RF Design of a TDMA Cellular/ PCS Handset, Part I

This article describes a typical RF architecture and focuses on key RF design considerations and trade-offs as applied to a typical dual-band TDMA transceiver RF section.

TIA/EIA/ANSI-136 is the applicable cellular/PCS standard in the North-American market for dual-band time-division multiple access (TDMA)-based handsets. Presently, the standards are being upgraded to provide a smooth transition into the third-generation (3G) line of wireless data/voice terminals. The Universal Wireless Communications Consortium's new UWCC-136 standard will allow multimedia capabilities and services such as wireless Internet access in future data-centric handsets. This added functionality translates into the need for higher bit rates and increased RF bandwidth.

The mobile station transmission frequency bands are presently allocated from 824.04 to 848.97 MHz for cellular service and from 1,850.01 to 1,909.95 MHz in the PCS band. The mobile receiver section operates in the 869.04- to 893.97-MHz frequency range for cellular and in the 1,930.05- to 1,989.99-MHz frequency range for PCS. The present RF channel bandwidth is 30 kHz, with plans to extend it to 200 kHz and even to 1.6 MHz. The 200-kHz bandwidth will feature higher-order digital modulation schemes like 8-PSK to make it compatible with future generations of TDMA-based, GSM-enhanced data rates for GSM evolution (EDGE) data-capable handsets.

Most of today's TDMA handsets accommodate dual-band tri-mode capabilities (FM-based analog service in the cellular band, and /4 DQPSK-based digital service in both cellular and PCS bands). The new standards set the requirements for multi-band, multi-bandwidth, and multi-timeslot operation, which will lead to more complex mobile and basestation designs. Three critical design variables ý cost, size, and power consumption ý will be greatly impacted. Most of the future handset designs targeted for the US market require dual-band operation as a single operator may own spectrum in cellular and PCS bands.

Requirements for such a dual-band handset include a receiver sensitivity of at least -116 dBm in analog mode (for 12 dB signal + noise + distortion/ noise + distortion [SINAD]) and -110 dBm for 3% bit error rate (BER) in digital mode (in static conditions). The maximum input level specification is -25 dBm for 3% BER.

The transmitter output power level for a Class IV handset is limited to 600 mW effective radiated power (ERP, referenced to a half-wave dipole). Error vector magnitude (EVM) is a quantitative indication of the digital modulation quality. The EVM root-mean-square (RMS) value is specified to be 12.5% maximum.

The adjacent channel (fc ý 30 kHz)/alternate channels (fc ý 60/90 kHz) power ratio (ACPR) specifications reflect the transmitter's linearity (fc is the desired RF channel frequency). Emission power levels should not exceed -26 dBc and -45 dBc for adjacent and alternate channels, respectively.

A block diagram that illustrates the main functional blocks in a typical handset is shown in Figure 1 . From a hardware standpoint, two main sections become apparent: the logic and the transceiver sections. The latter is the radio section of the handset.

The logic section includes a microprocessor, DSP baseband processor, memory, display, power management functions, and the vocoder. On the transmit side, the purpose of the logic section is to efficiently digitize the voice in terms of bit rate (this example uses a 7.4-kbps algebraic code-excited linear prediction [ACELP] algorithm), to provide channel-coding functions and to provide the digital modulation I&Q baseband signals for the transceiver. The logic section receive side provides channel filtering equalization/decoding and ACELP speech decoding. The resident software serves the call processing and user interface functions.

The RF module's basic objectives are to modulate high-frequency RF carriers with the analog and I&Q digital signals coming from the handset's baseband section, and to demodulate the received analog and/or digitally modulated RF signals.

Four functional sections can be identified in a typical dual-band RF module: the front end, the receiver, the frequency synthesizers, and the transmitter (see Figure 2 ).

Dual paths for the RF signal down-conversion to a common intermediate frequency (IF) in the cellular and PCS bands receive sections, and dual paths for the transmitting section can be clearly seen in the dual-conversion receive and transmit architecture depicted in Figure 2 .

A common antenna serves both receive and transmit operations in both bands. Design compromises for dual-band operation focus on antenna gain, radiation patterns, and a common matching network.

The received signal from the antenna is routed through the low-pass filter section of the input diplexer for the cellular band operation. The PCS band transmit and receive paths are served through the high-pass filter section of the diplexer.

The diplexer separates the cellular and PCS bands, while the duplexer separates the transmit and receive sections in the cellular band. This leads to different frequency selectivity requirements (such as different insertion losses, size, and cost).

In the cellular band, the duplexer's jobs are:

• Preventing transmitter noise in the receive band from desensitizing the receiver in the full-duplex analog mode.
• Attenuating the power amplifier (PA) output signal to avoid driving the low-noise amplifier (LNA) into compression.
• Attenuating the receiver's spurious responses (first image and others).
• Attenuating first local oscillator (LO) feed-through using the first mixer LO-RF ports.
• Attenuating transmitter output harmonics and other undesired spurious products.

The receiver's (broadband) frequency selectivity is shared between the duplexer and the image filter, with trade-offs in terms of size and insertion loss.

The PCS receiving path is different from the cellular path. Instead of a duplexer, a transmit-receive (T/R) switch is provided. This is feasible because of the present digital half-duplex mode of operation in the PCS band (nonsimultaneous receive and transmit functions). Future digital-mode multi-timeslot operation will mandate, in some operating modes, simultaneous transmit and receive operation. Thus, the use of a duplexer in the PCS band will be required.

The receiver is based on a double-conversion, superheterodyne architecture that provides, at a relatively low cost, the high dynamic range and selectivity required for this application.

The LNA's main purpose is to increase the level of the weak incoming RF signal without significantly degrading the signal-to-noise ratio (SNR) and without introducing nonlinearities that generate undesired intermodulation products. Two different LNAs with a one-step attenuation gain control are shown, one for each band (see Figure 2 ). Under strong-signal conditions, this reduction in gain prevents overloading the active stages.

The LNA output signal is routed through a 25-MHz-bandwidth bandpass filter that provides additional attenuation for the first image, signal image noise, and other receiver spurious responses in the cellular band. In the PCS band, the LNA is preceded and followed by two 60-MHz bandwidth image filters. Careful attention must be paid to the LNA output-mixer RF input isolation to avoid degradation of the first image rejection.

The received RF signals are then routed to the first mixers (M1-PCS and M1-CELL) where they are down-converted to a common first IF. In a practical scenario, only the selected band receiver front end is powered up, to minimize power consumption.

The first IF filter is generally a 30-kHz narrowband surface acoustic wave (SAW) filter with a center frequency typically above 100 MHz. Steep out-of-band attenuation and amplitude and phase linearity in the passband are important. The second image rejection is determined by this filter. Its out-of-band attenuation characteristics contribute to the alternate channel rejection (fc ý 60 kHz) and the receiver's overall intermodulation (IM) performance. The IF signal is then applied to the input of the second mixer (M2), which provides the final down-conversion to the last IF (typically 450 kHz).

The second IF filter is generally a 30-kHz narrowband ceramic filter with a 450-kHz center frequency. Its steep attenuation allows it to meet the adjacent/alternate channel rejection specifications. Good amplitude and group delay characteristics in the passband are required to avoid degrading the BER in digital mode.

The 450-kHz output signal is split and fed into:

• FM IF amplifier, limiter, quadrature demodulator, and received signal strength indicator (RSSI) sections.
• The automatic gain control (AGC) and I&Q demodulation stages.

In analog mode, the received signal is frequency demodulated to produce the baseband audio signals, supervisory audio tone (SAT), signaling tone (ST), and wideband data required for call setup and control. The RSSI provides a dc voltage output proportional to the received signal strength. This information can be used to determine such things as the strongest channel for mobile-assisted hand-off operations (MAHO) and signal strength.

In digital mode, the p/4 DQPSK digitally modulated signal is demodulated into its baseband I&Q components, respectively. The demodulated audio signals, RSSI, and the digital mode I&Q demodulated signals are digitized in the handset logic section for further processing in the digital domain.

Both front-end and second IF gain-controlled stages keep the receiving chain linear in digital mode across an input signal range of about -115 dBm up to -25 dBm (90 dB dynamic range). This configuration keeps the I&Q demodulated signal amplitudes to a near constant level that properly fits within the available ADC dynamic range.

The down-conversion to baseband can be implemented with a 450-kHz third LO signal provided by the synthesizer section (LO#3).

Design trade-offs for the receiver include:

• Discrete versus integrated front ends (LNAs, mixers, and LO buffers). The discrete option allows the designer to tailor the gain, noise figure (NF), third-order input intercept point (IIP3), and power consumption for optimum performance, at the expense of increased parts count and size. Printed circuit board (PCB) component placement also becomes more flexible in a discrete design approach. An integrated RF ASIC reduces development time, but the final cost may be higher because of lower RF IC yields, packaging limitations, RF isolation, and testing issues.

• The RF gain distribution is implemented in order to achieve an optimum dynamic range (IIP3/NF trade-off).

• A higher IIP3 leads to higher current consumption (shorter standby time). Good strong-signal performance is of paramount importance in a mobile cellular environment.

• Use of shared components in both frequency bands reduces parts count, size, and cost, but increases the difficulty of optimizing the performance in each band independently.
  
  

Two basic receiver architectures are available: superheterodyne and direct conversion (or low-IF). The latter architecture eliminates the IF stages, mixers, filters, and associated LOs, providing the flexibility to accommodate different bandwidths and standards.

Direct conversion has potential advantages in terms of reducing cost, PCB real estate, and power consumption. However, its present performance for TDMA applications (with the non-unity peak-to-average power ratio envelope p/4 DQPSK modulation format) lags behind the time-proven superheterodyne topologies that allow excellent dynamic range and selectivity, with shorter and less risky development cycles for a given performance level. Disadvantages of the superheterodyne approach are a high parts count and integration difficulties due to the high-Q filters, which should be placed off-chip.

A practical direct conversion approach requires the efficient resolution of several issues: time-varying dc offsets, LO leakage through the antenna, gain/phase matching and second-order nonlinear distortions in the down-converting quadrature mixers, and proper operation under the TDMA dynamics.

The main purpose of an RF receiver is to receive the desired signal while at the same time rejecting undesired (spurious) signals that can be present at the receiver input at much higher power levels.

The three most important receiver spurious responses are the 1st Image, 2nd Image, and Half-IF. Undesired strong RF signals at certain frequencies can cause serious degradation in the BER or SINAD. In severe cases they can lead to dropped calls. Their frequency location is explained in Figure 3 , Figure 4 , and Figure 5 . A numeric example is added to help clarify the topic. Two factors determine their frequency location: the receiver frequency plan and the desired tuned channel frequency. A significant part of the receiver design is eliminating or attenuating these undesired signals. The filters' attenuation requirements and the first mixer linearity specifications are dictated by the required level of spurious suppression.

In the 1st Image, a desired low-level RF signal at 1,930 MHz is down-converted to a 100-MHz first IF with a first LO set at 2,030 MHz (see Figure 3 ). An out-of-band strong undesired signal at 2,130 MHz is also down-converted to the 1st IF, because the difference between the LO and this interfering RF signal is also 100 MHz. The frequency location of the 1st Image is 2,130 MHz in this particular scenario. The only way to attenuate this strong undesired signal is by providing enough selectivity in the two front-end filters. A higher first IF will also help. In certain cases, a narrowband LNA also contributes to reject the 1st Image response.

In the 2nd Image, the same frequency plan (LO and IF frequencies) as illustrated in Figure 3 is used (see Figure 4 ). An out-of-band strong undesired signal at 1930.9 MHz is down-converted to the first IF stage as a 99.1-MHz signal (99.1 = 2,030 ý 1,930.9). The desired signal is down-converted to 100 MHz. Both the desired (100 MHz) and the undesired signal (99.1 MHz) will be mixed with the fixed LO#2 (99.55 MHz) down to 450 kHz. This will degrade the carrier-to-interference (C/I) ratio. The RF frequency location of the 2nd Image is 1930.9 MHz in this particular scenario. The only way to attenuate this strong undesired signal is by providing enough attenuation in the first IF filter (more than 60 dB at 99.1 MHz in this example). In most cases, this high level of attenuation is provided by a single SAW IF filter. To meet this difficult requirement, close attention must be paid to the PCB layout design around this filter to provide high isolation between the filter's input and output ports. Otherwise, the resulting IF filter + PCB total attenuation might not be enough to achieve the required attenuation level.

The Half-IF spurious response is sometimes very troublesome to attenuate (see Figure 5 ). Its RF frequency location is half the first IF frequency (50 MHz) away from the desired signal. In the numerical example, it is located at 1,980 MHz, because the desired signal is at 1,930 MHz and the IF is 100 MHz. The two RF filters cannot attenuate it because this frequency falls in their passband region.

The first mixer's LO is set at 2,030 MHz. Its second harmonic is 2 x 2,030 = 4,060 MHz. The RF undesired signal (Half IF) second harmonic is located at 3,960 MHz (2 x 1980). The mixing (difference product) of both LO and RF second harmonics leads to a new (undesired) 100-MHz IF signal that will interfere with the desired IF signal. The level of this (2,2) fourth-order product (2 x LO ý 2 x RF) is determined by the first mixer's second-order intercept point (IP2).

If the desired channel frequency falls in the middle of the PCS band (1,960 MHz), the receiver's half-IF response will be located at 2,010 MHz. Because the input RF filters have a 1,930 to 1,990 MHz passband, additional attenuation is provided by these filters.