Switching-regulator supply provides low-noise biasing for varactor diodes

Low-voltage systems often need a locally generated high voltage. Even for an application as noise-sensitive as varactor-diode biasing, a carefully planned switching-regulator-based design and layout can provide the necessary bias voltage.

Jim Williams and David Beebe, Linear Technology Corp -- EDN, 11/9/2000

Telecommunication, satellite links, and set-top boxes require tuning of a high-frequency oscillator. The actual tuning element is a varactor diode, which is a two-terminal device that changes capacitance as a function of reverse-bias voltage (see sidebar "Variable-capacitance diodes"). The oscillator is part of a frequency-synthesizing loop (Figure 1). A PLL compares a divided-down representation of the oscillator with a frequency reference. The circuit level shifts the PLL's output to provide the high voltage necessary to bias the varactor, which closes a feedback loop by voltage tuning the oscillator. This loop forces the VCO (voltage-controlled oscillator) to operate at a frequency determined by the frequency reference and the divider's division ratio.

The high-voltage bias is necessary to achieve wide-range varactor operation. Figure 2 shows varactor-capacitance versus reverse-voltage curves for a family of devices.A 10-to-1 capacitance shift is available, although a 0.1 to 30V swing is necessary. The curves in Figure 2 are characteristic of typical hyperabrupt devices. Response modification is possible with compromises in performance, particularly with linearity and sensitivity.

Designers traditionally meet the bias-voltage requirement using the existing high-voltage rails. However, the current trend toward low-voltage-powered systems means that you must locally generate the high-voltage bias. Local generation of a high voltage implies the presence of some form of voltage-step-up switching regulator. You can use a step-up approach, but varactor-noise sensitivity complicates the design. In particular, the varactor responds to any form of amplitude variation of its bias, which results in an undesired capacitance shift. Such a shift causes VCO-frequency movement, resulting in spurious oscillator outputs. The PLL's loop action removes dc and low-frequency shifts, but activity outside the loop's passband causes undesired outputs. Most applications require that any spurious oscillator outputs, or spurs, are at least 80 dB below the nominal output frequency.

All of these requirements necessitate a low-noise, high-voltage supply and mandate caution in the switching-regulator design. Switching regulators are often associated with noisy operation, which makes a varactor-bias application seem hazardous. Careful preparation can eliminate this concern and allows for a practical switching-regulator-based approach to varactor biasing.

Simple boost regulator

In theory, a simple flyback regulator works for this application, but component choice and attention to layout are critical to achieving low noise. Additionally, component count, size, and cost are usually considerations in varactor-bias applications. Figure 3a shows a step-up switching regulator that, properly incarnated, permits low-noise varactor biasing. The circuit is a simple boost regulator. L₁, in conjunction with the SW pin's ground-referred switching, provides voltage step-up. D₁ and C₂ filter the output to dc. D₂ clips possible L₁ negative excursions. The feedback resistor ratio sets the loop servo point and, hence, the output voltage. C₃ tailors the loop's frequency response, minimizing switching-frequency ripple components at the output. C₁ and C₂ exhibit low-loss dynamic characteristics, and the 1.7-MHz switching frequency of the regulator IC allows miniature, small-value components. The relatively high switching frequency also means that ancillary downstream filtering is possible with similarly miniature, small-value components.

Layout is the most crucial design aspect for obtaining low noise. Figure 3b shows a suggested layout. The layout distributes ground, V_IN, and V_OUT in planes to minimize impedance. The IC's GND pin, pin 2, carries high-speed, switched current; this current's path to the circuit's power exit should be direct and highly conductive at all frequencies. R₂'s return current should not mix with pin 2's large dynamic currents. The location of C₁ and C₂ should be close to pin 5 and D₁, respectively. The grounded ends of these capacitors should tie directly to the ground plane. L₁ has a low-impedance path to V_IN; the driven end of L₁ returns directly to pin 1 of the IC. D₁ and D₂ should have short, low-inductance runs to C₂ and pin 2, respectively. Also, the common connection of D₁ and D₂ should mate tightly with pin 1 and L₁. Pin 1 has a small area, which minimizes radiation. Note that planes operating at ac ground enclose pin 1, thereby forming a shield. The layout further shields the feedback node, pin 3, from switching radiation, preventing unwanted interaction. Finally, the layout should orient L₁ so that its radiation causes minimal circuit disruption.

The low-voltage PLL output in Figure 1 requires an analog-level shift to bias the varactor. Figure 4 shows some alternatives. In Figure 4a, the LT1613 regulator IC's 32V output powers the amplifier. The feedback ratio sets a gain of 10, resulting in a 0 to 30V output for a 0 to 3V input. Figure 4b is a noninverting common-base stage. The gain in this circuit is less well-controlled than in Figure 4a, but overall frequency-synthesizer loop action obviates this concern. Figure 4c's common-emitter circuit is similar to Figure 4b's except that it inverts the signal to the varactor.

Figure 5 combines the considerations mentioned above into a realistic test circuit. The 5V-powered design comprises the LT1613 regulator, an amplifier-based level shift, and a VCO operating in the gigahertz region. Using a filtered LT1004 reference and a gain of 10, the circuit biases the amplifier to a 12V output, which simulates a typical varactor-bias point. The regulator configuration's low-noise output receives additional filtering via the 100?, 0.1-µF network at the amplifier power pin and by the amplifier's PSRR (power-supply rejection ratio). The RC combination provides a theoretical, or unloaded, break below 20 kHz, and you can derive the amplifier's PSRR benefit from Figure 6. This graph shows PSRR versus frequency for a typical amplifier. There is a steep roll-off beyond 100 Hz, although almost 20 dB of attenuation is available in the megahertz region. This attenuation implies that the amplifier provides some beneficial filtering of the LT1613's residual 1.7-MHz switching components.

A final RC filter section sits directly at the VCO-varactor-bias input. Ideally, this filter's break frequency is far away from the 1.7-MHz switching rate for maximum ripple attenuation. In practice, the filter is within the PLL, which places restrictions on how much delay the filter can introduce. A PLL bandwidth of 5 kHz is usually desirable and dictates a filter point of about 50 kHz to ensure closed-loop stability. As such, the design sets the final RC filter—1.6 k? and 0.002 µF—at this frequency. It is worth noting that the varactor's input resistance is high—essentially that of a reverse-biased diode—and no filter buffering is necessary to drive it.

Analyzing noise performance

Careful measurements permit verification of circuit-noise performance (see sidebar "Preamplifier and oscilloscope selection"). Figure 7a shows ripple of approximately 2 mV at the LT1613's 32V output. Taken at the amplifier power pin, Figure 7b shows the effect of the 100?, 0.1-µF filter. Ripple and noise decrease to about 500 µV. Figure 7c, recorded at the amplifier output, shows the influence of amplifier PSRR. Ripple and noise further decrease to approximately 300 µV. The actual ripple component is approximately 100 µV. The final RC filter, located directly at the VCO varactor input, gives approximately 20 dB of further attenuation. Figure 7d shows ripple and noise inside 20 µV with a ripple component of about 10 µV.

The above results require good measurement technique and the use of a coaxial probing environment (Reference 1). Deviations from this regime produce misleading and pessimistic indications. For example, Figure 8a shows a 50% amplitude error over Figure 7a, even though the scope probe nominally monitors the same point. The difference between these two figures results from Figure 8a's use of a 3-in. probe-ground lead instead of Figure 7a's use of a coaxial ground-tip adapter. Similarly, the 500-mV measurement at Figure 7b's amplifier power pin degrades to Figure 8b's indicated 2-mV representation using the 3-in. probe-ground strap. The same ground strap causes error in Figure 8c's apparent 2-mV amplifier output unlike Figure 7c's correct 300-mV excursion. Figure 8d shows a 70-mV indication at the VCO varactor input using the 3-in. ground strap, which is different from Figure 7d's 20-mV data taken with the coaxial ground-tip adapter. (If you don't think 70 mV is a long way from 20 mV, you should consider your reaction to a 3.5-times income-tax reduction.)

When using the coaxial ground-tip adapter (Figure 8e), the VCO varactor input shows a blizzard of noise, compared with Figure 7d's orderly trace, because a 12-in. voltmeter lead connects to the input point. Pickup and stray RF act against the node's finite output impedance, corrupting the measurement. Figure 8f, also taken at the VCO input, is clearer than Figure 8e but still shows greater than 50% error. The culprit is a second probe, which on the LT1613 VSW pin and triggers the oscilloscope. Even with coaxial techniques in use at both probe points, the trigger probe dumps transient currents into the ground plane. This current introduces small common-mode voltages, resulting in a noise increase. One approach is to trigger the oscilloscope with a noninvasive probe (Reference 1).

Checking results

Although the varactor-bias noise-amplitude measurements are critical, it is difficult to correlate them with frequency-domain performance. Varactor-bias noise amplitude translates into spurious VCO outputs, which is the measurement of ultimate concern. Although it is possible to view the gigahertz-region VCO on an oscilloscope, a time-domain measurement lacks adequate sensitivity to detect spurious activity. You should use a spectrum analyzer. Figure 9a, a spectral plot of the VCO output, shows a center frequency of 1.14 GHz, with no apparent spurious activity within the '90-dB measurement noise floor. The marker at 1.7 MHz (3.5 divisions from center), corresponds to the LT1613's switching frequency. No distinguishable activity is apparent at approximately –90 dBc. Succeeding figures "sanity-check" this performance by systematically degrading the circuit and noting results. In Figure 9b, a direct connection replaces the VCO varactor input's RC filter. The 1.7-MHz spurious outputs are clearer at approximately –62 dBc. Connecting a 12-in. voltmeter lead to the measurement point results in a 4-dB degradation to approximately –58 dBc (Figure 9c). Figure 9d shows effects due to poor LT1613 layout. (A power-ground pin is routed circuitously, rather than directly, back to input common.) The figure also shows poor component choice, such as a lossy capacitor for C₂. In this case, spurious activity jumps to –48 dBc. Even with the proper layout and components, you can see problems in Figure 9e when you move the varactor-bias line close to switching inductor L₁. The bias line and RC-filter components are also farther away from the ground plane in Figure 9e than in the previous plots. The resultant electromagnetic pickup and increase in bias-line effective inductance cause 1.7-MHz spurs at –54 dBc. Additional harmonically related activity, although less severe, is also apparent. When you restore the bias line and RC filter to their proper orientation, the resultant plot is essentially identical to Figure 9a.

In sum, layout and measurement practices are at least as important as circuit design. As always, "the hidden schematic dominates performance," which is a favorite quotation of Charly Gullett of Intel Corp.

Variable-capacitance diodes Neil Chadderton, Zetex Inc The varactor diode capitalizes on the properties of the depletion layer of a p-n diode. Under reverse bias, the carriers in each region—holes in the p type and electrons in the n type—move away from the junction, leaving an area that is depleted of carriers. Thus, reverse bias creates a region that is essentially an insulator and comparable to the classic parallel-plate-capacitor model. The effective width of this depletion region increases with reverse bias, and, consequently, the capacitance decreases. Thus, the depletion layer effectively creates a voltage-dependent-junction capacitance that can vary between the forward conduction region and the reverse breakdown voltage. Manufacturers can produce varactor diodes with different junction profiles that exhibit different CV (capacitance-voltage) characteristics. Varactor types include those that exhibit a small range of capacitance to types that show a large change in capacitance for a relatively small change in bias voltage. This feature is particularly useful in battery-powered systems where the available bias voltage is limited. When you choose a varactor diode, you should consider numerous device characteristics. The most important characteristic is the CV curve that summarizes the range of useful capacitance and also shows the shape of the CV relationship, which may be relevant when a specific response is necessary. Factors to consider include the circuit's operational frequency range and, hence, the appropriate capacitance range, the available bias voltage, and the required response. The quality factor, or Q, at a particular condition is a useful parameter in assessing the performance of a device in tuned circuits. With respect to stability, the temperature coefficient of capacitance as capacitance changes may also be relevant. The reverse breakdown voltage, VBR , also has a bearing on device selection because this parameter limits the maximum reverse bias that you can use to achieve the minimum capacitance. REFERENCE Chadderton, Neil, "Zetex variable capacitance diodes," Application Note 9, Zetex plc, www.zetex.com.

Preamplifier and oscilloscope selection The low-level measurements this article describes require some form of preamplification for the oscilloscope. Current oscilloscopes rarely have sensitivities greater than 2 mV/DIV, although older instruments offer more capability. Table A lists representative preamplifiers and oscilloscope plug-ins suitable for noise measurements. These units feature wideband, low-noise performance. It is particularly significant that many of these instruments are no longer in production in keeping with current instrumentation trends that emphasize digital-signal acquisition as opposed to analog-measurement capability. The monitoring oscilloscope should have adequate bandwidth and exceptional trace clarity. High-quality analog oscilloscopes are unmatched in trace clarity. The exceptionally small spot size of these instruments is well-suited to low-level-noise measurement. In our work, we have found Tektronix's 454, 454A, 547, and 556 to be excellent choices. Their pristine trace presentation is ideal for discerning small signals of interest against a noise-floor-limited background. The digitizing uncertainties and raster-scan limitations of DSOs impose display-resolution penalties. Many DSO displays do not even register the small levels of switching-based noise.

Author info

Jim Williams, staff scientist at Linear Technology Corp (Milpitas, CA), has been writing technical articles for EDN for 25 years. He specializes in analog-circuit and -instrumentation design. He has served in similar capacities at National Semiconductor, Arthur D Little, and the Instrumentation Laboratory at the Massachusetts Institute of Technology (Cambridge, MA). A former student at Wayne State University (Detroit), Williams enjoys art, collecting antique scientific instruments, and restoring old Tektronix oscilloscopes.

David Beebe is associate design engineer at Linear Technology Corp (Milpitas, CA), where he has worked for five years. In his current position, he helped develop a line of bipolar, low-power, and micropower switching power supplies. He attended the College of San Mateo (San Mateo, CA), Mission College (Santa Clara, CA), and Community College of the Air Force (Montgomery, AL). Spare-time interests include acting as a mechanic and pit-crew member for Cliff Blackwell's No. 27B Sprint Car in Northern California.

REFERENCES

Williams, Jim, "Layout and probing techniques ensure low-noise performance," EDN, Feb 2, 1998, pg 141.

Williams, Jim, "Switching-regulator design lowers noise to 100 µV," EDN, Dec 4, 1997, pg 151.

Williams, Jim, "High-Speed Amplifier Techniques," Linear Technology Corp, Application Note 47, August 1991.

Hurlock, Les, "ABCs of Probes," Tektronix Inc, 1990.

McAbel, WE, "Probe Measurements," Tektronix Inc, 1971.