DESIGN VIEW is the summary of the complete DESIGN SOLUTION contributed article, which begins on Page 2.
One, zero, one, zero, one, zero... repeat forever. Building a clock should be one of the simplest engineering design challenges. And it is, unless you need it to be small, stable, and tunable. Traditional approaches involve a 555-type timer, or perhaps a comparator with a handful of discrete components. But these solutions consume board space, have limited stability over voltage and temperature, and yield poor accuracy beyond a few hundred kilohertz.
Resonant-element oscillators (crystals, ceramic resonators) boast superior accuracy and stability but are less robust. They're also not adjustable. However, a new class of device, the resistor-programmable oscillator (RPO), combines excellent accuracy and linearity, a small footprint, low power, and the ability to be swept over a wide frequency range. It uses only two components (three if you count the supply bypass capacitor): a tiny SOT-23 IC and a timing (set) resistor. They are the only oscillator ICs that can accurately generate an infinitely variable square wave without using a crystal, ceramic resonator, or external clock reference.
Viewed from the outside, the RPO is deceptively simple. Behind the curtain, though, a proprietary internal feedback loop works to maintain a linear relationship between RSET and the output frequency. Plus, there's a typical temperature coefficient of only 40 ppm/°C and stability over the supply-voltage range of 0.05%/V.
The article highlights the benefits of the RPO and discusses applications where it enhances design flexibility and performance. Some of these applications include timing circuits, synchronized power regulation (via a multiphase clock), and filter circuits. Tips on getting the most out of the RPO are provided as well.
| HIGHLIGHTS: | |
| RPO Basics | The output frequency of an RPO is set by one external resistor (RSET) and a pin-settable divider, which together provide a wide output span. It offers a number of benefits, such as infinite frequency resolution, a tiny footprint, shock immunity, and fast startup. |
| RPO Applications | The most obvious application for RPOs is the master digital clock in microprocessor-based systems. They provide a stable yet flexible clock signal, which is particularly useful if the processor must run at multiple frequencies. |
| Turning An RPO Into A VCO | A natural extension of an RPO is setting the frequency with an external voltage, thereby forming a voltage-controlled oscillator (VCO). One way to attain this is by steering a current in or out of the SET pin. |
| Instrumentation | Replacing RSET with a thermistor creates a predictable, albeit nonlinear, temperature-to-frequency generator. |
| Tips And Tricks | Paying attention to certain details when designing with an RPO can reap even more performance benefits. The "tips" addressed include "Choosing a Resistor," "Managing Jitter," "Watching the Layout," and "Using Remote Transducers." |
Full article begins on Page 2
One, zero, one, zero, one, zero…. repeat forever. Building a clock should be one of the simplest engineering design challenges. And it is–unless you need it to be small, stable, and tunable. Traditional approaches involve a 555-type timer, or perhaps a comparator with a handful of discrete components. However, these solutions consume precious board space, exhibit limited stability over voltage and temperature, and yield poor accuracy beyond a few hundred kilohertz. Although resonant-element oscillators, such as crystals and ceramic resonators, boast superior accuracy and stability, these mechanical devices are less robust than their solid-state brethren. Not only are they subject to mechanical wearout, but upon physical impact, they will cause errors in the output frequency and phase. They’re also not adjustable, and generating multiple frequencies means stocking multiple crystal values.
Recently, a new class of device emerged. It combines excellent accuracy and linearity, small footprint, low power, and the ability to be swept over a wide frequency range. These "resistor-programmable oscillator" (RPO) ICs use only two components (three if you count the supply bypass capacitor): a tiny SOT-23 IC and a timing (set) resistor. In fact, they are the only oscillator ICs that can accurately generate an infinitely variable square wave without using a crystal, ceramic resonator, or external clock reference.
RPO Basics The output frequency of an RPO is set by a single external resistor (RSET) and a pin-settable divider (÷1, ÷10, ÷100), which together provide a wide output span (Fig. 1). RSET is chosen by the simple formula shown in Figure 2.
Viewed from the outside, the circuit is deceptively simple. However, behind the curtain, a proprietary internal feedback loop works to maintain a linear relationship between RSET and the output frequency. Plus, there’s a typical temperature coefficient of only 40 ppm/°C, and stability over the supply-voltage range of 0.05%/V. Using a 0.1% resistor typically gives better than 0.6% accuracy at 25°C. Before delving into applications for this technology, it’s helpful to spotlight some of the more noteworthy benefits:
- Infinite frequency resolution: One resistor sets any frequency from 5 kHz to 33 MHz. Such flexibility makes RPOs ideal for sweeping over a broad frequency range, or providing nonstandard-value reference clocks, like those used in switched-capacitor filters. This adjustability also provides a "knob" for systems engineers to turn late in the design cycle. Alternatively, this same knob can be turned during production calibration by adjusting RSET as a final trimmer.
- Tiny footprint: The SOT-23 IC /one resistor combination lets the oscillator be placed at the "point of use" on a pc board, instead of routing fast clock signals over long distances. With a footprint as small as 9 mm2 and height under 1 mm, these circuits are typically 5 to 10 times smaller than the alternatives.
- Shock immunity: Automotive, military, and medical markets place a high value on frequency and phase stability, as well as on mechanical-component reliability. End products live in environments that are subject to prolonged vibration or high shock. RPOs have been tested to over 60,000 g without any measurable degradation in performance.
- Fast startup: RPOs power up quickly and predictably, typically settling to 1% in much less than 1 ms. This compares favorably with crystals, which can take 10 ms in the megahertz range and up to a full second below 100 kHz.
- Excellent frequency performance: While they don’t offer the precision of crystal oscillators, RPOs combine stability over temperature (40ppm/°C) and supply voltage (0.05%/V) that far exceeds RC-based oscillators—while offering the tunability that crystals can’t achieve.
- Low power: With typical supply currents pushing below 500 µA, power consumption is typically 10 to 100 times lower than competing solutions.
- Reduced stocking: Purchasing can buy one component in larger volume, rather than procure and inventory multiple crystal values. Stocking resistors is much less expensive than stocking crystals.
Currently, there are three RPOs on the market (see table, RPO Family Characteristics). The LTC1799 arrived on the market in 2001, and was followed up by the LTC6900, which cuts the power in half at the expense of frequency range. Third, there’s the LTC6902, intended for multiphase-power-supply synchronization.
Application The most obvious application for RPOs that comes to mind is the master digital clock in microprocessor-based systems. In fact, in moderate-speed microcontroller applications, RPOs provide a stable, yet flexible clock signal—especially useful if the processor must run at multiple frequencies (such as for sleep, standby, burst modes, etc.). Using I/O lines, the CPU can switch in various RSET values to program its master clock for each mode. However, limiting RPOs to this simple function ignores their impressive flexibility.Timing Circuits The broad frequency range of RPOs supports their use in a wide variety of timing circuits. Their wide span, tiny footprint, and low power, combined with near total immunity from shock, make these devices ideal building blocks to construct flexible, precise timing functions. Figure 3 shows an interval generator that’s 1% accurate over an 800-ns to 16-second range (that’s 20 × 106:1 of dynamic range!), which can be extended by using other counters. Conveniently, interval accuracy and stability of the circuit are dictated almost exclusively by the LTC1799’s programming resistor.Voltage-Controlled Oscillator A natural extension of an RPO is setting the frequency with an external voltage, thereby forming a voltage-controlled oscillator (VCO). One method of achieving this is by steering a current in or out of the SET pin. Figure 4 shows how to implement this concept by connecting a second resistor, RIN, and a ground-referenced voltage source VIN. The value of VIN will control the oscillator frequency. Excellent linearity is maintained over about a 2:1 frequency range using this technique. The modulation bandwidth of this circuit (the speed at which the LTC6900 can respond to VIN) is 25 kHz.Synchronized Power Regulation Via A Multiphase Clock In servers, communications, and large data systems, a variety of regulated supply rails are generated for the CPU, drives, I/O subsystems, and so forth. Synchronizing the timing of these switching regulators so that they run out of phase offers several benefits. Input and output ripple and capacitor stress is reduced because the current in one regulator increases while another decreases. Plus, beat notes, caused by mixing various clock frequencies and their harmonics, are eliminated. A third interesting benefit is that radiated EMI is diminished due to the smaller switching-current transients. In many applications, cutting EMI is a big deal, as it can reduce shielding-enclosure costs, reduce the need for complex ground and power planes, and lead to fewer design iterations caused by elevated readings at FCC Class B or CE testing.
The LTC6902 RPO was designed to solve this exact problem by supplying one-, two-, three-, or four-phase outputs, all derived from a common master oscillator. This device also offers another key advantage over the single-channel versions: the output can be spread-spectrum frequency modulated (SSFM) using a pseudorandom-noise (PRN) signal to spread the oscillator’s energy over a wide band.
\{\{A Brief SSFM History: Hedy Lamarr—a successful Hollywood actress (www.hedylamarr.org)—and composer George Antheil patented the SSFM concept in 1942 with the intention of cloaking radio-controlled torpedoes during WW2. While never used for that purpose, it made its way into countless military applications soon after that. Their patented "frequency-hopping" technique was based on the 88 piano keys, synchronizing them with perforated paper. This idea served as the genesis for the plethora of SSFM techniques now used in a wide variety of digital wireless communication systems.\}\}
Peak radiated EMI can be reduced up to 20 dB by properly selecting a modulation-setting resistor, which establishes the percent of down-spread modulation from 0% to 100% (although a practical range is approximately 10% to 40%). By selecting the proper percentage of modulation, the spectral profile of radiated EMI can be flattened out and spread out evenly throughout the frequency band.
Because SSFM is applied to the fundamental frequency, it should be noted that all harmonics are reduced. As part of a servo loop, the frequency "hops" are slowed via an internal 25-kHz low-pass filter. These well-controlled transitions ensure that the switching bandwidth of the regulator isn’t exceeded, and that regulation, efficiency, and load response are maintained.
Instrumentation Replacing RSET with a thermistor creates a predictable, albeit nonlinear, temperature-to-frequency generator (Fig. 5a). Thermistors with resistance values covering a wide temperature range that fit conveniently within a single divider setting of the master oscillator can be selected. By experimenting with a spreadsheet, additional series and parallel resistor values can be calculated to improve output frequency linearity for specific thermistors and temperature ranges. Even over temperature, the LTC1799 in this circuit will contribute less than ±0.5°C frequency error.
Humidity can be one of the most difficult environmental parameters to measure. Jim Williams’ novel approach uses an RPO in an AM heterodyne circuit as an interface for capacitive RH sensors (Fig. 5b). The sensor controls a variable oscillator, which is mixed with a reference frequency provided by the LTC1799 RPO. The demodulated difference frequency at the output is a 0- to 1-kHz signal corresponding to 0% to 100% RH. Using an RPO in this application makes calibration a snap. Plus, noise is minimized because the circuit allows one leg of the sensor to be grounded. The circuit contributes approximately 400 ppm/°C of error and provides a PSRR of less than 1% from 4.5 to 5.5 V.
Filter Circuits Switched-capacitor filters frequently require a nonstandard, or even tunable, reference frequency. Because the filter response will vary with this frequency, it’s critical that the source is stable over time, temperature, and voltage. RPOs are a natural fit for this application. Figure 6 illustrates a simple, precise, 60-Hz tunable notch filter. The LTC1062 and op amp provide the filtering, while the LTC1799 supplies the reference clock. Common notch frequencies are listed in the table. Not only is this circuit flexible, it also offers very high performance, with sharp slopes and over 45 dB of attenuation at the notch frequencyTips And Tricks Most circuits are simple, and are usually up and running in a few minutes. But keep in mind that the datasheets and application notes for RPOs provide a wealth of information for developers who adapt them for use in nonstandard applications–or looking to eke out the last drop of performance.
Choosing A Resistor: Because the period of the output signal has a linear relationship with RSET, it’s important to choose wisely. Errors in initial tolerance or in the temperature coefficient add 1 for 1 to the generated frequency error. Low tempco, precision metal-film resistors between 10 and 200 kÙ work well. This corresponds to a master-oscillator frequency of 0.5 to 10 MHz.
Managing Jitter: Jitter can be a problem in certain applications, especially when the frequency must be varied over a wide range. Each divider of the master oscillator has a jitter versus frequency signature. Because the frequency ranges for the dividers overlap, it’s a good idea to choose the one offering the least jitter over the required frequency span. In general, the divider should be set to obtain the lowest master oscillator frequency, as the IC will draw less power and offer better accuracy.
Watch The Layout: The set pin typically can’t tolerate lots of capacitance, so keep parasitics under 10 pF. This isn’t tough to do with a clean pc-board layout. But it can be a "gotcha" during debug if you have a poor layout, or use a high-capacitance scope probe. For best results, RSET should be placed close to the SET pin. Symptoms are inaccuracies due to supply bounce at high frequencies, and increased jitter.
Remote Transducers: In applications such as the thermistor interface discussed earlier, the transducer may be mounted at the end of a long cable, remote from the RPO. In this case, the effective capacitance can be minimized by "bootstrapping" the cable shield to the RSET voltage (Fig. 7).
Special thanks go to Doug LaPorte, Philip Karantzalis, Nello Sevastopoulos, and Jim Williams for their wisdom in preparing this article.
| Recommended Reading: |
|
Williams, J., "Instrumentation Applications for a Monolithic Oscillator," Linear Technology Corp., Application Note 93, February 2003. Sevastopoulos, N., "Using the LTC6900 Low Power SOT-23 Oscillator as a VCO," Linear Technology Corp., Design Note 293. Linear Technology Corp., LTC1799 Datasheet |