address the problems of wireless VoIP. Taken separately, each is powerful. When considered as a group, the prospect of a network operator’s dream — a cost-effective wireless mobile VoIP network that doesn’t require a dime being spent on spectrum — becoming reality is more promising than ever before. At the very least, these technologies represent new alternatives that could drive down the cost of wireless voice dramatically in the marketplace.
But there are a lot of problems to overcome. Even if quality of service (QOS) is activated through the 802.11e standard, Wi-Fi is a contention-based protocol that does not prioritize according to packet type. When two or more packets try to access the pipe at the same time, the packets can collide. In other words, a VoIP packet can fail to be delivered to an access point in a timely manner, which introduces unwanted latency to a VoIP call.
“That’s why you can only use a fraction of the available bandwidth to carry simultaneous VoIP calls,” said Joe Bobier, principal inventor of xG Technology’s xMAX solution. “By the time you throw a few VoIP calls on there, because they’re so sensitive, if you start getting some collisions — which happens very quickly as you add users — the VoIP calls become unusable.
“I’ve had that happen with as few as two users causing a noticeable degradation on a VoIP call.”
And Bobier’s scenario occurred indoors with a single Wi-Fi access point linked directly to a DSL line. When taking typical Wi-Fi systems outdoors, receivers have to deal with multipath — the echoing of signals as they bounce off buildings and other objects — which standard 802.11 equipment was not designed to do, Suter said. The result is “black” or “dark” spots in the network where there is little or no coverage, he said.
“Based on what we know, that’s not necessarily the result of a lack of energy,” Suter said. “That’s typically the result of the effects of harsh multipath. … In many cases, they are within range, but the receiver cannot handle the environment.”
In an effort to overcome this problem, as many as 30 or 40 nodes are being deployed per square mile, which can enable line-of-sight connectivity but greatly increases equipment, maintenance and spare inventory costs — often making the economics of the network untenable, Suter said.
By using iterative processing in its receiver-side technology, Cohda Wireless increases receiver sensitivity that alleviates many of the multipath problems and creates 6 dB gain in the receiver, Suter said. This allows increased range from each node, which means fewer nodes have to be deployed in a given area to achieve the same amount of coverage as a typical 802.11 deployment.
“A big part of Cohda’s value proposition is our ability to create a coherent channel in instances where there is harsh multipath,” Suter said. “When you couple that with the increased range of our devices, what you’re seeing is the ability to provide coverage with a greatly decreased number of nodes.”
Indeed, by using Cohda’s technology in fixed infrastructure nodes, network designers “conservatively” can reduce their node densities by more than 50%, according to Suter. In a network deployed in Adelaide, Australia, Cohda has deployed its technology in both the infrastructure and client devices. As a result, the Adelaide network requires just six nodes per square mile and supports high-speed mobile access at demonstrated speeds up to 135 mph.
“What we’re trying to explain is that the challenges that early deployment are facing are not intractable,” Suter said. “They are problems that we predicted and understand, and we believe we have a way in which we believe the challenges can be mitigated.”
Another problem in any wireless network is interference, and it’s one that is a greater concern in unlicensed bands. To help alleviate those concerns, TensorComm has developed Interference Cancellation Technology (ICT), which is designed to cancel interference from all sources except the signal of interest, to increase throughput, capacity and range, TensorComm founder and CEO John Thomas said (see above graphic).
“The impact is that, if you cancel all the interference, then the data transfer between the base station and this terminal goes up pretty significantly,” Thomas said. “Testing has shown we can increase the data rate anywhere from 50% to 80%.
“From the network operator’s perspective, you’re not resending packets that didn’t make it the first time, so you’re not being inefficient with your infrastructure and spectrum. In other words, you’re increasing the capacity of your network by doing this.”
Limiting the number of packets that have to be resent is especially critical in VoIP applications, which are very sensitive to latency. In addition, although VoIP packets are relatively small, they are more numerous than other data packets, so overhead can eat up bandwidth more quickly than might be expected at first glance.
“In some of those trials, we were doing up to a 6 dB performance improvement,” Thomas said. “We feel like getting a sustained 2 dB gain is very doable. As you know, for even a half a decibel, companies spend a lot of time and effort to get that. So what we’re offering is a very significant jump in performance relative to that.”
Currently available for licensed operation in CDMA, wideband-CDMA and high-speed downlink packet access (HSDPA) networks, the TensorComm solution also can be used in WiMAX and WiBRO solutions, based on a patent the company filed regarding its technology’s applicability to multiple input/multiple output (MIMO) networks. In a one-antenna Wi-Fi system, TensorComm’s ICT may not be able to distinguish the signals enough to help, but if the Wi-Fi system utilizes MIMO, TensorComm should provide significant benefits, Thomas said.
“We view [ICT and MIMO] as an additive,” he said. “If you’re going to incur the cost of a second antenna, we recommend that you put our solution in the handset to get the most benefit out of the cost of a second antenna.”
Of course, in many wide area networks, getting the signal to a node is just part of the battle. In mesh networks, a signal may have to be routed through multiple nodes before reaching an access point with high-speed backhaul. Even though routing typically is done in a split second, each hop represents a delay, creating latency and trafficking problems that can degrade data performance and make VoIP unusable.
“After a couple of hops, a linear mesh network is really dead in terms of bandwidth, and you have major issues with quality of service,” said Michael Nova, Kiyon Wireless CEO.
But Nova said Kiyon — a company rooted in building-automation technology — has spent two-and-half years developing a software-based routing solution that reduces the time it takes a router to switch a signal from the typical 100-plus milliseconds to just 1 millisecond. Such fast switching prevents the degradation of data throughput when numerous clients access a given node and makes VoIP practical, even on a mesh network, he said.
“The human ear can discriminate individual sounds at about 150 milliseconds,” Nova said. “If you’ve got jitter and latency way below that, you won’t get a choppy call, no matter how the packets are split up and come back to the human ear.
“Standard 802.11 [switching] is at 150 to 200 milliseconds — that’s why VoIP over wireless is so bad. We switch so fast and are switching channels so fast that we’re way under that 150 millisecond level. So you get a really good quality VoIP call — not only that, you get a lot of people who are able to do it at the same time.”
Nova said Kiyon’s routing software can be embedded into any network, including WiMAX and WiBRO. Already being used by SkyPilot to enhance its 802.11 networks at 2.4 GHz and 5.8 GHz, the company has struck deals with Honeywell and Siemens in the building-automation industry.
Meanwhile, Kiyon has conducted numerous demonstrations for Tier 1 and Tier 2 carriers interested in applying the routing solution in conjunction with ultrawideband (UWB), a low-power broadband technology that many believe will be used to distribute broadband signals in local area networks.
“UWB is pretty short range, that’s why you need a lot of hops,” Nova said. “We ran voice-over-IP over a ultrawideband network … using six hops, and it worked well. We were told that it was the first VoIP-over-UWB test.”
Although many companies provide solutions to specific problems inherent to wireless VoIP, xG Technology plans to license its xMAX technology to regional providers wanting to offer metro-area mobile wireless VoIP service during the next year. The Florida-based start-up, which went public in Europe last month, expects beta systems to be operating by the end of the year and full deployments to begin in 2007.
Utilizing xG Technology’s patented single-cycle waveform solution, the xMAX VoIP system will let an operator cover 1000 square miles and about 30,000 subscribers — based on existing mobile phone usage patterns — with a single $50,000 base station, said Rick Mooers, xG Technology chairman and CEO. Although typical costs for tower rental and provisioning lines will apply, the fact the xMAX gear leverages the unlicensed ISM band (902 MHz to 928 MHz) means operators will not have to pay for spectrum.
The xMAX VoIP phones will be able to function as a wired phone via an Ethernet port or wirelessly via a Wi-Fi hot spot or an xMAX network. The first generation of xMAX phones will not be able to hand off calls between the disparate networks, but future generations will, Mooers said.
While a typical Wi-Fi network is unable to handle even a dozen simultaneous VoIP calls from a relatively small area, the xMAX system can handle several hundred simultaneous calls from a single base station with an omnidirectional antenna, xG Technology’s Bobier said. Using directional antennas and other common capacity-enhancing techniques only will enhance those figures, he said.
The patented xMAX receiver technology ignores other waveforms, which greatly reduces interference issues. Meanwhile, by using TDMA-scheduled time slots instead of a contention-based carrier sense multiple access (CSMA) system such as Wi-Fi, latency-inducing packet collisions are avoided, Bobier said.
“We have the advantage of being able to build our MAC layer from the ground up, specifically for this application, so we don’t really have to worry about collisions and lots of different devices operating on the same Wi-Fi hot spot all the time. When the customer uses our equipment on our system, he is on a TDMA system and has enough dedicated bandwidth [to make a VoIP call].
“It’s kind of an advantage when you’re starting from scratch,” Bobier said. “You don’t have to worry about backward-compatibility because there’s nothing we can be compatible with.”
For network operators looking for a VoIP solution that uses technology that is more familiar to the marketplace, Israel-based Trans Con Mobile believes it has an answer. Like xG Technology, Trans Con Mobile’s system is designed to deliver low-cost, unlimited mobile VoIP today with the intention of also offering IP-based data services in the future.
While xG Technology has demonstrated that its signal can carry at least 18 miles with its a proprietary waveform, Tran Con Mobile promises ranges similar to existing cellular systems — as well as seamless handoffs at speeds up to 80 mph — using 802.11 technology and off-the-shelf Cisco System access points in the unlicensed 2.4 GHz band.
“The transmission power and the reception sensitivity are dramatically higher than what is common in Wi-Fi phones that you can find on the market today,” said Haim Yashar, Trans Con Mobile’s director of the patent-pending technology.
By transmitting signals 1.4 to 1.6 miles in open areas, Trans Con Mobile’s solution requires less than 25% more base stations required for a cellular system, but the base stations’ light weight enables greater flexibility in deployment, said Avi Shani, Trans Con Mobile chairman. And the cost savings associated with the base station — about 10% of the price associated with a cellular base station — more than compensate for any density differences, he said (see table on page 48).
“We reached the point where it’s so cheap that I wouldn’t say it’s a joke, but it’s really a dramatic change in what the cellular industry knows these days,” Shani said.
Although the Trans Con Mobile base stations are available today, the handsets are in the final design stage for production. Prototype handsets used with Trans Con Mobile’s test network in Israel are inexpensive to build and deliver impressive performance when compared to existing Wi-Fi phones, Shani said.
“Just to give you an example, the battery power on a regular Wi-Fi phone lasts a maximum of 30 minutes and then you have to recharge — and the distance is 30 feet,” he said, noting that future generations of handsets also will be able to access GSM networks. “Our battery power lasts six hours talk time and 120 hours standby. And we achieved it through software, not a special battery. It’s obviously a bigger battery, but you don’t see it on the handset because we wanted to keep the handset the same size as a regular cellular phone.”
Such solutions sound extremely promising, but service providers should approach them with a healthy dose of skepticism, said Roger Entner, a wireless analyst for Ovum. These companies will bear a large burden of proof before the marketplace will be convinced that building a business plan around unlicensed spectrum is a good idea.
“The moment that stuff gets popular, all hell will break loose,” Entner said, noting his own unfavorable experience when a Wi-Fi data connection at a trade show failed because too many users tried to access it simultaneously. “That’s the way the cookie crumbles with unlicensed spectrum. …. I wouldn’t bet my farm on it.”
But if they work, the market demand for such systems should be enormous. By operating in unlicensed bands, the systems such as those offered by xG Technology and Trans Con Mobile give service providers the opportunity to enter the lucrative wireless market without the uncertainty attached to auctions that require billions of dollars just to secure the spectrum necessary to build out a nationwide network.
With relative certainty regarding deployment cycles and spectrum availability, the savings promised by unlicensed wireless VoIP networks should generate low-cost mobile voice alternatives — or redundant networks — that could dramatically impact pricing throughout the wireless marketplace.
“If you know any carrier or any future carrier that is not yet in the market and wants to save itself a lot of money, it would be our pleasure to meet with them,” said Trans Con Mobile’s Shani. “It’s much easier to build, much quicker to build, much cheaper to build — there’s no need to buy spectrum and no preparation for the tender, which will cost each company about $20 to $25 million in paperwork.”
|Traditional cellular||Trans Con Mobile|
|Spectrum availability||By FCC auctions only||Unlicensed|
|Spectrum cost||Billions of dollars||Free|
|Regulation||Operation license required||No license required|
|Network infrastructure||High cost: $200K-$300K per base station||Low cost: $30,000 per base station|
|Charges for calls||Traditional way: per minute; charges for incoming calls; high cost for international calls||VoIP: Flat monthly rate (unlimited); no charges for incoming calls; low cost for international calls|
|Source: Trans Con Mobile|