Multiple techniques power signal sources

High-speed wireless and data-communications products require high-frequency signal sources during laboratory and production test. Vendors employ a variety of techniques to generate test signals, ranging from mechanically driven approaches that alter oscillator parameters such as capacitance to digital synthesis techniques that derive variable output frequencies from a single reference frequency.

Mechanical approaches have largely given way to solid-state approaches because of the latter's ruggedness and programmability. With programmable systems, you can automate microwave-component test applications that once required multiple test sequences, and, perhaps, multiple pieces of test equipment, all manually operated by a technician. Whether for bench-level R&D tasks or for production lines, today's signal sources excel at providing the signals that can help you characterize the components that go into products that operate in the 1-GHz-and-up RF and microwave ranges.

Of course, stimulus signals are only part of the test process. You'll want to measure how your DUT responds to an RF/microwave signal, too. If the response is an RF/microwave output, you might want to employ a vector or scalar network analyzer (Ref. 1); such equipment combines source and stimulus instruments in one system. But for many applications, a stand-alone benchtop signal source provides the needed performance and flexibility.

As you peruse literature describing vendors' RF/microwave sources, you'll find such terminology as signal generator, signal synthesizer, synthesized signal generator, and synthesized sweeper. The first—signal generator—can mean almost anything, but it generally denotes the presence of a voltage-controlled oscillator as the primary source of RF/microwave frequencies. Such units tend to have low costs, are compact and lightweight, and are adequate for many applications. They may not, however, offer the frequency stability, spectral purity, or accuracy you need.

A higher performance alternative includes the YIG-tuned oscillator (YTO), in which electrons oscillate within a magnetic field established within a yttrium-iron-garnet (YIG) alloy by an electromagnetic coil, whose current (and, hence, magnetic strength) determines the oscillation frequency. Firms including April Instrument (Sunnyvale, CA; www.aprilinstrument.com) offer YTO-based RF/microwave signal sources. APA Wireless (Fort Lauderdale, FL; www.apawireless.com) has proposed what it calls a YRO (an APA Wireless trademark), for YIG replacement oscillator, as the basis for a frequency synthesizer; YROs are silicon replacements for the heavy, expensive YIG alloy used in YTOs (Ref. 2).

The presence of a fixed-frequency source is the key to the "synthesizer" nomenclature. The US Patent and Trademark office describes a synthesizer as an instrument ". . . in which a single stable oscillator is used as a source of one or more frequencies derived therefrom by frequency multiplication or division . . . The stable oscillator is often crystal controlled or phase-locked" (Ref. 3).

Traditional YTO-based instruments are not synthesizers. To vary their frequency, you alter the current through an electromagnet surrounding the YIG alloy. Such YTO-based instruments can generate clean RF outputs, yet they often lack the ease of control and frequency agility of synthesis-based instruments—they can take milliseconds rather than microseconds to switch frequencies because of the lengthy time constants imposed by the electromagnet's inductance. In contrast, synthesizers derive all their output frequencies by performing analog or digital operations on a single reference frequency and can change output value rapidly. That's important for generating fast sweep rates in synthesized sweepers and can be important in other applications as well—for example, when you need to simulate the operation of frequency-hopping components. What's more, synthesizers can readily accept extremely stable reference oscillators, including temperature-compensated crystal-controlled oscillators (TCXOs) and oven-controlled oscillators. Nevertheless, fixed-frequency YTOs or YROs can serve as the fixed-frequency reference of a synthesized signal source.

Synthesizers come in several flavors. Analog synthesizers employ multipliers and dividers as well as mixers, filters, and RF switches in what Programmed Test Sources (Littleton, MA; www.programmedtest.com) calls a "mix and divide" analog direct-synthesis approach (Ref. 4). In contrast, indirect digital synthesis substitutes phase-locked loops for the multiple filters and mixers of the analog approach.

Indirect digital synthesis approaches include the "fractional-N" approach (Fig. 1), in which a divider's integer division ratio changes periodically based on the value of a fractional-frequency-instruction input. For example, if the fractional input value corresponds to 0.1, then the divider ratio will change by one integer value for every tenth cycle. The fractional-N approach introduces phase perturbations that can require the addition of correction circuitry. This circuitry is often analog, yet IFR Systems (Wichita, KS; www.ifrsys.com) holds a patent on a digital correction technology. The digital approach avoids the drift problems that can plague analog components.
    

Figure 1. A basic fractional-N synthesis technique employs a phase-locked loop to derive an output frequency from a reference. Courtesy of IFR Systems.

Note that the frequency of an indirect fractional-N synthesizer can't change instantaneously: frequency shift can only occur on accumulator overflow, a condition whose occurrence depends on the applied fractional-frequency instruction. A third synthesis approach—direct digital synthesis—overcomes this limitation. It is favored by Programmed Test Sources and Tektronix (Beaverton, OR; www.tek.com). Direct digital synthesis (Fig. 2) makes use of adder circuitry, which accumulates phase information—from 0° to 360°—based on a frequency-control input (Ref. 5). The phase information in turn serves to call up a particular point of stored waveform values.
     

Figure 2. In direct digital synthesis, a phase accumulator accepts a frequency-control input and addresses discrete values in waveform memory. Successive waveform values drive a DAC, which generates the analog output signal. Courtesy of Tektronix.

If a phase accumulator is 30 bits wide, for example, then 360° will correspond to 2³⁰, or about 1 billion. A frequency-control input of 1 would then require about 1 billion cycles to address all 360° worth of phase information stored in waveform memory, resulting in the lowest frequency available from this direct synthesis approach. (Note that at a quarter of the way through the 1 billion cycles, the delta register output would address the waveform memory value corresponding to 90°.) Higher control-input values would cause the 360° worth of phase information to accumulate more rapidly, resulting in a higher output frequency.

Real-world signal sources don't necessarily fit smoothly within the categories outlined here. Giga-tronics (San Ramon, CA; www.gigatronics.com), for example, combines a fixed-frequency reference with a tuned YIG oscillator plus high-speed logic circuitry to implement microwave signal synthesis (Fig. 3).
    

Figure 3. Synthesizers can combine direct and indirect synthesis techniques as well as accommodate modulation inputs. A DSP handles sweep-frequency chores. Courtesy of Giga-tronics.

Ultimately, your choice of signal source will depend on whether an instrument meets the specifications you need at an affordable price. And as this magazine has already advised, don't overspecify—you'll be wasting money. The synthesis techniques employed internally to an instrument don't matter if the instrument's guaranteed performance meets your needs. In fact, Agilent Technologies (Palo Alto, CA; www.agilent.com) and Anritsu (Morgan Hill, CA; www.anritsu.com) don't advertise their synthesizers' internal workings.

Working from data sheet values alone when comparing RF/microwave signal sources, though, can be difficult. The operating-frequency ranges are usually provided in a straightforward manner, but when it comes to spectral-purity and phase-noise specs, manufacturers often offer inconsistent specification methods, failing, for example, to specify noise levels at consistent offsets from carrier frequency. Graphical descriptions of performance can help (Fig. 4), but they are often not available or else they specify typical—not worst-case—performance. In short, published specs don't tell the whole story, and the more you know about how a particular signal source works, the better chance you'll have of getting the instrument you need.
    

Figure 4. Graphical depictions of specifications such as phase noise can help you compare signal sources from various vendors. Courtesy of Agilent Technologies.