While 3G hasn't quite arrived, designers are already thinking about 4G technology. With it comes challenging RF and baseband design headaches.
Cellular service providers are slowly beginning to deploy third-generation (3G) cellular services. As access technology increases, voice, video, multimedia, and broadband dataservices are becoming integrated into the same network. The hope once envisioned for 3G as a true broadband service has all but dwindled away. It is apparent that 3G systems, while
maintaining the possible 2-Mbps data rate in the standard, will realistically achieve 384-kbps rates. To achieve the goals of true broadband cellular service, the systems have to
make the leap to a fourth-generation (4G) network.
This is not merely a numbers game. 4G is intended to provide high speed, high capacity, low cost per bit, IP based services.
The goal is to have data rates up to 20 Mbps, even when used in such scenarios as a vehicle travelling 200 kilometers per hour. New design techniques, however, are needed to make
this happen, in terms of achieving 4G performance at a desired target of one-tenth the cost of 3G.
The move to 4G is complicated by attempts to standardize on a single 3G protocol. Without a single standard on which to build, designers face significant additional challenges. Table 1 compares some of the key parameters of 3G and 4G (4G does not have any solid specification as of yet, so the parameters rely on general proposals). It is clear that some
standardization is in order.
| TABLE 1: Key Parameters of 3g and 4 G Systems |
|
|
3G |
4G |
| Frequency band |
1.8 - 2.5 GHz |
2 - 8 GHz |
Bandwidth |
5 - 20 MHz |
5 - 20 MHz |
| Data rate |
Up to 2 Mbps (384 kbps deployed) |
Up to 20 Mbps |
| Access |
W-CDMA |
MC-CDMA or OFDM (TDMA) |
| Forward error correction |
Convolutional rate 1/2, 1/3 |
Concatenated coding scheme |
| Switching |
Circuit/packet |
Packet |
| Mobile top speeds |
200 km/h |
200 km/h |
Multicarrier modulation
To achieve a 4G standard, a new approach is needed to avoid the divisiveness we've seen in the 3G realm. One promising underlying technology to accomplish this is multicarrier
modulation (MCM), a derivative of frequency-division multiplexing. MCM is not a new technology; forms of multicarrier systems are currently used in DSL modems, and digital
audio/video broadcast (DAB/DVB). MCM is a baseband process that uses parallel equal bandwidth subchannels to transmit information. Normally implemented with Fast Fourier
transform (FFT) techniques, MCM's advantages include better performance in the intersymbol interference (ISI) environment, and avoidance of single-frequency interferers.
However, MCM increases the peak-to-average ratio (PAVR) of the signal, and to overcome ISI a cyclic extension or guard band must be added to the data.
Equation 1, describes peak to average adjustment - the difference of the PAVR between MCM and a single carrier system is a function of the number of subcarriers (N) as:
(1)
Any increase in PAVR requires an increase in the linearity of the system to reduce distortion. Proposed approaches to reduce PAVR have consequences, however. One such technique
is clipping the signal; this results in more non-linearity. Linearization techniques can be used, but they increase the cost of the system, and amplifier backoff may still be required.
Cyclic extension works as follows: If N is the original length of a block, and the channel's response is of length M, the cyclically extended symbol has a new length of N + M - 1. The
image presented by this sequence, to the convolution with the channel, looks as if it was convolved with a periodic sequence consisting of a repetition of the original block of N.
Therefore, the new symbol of length N + M - 1 sampling periods has no ISI. The cost is an increase in energy and uncoded bits added to the data. At the MCM receiver, only N samples
are processed, and M - 1 samples are discarded, resulting in a loss in signal-to-noise ratio (SNR) as shown in Equation 2.
(2)
Two different types of MCM are likely candidates for 4G as listed in Table 1. These include multicarrier code division multiple access (MC-CDMA) and orthogonal frequency division
multiplexing (OFDM) using time division multiple access (TDMA). Note: MC-CDMA is actually OFDM with a CDMA overlay.
Similar to single-carrier CDMA systems, the users are multiplexed with orthogonal codes to distinguish users in MC-CDMA. However, in MC-CDMA, each user can be allocated
several codes, where the data is spread in time or frequency. Either way, multiple users access the system simultaneously.
In OFDM with TDMA, the users are allocated time intervals to transmit and receive data. As with 3G systems, 4G systems have to deal with issues of multiple access interference and
timing.
Differences between OFDM with TDMA and MC-CDMA can also be seen in the types of modulation used in each subcarrier. Typically, MC-CDMA uses quadrature phase-shift keying
(QPSK), while OFDM with TDMA could use more high-level modulations (HLM), such as, multilevel quadrature amplitude modulation (M-QAM) (where M = 4 to 256). How-ever,
to optimize overall system performance, adaptive modulation can be used; where the level of QAM for all subcarriers is chosen based on measured parameters.
Let's consider this at the component level. The structure of a 4G transceiver is similar to any other wideband wireless transceiver. Variances from a typical transceiver are mainly
in the baseband processing. A multicarrier modulated signal appears to the RF/IF section of the transceiver as a broadband high PAVR signal. Base stations and mobiles are
distinguished in that base stations transmit and receive/ decode more than one mobile, while a mobile is for a single user. A mobile may be a cell phone, a computer, or other
personal communication device.
The line between RF and baseband will be closer for a 4G system. Data will be converted from analog to digital or vice versa at high data rates to increase the flexibility of the
system. Also, typical RF components such as power amplifiers and antennas will require sophisticated signal processing techniques to create the capabilities needed for broadband
high data rate signals.
Figure 1 shows a typical RF/IF section for a transceiver. In the transmit path inphase and quadrature (I&Q) signals are upconverted to an IF, and then converted to RF and amplified for transmission. In the receive path the data is taken from the antenna at RF, filtered, amplified, and downconverted for baseband processing. The transceiver provides power
control, timing and synchronization, and frequency information. When multicarrier modulation is used, frequency information is crucial. If the data is not synchronized properly
the transceiver will not be able to decode it.
From a high level, the structure of the RF/IF portions of the mobile and base station are similar, however, there are significant differences in their architectures and performance
requirements. Key drivers for both are performance and cost; mobiles also need to consider power consumption and size.
4G processing
Figure 2 shows a high-level block diagram of the transceiver baseband processing section. Given that 4G is based on a multicarrier technique, key baseband components for the transmitter and receiver are the FFT and its inverse (IFFT). In the transmit path the data is generated, coded, modulated, transformed, cyclically extended, and then passed to the
RF/IF section. In the receive path the cyclic extension is removed, the data is transformed, detected, and decoded. If the data is voice, it goes to a vocoder. The baseband subsystem
will be implemented with a number of ICs, including digital signal processors (DSPs), microcontrollers, and ASICs. Software, an important part of the transceiver, implements the
different algorithms, coding, and overall state machine of the transceiver. The base station could have numerous DSPs. For example, if smart antennas are used, each user needs
access to a DSP to perform the needed adjustments to the antenna beam.
Receiver section
4G will require an improved receiver section, compared to 3G, to achieve the desired performance in data rates and reliability of communication. As shown in Equation 3, Shannon's
Theorem specifies the minimum required SNR for reliable communication:
(3)
where C is the channel capacity (which is the data rate), and BW is the bandwidth.
For 3G, using the 2-Mbps data rate in a 5-MHz bandwidth, the SNR is only 1.2 dB. In 4G, approximately 12-dB SNR is required for a 20-Mbps data rate in a 5-MHz bandwidth.
This shows that for the increased data rates of 4G, the transceiver system must perform significantly better than 3G.
With any receiver, the main issues for efficiency and sensitivity are noise figure, gain, group delay, bandwidth, sensitivity, spurious rejection, and power consumption. 4G is no
exception; the sensitivity can be determined as shown in Equation 4 :
(4)
where KTo is the thermal noise (for this equation it is -174 dBm), BW is the receiver bandwidth, NF is the receiver noise figure, and SNRavgMCM is the average SNR for a MCM
system needed for an expected bit error rate.
For a 4G receiver using a 5-MHz RF bandwidth, 16 QAM modulation and NF of 3 dB, the receiver sensitivity is -87 dBm. For 3G, the receiver sensitivity needs to be -122 dBm; the
difference is due to the modulation and PAVR. This illustrates the need to reduce PAVR by clipping or coding. Also the gain is required to be linear, and the group delay must be flat
over the bandwidth of the signal.
The receiver front end provides a signal path from the antenna to the baseband processor. It consists of a bandpass filter, a low-noise amplifier (LNA), and a downconverter.
De-pending on the type of receiver there could be two downconversions (as in a super-hetrodyne receiver), where one downconversion converts the signal to an IF. The signal is
then filtered and then downconverted to or near baseband to be sampled.
The other configuration has one downconversion, as in a homodyne (zero IF or ZIF) receiver, where the data is converted directly to baseband.The challenge in the receiver design is
to achieve the required sensitivity, intermodulation, and spurious rejection, while operating at low power.
The first line of defense
The receiver bandpass filter is the first line of defense to eliminate unwanted interference and noise. This filter must be able to achieve the cutoff needed for each bandwidth. In a 4G
implementation, the bandwidth could be as low as 5 MHz and as high as 20 MHz. If the filter were to be only 5-MHz wide, it would not have the capabilities to use the 20-MHz
bandwidth. However, if the filter is 20-MHz wide and the signal is only 5-MHz wide, the extra interference would increase the noise and reduce sensitivity. This means that a
tunable filter is needed. One option would be a bank of filters with different bandwidths, where selection is made based on the need.
A typical LNA has a noise figure of approximately 1 dB and a gain of about 20 dB. A trade-off is made between gain and noise to provide the best solution. The LNA sets the noise figure
of the overall receiver, since it is one of the first components of the receiver. Because of the high PAVR of the signal, the LNA will also have to be very linear to minimize any extra
distortion.
The downconverter section of the receiver will have to achieve good linearity and noise figure while consuming minimal power. A measure of the linearity in the mixer section is the
spurious free dynamic range (SFDR). This is directly related to the second and third order intermodulation products also known as IP2 and IP3.
The analog-to-digital converter (ADC) is the key component that can break the new system. System issues of the ADC concern whether or not to use undersampling, the PAVR of the
signal, the bandwidth, and the sampling rate. For a 5-MHz bandwidth signal a typical sampling rate would be 20 MHz. If IF sampling is used, the aperture uncertainty or jitter must
be low enough to prevent errors.
The next requirement is the dynamic range. For an MCM system using the theoretical PAVR for a 512-point IFFT, the dynamic range required would be 80 dB, which is equal to 13
bits. This relationship is demonstrated in Equation 5, which shows quantization noise,determined from the link budget as follows:
(5)
The desired quantization noise is determined by the average ratio of average signal power to average noise spectrum density measured in dB (Eb/No) for the subcarriers, the data
rate (DR), and backoff (which is generally 15 dB). The constant 20 dB is added to the end to put the quantization noise 20 dB lower than the system noise. The number of bits can be
calculated as shown in Equation 6.
(6)
In this equation, fs is the sampling rate. If the signal has interference or blocking, the ADC requires additional bits. The required dynamic range of the ADC could increase from 15 to
17 bits.
Baseband processing
The error correction coding of 4G has not yet been proposed, however, it is known that 4G will provide different levels of QoS, including data rates and bit error rates. It is likely
that a form of concatenated coding will also be used, and this could be a turbo code as used in 3G, or a combination of a block code and a convolutional code. This increases the
complexity of the baseband processing in the receive section.
4G baseband signal-processing components will include ASICs, DSPs, microcontrollers, and FPGAs. The receiver will take the data from the ADC, and then use it to detect the proper
signals. Baseband processing techniques such as smart antennas and multi-user detection will be required to reduce interference.
MCM is a baseband process. The subcarriers are created using IFFT in the transmitter, and FFT is used in the receiver to recover the data. A fast DSP is needed for parsing and
processing the data.
Different algorithms can be used to create a smart antenna; the goal is to improve the signal by adjusting the beam pattern of the antennas. The number of DSPs needed to implement
an smart antenna depends on the type of algorithm used. The two basic types of smart antenna are switched-beam antennas and adaptive arrays. The former selects a beam pattern
from a set of predetermined patterns, while the latter dynamically steers narrow beams toward multiple users. Generally speaking, SA is more likely be used in a base station than
a mobile, due to size and power restrictions.
Multi-user detection (MUD) is used to eliminate the multiple access interference (MAI) present in CDMA systems. Based on the known spreading waveform for each user, MUD
determines the signal from other users and can eliminate this from the desired signal. Mobile devices do not normally contain the spreading codes of the other users in the cell, so
MUD will likely be implemented only in base stations, where it can improve the capacity of the reverse (mobile-to-base) link.
Transmitter section
The purpose of the transmitter is to generate and send information. As the data rate for 4G increases, the need for a clean signal also increases. One way to increase capacity is to
increase frequency reuse. As the cell size gets smaller to accommodate more frequency reuse, smaller base stations are required. Smaller cell sizes need less transmit power to
reach the edge of the cell, though better system engineering is required to reduce intra-cell interference.
One critical issue to consider is spurious noise. The regulatory agencies have stringent requirements on the amount of unwanted noise that can be sent out of the range of the
spectrum allocated. In addition, excess noise in the system can seriously diminish the system's capacity.
With the wider bandwidth system and high PAVR associated with 4G, it will be difficult to achieve good performance without help of linearity techniques (for example, predistortion
of the signal to the PA). To effectively accomplish this task, feedback between the RF and baseband is required. The algorithm to perform the feedback is done in the DSP, which is
part of the baseband data processing.
Power control will also be important in 4G to help achieve the desired performance; this helps in controlling high PAVR - different services need different levels of power due to the
different rates and QoS levels required. Therefore, power control needs to be a very tight, closed loop. Baseband processing is just as critical whether dealing with the receiver or
transmitter sections. As we've seed, RF and baseband work in tandem to produce 4G signals. The baseband processing of a 4G transmitter will obviously be more complicated than in
a 3G design. Let's consider the chain of command.
The digital-to-analog converter (DAC) is an important piece of the transmit chain. It requires a high slew rate to minimize distortion, especially with the high PAVR of the MCM
signals. Generally, data is oversampled 2.5 to 4 times; by increasing the oversampling ratio of the DAC, the step size between samples decreases. This minimizes distortion.
In the baseband processing section of the transmit chain, the signal is encoded, modulated, transformed using an IFFT, and then a cyclic extension is added. Dynamic packet
assignment or dynamic frequency selection are techniques which can increase the capacity of the system. Feedback from the mobile is needed to accomplish these techniques. The
baseband processing will have to be fast to support the high data rates.
Even as 3G begins to roll out, system designers and services providers are looking forward to a true wireless broadband cellular system, or 4G. To achieve the goals of 4G,
technology will need to improve significantly in order to handle the intensive algorithms in the baseband processing and the wide bandwidth of a high PAVR signal. Novel techniques
will also have to be employed to help the system achieve the desired capacity and throughput. High-performance signal processing will have to be used for the antenna systems,
power amplifier, and detection of the signal.
Michael LeFevre is a system engineer in broadband communications in Motorola's Wireless Infrastructure Systems Group Division. He holds a MSEE from Brigham Young University
in Provo, Utah. He can be reached at michael.lefevre@motorola.com.
Peter Okrah received his Ph.D. in electrical engineering from Stanford University. He is the Manager of 4G systems technologies research of the Wireless Infrastructure Systems
Division of the Motorola Semiconductor Products Sector, Tempe, Arizona. He can be reached at at peter.okrah@motorola.com.
