HSDPA testing
Test methodologies must evolve as GSM and WCDMA combine with High-Speed Downlink Packet Access.
Dr. Salim Manji, Spirent Communications -- Test & Measurement World, 2/1/2006
 |
| SIDEBAR: What about field testing? READ OTHER FEBRUARY ARTICLES: Contents, February 2006 |
UMTS wireless network operators have recently taken a giant step forward by deploying WCDMA service alongside existing GSM services. This deployment was not a minor undertaking, as GSM and WCDMA are based on two very different technological foundations. Now, network operators are facing even greater challenges as they prepare to implement High-Speed Downlink Packet Access (HSDPA), a data-transmission technology that has evolved from WCDMA.
Like any CDMA technology, WCDMA is a spread-spectrum technology that introduces a new domain—the code domain—in which orthogonal spreading codes at the physical layer differentiate channels. Because the physical layers are different, the effects of air-interface phenomena ripple up the WCDMA and GSM protocol stacks differently.
The WCDMA service now offered is known as Release 99 service, named after the currently deployed version of the 3rd Generation Partnership Project (3GPP) UMTS/WCDMA technical specification. Network operators will soon be taking the next step by implementing HSDPA, a new service that will coexist with Release 99 and, in most cases, share space with older networks. Although early HSDPA deployments are likely to be on a par with current broadband speeds, HSDPA offers theoretical data rates of up to 14 Mbps.
Test engineers have just finished bridging the chasm between GSM and Release 99, and some are tempted to treat HSDPA as just a "faster Release 99." There are key differences between the two, though, and they are significant enough to require a closer look. As the name implies, HSDPA is meant for downlink data traffic only, and it offers several enhancements for handling this traffic. To ensure the HSDPA technology works properly in the field and is well received by users, the industry will need to devise new test methods that are designed specifically for data-centric transmissions.
HSDPA vs. existing technologiesRelease 99 is largely built on the traditional CDMA principle of separating users in the code domain. With the Dedicated Physical Channel (DPCH) technology in Release 99, network operators are able to manage the user population in a cell. The use of fast power control ensures equality among users, which allows for maximum cell capacity. Mobility can be managed effectively through the use of soft handovers.
This configuration works well for moderate data rates, but to support high-data-rate users, a provider must allocate a large portion of the code space, which diminishes the available cell capacity for other users. Since data traffic is often bursty, physical-layer resources may not be in use all the time. Quick reallocation of resources to support the "instantaneous" demands of the users in a cell is difficult to achieve in Release 99.
In contrast, HSDPA was designed with data in mind. The technology uses the benefits of the CDMA air interface while also optimizing network operation for data users. The network allocates one "fat pipe" for high-speed data, and the remaining cell resources can still be used for voice and low-rate data traffic. The network operator chooses how to allocate resources among high-speed data and all other users.
HSDPA's "fat pipe" is known as the High-Speed Downlink Shared Channel (HS-DSCH), and it may be shared among multiple users. The HS-DSCH is relatively static in terms of CDMA resources. For example, the spreading factor (the factor by which conversion to the code domain "spreads" the signal spectrum), maximum number of code channels, and code powers are generally constant, unlike in Release 99. Therefore, the network has a fixed resource to work with and can efficiently serve the high-speed data user population. Users with the best chance of getting higher-rate data are assigned more of the shared resource, based on proximity to the cell, lack of other users in the area, and signal strength.
HSDPA offers several ways to optimize resource allocation, including Adaptive Coding and Modulation (ACM), Medium Access Control high-speed (MAC-hs) scheduling, and the Hybrid Automatic Repeat Request (HARQ).
Fast ACMFast ACM replaces the Release 99 DPCH's reliance on "fast" or inner-loop power control as a solution to the "near-far" problem. If a Release 99 device is close to a Node B base station, the Node B base station can adjust power so as to equally share resources among subscribers. In contrast, in HSDPA the Node B base station implements ACM in such a way that it adjusts the data rate, and not the associated power, to make the best use of network resources.
 |
| Figure 1. Constellation diagrams indicate the relative complexity of 16QAM vs. QPSK modulation schemes. (a) A QPSK symbol can be demodulated as long as its phase can be determined to within 90°. (b) In contrast, a 16QAM symbol can be no more than 26.6° out of phase from another symbol. |
The data rate is adjusted by two methods. The first is adaptive modulation. An HSDPA signal can employ quadrature phase-shift-keying (QPSK) modulation, as in Release 99, or it may employ the more efficient 16QAM (quadrature amplitude modulation with 16 target symbols), but only if the receiver provides sufficiently good reception, because 16QAM signals are more difficult to demodulate than QPSK signals. A receiver can demodulate a QPSK symbol as long as its phase can be determined within 90°, while demodulation of a 16QAM signal requires a little more processing (Figure 1). In fact, 16QAM phase detection must be better than QPSK phase detection by a factor of three for proper demodulation; a 16QAM symbol may be only 26.6° out of phase from another symbol. And 16QAM poses an additional challenge: Amplitude detection is unnecessary in QPSK reception, but it is a key part of QAM demodulation.
HSDPA's second data-rate adjustment method is adaptive coding. To correct errors on the fly, WCDMA uses redundancy coding, which sacrifices bit rates to gain error-correction capability. If the mobile unit's reception will support it, the data rate can be adjusted by optimizing the coding rate and the number of code channels for a user.
MAC-hs scheduling and HARQIn Release 99 networks, a Radio Network Controller (RNC) handles scheduling. This approach works well enough for voice networks, but for efficient data transmission, HSDPA uses an alternative approach. Since the Node B base station is always the first link in the chain from phone to network, the Node B itself handles scheduling in HSDPA, taking advantage of the fact that bursty data scheduled for a number of users will most likely not be transmitted at the same time. Data traffic targeted for a number of users can be statistically multiplexed. By moving the MAC-hs into the Node B, HSDPA traffic can be scheduled at a time resolution of 2 ms, much tighter than in Release 99.
The MAC-hs handles not only the scheduling of transmitted data, but also the scheduling of HARQ retransmissions. While the ACM process performs some error correction, the HARQ function is responsible for recovering from errors where channel coding alone is not sufficient. In other ARQ processes, the receiver throws an erroneous packet away and sends a NACK message to the transmitter, which retransmits the packet. In the HARQ process, the receiver saves, rather than discards, an erroneous packet, making use of its residual information in combination with the retransmitted version in a seemingly complex process that actually improves performance in wireless applications, where impairments are a fact of life.
Isn't 3GPP testing enough?3GPP offers a set of test specifications designed for testing UMTS mobile devices, including HSDPA devices. These tests ensure that a mobile device meets a set of minimum acceptable standards. Network operators have to balance operational costs in order to offer attractive pricing, and the 3GPP will make sure that they have that chance. What the 3GPP does not do is perform design verification or competitive analysis, or predict loads on network resources.
Consequently, a large body of important testing is not defined by the 3GPP. To please wireless consumers who use ever-increasing amounts of bandwidth to receive and transmit data, network operators are relying on engineers in labs all over the world who are pursuing new, effective ways to use performance testing to test critical cellular characteristics.
Most performance testing uses 3GPP-defined standard tests as a starting point. For example, the 3GPP defines a test with hard pass/fail limits to verify that HS-DSCH reception by a mobile device meets a minimum set of throughput requirements. As defined, this test includes a number of variations based on the stated capabilities of the mobile device. It does not characterize the device, but it does provide an excellent starting point for doing so.
One portion of the 3GPP document describes a test called Single Link Performance (Ref. 1). This test procedure includes a lot of detail regarding different conditions that must be tested: multipath (signal reflection) conditions, throughput targets based on modulation schemes, and references to different sets of parameters. Understanding the pass/fail test requires some knowledge of multipath and the fading environment, and it requires a lot of poking around through appendixes in the specification as well as related specifications.
Once you understand the test, you can create many variations. For example, suppose you want to know how throughput rates vary with fading or noise. Or suppose you want to find out what happens if Node B isn't behaving ideally. What if the fading models defined for this test don't accurately reflect the fading conditions that the customer will experience? All of these are opportunities for performance tests.
It is impossible to define the "perfect" set of performance tests for HSDPA, which is another reason why the 3GPP doesn't define detailed performance tests. There are a seemingly infinite number of options, and these must be winnowed down to those that apply in each case. If history is any guide (and it usually is), there will be surprising variations in the performance of commercially deployed devices.
Performance testing on Release 99 handsets, for example, has shown dramatic differences among a sample set of phones, all of which passed the tests defined in the standard. These results show that consumers will have different experiences, that network operators will bear different deployment costs, and that handset manufacturers will enjoy or endure the long-term effects of brands associated with performance disparities.
Field testing may seem to be the ultimate in realistic testing. But while in-field beta testing should be done before any product is released, field tests present very real limitations (see "What about field testing?"); they can interfere with paying consumers, for example, and they are rarely repeatable.
End users will perform the ultimate long-term field test. Their perception of HSDPA will be the one that counts. Yet, the quality of their experience will depend on the quality of performance testing done beforehand. These "real-world testers" will be the ones who vote with their wallets and eventually decide on the futures of device makers, network operators, and HSDPA itself.
| Author Information |
| Salim Manji is a product manager at Spirent Communications in Eatontown, NJ. He earned a PhD (2004) from Rutgers University's Wireless Information Network Laboratory (WINLAB), where his thesis concerned image transmission over high-speed cellular networks in the presence of fading. His work at Lucent's Bell Labs involved early research on such topics as MIMO over UMTS and HSDPA. |
| REFERENCE |
|
|
