This year’s virility symbol for equipment vendors? Maybe, but can HSDPA really deliver at the low cost its proponents are claiming — and what about the handsets? Keith Dyer looks at the development and impact of HSDPA.
This time last year no equipment manufacturer’s strategy presentation for 3G was complete without five letters somewhere towards the right hand top corner of the powerpoint slide. HSDPA were the letters, not as easy to spell by waving your arms about your head as Village People’s YMCA. But the idea was the same. HSDPA — “You can get yourself cleaned, you can have a good meal, you can do whatever you feel”. Well, sort of. But the point was, HSDPA did and will offer the sort of shift in user experience that broadband offered compared to dial-up in the fixed world. Whereas 3G, as we know, started off promising 2Mbs and ended up delivering 384kbps if you were sitting still in a cell with not too many other users nearby, HSDPA offers the potential for operators to provide much higher data rates — up to a maximum of 14Mbps according to the standard but much more likely in the real world to be in the 1-2 Mbps range, perhaps higher in good cell conditions.
But the powerpoint slides were little more than that. Operators were all-out to get their first 3G networks up and running commercially and, in fact, even the Tier One operators have only just met that target. Some, indeed, have not launched yet.
But this year at 3GSM things will be a little different. Things are beginning to happen. The equipment vendors all have trials they can talk about with customers. Real numbers are starting to come in, and massaged numbers be fed out to the public. There seems real agreement that this year, HSDPA will be hot at 3GSM, and that by the end of this year there may well be some viable pre-commercial services in a few locations — and that in 2006 there will be real action with some commercial launches by Christmas. These are timelines from those in an industry that have learnt to stop making outrageous predictions about either when or how things will happen.
Nortel’s Scott Wickware says, “All our major customers are very focused on HSPDA. It has become a major focus in the past six to nine months for the longer and shorter term people. Whereas 12 months ago it was viewed as a way down the road.”
Just as real are the reasons for the movement in HSPDA, although not all of them are necessarily good news for the industry. The first is the realization that the user experience in R’99 UMTS is, while something of a step forward for the mobile industry, still causing some problems with early users and with trialists. Users simply don’t want to wait for downloads, or for a web page to download. HSPDA offers the ability to up that downlink speed by perhaps ten times the current rate — taking 364kbps and turning it into something nearer 3Mbps. As this is faster than most Europeans get at home on a broadband modem, then the service potential of HSDPA is obvious.
In the vanguard
One of the fist manufacturers to give actual numbers from actual network trials was Motorola. Of course, there is a need for Motorola to be seen to be in the vanguard with HSDPA, because it has had little visible 3G success so far. But Motorola has said that UMTS cell sites equipped with HSDPA technology will probably offer users average data rates of between 500kbps and 1.5Mbps when they are working at full capacity.
Motorola has been conducting HSPDA trials with five European operators, simulating a variety of different network conditions, including that of a cell working at peak capacity. HSPDA is labelled by the 3GPP standard as having a peak of 3.6Mbps on the downlink (and 14.4Mbps with the most powerful modulation), but Raghu Rau, corporate vice president for marketing for Motorola’s networks business, says that, in reality, the throughput will be tempered by the amount of users in a cell, and the extent of their use.
“Motorola estimates that average user throughput will be approximately between 500kbps and 1.5Mbps during the download. Overall HSDPA will appear to the user between 3 and 10 times faster than UMTS. This differential will increase as the cell size gets smaller,” Rau said.
Motorola did clock a speed of 2.9Mbps in one of its trials, at the edge of a cell using a single HSDPA test device. Other vendors have announced speed of 3.6Mbps but these Motorola trials were designed to help the mobile operators concerned to build optimised HSDPA-enabled networks. Motorola set up a menu of test options for the participating operators to choose from — to mimic the individual operating conditions of each network — with different access options to measure performance, compatibility and interoperability. During the trials, services including e-mail, video streaming, music downloads and web browsing are being tested for speed, capacity and data quality from
normal to high-traffic conditions. This also contrasts with some other triallist’s vague announcements of ‘data calls’.
Nortel’s Wickware says that, in the trials and the specification, it is always the higher data rates which get the most press attention. But also in the spec is the increase in capacity, and the reduced latency. “The three together are the things we are demonstrating. We have demo’ed 3.6Mbps with O2 before Christmas in good clear cell conditions, but it is the triple play of increase speed, reduced latency and increase in capacity that customers have liked.”
A second reason for the HSDPA buzz is the, as yet, non-existent — but soon to be pressing — need for 3G operators to offer more capacity. It is true that any one 3G cell at the moment will struggle to cope with more than a handful of users all running, say, a video application at the same time. Now, if the operators only had that as a real problem, the cynics might say. But it is something that operators, including 3, have thought about very hard.
Sterling Essex of Ubinetics — a company that has worked in a test capacity with NTT DoCoMo on its development of HSDPA — says that 3 is looking at a bill of billions of pounds if it wants to meet predicted traffic growth with suitable network expansion. If you take this at face value, that means that adding more of the same capacity will cost nearly what it did to roll out the first network. And that’s not an option. So HSDPA steps in as a relatively cheap way to get more capacity in the network.
“Operators are looking for HSDPA to make a visible difference to their service capability, and European operators are keen because it does not involve a lot more capacity expenditure and it may get them out of a hole on congested cells,” Essex says.
Wickware agrees, up to a point.
“If you are in a one-carrier scenario then that’s a fair statement. It will not take long to exhaust a cell if you are offering 384kbps per user. So either you must add more carriers, which is what happens in a CDMA environment, or the other way is to get more capacity out of HSDPA. The difference is that, to kick start data use from your customers, the cost of capacity is significantly lower compared to R99.”
This lower cost of capacity is equally important to Wickware in developing operator business cases. “At a time when the business case it not quite figured out, it is definite that HSDPA can have some effect,” he says.
“HSDPA’s primary focus is to allow increased payload. The other thing is to improve latency, and the hope is that by making it faster, lowering latency and including a large capacity improvement the cost point on some applications which have not had a business case before will now work. A 64kbps video clip at an R99 cost base had to be around â‚-1 but on HSDPA the same service is much less costly to offer and still offers a good margin at around 10 or 8 Euro cents.”
Another reason for HSDPA’s likely dominance of radio network discussions this year is it is becoming the new status symbol for the network equipment vendors. Again, Essex says that real, live working HSDPA will be this year’s ‘must-have’ accessory for vendors.
As Essex says: “Talk of systems availability is the next virility symbol.” But he also counsels caution when viewing vendor claims. His view is that operators must look particularly closely at vendors’ claims that HSDPA is a software only upgrade.
“It is not always true that the Node B’s only need a software upgrade,” he says. “There is often a need for a board swap in the Node B, though it is the same Node B.” The reason for this is that HSDPA works with a lot of powerful processing required within the base station. The base station dynamically allocates channel resources to users, scheduling calls depending on cell quality, signal strength and what data queues there are to deal with. This means an upgrade in the baseband processing power of the base station, which is the board swap Essex mentioned. Stress testing the scheduler, Essex says, is key. Essex also says he knows of some Ubinetics tests in which the Node B’s have been compromised.
Another possible cause for caution, according to Essex, is the real lack of chipsets at the handset level. Devices lagging behind a network has been an issue since the early days of GSM, and HSDPA may be no different.
“For terminal makers things are even more tricky,” says Essex. “You’re talking about a handset that can clock ten times the 384kbps rate, which means the baseband chip needs a much higher bandwidth. The chipset manufacturers are important because the baseband processor has to do the work, and for HSDPA the chipsets are not near ready.” The chipset develeopers also have to design a chipset that can deal with the bursty nature of the data downlink, and do so at half a frame’s notice. As HSPDA works on extremely short frames, this means about 1ms.
For Essex, chipset development means that 2006 will be the real peak for ODMs in 3G handsets, and then HSDPA may have it’s turn. This puts handset development right on the margin of the vendors’ predictions of an end of 2006 commercial launch.
Alcatel’s Jean-Louis Hurel, product marketing director, mobile radio division, is slightly more positive, as you might expect. For him, the requirement for HSDPA is not about companies pushing technology at an unwilling market. Instead, it’s about a technology answering a market need for faster, more efficient data.
“We see HSDPA as an integral part of Alcatel’s user-centric broadband vision, and believe me, today we are no more in a technology push mode. Technologies now can only be created and successful as long as there is demand from the market. The 384kbps bit rate of today is quite high in terms of its contrast to 2G but another step is needed. HSDPA fulfils this, and can be successful now.”
“The operator challenge” he says, “is to be able to deploy HSDPA at minimum cost.” With that in mind, he says, the issue of a software-only upgrade has been misunderstood.
“Alcatel’s Evolium UMTS solutions have the capability to include a software upgrade and all Node B platforms are HSDPA hardware ready completely. Some equipment competitors that rolled out Node B’s some time ago had to change the major part of the old cards. But the more recent ones in place need no hardware upgrade. Some do need to add additional baseband boards to add processing capacity before they can do HSDPA. This is so the RF part can host the new modulations, and if you do have to change that then that represents a high cost. Evolium offers enough processing capacity and memory to host new technologies,” he claims.
Nortel’s Wickware says, “The thing people mix together is the requirement to add capacity to the base stations in the form of equipment. If there is ten times the data rate then users will need extra processing capacity. But HSDPA itself on the base station is a software upgrade.”
Some ‘competitors’, Hurel says, may also need to change the power amplifiers because the existing ones can’t support the new modulation part.
As for the terminal side, “We are working with chipset and PC Card and handset suppliers, in order to prepare for use as soon as they are available. Chipsets do not all propose 16QAM modulation. But in the short term time frame we are planning tests in order to be ready.”
For Wickware, perhaps of more interest is the impact that HSDPA may have on network design. “When you do the upgrade, what is the business case and what do you intend to offer with that data capability? If you decide to change the uplink offer — ie, 16 on the uplink and 32/64 on the downlink — there is an impact on the cell design. An HSDPA high downlink speed may mean an operator decides to offer an increased uplink — now that’s an impact on the cell size. That’s not the technology itself requiring that but it may drive different attitudes for network design.”
So in the short term, the answer to “When will we see volume deployments of HSDPA?” may or may not be “Around the end of 2006.” But in the longer term the questions need to move on to how HSDPA affects the business case of operators’ data services, and indeed the networks themselves.
HSDPA — THE FACTS
So, what is HSDPA? High Speed Downlink Packet Access (HSDPA) is a packet-based data service introduced in Release 5 of the 3GPP W-CDMA radio format. As its name suggests, it is intended to provide a higher speed downlink — with data transmission of anything up to 14Mbps over a 5MHz bandwidth in the WCDMA downlink. HSDPA implementations includes Adaptive Modulation and Coding (AMC), Multiple-Input Multiple-Output (MIMO), Hybrid Automatic Request (HARQ), fast cell search, and advanced receiver design.
In 3rd generation partnership project (3GPP) standards, Release 4 specifications provide efficient IP support enabling provision of services through an all-IP core network, and Release 5 specifications focus on HSDPA to provide data rates up to approximately 10 Mbps to support packet-based multimedia services. MIMO systems are the work item in Release 6 specifications, which will support even higher data transmission rates up to 20 Mbps. HSDPA is evolved from and backward compatible with Release 99 WCDMA systems.
HSPDA is principally different from earlier iterations of W-CDMA because it adds a new shared data channel on the downlink. The ‘shared aspect’ means it allows multiple dedicated channels to be multiplexed into one channel, with usage dynamically allocated between them. What this means is that, instead of one user’s data being carried over a dedicated transport channel — and another over another — both users can be time multiplexed together to have data transmitted over a shared channel. This is useful because it allows the network to take account of bursty data — when one user is not using the channel then another can.
Additional intelligence is added to the network to manage the channel at each instant in time, and a variety of techniques are used to improve response times and accuracy. HSDPA increases user rates in the downlink by three main means. The first is the use of adaptive modulation and coding, which means the coding and modulation is adapted according to the link conditions. For instance, users at the edge of a cell will work on a lower coding rate than users close to the base station. The second is a fast scheduling instruction which is controlled within the Node B base station, rather than the radio network controller. The scheduler within the base station takes account of the channel quality, the terminal capability, any quality of service class and code availability.
The third is fast retransimissions, with incremental redundancy. Fast retransmissions mean that failed transmissions are re-sent by the base station, again closer to the user than the RNC, which controls this function in R’99 W-CDMA networks. Incremental redundancy refers to the ability to select correctly transmitted bits, reducing the need for further repeat requests in transmissions with multiple errors.
The new release provides for two delivery systems (or modulations). One of these, with a maximum downlink speed of 3.6 Mb/s, is mandatory. The other, with a maximum downlink speed of 14.4 Mbps, is optional.