Application Report

SPRA650 - March 2000

Evolving Cellular Handset Architectures but a

Continuing, Insatiable Desire for DSP MIPS
Michael L. McMahan Texas Instruments Incorporated

ABSTRACT

The year is 2005. Wireless technology has become so pervasive that kindergarten children
and family pets are tracked with GPS receivers; businessmen work in virtual offices with
high-speed wireless connections; shopping, stock trading, and ticket purchases are all done
routinely using wireless terminals that are worn—not carried. Utopia—or your worst
nightmare—whatever your perspective, this is a future which is just over the horizon. It follows
down a path which began with Hertz and Marconi experimenting with radio transmission
through the “ether” in the late 19th century and continues today with the explosion of wireless
products and services.

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Cellular Standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
3 Handset Architectures Today . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4 And in the Future . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

List of Figures
Figure 1. Evolution of Cellular Standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Figure 2. Multiaccess Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Figure 3. Generic 2G TDMA Baseband Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Figure 4. Mobile Data Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

List of Tables
Table 1. 2G Cellular Standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

This document was originally published in the Texas Instruments Technical Journal, January-March 2000, Volume 17,
Number 1.

1
SPRA650

1 Introduction
Twenty years ago, few would have predicted the revolution which resulted from the development
of the personal computer in the late 1970s. It is easy to forget that cellular telephony is an even
more recent phenomenon. The first commercial cellular system in the United States was turned
on in Chicago in 1983. As recently as 1988, George Calhoun, in his book, Digital Cellular Radio,
talked about the reasons for the failure of cellular technology in the marketplace. Today, a
decade later, cellular telephony is one of the largest, fastest growing markets in the high-tech
marketplace. In the last two years, the worldwide consumption of cellular handsets has tripled
from 48M units in 1996 to 153M units in 1998. Cellular penetration has experienced a
compounded growth rate of approximately 60%–70%. This paper gives a short tutorial
background on cellular standards, describing the current evolution of second to third generation
systems, and discusses how handset architectures, in general, and digital signal processor
(DSP) technology, in particular, must evolve to meet the needs of these future systems.

2 Cellular Standards
Figure 1 describes the evolution of cellular standards from the first generation (1G) of many
incompatible analog standards to a third generation (3G) that promises to be all digital, but
almost as confusing.

JTACS PDC ARIB

WCDMA

GPRS

Harmonized WCDMA
TACS- GSM ETSI FDD mode (WCDMA)
NMT WCDMA MC mode (CDMA2000)
TDD mode

EDGE

IS-54/IS-136 UWC-136

AMPS

IS-95 CDMA2000

IS-95B

1G 2G 2.5G 3G

Figure 1. Evolution of Cellular Standards

2 Evolving Cellular Handset Architectures but a Continuing, Insatiable Desire for DSP MIPS
 SPRA650

Analog cellular systems use a frequency division, multiple access scheme to support multiple
users within a finite geographical area (see Figure 2a). The AMPS standard divides the
allocated spectrum into 30-kHz channels—one user per channel. TACS, NMT, and JTACS used
a 25-kHz spacing. Within a single channel, all of these standards use frequency modulation
(FM) to represent the speech waveform. AMPS, for example, modulates the carrier up to
±12 kHz to represent the amplitude of the waveform.
The 1990s saw the emergence of the second generation (2G) of cellular systems based on
digital technology. Digital technology was enabled by the emergence of cost-effective,
high-performance DSPs—particularly the Texas Instruments (TI) TMS320C54x. Initially, the
transition was driven by cost. Because digital systems were able to multiplex more users in a
given block of frequencies, it was cheaper to support a digital user than an analog one.
Moreover, digital technology could support a wide range of services beyond voice telephony.
After all, bits are bits. It does not matter whether those bits represent voices or stock quotes.
This fact gave the service provider the ability to begin to differentiate his service from his
competitors. From the user perspective, the key advantage of digital standards is increased
battery life. Analog phones have to constantly monitor the channel to check for incoming phone
calls. This burns precious power. Digital handsets on the other hand, support the ability to enter
into a sleep mode, which consumes very little power, and then wake up periodically to check the
channel for incoming calls. The standby specification for analog phones is measured in hours. It
is not uncommon for modern digital cell phones to exhibit standby times of days or even weeks.
The second advantage of digital technology for the consumer is security. Analog cellular
telephony is notoriously easy to monitor or spoof. Digital technology is amenable to
incorporating the latest encryption technology and is consequently much more secure.
Second-generation cellular telephones are classified based on the technique which is used to
multiplex multiple users onto a single carrier frequency (see Figure 2b, Figure 2c, and Table 1).
GSM and IS-136 are both examples of time division multiple access (TDMA) systems. This
technology allows “n” users to share a single carrier frequency by giving them unrestricted
access during a fraction (n/T) of each frame (see Figure 2b). IS-136 uses a channel width of
30 kHz (compatible with AMPS), divides time into 40-ms frames, and shares that time among 3
users. GSM uses a 200-kHz channel spacing, divides time into 4.62-ms frames, and divides
each frame into 8 different user slots.
IS-95 is based on a spread spectrum technology call Code Division Multiple Access (CDMA).
This standard divides the spectrum into 1.25-MHz channels. The standard supports up to 64
users (though practically only about 20 can operate simultaneously) in each channel. Separation
among these users is accomplished by spreading the data signal of each user with a unique
orthogonal digital code (see Figure 2c). The receiver then correlates the received bit stream with
that digital code to extract a specific user’s data. Since the spreading codes are orthogonal,
users whose data is spread using a different code appear as white noise. Consequently, as
additional users are added, the signal-to-noise ratio (SNR) experienced by any individual
receiver will rise and the quality of a call (as measured by bit error rate (BER)) will gradually
degrade. This is different from TDMA systems, which have a hard limit on the number of users
who can simultaneously use a specific cell. When that number is exceeded, new calls are
blocked.

Evolving Cellular Handset Architectures but a Continuing, Insatiable Desire for DSP MIPS 3
SPRA650

(a) Frequency division multiple access

User #1

Frequency
User #2

User #n

Time
(b) Time division multiple access
Frequency User User User
User User User
User User User
User User User

Time
(c) Code division multiple access

User 1 * Code 1
Frequency

User n * Code n
User 2 * Code 2

Time

Figure 2. Multiaccess Technologies

Table 1. 2G Cellular Standards

GSM IS-136 IS-95 PDC

Frequency (MHz) GSM900 Cellular Cellular
Forward 935-960 869-894 869-894 940-956
Reverse 890-915 824-849 824-849 810-826

Forward GSM1800 1429-1453

Reverse 1805-1880 1477-1501
1710-1785
Forward GSM1900 PCS PCS
Reverse 1930-1990 1930-1990 1930-1990
1850-1910 1850-1910 1850-1910
Carrier Spacing (kHz) 200 30 1250 25
Access Method FDMA/TDMA FDMA/TDMA FDMA/CDMA FDMA/TDMA
Users/Carrier 8 3 ~20 3
Duplexing FDD FDD FDD FDD
Modulation GMSK π/4 DQPSK BPSK/ODQPSK π/4 DQPSK
Speech Data Rate 13 Kbps 7.95 Kbps 13 Kbps (variable) 6.7 Kbps

4 Evolving Cellular Handset Architectures but a Continuing, Insatiable Desire for DSP MIPS
 SPRA650

Today, at the end of the 20th century, the transition from analog to digital is largely complete.
The biggest trend today is from voice-centric to data-centric systems. Remember that one of the
key advantages of digital technology is its ability to support a wide variety of services. However,
1G and even 2G telephones are relatively poor data devices because of their limited bandwidth.
Most digital standards only support about 9.6–14.4 Kbps. While this may be sufficient for
services like GSM’s Short Message Service (SMS) or very basic data services like stock quotes,
sophisticated services involving internet access need more. That is what is driving the next
transition—to 2.5G and even 3G standards.
Each of the 2G standards has some kind of plan to support higher-bandwidth services. Most of
these beong to a class of systems which has been categorized as 2.5G. Most of these also
include a transition from circuit-switched to packet-based communications. The value of
packet-switching is that it gives the consumer an “always-on” service, which eliminates the long
call setup time that is built into any circuit-switched connection. Generalized Packet Radio
Services (GPRS) is probably the best and most advanced example. GPRS combines GSM slots
to provide as much as 115 Kbps in packet-switched, always-on bandwidth. Another 2.5G
standard is called Enhanced Data rates for GSM Evolution (EDGE). EDGE will use higher-order
modulation (either GMSK or 8 PSK) to support data rates of up to 384 Kbps—basically the same
rates as advertised for 3G pedestrian-class applications. Significant GPRS rollout should begin
in 2000. Deployment of EDGE is much less certain, but is probably still several years away.
Third-generation standards were developed specifically to support high-bandwidth data services.
The dominant 3G standard, Wideband CDMA (WCDMA), is based on the same class of
technology as IS-95. At a high level, the major difference is that IS-95 spreads the signal to 1.25
MHz with a chip rate of 1.2288 Mcps, whereas WCDMA spreads to 5 MHz with a chip rate of
3.84 Mcps. WCDMA will support data rates of 144 Kbps for high mobility applications, 384 Kbps
for more pedestrian-class mobility, and 2Mbps for fixed, office environments. The whole thrust
for both 2.5G and 3G cellular system is pointed at data applications. That fact is important to
keep in mind as we look at the implementation architectures that are used to deliver today’s 2G
handset and which will be required to support 2.5G and 3G in the future. 3G deployment will
begin in Japan in late 2001 with a significant ramp in 2002.

3 Handset Architectures Today

Figure 3 is a functional block diagram of a generic 2G voice-centric cellular phone. The
architecture which implements this system is based on two processors (generally integrated
onto a single integrated circuit).
The heart of all digital cellular phones is a DSP. The dominant DSP in this market is the Texas
Instruments TMS320C54x, which can be found in about 65% of modern cell phones. This
processor is responsible for modulating and demodulating the data stream, coding and decoding
to maintain the robustness of the transmission in the face of transmission bit errors, encrypting
and decrypting for security, and compressing and decompressing the speech signal. In early 2G
TDMA phones these functions could be accomplished with 30–50 DSP MIPS. As vocoders
have become more complex and as data rates have risen in 2.5G phones, the total DSP load
has risen past 100 MIPS. The CDMA standard requires a somewhat different functional
partitioning because of the data rates generated by spreading. While the DSP can still be used
to process at the basic data rate (functions like forward error correction, encryption, or voice
compression), ASIC hardware operating under the control of the DSP must be used to process
and modulate/demodulate the spread signal.

Evolving Cellular Handset Architectures but a Continuing, Insatiable Desire for DSP MIPS 5
SPRA650

Antenna

Analog baseband

RF interface Receive
Audio
interface
I/Q tx & rx
ADC & DAC
microphone DSP filters Synthesizer Power
ADC speech code/ aux amp
speaker DAC decode ADCs/DACs
error correction synth ctrl
channel code/
decode Modulator
ASIC backplane

equalization
demodulation
encryption
Radio subsystem

User display
Micorcontroller
level 2/3 protocol Battery
Flash radio resource Keyboard interface &
memory mgmt. power control
short message
service SIM card
man & machine
interface
O/S software

Digital baseband

Figure 3. Generic 2G TDMA Baseband Architecture

The second processor that is found in a 2G phone is a general-purpose microprocessor which

supports the user interface of the handset and which handles the upper layers of the
communications protocol stack. In most modern 2G phones, this processor is implemented with
a 32-bit RISC engine (generally operating in a 16-bit mode to conserve memory). For a
voice-centric handset, the functions implemented by this RISC processor require less than
10 MIPS. Even for many data-intensive 2.5G phones, the MIPS requirement of this processor
will be minimal—probably less than 40 MIPS.

4 And in the Future

Future handsets will change along multiple dimensions. First, radio architectures are evolving in
directions that could dramatically change the partitioning illustrated in Figure 3. The goal is to
support more flexibility in the radio frontend and to reduce the number of components required.
In many cases, this means that some of the functionality that is implemented with analog
components today will shift to the digital signal processing domain. For example, current
narrowband analog receivers could be replaced with wideband analog frontends that use
programmable digital techniques to filter and select channels. Other examples include using
adaptive digital techniques to linearize transmit power amplifiers or to eliminate the DC offset in
direct conversion receivers.

6 Evolving Cellular Handset Architectures but a Continuing, Insatiable Desire for DSP MIPS
 SPRA650

On a different dimension, as data applications become more important, the nature of a mobile
handset will change. Today, a handset is the perfect example of a closed, static, embedded
system. Its function is defined when the phone is manufactured. Although it may support a few
simple applications beyond voice communications (like SMS or even limited information
services), the system exists to implement a voice communications channel—a fixed, real-time
DSP-intensive task. In the future, it seems likely that the market for mobile handsets will
fragment. We will probably always have a segment of the market for the traditional, voice-centric
mobile phone. However, we will also have hybrid devices where the modem communications
function is a necessary component, but one or more data applications may well be the reason
the device is purchased. In some cases, these applications may be fixed, but given the advent of
the Internet, it seems more likely that applications will be downloadable and will change at the
whim of the consumer. Today, I might choose to download an MP3 player. Tomorrow, I might
want my phone to support a video conference or act as a GPS navigator. In a sense, the
handset has taken on a very PC-like character. It has become an applications platform (see
Figure 4). What are the implications for handset architectures?

Image mail

Navigation
Personal
information Multimedia
management communications

Personal
Voice video
communications Personal
audio

Figure 4. Mobile Data Applications

As long as the processing load of a cellular handset was fixed, it was relatively easy to size and
configure the processing resources required to implement the communications modem. In this
new environment, the requirement for a cellular modem is unchanged (though the data rates
may be substantially higher). However, the applications requirements are potentially unbounded.
The question is how to best add processing resources to handle these additional tasks. The
answer is complicated by the fact that applications—like the modem—will have both control and
DSP requirements. As in most complicated engineering problems, there is no single best
answer, but the major issues are clear:
1. How do you supply the processing resources necessary to handle an open-ended,
unspecified, and potentially unbounded load ?
Because the platform will need to support application flexibility, its basic architecture will be
based on one or more programmable processors. This was true in 2G phones because of

Evolving Cellular Handset Architectures but a Continuing, Insatiable Desire for DSP MIPS 7
SPRA650

ever-changing communications standards and time-to-market issues. This will continue to be

true in the future, but the nature of these processing elements and the software they execute
will change.
Second-generation handsets were very DSP-centric. That fact will remain true in the future.
However, because of the need to support sophisticated applications, 2.5G and 3G handsets
will require a more balanced set of processing resources. Either the current RISC engine
must be enhanced or a separate RISC platform must be added to the architecture. This
processor will need to operate at >100 MIPS and will need to incorporate features to allow it
to support much more significant operating systems (OSs) than we have seen in traditional
cell phones. Applications-centric OSs such as EPOC from Symbian, Windows CE from
Microsoft, or the PalmOS from 3COM will become the order of the day. The software
environment of the DSP will also have to change in order to support downloadable
applications (which is likely considering the spectrum of applications shown in Figure 4).
Since the DSP resource will have to support the dramatically increased communications load
imposed by high-bandwidth 3G standards (>200 MIPS) and also satisfy the demands of a
range of applications-specific DSP functions (speech recognition, image/video coding and
decoding, etc.) which themselves will drive a significant MIPS load, it is possible that these
future handsets will require more than one DSP.
2. How do you provide substantially increased processing resources without decreasing
battery life?
The consumer is quite demanding. Now that he has experienced operating battery life in
excess of 4 hours and standby times measured in days or weeks, he will never be willing to
accept anything less. The challenge will be to supply an order-of-magnitude increase in
processing power without no decrease in battery life. Part of this challenge will be met with
advances in silicon technology, but part will have to be achieved with architectural
innovation. Examples could include:
a. More care in matching processing resources to the type of algorithm required. This
implies a mix of general-purpose RISC and DSP resources. Although many simple DSP
tasks can be accomplished using a RISC instruction set, it is often more power efficient
to use a DSP.
b. Innovative use of attached, programmable coprocessors. These can often supply more
power-efficient execution than a fully programmable DSP, yet retain much of the
flexibility.
c. More extensive use of multiprocessor architectures to minimize clock rates required to
accomplish task. This would have the further benefit of allowing unused functions to be
turned off to minimize standby power consumption. Such architectures will also drive a
need for new operating system and software technology.
3. How do you integrate the requirements of hard real-time communications tasks with the
requirements of non- or soft-real-time applications ?
The answer may well be that you do not. In order to maximize system reliability and
minimize the software development complexity of any platform, it may be necessary to
segregate real-time and non-real-time tasks. None of the OS environments mentioned above

Windows and Microsoft are registered trademarks of Microsoft Corporation.

PalmOS is a registered trademark of Palm Computing, Inc.
3COM is a registered trademark of #COM Corporation.
Other trademarks are the property of their respective owners.

8 Evolving Cellular Handset Architectures but a Continuing, Insatiable Desire for DSP MIPS
 SPRA650

are particularly well-suited for supporting the predictable, deterministic requirements of

high-speed wireless modems. Although all of these software platforms are moving to
incorporate real-time features, it will take time for designers to become comfortable with their
ability to achieve the reliability and performance which these systems require.

5 Conclusions
Future data-centric, 2.5G and 3G mobile handsets will require a new approach to the
architectures which implement those devices. Care must be taken to consider application and
user interface performance, system-level power consumption, hardware costs, software
complexity, and time to market. The winning system will likely incorporate multiple processors
tuned to fit the tasks for which they were designed. You will see high-end RISCs, which
implement protocol stacks, user interfaces, high-end OSs, and downloadable applications. You
will also see programmable DSPs and coprocessors that can provide power-efficient media
processing and support needed application flexibility and upgradeability. The biggest open
question is the extent to which the applications and communications functions of these devices
can be effectively combined into one programming environment. In the near future, it is likely
that they will be separate.

6 References
1. U. Amin and S. Chennakeshu, “The Evolution of TDMA to 3G,” guest editorial, IEEE Personal
Communications, Vol. 6, No. 3, Jun. 1999, pp. 6–7.
2. P. Bahl, “Supporting Digital Video in a Managed Wireless Network,” IEEE Communications, Vol.
36, No. 6, Jun 1998, pp. 94–102.
3. G. Calhoun, Digital Cellular Radio, Artech House, 1988, pp. 1–54.
4. A. Gatherer, T. Stetzler, M. McMahan, and E. Auslander, “DSP-Based Architectures for Mobile
Communications: Past, Present, and Future,” to be published in IEEE Communications, Jan.
2000.
5. U. Ko, M. McMahan, and E. Auslander, “DSP for the Third Generation Wireless
Communications,” to be published in ICCD ’99.
6. Z. Kostic and S. Seetharaman, “Digital Signal Processors in Cellular Radio Communications,”
IEEE Communications, Vol. 35, No. 12, Dec. 1997, pp. 22–43.
7. M. McMahan, A. Khatzibadeh, and Pradeep Shah, Wireless Systems and Technology Overview,
internal Texas Instruments white paper, 8/95.
8. M. Oliphant, “The Mobile Phone Meets the Internet,” IEEE Spectrum, Aug 1999, p. 20-28.
9. N. Sollenberger, N. Seshadri, and R. Cox, “The Evolution of IS-136 TDMA for Third-Generation
Wireless Services,” IEEE Personal Communications, Vol. 6, No. 3, Jun. 1999, pp. 8–18.

Evolving Cellular Handset Architectures but a Continuing, Insatiable Desire for DSP MIPS 9
 IMPORTANT NOTICE

Texas Instruments and its subsidiaries (TI) reserve the right to make changes to their products or to discontinue
any product or service without notice, and advise customers to obtain the latest version of relevant information
to verify, before placing orders, that information being relied on is current and complete. All products are sold
subject to the terms and conditions of sale supplied at the time of order acknowledgment, including those
pertaining to warranty, patent infringement, and limitation of liability.

TI warrants performance of its semiconductor products to the specifications applicable at the time of sale in
accordance with TI’s standard warranty. Testing and other quality control techniques are utilized to the extent
TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily
performed, except those mandated by government requirements.

Customers are responsible for their applications using TI components.

In order to minimize risks associated with the customer’s applications, adequate design and operating
safeguards must be provided by the customer to minimize inherent or procedural hazards.

TI assumes no liability for applications assistance or customer product design. TI does not warrant or represent
that any license, either express or implied, is granted under any patent right, copyright, mask work right, or other
intellectual property right of TI covering or relating to any combination, machine, or process in which such
semiconductor products or services might be or are used. TI’s publication of information regarding any third
party’s products or services does not constitute TI’s approval, warranty or endorsement thereof.

Copyright  2000, Texas Instruments Incorporated