The Wireless Edge: High Speed Downlink Packet Access

First, an important correction: In my last column on wireless datauptake, I reported 500,000 EV-DO customers for Verizon--the numberthat has circulated in various stories. Mobile Pipeline Editor DaveHaskin questioned this number and obtained a clarification fromVerizon that the number refers to data subscribers for both 1xRTTservice and EV-DO. Most likely then, the majority of the customers areusing the earlier 1xRTT service. Now, on to HSDPA (High Speed DownlinkPacket Access).

Vodafone, one of the world's largest cellular operators, and Nokiajust completed testing of HSDPA, a new version of WCDMA (WidebandCDMA), in Italy using Nokia infrastructure. The two companies reportedthroughput rates of 1.5 Mbps. Various other companies havedemonstrated throughput speeds of over 1 Mbps for this technology, andpeak speed claims of over 14 Mbps are also common. But what speeds canyou realistically expect and what services will be available when?Also, how does this technology line up with the competition, includingEV-DO, Flash OFDM and WiMAX?

First, let's look at what the technology is. HSDPA is an enhancementto WCDMA, a technology that is also referred to as UMTS (UniversalMobile Telecommunications System), the 3G path chosen by most GSMoperators around the world. Today, there are already some 75 UMTSnetworks in operation around the world. In the United States, Cingular(via its purchase of AT&T Wireless) has UMTS available in six cities.However, the company is planning an aggressive deployment of HSDPA,with 15 to 20 cities planned by the end of 2005 and most majormetropolitan areas by the end of 2006. The six cities with currentUMTS service will be upgraded to HSDPA as well. Globally, many UMTSoperators are planning on the HSDPA upgrade, and operators that havenot deployed UMTS yet are likely to go directly to HSDPA.

HSDPA is important for operators because current UMTS technology,which based on 3GPP WCDMA Release 99 specifications, is extremelyefficient for voice service, but it is not optimized for dataservices. The HSDPA upgrade, based on 3GPP Release 5 specifications,keeps the same voice mechanisms but adds highly efficient datacapability. It does this through radio mechanisms such as higher ordermodulation (16QAM in addition to QPSK), improved error correction,dynamic adaptation of modulation and coding based on radio conditions,and a tighter link between mobile and base station. It also adds apacket scheduling approach that favors downlink transmissions to userswith the best radio conditions, thus allowing them to receive data athigher throughput rates. Since conditions vary by users over time,this results in what is called user diversity. All these approachesworking in combination result in at least a doubling of spectralefficiency, and with later improvements on the roadmap, a furtherdoubling. Bottom line: The network can support many more users ata higher speed.

HSDPA is not the first technology to invent these approaches. You seethem in most new wireless technologies, including CDMA2000 EV-DO andWiMAX, as well as in Wi-Fi. As for speeds, that's where things get abit tricky. The network does indeed support a peak rate of 14 Mbps,but this is what you would get with the full capacity of the cellsector at the highest modulation level and no error correction. Inother words, you'll never actually experience this speed, as you'llnever experience the peak rate of 2.4 Mbps quoted for EV-DO. Initialdevices are likely to have peak rates of 1.8 Mbps or 3.6 Mbps,depending on their designs. Subsequent devices will be faster. As fortypical speeds you can expect, it will depend on the number of usersactive in the network, but average speeds in the 500 kbps to 1 Mbpsrange should be achievable in relatively lightly loaded networks.However, speeds may go down a bit if people flock to the technology,as is the case for all the 3G technologies. Of course, with thesomewhat slow adoption of cellular data as discussed in my lastcolumn, this is a problem that operators would love to have, and onethey can manage through additional cell sites and new spectrumallocations. HSDPA latency goals are also aggressive, and in initialnetworks, latency will likely be in the 100 to 200 msec range.

As for how this lines up with EV-DO that Sprint and Verizon aredeploying, HSDPA will likely meet or exceed EV-DO performance. HSDPAalso allows simultaneous voice and data on the same device. However,EV-DO operators will have broader coverage in the near term. MobileWiMAX specifications point to somewhat higher performance, but thereare a series of further enhancements planned for HSDPA.

In my view, HSDPA is a great thing for the wireless industry for anumber of reasons. First, it will offer a highly compelling broadbandwireless service over very large coverage areas. Second, it willcompete head to head with EV-DO, resulting hopefully in competitivelypriced service plans. Third, it will keep the WiMAX people on theirtoes, forcing them to deliver the best possible technology; otherwise,it won't be competitive. The reciprocal case is also present, as WiMAXhas accelerated 3GPP (the organization that specifies GSM-UMTStechnology) efforts to start designing what comes after HSDPA,currently called 3GPP Long Term Evolution.

From a usage point of view, even by the end of 2006, HSDPA won't beeverywhere. Coverage is likely to emphasize urban areas, and in lessdensely populated areas, the fallback will be to EDGE for quite sometime. This means there's no reason to wait for HSDPA, as you shouldprobably plan for most applications you deploy to work over both HSDPAand EDGE. The same is true for EV-DO, where the fallback is to 1xRTT.

As HSDPA gets closer, you're likely to see lots of hype about thetechnology. But if you keep in mind its true capabilities, there's noreason to not start incorporating HSDPA into your wireless planning.