Software Radio

New SDR architecture enables ubiquitous data connectivity

While WiFi in the laptop has enabled unheralded mobility for many, WiFi hotspot coverage is not ubiquitous, limiting where laptops can be connected. Other emerging global standards for high-speed data, like WiMAX, WiBro, HSDPA and EVDO, presently provide islands of coverage. What is required is a laptop wireless modem that can seamlessly interoperate amongst these various standards. Although, software-dened radio (SDR) is the right technology to stitch together these services for ubiquitous global laptop connectivity, the high-power ADCs found in common SDR architectures make it uneconomical. A new approach to SDR is required, which is described in this article. By John A. Kilpatrick; Russell J. Cyr; Erik L. Org and Geoffrey Dawe
ervasive wireless connectivity has clearly catalyzed change in todays high technology businesses as well as in personal lives. When the rst mobile phones were being developed back in the 1980s, few foresaw the impact that mobile communications would have on our daily lives. The growth in all types of mobile connectivity has exceeded expectations. On March 6, 1983 when Motorola introduced the worlds rst commercial portable cellular phone, the DynaTAC 8000X, none of us would have thought that by 2005 more than 80% of the population in Europe and in the United States would subscribe to a cellular service provider; more than 250 billion SMS messages would be sent in 2004; sales of mobile computing devices (laptops) would exceed those of desktops and that almost each laptop sold would have an 802.11 b/g adapter built in. Although many protocols for wireless data have been proposed and implemented, WiFi (802.11) stands out. It has managed to cross a tipping point similar to the way cellular services became ubiquitous in the 1990s. The nationwide availability of hotspots has reached a point where the network effect has begun to drive WiFis acceptance and inclusion in a range of mobile devices. Although hotspots are not yet as ubiquitous as many road warriors would like, it has become accessible enough that most users are condent that they can locate a hot spot quickly and easily. In spite of its marketing success, the business model for WiFi

servicesoffering speeds up to 54 Mbps within several hundred feet of an access pointhas been difcult. Potential users have been turned off by roaming and coverage problems. Prices have also seemed high given the inconsistent coverage. However, when considering how to best implement mobile data services, it would be incorrect to focus on just WiFi. What users want is seamless mobility. We realize that a single standard is incapable of providing global mobile data services. Engineering trade offs must be made to balance battery life, transmit power, sensitivity and throughput. Thus, in reality, WiFi is only one part of a successful global mobile data solution because of the inevitable need for true mobility provided by multiband multistandard devices.

Multiple data standards

When users consider which mobile data standard to choose, they intuitively trade off cost with range and throughput. Efforts to improve the WiFi standards have focused on just that. However, the wireless world doesnt only exist within the local area network (LAN), it can be compressed and stretched to match the geographic scale of business and personal lives. Users require laptop data connectivity in wide area networks (WANs) as well as in metropolitan area networks (MANs),
Industrial Measurement Voiceband, Audio Data Acquisition High Speed: Instrumentation, Video, IF Sampling, Software Radio, Etc.

Segmenting the wireless world

Resolution (Bits)

24 22 20 18 16 14 12 10 8 10 100

30 feet

PAN

300 feet

LAN

3000 feet

MAN

30000 feet+

WAN

SAR

PAN IR, Bluetooth, UWB LAN UWB, WiFi, WiMax MAN WiMax, WiBro WAN 1xRTT, EVDO, EDGE, HSDPA

Current State of the Art (Approximate) 1k 10 k 100 k 1M

PIPELINE

10 M 100 M

1G

Sampling Rate (Hz)

Figure 1. Segmenting the wireless world. Figure 2. ADC architecture, applications, resolution and sampling rates.

32

www.rfdesign.com

January 2006

Technology
Industry Body Commercial Rollout Cell Range Average Throughput Spectrum Allocation

Bluetooth
IEEE

WiFi
IEEE

MediaFLO
Qualcomm

EV-DO
3GPP2

TD-CDMA
3GPP

WCDMA
3GPP

HSDPA
3GPP

HSUPA
3GPP

HSOPA
Nortel

Flash OFDM
Flarion

802.16d
IEEE/WiMAX Forum 2005 2-5 km 300500 Kbps 2500-2690, 2400-3600, 5725-5850 OFDM architecture, Deals with interference issues

802.16e
IEEE/ WiMAX Forum 2007/2008 Rollout 2-5 km 1-2 Mbps 2500-2690, 2400-3600, 5725-5850 Non-line of sight technology, Mobility

2001 10m 400Kbps 2.4-2.483

2000 100m 2Mbps 2.4-2.4835 (U.S.) 2.4712.497 (Japan) Low cost rollout

2007 25 km 6 Mpbs 700

2002 3-5 km 300-500 Kbps 450, 800, 1900, 2100

2004/2005 3-5 km 600-700 Kbps 1900-1920, 2010-2025

2003/2004 3-5 km 150-200 Kbps 19001920, 2010-2025 Increase in voice capacity and improved data throughput Bandwidth capacity for true broadband, Upgrades needed Ericsson, Lucent, Nokia, Siemens, Nortel

2005/2006 3-5 km 900 Kbps1 Mbps 1920-1980, 2110-2170

2007/2008 3-5 km 900 Kbps1 Mbps 19201980, 2110-210 Improves HSDPA uplink speed

2007 3-5 km 900 Kbps1 Mbps 1920-1980, 2110-2170

2005/2006 3-5 km 900 Kbps1 Mbps 450, 900, 1800, 2100

Strengths

Low-cost rollout, low power consumption

Low cost rollout, dedicated content channel

Competitive with HSDPA, High data throughput

Ability to allocate resources for network demands

Software upgrade to WCDMA for higher bandwidth

Integration of OFDM framework

Increased QoS, Deals with interference issues

Weaknesses

Security

High power consumption

Proprietary, lack of ecosystem,

Uplink capacity, further upgrades needed for true broadband Lucent, Nortel, Motorola Samsung

Vendor support, Interference issues

Uplink constrained

Question of further evolution path

Limited vendor support

Proprietary vs. lack of ecosystem

Lack of mobility

Timing uncertainty, Standard still in development

Key equipment suppliers

CSR

D-Link. Linksys

Qualcomm

IP Wireless, UTStarcom, Alcatel

Ericsson, Nokia, Siemens, Nortel

Ericsson, Nokia, Siemens, Nortel

Nortel

Flarion, Siemens

NA

NA

Table 1. A complex array of wireless data standards.

LANs and personal area networks (PANs) as Softransceiver illustrated in Figure 1. Software Controlled Transceiver Many new emerging data protocols are on the horizon. WiMAX, which can transfer BB FEM 70 Mbps over a distance of up to 30 miles to Configurable Digital LNA ADC Receiver Processing thousands of users from one base station, is undergoing numerous trials around the world, in countries such as the United Kingdom, the Philippines, Indonesia, Malaysia and in the Freq Synth. United States. Software Transceiver CONTROL Controller CONTROL In todays mobile world, it has become commonly accepted that mobile communication solutions have been the foundation for the productivity increases of the last decade. An Digital overly simple approach to ubiquitous coverage Digital Configurable Processing Driver DAC Processing Transmitter is to choose one standard as a global standard; AMP attempt global regulation; then rollout a dense BitWave infrastructure that provides seamless connec Single component chain tivity. This is seen as an idealistic vision. Most RF/IF Tunable Performance users intuitively accept that competition in Variable Component Layout Mixed Signal heterogeneous mobile data markets will create Digital Control Digital DigRF Digital I/O a heterogeneous set of mobile data standards. Dynamically Reconfigurable Components And naturally, users and developers have asked themselves whether it would be possible to build a mobile data modem that could communicate using that set of heterogeneous mobile data technologies such as EVDO, TD-CDMA, HSDPA, OFDM, WiMAX, and Figure 3. BitWaves Softransceiver functional architecture. WiBro (Table 1[2]). data solution; the next step in delivering true mobility is the successMany attempts have been made to address this problem and commer- ful development of a software-dened transceiver suitable for mobile cialize congurable mobile radios. Vanu Inc. has recently received FCC applications. approval for its Anywave base station. This is the rst FCC-approved software radio on the market. Although a successful SDR application, Current approaches it is not a mobile radio. Other companies such as Icera, Sandbridge, The practical approach that todays transceiver vendors have Intel and Texas Instruments are developing software-dened radio defaulted to is providing multiple analog transceivers. Theyve built (SDR) or SDR-like modem solutions for mobile radio. multichip modules, multidie packages, as well as multiple transceivers But those software-dened modems are only one part of a mobile on die. Unfortunately, these approaches all come with limitations. They
Programmable Antenna, PA, Band Select & Duplexer

34

www.rfdesign.com

January 2006

Multi-Standard Modem

DigRF Interface + Extensions

Analog Low IF Architecture with optional ZIF Output

Analog Digital

An intermediate step to a Softransceiver

Analog Digital

Zero IF (bypasses IF)

LNA

Zero IF
ADC

RF

BPF

PGA

BPF

PGA

Complex Digital Filter

LNA

RF

BPF

PGA

BPF

PGA

ADC

LO1

LO2

Low IF (uses IF stage)

LO1

LO2

Digital Down Converter + Complex Digital Filter

Figure 4. An analog low IF (LIF) architecture with optional ZIF output.

all require more die area for each additional band. Each additional transceiver draws additional power. Also, each additional transceiver may require an additional antenna and matching network. Designers are thus forced to make trade offs for cost, size, performance and power, yet the resulting designs are still inexible, large, expensive and power hungry. On the other extreme, an SDR takes a completely different approach to multiband yet still runs the same problems that the multitransceiver/multidie/multichip module approaches all do. The classical SDR architecture requires a high sampling rate, wide bandwidth and power-hungry analog-to-digital converter (ADC), as well as a highperformance, high-power digital drop receiver (DDR). For mobile applications, this creates unacceptable demands on the battery and so far has precluded its use in mobile devices. The trick is to come up with a transceiver architecture that is exible enough to support a variety of signal bandwidths, modulation formats, signal levels and blocking specications. As an example, cellular standards have low to medium bandwidths, but have very high dynamic range requirements and challenging blocker environments. WLAN standards, on the other hand, tend to operate in unlicensed spectrum and thus have lower power levels, less dynamic range, fewer blocker considerations, but have high signal bandwidths and high-order modulation (and thus require higher signal-to-noise ratio or SNR). So one can either design multiple low noise ampliers (LNAs), multiple baseband lters and several ADCs (which doesnt seem like an improvement over the multichip or multidie approach), or one can design the circuits to be congurable. The LNAs must be tunable over a wide bandwidth and support high linearity; the bandwidth, dynamic range and order of the baseband lters must be congurable; the ADC must adapt its power consumption to the dynamic range and bandwidth required. Finally, low-power digital processing is necessary to achieve the necessary decimation, downconversion, channel selectivity and gain/phase compensation. On the transmitter side, one requires an architecture that supports narrow and wideband signals, constant and non-constant envelope modulation schemes, various output powers and covers a wide frequency tuning range. While polar modulators have proven efcient for narrowband signals, they have yet to be successful for wideband signals such as WiFi. Direct IQ modulators have signicant exibility as long as the digital-to-analog converter (DAC) sampling rates and lter bandwidth can be congurable. Again, highly congurable digital processing can be used for upsampling, digital upconversion, and even pulse shaping and modulation. Since some data standards are full duplex while the WiFi and WiMAX protocols are half-duplex, the multiprotocol transceiver faces additional challenges. Whereas half-duplex designs typically share the synthesizer between transmit and receive, the full-duplex architecture requires that two local oscillators (LOs) be generated. To reduce power in the half-duplex mode of operation, a single LO can be used to drive the transmit and the receive mixers. However, this must be done in such a way that avoids leakage (from the transmit mixer LO to the receive mixer) when in full-duplex operation. Based on the challenges, its clear a parallel hardware architecture is

The programmable Local Oscillator (LO) enables both LIF and ZIF functionality fLO (ZIF) = f (RF) fLO (LIF) = f (RF) f Remove a large analog IF block and implement a digital down converter (DDC) in low cost digital CMOS
Figure 5. Intermediate step to the Softransceiver.

BitWaves Softransceiver Receiver Chain

Analog Digital Low IF
LNA

RF

BPF

PGA

LPF

PGA

ADC

Digital Down Converter Complex Digital Filter

Considerations in designing a multiband solution

LO1

ZIF

Switch between ZIF and Low IF by adding/removng DDC using low cost digital circuit Programming new LO frequency with low cost digital control
Figure 6. Softransceiver receiver chain.

not the economical solution. To enable this Holy Grail of true mobile multiband multistandard devices, an SDR-like solution is required. Only SDR can offer the adaptability, the low cost (programmability) and future proong needed to convince operators, laptop designers and customers to accept SDR as a necessary part of the mobile world. A traditional SDR solution places the ADC as close to the antenna as possible. For low-cost, low-power consumer devices, this has to be after the downconversion, either at baseband or at an intermediate frequency (IF). Once the entire signal band has been digitized with sufcient dynamic range to capture the maximum blocker levels as well as the minimum desired signals, the signal-processing algorithms can quite easily extract the signal of interest. One example was the Steinbrecher MiniCell base transceiver station, which digitized the 12.5 MHz of cellular bandwidth and extracted 30 kHz AMPS or TDMA channels using DSPs. A multistandard SDR would require that the ADC dynamic range be variable (or designed for the worst case) and that the signal processing be programmable. This precludes the use of application-specic ICs (ASICs), and requires the use of eld-programmable gate arrays (FPGAs) or DSP technology, both of which are expensive for consumer devices. For a 10 MHz band of interest and a 100 dB of dynamic range (-110 dBm to -30 dBm signals with 20 dB allocated for headroom, modulation crest factor and quantization noise), the ADC has to sample at greater than 20 Msamples/s with 17 effective number of bits (ENOB). Alternatively, one could also use an oversampling sigma-delta architecture with a 320 Msamples/s clock rate and reduce the ENOB to a more manageable 10 bits to12 bits (depending on the order of noise shaping). Both of these alternatives are about on the edge of current technology.[1] (Figure 2.) As an example, an advanced ADC from TelASIC,

Traditional SDR

36

www.rfdesign.com

January 2006

LNA Example

the TC1411, claims to support 14 bits, 250 Msamples/s over a 75 MHz bandwidth. While this is sufcient for many SDR applications, it requires 1.9 W, which is too power hungry for mobile applications. As a consequence, the optimum transceiver design must include the channelization and AGC functions before the ADC for the foreseeable future.

increasing performance. (Figure 3). The Softransceiver can effectively move V1 (FC, Q) its operating characteristics in real time by software commands. It can shift the center Z1 (FC, Q) V2 O/P frequency, modify the bandwidth and samZ2 pling rate, and change the linearity and noise I/P gure of a transceiver channel in real time. Thus, one programmable transceiver can V3 (W) replace the many xed transceivers now found in a typical cell phone or data modem. This A new approach recongurable transceiver technology also BitWaves long experience with SDR has V4 (F. Q) dramatically reduces the size and power of enabled it to develop a new solution to this Z3 the transceiver chip and increases exibility complex problem. To convince the wireless when compared with a multitransceiver on industry to accept and adopt a new implemenV5 (I) die solution. The Softransceiver stands apart tation of SDR, one must rst step back and from other multiband solutions in that it is consider what the market requires. Its apparent competitive in performance even against single that a low power solution is required. This led the decision to base all components on existing standard transceivers. analog architectures. A software-dened trans- Figure 7. Component diagram of a dynamically One of the ways in which the Softransceiver compresses the required area is through a ceiver must also be able to optimize itself for recongurable low-noise amplier (LNA). each unique protocol; i.e., be easily controlled programmable LO. Each standard that the with digital technology by the baseband through a simplied interface. Softransceiver supports has an optimal architecture for implementation. Finally, the dynamic optimization required for each potential operating For example, most GSM standards can be more easily implemented in a environment of the device led to innovation for individually controlling low IF (LIF) architecture while other standards such as WiFi or WiMAX each RF, analog and mixed-signal blocks within the transceiver. may be more easily implemented with a zero IF (ZIF) architecture. The Hence, a Softransceiver technology was created with the preced- Softransceiver allows the selected standard to be demodulated with ing ideas in mind: it provides a way to increase exibility while whichever architecture works best. at the same time reducing cost, decreasing power consumption and In an analog architecture such as that shown in Figure 4, implementing both LIF and ZIF is possible, however, it requires a lot more analog die area since the designer needs to provide a second analog stage. In Figure 5, the combination of a widely programmable LO and the inclusion of a digital downconverter allows for the complete elimination of the second analog stage. In this manner, the Softransceiver achieves the desired combination of low

LNA

ABOUT THE AUTHORS

John A. Kilpatrick is director of systems engineering at BitWave Semiconductor Inc., Lowell, Mass. He has more than 30 years of experience in wireless communications. Kilpatrick received his BA in Applied Mathematics from Gordon College and his MSSE from the University of Lowell. Russell J. Cyr is CMO and co-founder of BitWave Semiconductor. A 20-year veteran of the wireless communications industry, Cyr has published numerous articles and has been granted several patents in radio technology. He received his BSEE and MSEE from the University of Massachusetts and his MBA from Rivier College. Erik L.Org is marketing manager for BitWave Semiconductor Inc. He has more than 10 years of experience with wireless markets and technology. He earned an MBA from Columbia Business School, as well as a BS in Electrical Power Engineering from Rensselaer Polytechnic Institute. Geoffrey Dawe is CTO and co-founder of BitWave Semiconductor. Dawe is a 21-year veteran of the RFIC industry. He is an author/co-author of more than 35 publications in the eld of RFIC technology. He has one issued patent and several others pending. Dawe received a BSEE from Norwich University.

38

www.rfdesign.com

January 2006