This chapter reviews the progress that has been made by the Army over the past 5 years since the publication of Energy-Efficient Technologies (NRC, 1997). It discusses reasonable expectations for the near- and far-term Land Warrior (LW) ensemble based on improvements observed from prototype through Stryker Interoperable (SI) to the Objective Force Warrior-Advanced Technology Demonstration (OFW-ATD). The chapter also reviews and extends predictions of the previous report and describes significant changes in commercial development trends.
Overall the Board on Army Science and Technology (BAST) committee was impressed by the amount of effort that has been put into power and energy concerns for the Objective Force Warrior-Advanced Technology Demonstration (OFW-ATD) Program. The Army, through its contracted lead technology integrator (LTI), now has a process that allows developing prototype technology to be inserted into the LW program. Initially, the focus is on relatively near-term technologies that can be demonstrated in the OFW-ATD and later inserted into a new version of Land Warrior, Advanced Capability (LW-AC). A snapshot of the concepts being considered for the next LW by the LTI showed the committee the direction of Army power solutions and soldier requirements.
The OFW design team is focusing on a 24-hr autonomous mission. It is assumed that soldiers will be resupplied at least once every day and that each unit will have the means (possibly a robotic vehicle) to carry extra batteries, rechargers, or fuel needed for 72-hr missions. These assumptions led to a goal for the future soldier fighting load as low as 50 pounds, including no more than 2 pounds for a power system providing 12 W.
The weapon for the OFW will be cabled for recharging through the centralized source and detached when fighting. A hybrid power concept is being considered for the power source to support extended missions. It involves a high specific energy/low specific power system for steady loads and a high specific power/low specific energy system for peak demands.
Peak power demand anticipated for the LW ensemble will not change appreciably under the initial OFW-ATD concept. Table 5-1 also shows that estimates for LW-SI peak power did not change substantially from the original LW estimates in Energy-Efficient Technologies (NRC, 1997). Similarly, average power estimates for the ensemble have remained relatively constant. As shown in Table 5-2, the peak power, average power, and average/peak ratios (without radios) for three generations of LW (LW, LW-SI, and OFW) are within 20 percent of one another.
The committee’s analysis was organized around four categories of functions that make up the suite of LW electronics, including displays, computer subsystems, sensors, and communications.
There has been a substantial reduction in the power requirement of the helmet-mounted display from the 1997 LW to the proposed 2007 OFW. The other displays consume approximately the same power in spite of power reduction progress made in reducing power needed for displays. This is probably the result of added capabilities (e.g., color and higher resolution) anticipated for the recent LW versions.
Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 46
Meeting the Energy Needs of Future Warriors 5 Progress This chapter reviews the progress that has been made by the Army over the past 5 years since the publication of Energy-Efficient Technologies (NRC, 1997). It discusses reasonable expectations for the near- and far-term Land Warrior (LW) ensemble based on improvements observed from prototype through Stryker Interoperable (SI) to the Objective Force Warrior-Advanced Technology Demonstration (OFW-ATD). The chapter also reviews and extends predictions of the previous report and describes significant changes in commercial development trends. OBJECTIVE FORCE WARRIOR-ADVANCED TECHNOLOGY DEMONSTRATION Overall the Board on Army Science and Technology (BAST) committee was impressed by the amount of effort that has been put into power and energy concerns for the Objective Force Warrior-Advanced Technology Demonstration (OFW-ATD) Program. The Army, through its contracted lead technology integrator (LTI), now has a process that allows developing prototype technology to be inserted into the LW program. Initially, the focus is on relatively near-term technologies that can be demonstrated in the OFW-ATD and later inserted into a new version of Land Warrior, Advanced Capability (LW-AC). A snapshot of the concepts being considered for the next LW by the LTI showed the committee the direction of Army power solutions and soldier requirements. The OFW design team is focusing on a 24-hr autonomous mission. It is assumed that soldiers will be resupplied at least once every day and that each unit will have the means (possibly a robotic vehicle) to carry extra batteries, rechargers, or fuel needed for 72-hr missions. These assumptions led to a goal for the future soldier fighting load as low as 50 pounds, including no more than 2 pounds for a power system providing 12 W. The weapon for the OFW will be cabled for recharging through the centralized source and detached when fighting. A hybrid power concept is being considered for the power source to support extended missions. It involves a high specific energy/low specific power system for steady loads and a high specific power/low specific energy system for peak demands. Comparison of OFW Concepts with Land Warrior Peak power demand anticipated for the LW ensemble will not change appreciably under the initial OFW-ATD concept. Table 5-1 also shows that estimates for LW-SI peak power did not change substantially from the original LW estimates in Energy-Efficient Technologies (NRC, 1997). Similarly, average power estimates for the ensemble have remained relatively constant. As shown in Table 5-2, the peak power, average power, and average/peak ratios (without radios) for three generations of LW (LW, LW-SI, and OFW) are within 20 percent of one another. The committee’s analysis was organized around four categories of functions that make up the suite of LW electronics, including displays, computer subsystems, sensors, and communications. Displays There has been a substantial reduction in the power requirement of the helmet-mounted display from the 1997 LW to the proposed 2007 OFW. The other displays consume approximately the same power in spite of power reduction progress made in reducing power needed for displays. This is probably the result of added capabilities (e.g., color and higher resolution) anticipated for the recent LW versions.
OCR for page 46
Meeting the Energy Needs of Future Warriors TABLE 5-1 Comparison of Estimated Power Requirements of Land Warrior System, by Function (All Peak Power) Function Land Warrior, 1997a (W) Land Warrior (Stryker), 2004b (W) Objective Force Warrior, 2007c (W) Communications Soldier radio 7.4 5.97 6.2 Squad radio 14 7.8 7.8 UAW/robotic vehicle 6 Computer displays Handheld flat panel 6.4 7.04 7.05 Helmet-mounted 4.9 1.4 0.5 Integrated sight—module display 2.6 2.65 3 Sensors 7.9 16.75 9.5 Computer 14.8 15.7 17.42 Total 58 57.31 57.97 aEstimates from NRC (1997). bBreakdown of Stryker interoperable wattages (Brower, 2003) • Soldier radio, 2.5 (WLAN card) + 0.17 (WLAN digital radio) + 3.3 (WLAN antenna and amplifier); • Squad radio, 7.8 (leader radio); • Handheld flat panel, 6.29 (display) + 0.75 (keyboard); • Helmet-mounted display, 1.4; • Integrated sight display, 2.15 (thermal weapon sight) + 0.5 (daylight video sight); • Sensors (everything but the 25-W chemical agent detector), 4 (multifunction laser) + 0.2 (weapon user input device) +1.17 (card reader) + 0.17 (GPS interface) + 2.4 (GPS card) + 0.25 (dead reckoning) + 0.5 (microphone) + 0.76 (helmet integrated assembly) + 0.6 (laser detector) + 2.5 (slave hub processor) + 0.76 (computer USB hub) + 0.32 (slave hub) + 2.97 (weapon hub) + 0.15 (chemical agent detector); and • Computer, 12 (computer processing card) + 3 (DRAM and radio frequency conversion) + 0.7 (computer/master hub subsystem). cBreakdown of OFW wattages (Erb, 2003): • Soldier radio, 7.8 (JTRS numbers not available; assumed the same as Stryker MBITR radio); • Squad radio, 4.4 (communications processor card) + 0.6 (WLAN card) + 0.6 (VoIP processor) + 0.6 (WLAN antenna); • UAW/robotic vehicle, 3 to 10 W for como-crypto interface (Brower, 2003); • Handheld flat panel, 6.3 + 0.75 (handheld KB and cable); • Helmet-mounted display, 0.5; • Integrated sight, 3 (HIA, module including breakaway connection to body PAN); • Sensors, 2.15 (thermal weapons sight) + 1.1 (daylight video sight) + 4 (multifunction laser) + 1.5 (GPS) + 0.25 (dead reckoning module) + 0.5 (microphone/speaker assembly); and • Computer, 2.1 (computer assembly) + 10.9 (computer processing card) + 3.42 (body PAN hub) + 1 (PAN weapon hub). NOTE: DRAM, dynamic random access memory; GPS, global positioning system; HIA, high integration actuator; JTRS, Joint Tactical Radio System; KB, kilobyte; MBITR, multiband intra/inter team radio; OFW, Objective Force Warrior; PAN, primary area network; UAW, universal access workstation; WLAN, wireless local area network; USB, universal serial bus; VoIP, Voice over Internet Protocol. TABLE 5-2 Comparison of Estimated Peak and Average Power and Their Ratios for Land Warrior Systems System (Without Radios) Land Warrior, 1997a Land Warrior (Stryker), 2004b Objective Force Warrior, 2007c Peak power (W) 35.3 43.59 37.97 Average power (W) 15.35 19.5 15.8 Average/peak ratio 0.435 0.447 0.366 aEstimates from NRC, 1997. bEstimates from Brower, 2003. cEstimates from Erb, 2003.
OCR for page 46
Meeting the Energy Needs of Future Warriors Computers The computer subsystems evolved in a different way. As opposed to the single processor in the earliest LW system, the OFW design includes a number of processors that are interconnected through multiple high-speed local area networks (body LAN). To assure longevity, the OFW design has gone to an open architecture with several standardized buses (e.g., Firewire, gigabit Ethernet). While a variety of buses enhances the number of modules that could connect to the OFW electronics, it does exact a premium for power to keep all the buses energized, even if there is only one transaction type per bus. The result has been that computer power demands of the three generations of LW are about constant in spite of the significant improvement in energy efficiency of the underlying computer system technologies. Sensors Power demand estimates for the sensor suite have increased slightly over the three generations. The number and types of sensors are similar, but there have been significant improvements in functionality. Communications The OFW-ATD is working to develop power-aware applications and an intelligent middleware layer that will efficiently manage bandwidth usage. It will use a radio (Joint Tactical Radio System (JTRS) Cluster 5 SLICE radio) that creates a peer-to-peer network architecture, but the software-based design solution for JTRS may not allow for reductions in power demand. Initial OFW estimates take an optimistic approach to what will be available in 2007 by using power numbers no worse than those for the MBITR radio with LW-SI, thus enabling a complete high-level comparison of the overall power demands of LW electronics. The importance of soldier communications-electronics to reductions in power demand is discussed further below. LAND WARRIOR POWER IMPROVEMENTS The Army Program Executive Office (PEO-Soldier) and the LTI provided briefings on facets of the OFW-ATD relating to power, including power sink technologies, system design, doctrine, networking, and power sources. For each recommendation derived from the five conclusions in Energy-Efficient Technologies (NRC, 1997), the LTI provided specific examples of how the OFW-ATD would improve the energy effectiveness of the LW system. Application of Energy Efficient Technologies to the OFW-ATD Program Efficient power usage is understood to be critical to the success of the OFW-ATD and has been identified as a key performance metric. Every energy-consuming capability must earn its way onto the system. The OFW-ATD is using state-of-the-art technology developed by both commercial entities (for main computers) and government programs (for radios) to reduce system power demand. The OFW-ATD system architecture is intended to be flexible enough to incorporate new technology as it is developed. The OFW-ATD is also using computer-aided design, simulation, and profiling tools to perform power analysis on proposed designs. All applications to be developed will be power-aware. A custom Linux kernel is being created, tuned for power and security. To synchronize doctrine with technology and minimize soldier communications transmissions, the OFW-ATD software team is working closely with the operational effectiveness team to analyze and prioritize the data that need to be transmitted across the Army force structure. The OFW-ATD is also studying the operational utility of unmanned vehicles, both air and ground, as nodes in the peer-to-peer network architecture that would have more powerful reach-back capabilities. In response to the 1997 recommendation for research and development in rechargeable batteries, the OFW-ATD is working with a commercial battery supplier to devise a rechargeable battery with a specific energy on the order of 200 Wh/kg. The OFW-ATD is also tracking advances in energy sources and is exploring hybrid systems. It also supports continued research into advanced energy sources and is particularly interested in direct methanol fuel cells. Committee Observations on Initial OFW-ATD Concepts The assumption that future soldiers will be resupplied every 24 hr does not necessarily modify the goal of 72-hr self-sufficiency. To justify the 24-hr assumption, OFW makes the further assumption that each unit will use a vehicle for resupply (such as a robotic Mule). This assumption, endorsed by the Army, is important because it has substantial logistical implications, for including procurement and transportation of Mules, spare parts, and fuel. The A123Systems battery technology is high payoff/ high risk. The high power capability of this approach is based on doped LiFePO4 as extrapolated from laboratory experiments. The chemistry is inherently safe, and the raw material is not expensive. Nevertheless, there are other rechargeable alternatives, and the OFW-ATD will probably need to pursue these concurrently to ensure success. OFW plans to embed data-logging capability in the system to track energy and power demand. This is an excellent idea that could provide an initial basis for subsequent development of needed models. Actual power usage profiles can also be used to evaluate future design trade-offs. This should provide substantially more accuracy for power modeling than the current estimates of subsystem duty cycles. The committee notes that hybrid techniques, such as
OCR for page 46
Meeting the Energy Needs of Future Warriors charging an electrochemical capacitor from a battery, can be used to provide pulses of energy for low-duty-cycle (1 percent or less) devices without compromising battery capacity. However, if devices like the multifunction laser (estimated at 4 W peak power) are used several times in succession, the local duty cycle will either not allow enough time for the hybrid system to recover or else appear as a higher duty cycle peak demand with adverse affects on capacity, especially if batteries are the main power source. The committee had additional substantive observations on the length of the LW procurement cycle, incentives for saving power, and the important area of soldier communications. Length of the Procurement Cycle The time horizon of the LTI contract is probably not long enough to collect effective feedback even though the program has adopted an iterative development and improvement cycle. Further, the relatively short duration of the cycle militates against there being time to bring energy-efficient system-on-a-chip (SoC) technology into play and significantly reduce power demand of one or more of the LW subsystems. Although SoC technology targeting the OFW application was recommended in Energy-Efficient Technologies (NRC, 1997), OFW-ATD will use off-the-shelf electronics componentry. The Army should begin the development of SoC technology that can evolve with requirements as they are understood and with new developments in algorithms and protocols. The current approach to designing and procuring Army soldier systems should be contrasted with the approach to designing commercial products such as cell phones. Each generation of an Army system starts with a new contractor and a clean sheet of paper, allowing only an after-the-fact, lessons-learned critique of the previous generation. There is not a lot of learning transferred from one generation to the next, leading to a lack of continuity in design concepts. In the commercial world, by contrast, there is continuity between products over multiple generations. Commercial electronics developers aim for progressive improvements to design, with successive generations of SoCs containing capabilities better than those of the previous generation. By building on earlier SoC designs, the cost and risk of the later generations are substantially less than the cost and risk of the first generation. Another cost of the standards-based plug-and-play strategy of the OFW is that standardized USB and Ethernet hubs for the wired soldier body LAN use considerably more energy and do not directly enhance the effectiveness of the soldier. There should be investments in developing low-power interconnect technology. For example, a fundamentally different approach would use a high-speed, short-range, wireless body LAN that has an undetectable emission signature. One approach to this goal would use ultrawideband (UWB) radio transmission, which is being developed commercially to transmit in the 3 to 5 GHz range and (from the 802.11.3a proposed standards) at a rate of 100 Mbits per second at less than 100 mW total power demand; receive at 200 mW; and, most important, have sleep modes that are three orders of magnitude lower (Batra et al., 2003). The very low transmit power and secure characteristics of the transmission signal would provide a radio frequency (RF) signature that would not be detectable beyond 10 m (or even less if the transmit power is constrained further). Chip sets for this network approach are projected to be available from Motorola, Intel, and other vendors for home video networks in the next 3 to 4 years at a cost of less than $5 per node. Assuming that internal or local operational interference is not a problem, such wireless technology could obviate the tethering of data buses on a soldier’s system. Incentives for Reducing Power A lesson from the original LW integration program is that there was not enough time or money to fully optimize energy efficiency. Due consideration must be given not only to the various power sources and sinks, but also to designs for electronics integration and power management. The OFW LTI will propose systems for integration, but neither the LTI nor the Army PEO have enough influence over concurrent acquisition efforts to effectively reduce power demand in the main electronics subsystems. An incentive structure would be one way to achieve innovation at the subsystem level. The shipping cost of batteries in the first Iraq conflict is estimated at more than $500 million, an indication of the savings that are possible in logistics alone. This includes only the cost of the logistical support, not of the batteries themselves. By reducing the average power demand by only 10 percent, a saving of $50 million could have been realized. This would be enough to develop five chips at $10 million apiece. The cost per soldier of providing batteries and related power source hardware, such as chargers, will be substantially higher as the OFW electronics suite is introduced. It is also possible to have reductions of much more than 10 percent. Finally, if the power requirements are reduced, the type of energy sources can be changed, which could also save costs. The committee therefore recommends that the Army should undertake a complete life-cycle cost analysis to determine the overall savings achievable by a substantial increase in development activity solely targeting power reductions (such as SoC design). Soldier Communications OFW is assuming the JTRS radio. Little information was available about the progress of JTRS design, but committee members familiar with software radio research suggested several areas for the Army to investigate for possible improvements in energy efficiency.
OCR for page 46
Meeting the Energy Needs of Future Warriors The JTRS has defined a standard based on the CORBA infrastructure, which was developed by the commercial sector for sharing software application modules over the Internet. While this has the advantage of defining a standardized interface, the power requirements associated with such an interface are likely to be very high, since it implicitly requires a software processor for the radio implementation. As shown in Figure 5-1, this approach will result in energy efficiency that is many orders of magnitude less than would otherwise be attainable. The commercial sector is also developing multiband and multistandard radios, but it uses software to reconfigure the hardware and is not considering approaches similar to the JTRS CORBA approach for its commercial applications. The fully flexible JTRS radio will need an analog front-end that is tunable over a broad range of frequencies, from low megahertz to high gigahertz. This will require a considerable breakthrough in RF design and is potentially very inefficient with respect to power. The commercial approach is to use a number of duplicate RF chains, each of which is optimized for a single frequency band. It is suggested that careful consideration be given to how the analog portions of the radio are designed, since Moore’s law technology scaling, which benefits the digital computation, will increase the power required for the analog circuitry. To reduce the required transmit power, the soldier JTRS radio is being designed as part of a mesh network where each radio serves as a repeater to extend transmission coverage. However, care should be used in designing the ad hoc network protocols needed in such a network. Simulation studies have shown that the greedy approach—increasing transmission power only to the point of communication with at least one other node—leads to rapid depletion of the batteries of the nodes in the center of the formation as well as to added latency. Nodes in the center of the formation have to store and forward communications between nodes on the periphery, increasing local power demand and accumulating more latency for all messages. Recognizing and broadcasting over congested areas more evenly distributes power demand among all the nodes. Radio network simulations have been developed that accurately model radio power demand and these could be used to evaluate the energy efficiency of various protocols. COMMERCIAL TRENDS This section discusses new commercial trends and whether the trends highlighted in Energy-Efficient Technologies (NRC, 1997) are still valid and relevant to the Army. It also discusses the validity of projections made in that report and updates the original LW predictive model to reflect more recent goals and requirements. Continuation of Moore’s Law The National Technology Roadmap for Semiconductors (NTRS) projections are still valid, as Moore’s law is expected to hold true for at least another 8 years. The predictions of the 1994 NTRS tables regarding feature size and voltage reduction have been realized and even slightly exceeded (SIA, 1994). On the other hand, neither chip sizes nor the number of bits per chip on dynamic random access memory (DRAM) have grown at the projected rate, because lithography and manufacturing techniques at and below 0.1 micron are very expensive. Circuit development cost, particularly mask development, is rising with progressively smaller integration scales: from less than $1 million for micron-scale to $3 million for nanometer-scale. The demand for ever denser circuitry has slowed, while demand for less complex application chips is growing exponentially. For example, a 90-nm complementary metal-oxide semiconductor (CMOS) can provide 8 to 64 MB/cm2 (SRAM or DRAM) and 25 M logic gates. However one of the main sources of chip demand, the cell phone, typically requires only 2 M gates. Even most complex applications require on the order of 8-10 M gates. One of the largest commercial drivers is the personal computer (PC), but PC sales have declined since 1997. Industry is looking for new growth applications, but these new applications are likely to require fewer gates than cutting-edge technology. Even though application chips are less complex than state-of-the-art chips, costs and prices have still benefited from the feature size reduction. Decreasing demand for traditional masks in development may well combine with increasing demand for application chips to reduce costs. Low-Power Electronics Technology Since Energy-Efficient Technologies was written, there has been rapid progress in low-power technology, outstripping the Semiconductor Industry Association (SIA) roadmap, with increased performance and less power per circuit expected from scaling the transistor and wiring dimensions and to some degree the voltage. Since 1997, the energy efficiency of circuits has improved by a factor of at least 5. By one measure, the reduction in demand is greater than the improvement in rechargeable batteries, since time between recharges has only increased 20 percent. With increasing functionality, the net power demand for consumer electronics has remained essentially constant. However, the exponential growth that has characterized improvements in commercial microprocessor performance may dramatically slow in the next few years. There are numerous barriers to progress after one or two more generations. The most relevant to low-power electronics are the various leakage currents that cause a passive power component, which is becoming increasingly significant compared to the active power.
OCR for page 46
Meeting the Energy Needs of Future Warriors It has been anticipated for some time that we cannot reduce the supply voltage much below 1 V for high-performance applications because the sub-threshold leakage current would go up greatly if the threshold voltage is scaled much below 0.3 V, as would be needed to maintain performance. For low-power applications at reduced performance, optimization studies for minimizing power at a given performance show that this threshold voltage needs to be increased as the voltage is reduced to maintain a balance between the static power and the reduced active power. Reducing the supply voltage down to about 0.5 V reduces the power more rapidly than the speed, with the result that energy/computation is improved, an important result for low-power electronics but made less palatable by the need to convert the voltage somewhat inefficiently from a higher voltage source. Generally the energy-efficient design principles set forth in Energy-Efficient Technologies (NRC, 1997) are still good guidelines. The most important new trend is toward changes in technology and design to deal with the growth in static power relative to active switching power. This increased static power is from the increased transistor “off” current as the threshold voltage has been reduced, along with the operating power-supply voltage, to maintain performance growth. Static power also increases greatly with operating temperature. Less important, but rapidly becoming more so, tunneling currents in the gate insulator and at the drain-body junction are now limiting further transistor scaling, which requires thinner gate insulators and heavier body doping to maintain current trends. These limits are even more imposing for some applications such as static random access memory (SRAM) development, which is proving very difficult owing to threshold fluctuations in small-width devices. As a result, several commercial chip design trends have developed: Technologies are being developed with multiple threshold voltages (Vt), allowing designers to sprinkle in low Vt transistors in performance-critical areas while using higher Vt transistors to save static power in less critical areas. The use of large switches in series with the power supply (header or footer devices) to turn off leakage current in inactive blocks of circuits. This is only effective after the stored energy in all the affected circuit capacitances leaks away. Reducing the power supply voltage to inactive circuits to reduce leakage. This is less effective than the use of large switches but allows the latched state to be retained. The use of adaptive body bias to raise Vt in inactive blocks or to compensate for Vt spreads due to process variations. Except for multiple threshold voltages Vt (the first technique above), these techniques have been implemented mainly in low-power, battery-operated applications with significant standby time. Other trends aimed mainly at active power are (1) clock gating to turn off idle circuit blocks and (2) voltage-frequency scaling to reduce power (and energy/ operation) when peak performance is not needed (Burd and Brodersen, 2002). There are efforts in the industry to develop design approaches using nonvolatile memory, so as to have no leakage current in the memory elements themselves. EEPROM and FLASH, for example, are more storage devices than memory and can be used in special applications with relaxed requirements, such as write-cycle endurance. While nonvolatile memory is not likely to take the place of SRAM or DRAM embedded on processor chips, the nonvolatile memory used in the electronics and miniature hard drives for such products as digital cameras and personal digital assistants (PDAs) is very likely to find its way into some components of future soldier systems. Changes in Commercial Development Trends There is beginning to be a divergence between low-power personal computers and handheld devices (cell phones, PDAs, and the like). While computers emphasize higher processing speed with relative disregard for power leakages, handheld devices seek low-power implementations. Radio hardware chips were universally proprietary in 1997, but many more chips are being made openly available today. This may represent an opportunity for the Army but would require the Army to evaluate the use of commercial waveforms for Army applications. Indeed, wireless LANs using 802.11 waveforms have already been used in the LW-Stryker generation for the soldier intercom. Molecular electronics (e.g., nanotech applications such as individual transistors and carbon nanotube batteries) are viewed as follow-on technologies once the use of silicon plateaus. New materials, such as plastic semiconductors, are being developed for gates and displays and have the potential to one day be very inexpensive. The Army Institute for Soldier Nanotechnology is investigating the integration of circuitry into fabrics, so that uniforms for Future Warriors might well incorporate much of the electronic circuitry for LW functions. Trends in Commercial Cell Phone Development The development of dedicated applications such as cell phones emphasizes the system-on-a-chip approach, coupled with low-power architectures and aggressive powerdown strategies (see Chapter 6, “Power Management Approaches”). Power drain for traditional cell phones (voice only) is now considered to be under control. Digital logic originally required more power than analog, but both now require about the same amount. In the future, it is very likely the analog portion will become more dominant. The evolution of multimedia on cell phones has renewed
OCR for page 46
Meeting the Energy Needs of Future Warriors the industry concern for power drain as a competitive factor. Soldiers will also require multimedia, but it is not clear how advances in consumer-electronics architectures will benefit the Army, since the applications are being optimized for consumer applications. Experience tells us that the Army will continue to use commercial developments more as a menu of prospective applications and functions, then engage a contractor to independently develop or integrate the most relevant functions. Energy Efficiency of Integrated Circuits In Energy-Efficient Technologies (NRC, 1997), projections were made of the evolving energy efficiency of processors and programmable digital devices. Those projections are still valid, but other factors that are critically important in determining energy efficiency were not taken into consideration. An important aspect of evolving LW design has been the move toward an open architecture that can provide as much flexibility as possible. Among other things, such flexibility will make it possible to extend system capability into the future, to rapidly deploy prototypes, and—particularly in the communications area—to adapt to a wide variety of legacy systems. While it is clear that such a flexible, open architecture can be achieved, there is a wide range of alternatives that provide equivalent flexibility with greatly varying degrees of energy efficiency. Such alternatives range from the use of software on a conventional processor, through the use of interconnect reconfiguration, as in a field-programmable gate array (FPGA), to unique SoC circuitry. The challenge is to come up with an SoC design that has sufficient generality to provide all the necessary modes but also specificity to the OFW problem domain. Solutions that are optimized for specific applications can have an energy efficiency that is several orders of magnitude greater than solutions providing for relatively unlimited flexibility. It is therefore critical that this cost of flexibility be understood, so that an informed decision can be made about how much flexibility is needed and to what degree specific tasks can be optimized for. Energy Efficiency Metric To quantify the cost of this flexibility, the committee considered the amount of energy, in millijoules,1 required to execute an average operation. An operation is defined as a basic execution element, which for a software processor is an instruction and for other architectures would be an arithmetic or memory function, such as an add, subtract, or delay. The metric is millions of operations, MOP, per millijoule, or MOP/mJ. Energy efficiency can also be a more familiar ratio of rates: MOPS, which is million operations per second, divided by milliwatts, which is millijoules per second. Thus the energy efficiency metric is MOPS/mW, equivalent to MOP/mJ. To see the relationship between this energy efficiency metric and the various levels of flexibility, a variety of designs will be compared that range from completely flexible, software-based processors (including both general-purpose processors and those optimized for digital signal processing) to inflexible designs using hardware dedicated to a single application, which it termed SoC solutions. Metric Comparisons In fixed-function designs, the operation count is straight-forward, which is not the case for comparisons with processors that are flexible. In these cases different throughputs are possible, depending on the benchmark. When comparing architectures for different applications, the committee used the highest achievable throughput numbers. To determine the state of the art for commercial circuits, the efficiency metrics for a number of chips chosen from the International Solid State Circuits Conferences from 1998 to 2002 are illustrated in Figure 5-1. (The chips selected had to be for a technology that ranged from 0.18 to 0.25 microns, and enough information had to be available to do a first-order technology scaling and to calculate the energy and area efficiencies) (IEEE, 1998; IEEE, 1999; IEEE, 2000; IEEE 2001; IEEE 2002). Though this is a relatively small sample of circuits, it is believed that the trends and relative relationships are accurate representations of the various architectures being compared because of the remarkable consistency of the results. Table 5-3 summarizes all the circuits that were used in the comparison. In the table, the designs are sorted according to their energy efficiency, and—very surprisingly—this sorting results in their being grouped into three basic architectural categories, which are differentiated by degree of flexibility. Chips 1-9 are general-purpose microprocessors and are fully flexible without any optimization for a given task. Chips 10-15 are software processors optimized for digital signal processing functions such as required by many of the OFW applications. Chips 16-20 are dedicated application SoCs, with very limited flexibility. The medium-term estimates in Energy-Efficient Technologies (NRC, 1997, p. 137) predicted energy efficiencies of 2 MOPS/mW in 2001 for software programmable digital signal processors and 10 MOPS/mW for dedicated solutions. As seen now in Figure 5-1, these turn out to be conservative in comparison to the actual chips in 2001, which had efficiencies up to 10 MOPS/mW for the software digital signal processors and 100 MOPS/mW for dedicated designs. Figure 5-1 also illustrates that energy efficiency varies three to four orders of magnitude between the most flexible solutions and the most dedicated. It is not surprising that efficiency decreases as the flexibility is increased, but the flexibility is 1 One millijoule (mJ) = 10−3 J.
OCR for page 46
Meeting the Energy Needs of Future Warriors FIGURE 5-1 Energy and area efficiency of different chips from 1998 to 2002. Chips 1-9 are software programmable microprocessors, chips 10-15 are software-programmable digital signal processors, and 16-20 are dedicated designs. (See Table 5-3.) TABLE 5-3 Description of Chips Used in the Analysis Type Chip No. Year Paper No. Description Software-programmable microprocessors 1 1997 10.3 μP—S/390 2 2000 5.2 μP—PPC (SOI) 3 1999 5.2 μP—G5 4 2000 5.6 μP—G6 5 2000 5.1 μP—Alpha 6 1998 15.4 μP—P6 7 1998 18.4 μP—Alpha 8 1999 5.6 μP—PPC 9 1998 18.6 μP—StrongARM Software-programmable digital signal processors 10 2000 4.2 DSP—communications 11 1998 18.1 DSP—graphics 12 1998 18.2 DSP—multimedia 13 2000 14.6 DSP—multimedia 14 1998 18.3 DSP—multimedia 15 2002 22.1 DSP—MPEG decoder Dedicated designs 16 2001 21.2 SoC—encryption processor 17 2000 14.5 SoC—hearing aid processor 18 2000 4.7 SoC—FIR for disk read head 19 1998 2.1 SoC—MPEG encoder 20 2002 7.2 SoC—802.11a baseband NOTE: μP, microprocessor; PPC, power personal computer; SOI, signal operating instructions; DSP, digital signal processing; MPEG, moving pictures expert group; SoC, system-on-a-chip; FIR, first-impressions report.
OCR for page 46
Meeting the Energy Needs of Future Warriors gained at enormous energy cost. With continued reductions in scale, from, say, 180 nm to 45 nm, an estimated 18-fold net advantage in efficiency can be expected. For functions requiring a low number of operations to be executed (one such is the user interface), the energy cost of solutions providing high flexibility will not be a significant component of the overall system power demand. For functions with high processing rates, such as video processing and communications, solutions should be more dedicated to take advantage of the multiple orders-of-magnitude reductions in power that can be achieved. Combining applications that possess both low and high processing demands leads to the most generalized SoC approach, a chip design that provides software programmability where needed for functions that must have full flexibility and more dedicated, power-efficient solutions for high-performance signal processing. It is believed that more than an order of magnitude reduction in the power demand of the digital computation would be achievable in the OFW system if an SoC approach is taken. On the other hand, it is clear that specification decisions that simply mandate a fully software-based system would result in crippling requirements from an energy-cost perspective, as appears to have happened in the JTRS program. For example, compatibility with multiple waveforms can be achieved by several strategies. Multiple dedicated radios could be placed onto a single chip, and since there is an area advantage similar to the power advantage for each dedicated radio using the SOC approach, a number of radios could easily be implemented in the same area that a software programmable solution would require. For the special case of radio designs, this is consistent with an approach that requires specialization in any case, because of the analog RF circuits that will require optimization if reasonable power levels are to be achieved. The energy efficiency of multiple dedicated radios on a single chip would easily be more than an order of magnitude better than that of a software-programmable solution. Another development in the commercial arena that provides increased flexibility is use of reconfiguration as opposed to software programmability, such as used in field-programmable gate arrays (FPGAs). Computation is implemented on these chips using an architecture that is essentially the same approach as that used in dedicated SoCs, giving them an inherent advantage over a software processor-based solution. FPGAs are able to exploit the improvements in the underlying technology better than the software processors, so in the future it is likely that for high-performance computation requiring energy efficiency and full flexibility, the approach of choice will be based on reconfiguration. At present, even though they are not optimized for energy efficiency, commercial FPGAs are still more efficient than software processors. The OFW soldier radio, for example, is being prototyped using FPGAs. An investment by the Army that would develop an energy-efficient, reconfigurable processor could achieve the dual goals of flexibility with reasonable energy efficiency. However, this solution will probably always be more than an order of magnitude less efficient than a more dedicated solution. If all of the high-performance computation were integrated onto one OFW SoC chip the power demand of these functions could be reduced by more than an order of magnitude. Design of this chip using reconfigurable architectures could achieve these gains without compromising flexibility. In the OFW scenario, this flexibility could include radio and communication processors, Voice over Internet Protocol (VoIP) processing, as well as processing for video compression and decompression. For low-rate human interface processing, a software processor could be integrated onto the chip to provide additional flexibility to meet evolving future requirements. Chapter 6 discusses design concepts for such a Future Warrior system. FINDINGS The Army has come a very long way since Energy-Efficient Technologies (NRC, 1997) in understanding the soldier as a system and in taking appropriate actions that result from this understanding. Though in some cases there have been impressive reductions in the power demand of individual items, the reductions are being more than offset by the demands of new and more capable devices as well as the desire to have a highly flexible open architecture. Based on its observations of the overall evolution of the LW and OFW-ATD programs, the committee made six findings, which are discussed next. Constraints on Reducing Power For the LW systems the average power has been 20 W and the peaks have been 60 W over all three generations. The energy savings made possible by technology improvements have been traded for improvements in combat effectiveness as well as to allow the use of plug-and-play architecture to support future evolution. While the desire for such flexibility is understood, turning plug-and-play into a basic requirement comes at a high energy cost and will restrict the use of solutions that could reduce power demand by more than an order of magnitude. Technology Time Horizon The time horizon for OFW is too close. The LW program needs enough time to develop a SoC solution for the OFW and not be constrained to off-the-shelf component solutions. Increasing the development time horizon would allow the program to build on prior programs by evolving the SoC to meet new needs and requirements, similar to the successful approach taken for commercial cell phone evolution, in which each new generation is an enhancement of the last generation with new capabilities.
OCR for page 46
Meeting the Energy Needs of Future Warriors Life-Cycle Costs The full life-cycle cost of providing power for soldier electronics is not being taken into account by the Army. The serious cost consequences of not using energy-efficient technology to design LW must be considered when determining the investments needed for reducing the power demand. The cost saving from requiring fewer batteries and other energy sources over the lifetime of the OFW system will more than pay for the development of highly optimized, low-energy solutions. For example, a 10 percent savings in power could be expected to reduce the number of batteries required by a comparable amount. This would have reduced the logistics cost of delivering batteries in Operation Iraqi Freedom, saving an estimated $50 million. Soldier Communications Power requirements for soldier communications are too great to ignore. As pointed out in Energy-Efficient Technologies (NRC, 1997), the requirements for soldier communications account for a considerable fraction of the overall energy consumption. The absence of reliable, more definitive estimates for the energy expected to be consumed by future OFW radios is thus of considerable concern. The solution being pursued by the OFW LTI is to use whatever radio is available in the time frame required for integration with OFW-ATD without any particular direct control of the radio design or its power requirements. In addition, the actual communication requirements and mission scenarios planned for JTRS are not necessarily synchronized with emerging requirements for network-centricity, and their lack of definition further obscures an already murky picture of what the OFW power demand will be. In particular, the JTRS program, while certainly visionary in its goal of achieving compatibility with multiple communication waveforms, is, perhaps, overly ambitious. The software-defined radio on which the soldier radio is based will require advanced use of integrated circuit technology (highly integrated, mixed-signal SoC chip designs) as well as breakthroughs in protocols and architectures. Further, the design approach is unique to DOD, which effectively prevents the military from leveraging the gains in energy efficiency that are expected in commercial communications gear. The OFW-ATD and LW program place considerable reliance on these JTRS developments with no guarantee that power reduction enjoys an equally high priority in the JTRS program. Without question, the power budgeted for communications is excessive considering the state of the art in communications electronics. The power sources community understands well the inherent limitations to achieving large, short-term, step function improvements in the energy:weight ratio of power sources. However, on the demand side, the communications-electronics (circuits and systems) community continues to use traditional military approaches to circuit and systems design that are based on mechanical transport and modular interchangeability. These approaches lag well behind the capabilities of private industry and will prevent the Army from reducing energy consumption for soldiers who must communicate to survive on the battlefield. Design Approaches Use of application-specific integrated circuits (ASIC) and SoC design techniques are essential and could reduce power by more than an order of magnitude for digital computing and communications processing, making it negligible in comparison to analog sensor and display demand. There has been no effort in this direction in spite of the recommendations of the earlier study. Effort is also needed in reducing the power required by the analog portions of OFW electronics, particularly in communications devices. Incentives for Reducing Power The present focus on improving combat effectiveness may not result in net power reductions. While development, integration, and procurement contracts may contain goals for power, there are no financial (or other) incentives to make improvements beyond requirements. In fact, there is a disincentive to reduce power—namely, the potential for increasing risk and near-term cost. The Army should turn the potential for logistics savings into an incentive. For example, a design saving 1 W could result in a savings of 864 Wh per LW system if the soldier participates in one 72-hr mission a month. Assuming 200 Wh/kg batteries, this would eliminate the need for almost 10 pounds of batteries per year. At $35,000 per ton to deliver supplies to a combat area, the transportation savings is over $100 per system per year. Assuming a 10-year system life and 1,000 soldiers, such a design that saves 1 W is worth $1 million in transportation savings alone. It took an estimated $500 million to ship (not buy) the batteries used in the first Iraq conflict. By reducing average power demand 10 percent, a $50 million savings could be realized. This would easily cover the development cost of five chips at $10 million per chip or pay for several iterations of a custom ASIC design. If the cost of batteries and related power source hardware, such as chargers, envisioned for the OFW electronics suite is added to the calculation, the savings per soldier would be substantially greater than 10 percent. Perhaps most important, reductions in power demand may well reduce the complexity of energy sources needed and provide additional dollar savings. Therefore, a reasonable recommendation is for the Army to perform its own life-cycle cost analysis before deciding what it can and cannot afford in the way of development costs. The committee believes the savings revealed by such an analysis would easily justify paying contractor incentives and increasing development activity on energy-efficient design approaches to future Land Warrior systems.