Breakthrough Ideas (continued, with some additions, given increasing insight)
The Second Breakthrough Idea
As if this ceaselessly evolving spread spectrum advantage were not already enough, then Viterbi had another breakthrough idea. gte.com
Breakthrough Idea # 2: By separating voice bits from data bits, the spectral efficiency of data throughputs more than triples the total throughput of voice, permitting high data rates and flexible time-dynamic need-based shifts between voice and data carriers.
Viterbi believed the challenge in wireless service provision was to understand the nature of the traffic along with the technological capabilities and tradeoffs available. He described two significant misconceptions in the industry: “(1) that with digital voice, the co-existence of telephony and data services in the same spectrum are a natural consequence; and, (2) that wider is better in the sense that efficiency grows significantly with increased bandwidth.”
Viterbi called the first misconception the co-existence fallacy, “failing to recognize that all bits are not created equal.” Because voice bits must be delivered reliably at a fixed rate to maintain a common quality of service, disproportionate resources must be allocated to disadvantaged users who are in shadow or far away. However, data delivery permits variable latencies and flexible requirements without overly penalizing data users. He concluded, “By using advanced techniques for network measurement and allocation, overall data throughputs can triple the throughput for voice.”
A communication system should be optimized for either voice or data, but co-mingling of voice and data in the same channel inadvertently wastes potential gains in data throughput because the algorithms in 2G CDMA (and in most 3G design rules) were optimized originally for voice. The nature and norms of voice traffic, for instance, require universal and immediate access to 911-emergency services. Whereas the nature and norms of data traffic permit levels of quality of service (QoS) within an Internet norm facetiously called the “World Wide Wait.” Viterbi realized, “By judicious management of latency, resources and requirements, along with some technological improvement, forward throughput can be increased by a factor of three to four times, resulting in average throughput in excess of 600K bits/sec and peak data rates of 2.4M bits/sec, all within the current CDMA bandwidth.”
Viterbi continued, “The second belief, that increased bandwidth makes for increased efficiency, is also overestimated.” The 3G standard favored by Europe approximately triple the bandwidth of 1.25 MHz by raising the clock rate to 3.84 MHz. Narrow guardbands permit Qualcomm to fit 3.69 MHz into three 1.25 MHz carriers, but UMTS’s 3.84MHz required an expensive 5MHz allocation of spectrum for each duplex channel. Viterbi contended that the only advantage in efficiency using a wider band would be due to increased trunking, that is, to combining a larger population of users by multiplexing messages.
However, he contended that once you isolate and segregate voice or high-speed data carriers, even this small advantage was illusory. This is so because data service acceptably permits a variety of data rates, depending on the user’s needs, location, and available resources. Algorithms that takes this into account can more than make up for the slight difference in bandwidths. If you will, data can wait its turn. But impatient voice, when it shares the data channel, must be privileged or lost. Also, such a small potential advantage from a shared voice/data channel is more than offset by the loss in flexibility that is required to allocate bandwidth in 5 MHz chunks rather then 1.25 MHz slices, especially when throughput can be tripled by optimizing a data only channel. The Qualcomm approach gives more bang for the buck. Viterbi concluded:
“In summary, just as a decade ago CDMA was proposed against mainstream opinion, and it has since prevailed, so today the common wisdom of voice-and-data coexistence coupled with wider spreading requirements is being questioned. In contrast, we propose that in tripling the bandwidth allocation, rather than supporting co-mingled voice and data in the entire band, we employ only the current bandwidth spreading for a data-only service, which can be tripled in throughput by techniques applicable only to data, while saving the other two-thirds of the tripled bandwidth for voice-only service which is likely to remain a major requirement for some time to come. When data becomes dominant, the voice service can flexibly be reduced back to make room for the increased data requirements.”
Thus, from a fresh second breakthrough idea, a beautiful sister to already beautiful1x was born: Qualcomm’s elegant 1xEV-DO.
The Machine Beauty of 1xEV-DO.
The “1x” stands for one-times bandwidth, meaning it uses one 1.25 MHz carrier. The “EV” stands for evolutionary, and “DO” stands for data optimized or, to contrast with a mixed voice/data carrier, data only. Although the CDMA RF technology is continuous or evolutionary, the architectural simplification of decoupling data from voice into separate and optimized carriers when combined with the optimized high data rate (HDR) algorithms in the DO architectural module itself is discontinuous or revolutionary. Thus, the beauty of High Data Rate technology dwells in its marriage of superior functioning HDR design rules and new coding parameters to the architectural simplicity created by separating the optimized strengths within a separated data carrier from the optimized voice algorithms in the RF architectural design that are still compatibly shared, Put another way, the principle of optimization requires a set of design rules and parameters that include new coding methods and algorithms optimized for ideally coding data to use the full potential of spread spectrum modulation theory and coding methods to maximize the potential spectral efficiency of mobile wireless data. Thus, following an uncontested, but decoupling, divorce from Voice, 1xEV-DO marries an Only-Yours-Carrier-Bride to a High-Data-Rate-Groom!
Optimized for packet data, the 1xEV-DO airlink was designed to enable high-capacity/high-speed data and Internet access. This architectural design requires a separate 1.25 MHz carrier to optimize efficient data transmission, but also it simplifies system software development, avoids difficult load-balancing tasks, and permits flexible RAN management through dynamic shifts in channel usage between voice and data to meet varying user needs by the minute, hour, day, or year. Given the increasing move toward data networks in telecommunications generally based upon IP packet-data technology, it marks the beginning of a set of new performance trajectories for evolving optimized mobile data networks
The ITU and 3GPP2 recognized 1xEV-DO as an international standard. In April 2002, the ITU declared it to be a third-generation standard, just as the ITU had decided that 1x, in fact, met their 3G requirements in October 2001, one year after its initial commercial rollout. Still, to manage expectations, their rivals and the press often call them 2.5G standards as a means of urging their installed bases to wait for the real thing, UMTS. And, in China calling these networks generation 2.75 permits China Unicom to gain a vast head start on UMTS, if it is ever chosen by the MII.
Following the injunction to simplify, simplify, simplify, 1xEV-DO uses the standard Qualcomm RF waveform, with the same chip rate, network plans, and RF designs for Access Terminals and infrastructure to retain CDMA compatibilities and to encourage mounting economies of scale. Moreover, it gives consumers the ability to receive an incoming 1x or IS-95 (with other modes and bands following soon) voice call while downloading data.
Nonetheless, because it uses a separate 1.25 MHz band, 1xEV-DO also can be deployed in TDMA/GSM/UMTS networks by using a separate carrier. And, it almost surely in the MSM6500 can become an upgrade of the NSM6300 or GSM1X chip to include HDR. Also, 1xEV-DO someday may find use in MMDS, LANS, or other available bands of spectrum set aside for mobile data usage.
The move from mobile-to-fixed use is a small engineering exception (re-setting the ASICs) compared to the general rule that wireless progresses historically by innovation from fixed-to-mobile wireless implementations. The principal advantage of fixed (even limited-mobility) CDMA is that a fixed environment permits power control to very accurately track power, thereby, reducing overall power transmission, increasing capacity and battery life. Nonetheless, above all, CDMA spread spectrum remains optimized for low-power and small-size mobility. This is not to say that Qualcomm, who is dedicated to quality communication, not just CDMA, will not develop fixed environment solutions so long as it does not slow the spread of its mobile architecture. Sprint intends to offer CDMA campus LANs, and I believe that “engineer” implied that “Redmond” was using a CDMA LAN now.
Spectrally efficient, 1xEV-DO provides 7.4 Mbps/cell (3 sectors) aggregate forward peak throughput. Because the majority of data applications receive more data than they transmit, the 1xEV-DO data rate is asymmetric, with a peak data rate on the forward link of 2.457 Mbps/sector, and 153.6 Kbps/sector on the reverse link. Because it uses network resources efficiently, in a high-speed mobile environment, the average forward link throughput in a three-sector cell is 2.0 Mbps/cell with a single receive antenna and 3.1 Mbps/cell with diversity receive antennae.
According to the founder and CTO of Airvana, Vedat Eyuboglu, “A well-engineered 1xEV-DO network delivers average download data rates between 600 to 1200 kbps during off-peak hours, and between 150-300 kbps during peak hours.” Robust Internet use begins around 128 kbps, permitting large downloads and multimedia; whereas almost full network transparency, which is close to the office LAN experience, begins at around 384 kbps. Not only is 1xEV-DO capable of supporting high-speed Internet access with full mobility at pedestrian or vehicular speeds, it can equally supply the home or hot spots, such as hotels or airports. Eyuboglu believes public 802.11 wireless data services can only be successful if integrated with the high-speed wide-area-network services of 1xEV-DO. (“Engineer” once claimed a TV cable company could use its cable to interface with the Qualcomm Hornet micro-BTSs to supply economical Internet access to a neighborhood.)
Eyuboglu stated, in what I would call a description of machine beauty, that recent advances in wireless communications, such as adaptive modulation system, advanced turbo coding, multi-level modulation, and macro-diversity, “allows 1xEV-DO to achieve download speeds that are near the theoretical limits of the wireless channel. … 1xEV-DO also takes advantage of a new concept called ‘multi-user diversity.’…Multi-user diversity can improve the overall throughput of a base station by a factor of two.” According to Eyuboglu, “Combined with Diff-Serv based QoS mechanisms, flexible 1xEV-DO packet schedulers can enable QoS within the entire wireless network.” Because the air interface is the bottleneck in wireless access, the breakthrough capabilities of a 1xEV-DO network are required to flexibly support both user- (premium services) and application-level (allocating resources based on the application’s specific need for data rates) quality of service. Per-subscriber costs are reduced because data subscribers using text-based application do not need the higher speeds required for digital imagery, which can be reserved for demanding applications like media streaming.
The nature of the logic and physics of radio wave transmission require different design rules to optimize data throughput because it is spectrally efficient. The machine beauty of 1xEV-DO is an elegant marriage of optimized and spectrally efficient coding with the simplicity of segregating this optimized data architecture, including the data only carrier, from the competing demands of instant universal access and maintaining quality of service even at the periphery of the RAN cell.
wirelessfuturemagazine.com
Architectural Advances In 1xEV-DO.
Because the rapid growth of the Internet and Mobile Wireless define our times, their convergence not only shapes markets but also society itself. The ability to communicate at anytime, anywhere¾using diverse devices¾transforms how people will work and play, how they will live and learn. While operators agree that Mobile Internet services are critical to their future, their issue is: which 3G architecture promises the best choice? Qualcomm believes the answer lies in its 1xEV-DO because it offers a new optimized architecture specifically designed for delivering packet data that offers spectral efficiency that is 3-4 times greater than current 1x RTT or the proposed UMTS standard.
Given this technical advance in bandwidth efficiency, another revolutionary performance play, why 1xEV-DO? What are the architectural advantages built into its design rules? What is the business case?
The business case for Wireless Internet services vitally depends upon sufficient subscriber capacity and a high quality level of service that is ultimately a function of average data throughput. Total throughput is the sum of average capacity available to multiple users in a sector. Because most Internet applications, like web browsing or video streaming, are asymmetric, with downloads exceeding uploads by a factor of 6 to 8 times, optimizing the forward airlink is vital.
Two factors primarily determine forward-link performance: (a) Burst Data Rate, and (b) Multiplexing Efficiency. The burst data rate is a peak rate¾the maximum transmission speed for any individual under ideal conditions when using a shared channel. On the one hand, when the system is lightly loaded, data throughput approximates the burst data rate. On the other hand, when the system is heavily loaded, the actual throughput the subscriber sees is a function of the efficiency of the multiplexing used to divide the air resources among multiple users. In contrast to a peak rate that influences only a user’s ideal experience, throughput affects both the user’s everyday experiences and the operator’s costs. As throughput increases, the capacities of cells increase, meaning fewer cells are necessary, which takes less capital investment and reduces operating expenses. Less sunk and operating costs yield higher margins.
During busy periods, 1xEV-DO networks offer 3 to 4 times more data throughput compared to 1xRTT or proposed UMTS networks. Its data optimized architecture has increased bandwidth efficiency 2 times by increasing burst data rates and 1.5 to 2 times by increasing the efficiency of its multiplexing solution.
The 1xEV-DO forward link uses spectrum efficiently because of: (a) multi-user diversity¾the full power of the base station is dedicated to a single user at any given time, a user who is selected because of highly favorable RF conditions, and (b) adaptive modulation¾the Access Terminal (AT) continually gives informative feedback to the Access Point (a data BTS) through its requested data rate, what it needs and can receive under present channel conditions. Using this information to select a user, the Access Point (AP) always transmits at full power, thereby achieving very high peak rates for users. To maximize efficient data throughput, information from both the AT and AP conjointly and dynamically determine each user’s forward link data rate. Transmission- power is used more efficiently when dedicated to a single user at any given time¾another elegantly simple, but powerful, solution to the fundamental problem of how to use expensive limited spectrum for propagating data.
A proportional fairness-scheduling algorithm fairly provides for multi-user diversity by improving traffic flow and maximizing overall average subscriber throughput. To improve average throughput by taking advantage of data’s variable bit rate requirements, the algorithm maintains a running average of each user’s RF conditions, using its C/I (SNR) information to deliver data at peak rates when conditions are favorable, to avoid delivery when unfavorable, and to increase priority fairly by tracking the duration since the last turn for service. At any given moment, the dynamically assigned data rate adjusts as rapidly as every 1.67 mSec, providing every subscriber with the best possible rate.
In contrast, when making CDMA voice calls, data rates were designed to provide a constant bit rate, usually between 8-16 Kbps. Under adverse channel conditions, like, say, created by fading as the propagation wave decays with distance, advanced power control increases transmission power to maintain, nay, to preserve the requisite low fixed rate, or the call would be dropped. That is, mobiles with less ideal RF conditions are favored. Thus, voice requires a reliable, but fixed, low data rate system to ensure the integrity of the voice channel. All mixed voice/data channels have voice-centric architectures designed to preserve the QoS of voice. This means that a mixed system also sends text, image, or media data using its voice-optimized system that employs a fixed data rate that cannot adapt upward to use an adaptive-rate scheme. Necessarily, this reduction in degrees of adaptive freedom decreases spectral efficiency when sending packet data.
Furthering the contrast, the 1xEV-DO network is adapting to the present channel conditions of each active user every 1.67 milliseconds. All active terminals constantly measure their channel quality based on pilot signals from all surrounding BTSs, requesting whatever rate from whichever base station can best meet their need now, permitting the selected base station to transmit the highest data rate a terminal’s channel conditions will permit.
Depending on the requested data rate, which ranges from 38.4 Kbps to 2.45 Mbps, the BTS selects the appropriate multi-level modulation schema. These modulation schemata vary in their complexity, where increasing complexity of advanced modulation coding equates to higher bandwidth efficiencies. Although existing 3G voice standards are limited to QPSK, the 8-PSK and 16-QAM modulation formats were first commercially used in 1xEV-DO.
Advanced coding requires higher processing power, more efficient algorithms, or both. Error control coding advanced significantly with the invention (I believe Viterbi had a role here.) of turbo coding as a powerful error correction scheme that uses the principle of redundancy. In regards to receiver noise and interference, turbo coding permits a communication system to operate near the Shannon theoretical limit. On the forward link, HDR’s maximum redundancy is 33% greater than that of 1x.
Qualcomm first used macro-diversity in combining multipath signals, including those from multiple BTSs, to combat fading and to create soft handoffs in voice-centric systems. However, some of the efficiency of this form of macro-diversity is reduced by the voice-required-specification that multiple BTSs send the same air link frame. More important, when voice-centric systems send packet data, packet scheduling is pushed from the BTS to the BSC, which introduces delays and reduces the efficiency of scheduling.
In contrast, HDR architecture overcomes these shortcomings of voice-centric design by using macro-diversity based on radio selection diversity. In radio selection diversity, as described above, each active terminal tells the network which access point it wants to receive, based on the quality of the pilot signal from the surrounding radios. The radio network controller (BSC) ensures that packets are sent to the selected BTS, which, in turn, schedules the packets for transmission.
On the one hand, in 3G voice-centric systems resources are allocated among all calls; this includes packet data calls, continuing to use more power for any disadvantaged data-requestors. On the other hand, to improve spectral efficiency, HDR maximized transmission by using a packet-based time-division multiplexing scheme. Because it analyzes existing diverse channel conditions among multi-users, and then transmits at the maximum data rate in short bursts, using all of its resources at each instant, to meet the data needs of one user at a time, it creates unusually efficient multiplexing. A scheduler in the BTS determines the sequence with which resources are thereby used to create this efficiency. When the efficiency of a round robin or turn-taking scheduler is compared to the proportional fairness algorithm used by Qualcomm, average system throughput for multi-user diversity is significantly greater than that of round robin scheduling. Multi-user diversity scales rapidly with only a few active users to improve efficiency by 50-100%.
In the reverse link, HDR architecture continues to use much of the same architecture as 1x because it happened to provide an excellent reverse link for packet data. Not only does this design choice simplify terminal operation, but it also provides a low-delay means for carrying TCP acknowledgments. In addition, however, the HDR architecture permits adaptive rate control on the CDMA reverse link, supporting rates from 9.6 to 153.6 Kbps, to achieve throughput that can be as much as 50% faster than voice-centric 3G systems.
The “always on” experience is transparent to the fairness-scheduling algorithm. Any delay is transparent because it could as easily be from an unremarkable hop- or server-delay in the Internet. The IP address is maintained in dormant mode when not active¾always on, awaiting reuse. The two-states, active or dormant, are required to (a) allow an inactive terminal to go to sleep to conserve battery power; (b) reduce required overhead for forward air link frames, and (c) better manage the forward and reverse link performance. An important feature of 1xEV-DO networks is its ability to support both a quick connection set-up or teardown, using a very short connection request message compared to voice-centric requests, and a fast re-connect capability, features which are particularly useful in packet data applications involving the Internet.
The 1xEV-DO international standard ensures interoperability, including an interoperability specification that precisely defines the interfaces between the various modules in the network.
Taken together, these advanced features increase both the user’s experience and the overall system capacity. Simply put, because data bits and voice bits are not created equal, 1xEV-DO’s elegant marriage of the power of multi-user diversity and the simplicity of a decoupled data carrier enables faster burst data rates, greater average throughput, more efficient multiplexing, advanced modulation and error coding, differentiated quality of service, and cheaper data bits. In Dr. Jacobs view, the key point for operators to understand is that high data margins require greater throughput and cheaper data bits. I believe a second key point that traditional trajectories of performance, lower costs from growing economies of scale, and a growing set of applications will grow customer demand for data. Give customers what they want, and they will want more because the basic and universal human need for information is almost as pressing as their need for interpersonal connection.
For technical details see:
qualcomm.com
(Whether any DV channel designed to optimize a simultaneous trade-off between voice and data requirements within a single shared channel can equal the efficiency of isolated channels remains to be demonstrated in practice. Until it is so demonstrated, 1xEV-DV may simply be an expectations-management tactic intended to slow the roll-out of DO by competitors who have no hope of matching its power in the foreseeable future. The strategy of time-dynamic shifting between two separate carriers is clearly Qualcomm’s preferred solution because of its unequaled efficiency, given the nature of the logic and physics of spread spectrum transmission. But, in the meantime until the market decides, it will prepare it own DV entry to compete in that potential market. You cannot make your customers or competitors drink even when you lead them to the purest water, but if they choose to drink a mix of voice and data from the same glass, give them the best glass to drink if from. Remember, that is exactly what cdmaOne already had because it used a spread spectrum modulation architecture that Europe despised, said it would never work, that we had already invented it here, until they capitulated. We began with Viterbi claiming that, once again, errors were being repeated. If he has the science right, as I believe to be the case, then data optimization requires a data only carrier because the logic and physics of nature requires it as the ideal solution to using the limited resource of radio-wave spectrum efficiently, which is required to meet the growing needs for voice and burgeoning needs for data and to create a viable business design for the future.) |
