Radio transmitters can now be designed with direct digital modulation, due to the availability of high-speed digital-to-analog converters (DACs). This approach is already commonly used to generate multicarrier quadrature-amplitude- modulation (QAM) signals for data transmission over Hybrid Fiber Coax (HFC) links in cable-television (CATV) access networks, and is also used to synthesize microwave intermediate- frequency (IF) signals for instrumentation, radar, and high-speed military wireless communications links.

Advances in silicon CMOS with effective scaling to create smaller feature sizes have yielded low-cost, numerically controlled oscillators (NCOs) that can generate high-frequency local-oscillator (LO) signals with low power consumption. Silicon CMOS also supports the design of low-power quadrature modulators for generating the in-phase (I) and quadrature (Q) signals that form the RF digital modulation in many systems. In conjunction with high-speed RF DACs such as the MAX19692 from Maxim Integrated Products, designers now have the building blocks needed to implement highperformance digital RF transmitters and other systems.

The update rates of 12- and 14-b CMOS RF DACs are approaching 5 GSamples/s. In conjunction with nanometerfeatured CMOS signal-processing components, frequencies to 2 GHz and higher can be synthesized digitally, with low spurious and noise content. A flexible digital transmitter can be formed by teaming a high-speed DAC with a highperformance programmable digital modulator.

Digital transmitters and modulators overcome some of the limitations of analog transmitters. Digital synthesis can provide signals limited only by the finite digital word length. Digital synthesis can be performed without the temperature drift, conversion loss, and distortion of analog sources. Using digitally generated quadrature modulation, quadrature errors and LO feedthrough can be essentially eliminated.

A high-speed RF DAC allows instantaneous, phase-continuous frequency hopping. It can synthesize broadband signals with excellent gain flatness. For example, the downstream CATV spectrum from 50 MHz to 1 GHz can be synthesized by means of a single MAX5881 DAC from Maxim with better than 2-dB gain flatness across the full bandwidth. With digital modulators, the lossy broadband RF combining networks can be eliminated for increased density of CATV head-end equipment. Radar signals with GHz bandwidths can be synthesized without using frequency doublers, providing increased resolution. Multi-standard, multi-band transmitters can be implemented using programmable digital hardware. Compared to an analog modulator and LO, a digital transmitter is simpler and more predictable and the transmitter subsystem requires less board space and consumes less power.

A DAC provides a discrete time signal as an output. Images of the carrier that are created are aliased symmetrically around integer multiples of the DAC update (clock) rate. If the DAC is non-ideal, it creates harmonics of the output signal that are also aliased symmetrically at integer multiples of the DAC update rate. In the cases where harmonics or folded images of the harmonics may fall within the desired RF band, they can be filtered. However, if they fall in the desired RF band and if the power level is high enough, the band's spurious limit may be violated. One way to avoid this problem is to push the DAC update rate high enough that the dominant harmonics fall out of band.

For CATV distribution in the 50-MHz-to-1-GHz band, Nyquist theory states that a DAC with sample rate higher than 2 GSamples/s should be adequate for synthesizing the entire CATV band. The dominant harmonic product for a DAC is typically the third harmonic (HD3), assuming a differential output where second harmonic content is strongly attenuated. If a DAC with sampling rate of 2.5 GSamples/s is used Fig. 1(a)>, the third-harmonic image for a real 1-GHz carrier will be located at 500 MHz. But if the DAC's update rate is pushed beyond 4 GSamples/s Fig 1(b)>, this third-harmonic image will never fall within the 50-MHz-to-1- GHz band. Pushing the DAC update rate higher also simplifies the design of the reconstruction filter.

DACs such as the 12-b, 4.3- GSamples/s model MAX5881 can be used in CATV head-end equipment and can directly synthesize multiple QAM carriers in the 50-MHz-to-1- GHz band. The MAX5881 DAC features a doubled update rate in which the output update rate of the DAC is twice the DAC clock rate. It updates on both edges of the input clock to achieve twice the sample rate of conventional DACs. The device features excellent spurious, noise, and adjacent channel power (ACP) performance and can synthesize multiple carriers per the requirements defined in the Data-Over-Cable Service Interface Specification (DOCSIS, detailed in Downstream RF Interface Specification CM-SP-DRFI-110-10011, Cable Television Laboratories, January 6, 2006).

Another DAC from Maxim, the MAX19692, operates with 12-b resolution and 2.3-GSamples/s update rate with selectable frequency response. The default operating mode of the DAC is the conventional nonreturn- to-zero (NRZ) mode that has a familiar sinc shape, with zeroes at every multiple of the DAC update frequency as shown in Eq. 1:

ANRZ = A0|outT)>/(πfoutT)| (1)
where
T = the clock period,
fout = the output frequency,
A0 = the DC amplitude, and
ANRZ = the amplitude response in NRZ mode.

In RZ-mode, the DAC zeros itself in the time domain at every half clock cycle. The resulting frequency (amplitude) response, ARZ, is described by Eq. 2. This frequency response is flatter than the NRZ-response in the three first Nyquist zones, and particularly in the second and third Nyquist zone, providing useful performance for synthesizing wideband signals in the second and third Nyquist zones.

The third mode of operation is called RF-mode. This is similar to mixing the DAC output with the DAC update clock. In this mode, the amplitude response, ARF, is described by Eq. 3.

Continue to page 2

Page Title

As can be seen in Fig. 2, the NRZ mode provides highest power in the first Nyquist zone. The RZ-mode provides the flattest frequency response in the first and third Nyquist zones, and the RF-mode provides superior output power in the second and third Nyquist zones, as well as the flattest gain in the second Nyquist zone.

The use of high-speed DACs is particularly useful in broadband, multicarrier transmitters, where the system can modulate and combine a large number of carriers in the digital domain. The composite, multiple-carrier signal can then be amplified with a single signal chain, rather than using multiple amplifiers with lossy power combiners, resulting in a reduction in overall system power dissipation.

The downstream portion of a CATV access network, for example, has 32 carriers with 256QAM modulation, each 6 MHz wide, to provide high-speed data, video-on-demand, and switched-digital-video services. The carriers can be placed anywhere from 50 MHz to 1 GHz. In most existing systems, this is achieved by combining multiple RF ports from conventional QAM modulators that support a limited bandwidth (e.g., four 6-MHz wide QAM channels). Due to combiner losses, the output power of each QAM modulator must be approximately the same as the total power needed at the output of the combiner network Fig. 3(a)>. Hence in this case, eight power amplifiers are needed. Conversely, in Fig. 3(b), all QAMs are modulated and combined in the digital domain, with signals converted to the analog domain via an RF DAC. In this approach, no external combiners are needed to combine the different QAM sources and only a single power amplifier is needed. Therefore, the power dissipated in generating the QAM signal using the high-density modulator is much lower.

Figure 4 shows the digital evolution of a basic transmitter architecture. The quadrature modulator of the analog approach Fig. 4 (top)> has been replaced with its digital equivalent Fig. 4 (bottom)>. The digital modulator uses less power than its analog equivalent, and the trend for lower-power consumption will continue with continued CMOS scaling. The LO of the analog approach is replaced with an NCO in the digital architecture. In a multichannel analog transmitter, each channel requires its own PLL and VCO, although clocks can be shared among modulators. In a multichannel direct-digital transmitter, the clock can be shared among modulators and no analog LO signals are needed, so that only one PLL and VCO are needed. Eliminating the additional PLLs and VCOs saves cost, power, and board space.

Analog transmitters often limit the maximum allowable bandwidth for its carriers. As a result, multiple carriers are placed in a frequency block for upconversion, although this can limit the flexibility of the analog modulators used in a system such as for downstream CATV applications. Because CATV service providers offer multiple services, such as voice, data, and video, finding enough bandwidth to support an analog transmitter only capable of block upconversion can be challenging.

Digital transmitters, on the other hand, provide considerably more flexibility in managing bandwidth. With a wideband RF DAC like the MAX5881, multiple carriers can be split into smaller blocks or spaced widely enough in the digital domain, so that finding space in a crowded spectrum becomes irrelevant (Fig. 5). Modern RF DACs provide enough bandwidth to support full carrier agility across an available RF band, facilitating flexible placement of carriers anywhere in the band. If not for legacy equipment, the entire 50-MHz-to- 1-GHz downstream cable band could be populated with 158 (6-MHz) QAM carriers using a single MAX5881.

What kind of performance can be expected from such DACs when producing digitally modulated signals? Figure 6 shows a 64-carrier block of 256QAM signals for digital CATV or DOCSIS high-speed data transmission. The symbol rate for each carrier is 5.36 Msymbols/s. This digitally generated carrier meets DOCSIS spurious and noise requirements by more than a 5-dB margin.

Synthesis of single QAM carriers is typically the most challenging for the DAC's intermodulation performance. Figure 7 shows a single 6-MHz-wide 256QAM carrier synthesized at 300 MHz using a MAX5881 12-b DAC at 4 GSamples/s. The noise floor is more than 80 dB below the carrier power level, meeting the requirements of high-performance communications infrastructure applications.

These DACs also allow signal synthesis in higher Nyquist zones. A lower input-data rate and clock rate can then be used at the expense of output power and more stringent outof- band analog filtering requirements. For example, Fig. 8 shows a 72-MHzwide block of 12 OFDM carriers centered at 535 MHz using a MAX19692 DAC. The signals were synthesized in the second Nyquist zone at a DAC update rate of 780 MSamples/s.

Figure 9 shows a 460-MHz-wide block of 8PSK satellite IF carriers centered at 1.82 GHz. This signal is synthesized in the fourth Nyquist zone. The impulse response of the DAC is a low-duty-cycle RZ-response, tailored to achieve good gain flatness while operating at a clock rate of 1.076 GSamples/s. This results in fairly low output power, but still meeting the carrier-to-noise (C/N) requirement with good margin. As another example, Fig. 10 shows a four-carrier WCDMA signal synthesized with the MAX19692 and centered at 2.15 GHz. The signal was synthesized in the third Nyquist zone.