Oscillators are fundamental circuit building blocks. A substantial percentage of electronic apparatus use oscillators as timekeeping references, as clock sources, for excitation, and for other tasks. The most obvious oscillator application is a clock source in digital systems. A second area is instrumentation. Transducer circuitry, carrier-based amplifiers, sine-wave generators, filters, interval generators, and data converters all use forms of oscillators. Although various techniques are common, a simply applied, broadly tunable oscillator with good accuracy widens the design possibilities of many of these circuits.

Commonly employed oscillators are resonant-element based or RC types (Table 1). Quartz crystals and ceramic resonators offer high initial accuracy and low drift—particularly quartz types—but are essentially not tunable over any significant range. Typical RC types have lower initial accuracy and increased drift but are easily tunable over broad ranges. A problem with conventional RC oscillators is that considerable design effort is necessary to achieve good specifications. The LTC1799 is also an RC type, but its accuracy and drift specifications fit between resonator-based types and typical RC oscillators (see sidebar "A simple, high-performance oscillator"). The device's combination of simplicity, broad tuning range, and good accuracy invites use in instrumentation circuitry, as in the following examples.

Using a platinum RTD for R_SET of the LTC1799 oscillator results in a highly predictable OSC₁ output period versus temperature (Figure 1a). A series of counters scale the OSC₁ output and present the resultant signal to a clocked, period-determining logic network that delivers digital output data. Over a sensed-temperature range of 0 to 100°C, the circuit delivers 1000 counts with accuracy inside 1°C. You can extend the circuit's range for sensor limits of –50 to +400°C by using a monitoring processor to implement linearity correction in accordance with sensor characteristics. Linearity deviation over –50 to +400°C is several degrees (Reference 1).

If the RTD is at the end of a cable, IC₁ should drive the cable shield, as the figure shows. This action bootstraps the cable shield to the same potential as R_SET, eliminating jitter-inducing capacitive-loading effects at the R_SET node. The R_SET node of the LTC1799 is not unduly sensitive but does require management of stray capacitance (also see sidebar).

Waveforms show the circuit's operation (Figure 1b). The RTD determines OSC₁'s output (Trace A), which the circuit divides by 100 and produces in square-wave form (Trace B). The logic network combines with OSC₂'s fixed frequency to digitize the period measurement, which appears as output data bursts (Trace C). The logic also produces a reset output (Trace D), facilitating synchronization of monitoring logic.

The accuracy is approximately 1.5°C, primarily due to the initial error of the LTC1799. Obtaining accuracy inside 1°C involves simulating a temperature of 100°C, which is equivalent to 13,850Ω, at the sensor terminals and trimming R_SET for the appropriate output. A precision resistor decade box, such as the ESI DB62 (www.esi.com), allows convenient calibration.

A simple circuit also directly converts temperature to digital data (Figure 2a). In this case, a thermistor sensor, R_T, biases the R_SET pin. The LTC1799 frequency output is predictable although nonlinear. The inverse R_SET-versus-frequency relationship combines with the thermistor's nonlinear characteristic to give Figure 2b's data. The curve is nonlinear although tightly controlled.

An alternative circuit, which builds on the previous approach, galvanically isolates the thermistor from the circuit's power-and data-output ports (Figure 3a). The 3500V breakdown barrier between the thermistor and the power- and data-output ports permits operation at high common-mode voltages, which are common in industrial-measurement situations.

The pulse generator comprising IC₁ and associated components runs at approximately 10 kHz and produces a 2.5-µsec-wide output Trace A (Figure 3b). Q₁ and Q₂ provide power gain, driving T₁. (Trace B is Q₂'s collector.) T₁'s secondary responds by charging the 100-μF capacitor to a dc level via the 1N5817 rectifier. The capacitor powers OSC₁, which oscillates at the sensor-determined frequency. OSC₁'s output, which the circuit differentiates to conserve power, switches Q₄. Q₄, in turn, drives T₁'s secondary. T₁'s primary receives Q₄'s signal, and Q₃ amplifies this signal, producing the circuit's data output (Trace C). Q₃'s collector also lightly modulates IC₁'s negative input (Trace D), which synchronizes T₁'s primary drive to the data output. IC₂ prevents erratic circuit operation below 4.5V by removing Q₁'s drive.

The pulse generator's clocking, while maintaining OSC₁'s isolated dc power supply, generates periodic cessations in the frequency-coded output. You can use these interruptions as markers to control the operation of monitoring logic. Thermistor characteristics determine the output frequency as the Table 2 shows.

You can also design a circuit to convert the varying capacitance of a linearly responding RH (relative-humidity) sensor to a frequency output (Figure 4a). The 0-Hz to 1-kHz output corresponds to 0 to 100% sensed RH. Circuit accuracy is 2%, plus an additional tolerance dictated by the selected sensor's grade. Circuit temperature coefficient is 400 ppm/°C, and PSRR (power-supply rejection ratio) is less than 1% over 4.5 to 5.5V. Additionally, one sensor terminal attaches to ground, which is often beneficial for noise rejection.

This circuit is basically a heterodyne circuit. The circuit mixes two oscillators—one variable and one fixed—to produce sum and difference frequencies. The capacitive humidity sensor controls the variable oscillator. The demodulated difference frequency is the output. (Other examples have applied heterodyne techniques, which you usually associate with communications circuitry, to instrumentation. This circuit's operation is adapted from approaches in references 2, 3, and 4.) The heterodyne frequency-subtraction approach permits a sensed 0% RH to give a 0-Hz output, even though sensor capacitance is not zero at an RH of 0%.

IC₁ and associated components comprise a sensor-controlled variable oscillator that runs between the indicated output frequencies for the noted RH-sensor excursion. The RH sensor is ac-coupled in accordance with its manufacturer's data sheet; dc coupling introduces destructive electromigration effects (Reference 5). You use the RH trim to tune reference oscillator OSC₁ to IC₁'s nominal 0% RH-dictated frequency. The circuit mixes the two oscillators at Q₁'s base, Trace A (Figure 4b). Q₁ amplifies the mixed-frequency components, although collector filtering attenuates the sum frequency. The RH-determined difference frequency, appearing as a sine wave at Q₁'s collector (Trace B), remains. The circuit filters this waveform and ac-couples the result to zero-crossing detector IC₂. Hysteretic ac feedback at IC₂'s input (Trace C) produces a clean IC₂ output (Trace D). Counter-based scaling at IC₂'s output combines with slight sensor padding, via the 2-pF value across the sensor, to provide numeric output-frequency correspondence to RH. Calibration involves simulating the RH sensor's 25% value and trimming OSC₁ for a 250-Hz output. You can build the simulated value using known discrete capacitors or simply dial out the value using a precision, variable air capacitor (General Radio 1422D).

When evaluating the circuit's operation, it is useful to consider that the sensor oscillator's frequency changes inversely with sensor capacitance; oscillator period is linear versus sensor capacitance. This relationship would normally corrupt the desired linear output relationship between frequency and RH. Practically, because the sensor's excursion range is small compared with its 0% RH value, the error is similarly small. This term almost entirely accounts for the circuit's stated 2% accuracy.

A circuit that digitizes the capacitive humidity sensor's output also has better specifications than the previous circuit (Figure 5a). Circuit accuracy is 0.3%, plus the selected sensor grade's tolerance. Temperature coefficient is approximately 300 ppm/°C, and PSRR is 0.25% for 5V±0.5V. Compromises include a floating sensor and somewhat more complex circuitry.

OSC₁ (Trace A, Figure 5b) clocks an LTC1043 switch-array-based charge pump. This configuration alternately connects the ac-coupled RH sensor to a 4V-reference-derived potential and then discharges the potential into IC₁'s summing point. IC₁, an integrator, responds with a ramping output (Trace B). When IC₁'s output exceeds IC₂'s negative input voltage, IC₂'s Q output (Trace C) goes high, triggering Q₁ and resetting the ramp. AC feedback to IC₂'s negative input (Trace D) ensures long enough on-time for Q₁ to completely reset the ramp. This action's repetition rate depends on the RH sensor's value. OSC₁'s output path to IC₂'s latch input synchronizes the IC₁-IC₂ loop to the charge pump's clocking. In theory, if the charge pump, the offset term (25% trim current), and the ramp amplitude tie to the same potential, this circuit does not require a voltage reference. In practice, the sensor's extremely small capacitance shifts magnify the effect of charge-pump errors versus the supply, necessitating powering the LTC1043 from the 4V reference, which effectively ties all of these points to the 4V reference. Note that the 5V-powered OSC₁ output requires level shifting to drive the LTC1043.

A trimmed dc-offset current, via the 100-kΩ potentiometer, into IC₁'s summing junction compensates the RH sensor's offset term (without compensation 0% RH≠0 pF). The 20-kΩ trim at IC₂ scales the output frequency so that 0 to 100% RH equates to a range of 0 to 1 kHz. Trimming involves substituting capacitance for the sensor's known 100 and 25% values and trimming the appropriate adjustments. The adjustments are somewhat interactive, necessitating repetition until convergence occurs. A precision variable capacitor (General Radio type 1422D) is invaluable in this regard, although you can achieve acceptable results with built-up calibrated discrete capacitors.

Another RH digitizer features 1% accuracy, PSRR of 1% over 4.5 to 5.5V, temperature coefficient of 350 ppm/°C, and a ground-referred sensor (Figure 6a). Additionally, the circuit's trim scheme accommodates RH sensors with a wide tolerance grade. The circuit is basically a time-domain bridge; it subtracts time intervals representing sensor and sensor-offset values to determine the sensor value extrapolated to RH=0%. This measurement is digitized and scaled so that 0 to 100 counts corresponds to 0 to 100% RH at the output.

OSC₁'s nominal 12.77-MHz output, which the circuit conditions using a counter chain and an inverter-configured gate, presents a 12.4-kHz, 2.5-μsec pulse, Trace A (Figure 6b), to Q_1A and Q_1B. The transistors' collectors fall to 0V. (Trace B is Q_1A's collector, Trace C is Q_1B's collector.) When the base drive ceases, both collectors ramp toward 5V. The slope of Trace B's ramp varies with the RH sensor's capacitance; the slope of Trace C's ramp represents the sensor's offset value (0% RH≠0 pF). IC₁ and IC₂ switch when their associated ramp inputs cross the comparators' common dc-input potential. The comparator outputs (Trace D=IC₁, Trace E=IC₂) define a "both-high" time region proportional to the ramp slopes' difference and, hence, an offset-corrected version of the sensor's value. The circuit gates this time interval with OSC₁'s output to provide the data output (Trace F).

Circuit operation is fairly straightforward, although some details bear mention. Q₁, a dual transistor, promotes cancellation of the individual transistors' V_CE-versus-temperature terms, minimizing their error contribution. Q₁'s transistor is a two-die type to minimize crosstalk; do not substitute monolithic types. Similarly, do not substitute a dual comparator for the single types of IC₁ and IC₂. Also, the comparators operate at high source impedance relative to their input characteristics, but symmetry provides adequate error cancellation. Finally, the 5.6-kΩ resistor combines with the output gates' input capacitances, forming a lag of approximately 20 nsec. This delay prevents false output-data transients when the ramps are resetting.

The trimming procedure is similar to that of the previous RH circuit. Trimming involves substituting capacitance for the sensor's known 100 and 25% values and trimming the indicated adjustments. The adjustments are somewhat interactive, necessitating repetition until convergence occurs. As with the previous circuits, a precision variable capacitor (General Radio type 1422D) is invaluable for this work, although acceptable results are possible with calibrated discrete-capacitor assemblies.