The advent of synchronous digital systems turned the humble oscillator into the heartbeat of modern microprocessor-based digital systems. Its thousands of applications have fostered an extremely wide range of oscillator sources and configurations using manifold resonator structures.
Still, the choice of an oscillator is often given short shrift without a good understanding of their use due to a variety of resonators, many differing internal amplifiers, and several different temperature stabilization schemes. All these affect the size, accuracy, stability, and cost of the device, as well as how they should be applied in a design.
This article will help designers better understand the operation and structure of oscillators, as well as their critical specifications and how they match up with a design’s requirements.
In the process, it will investigate output waveshape, frequency accuracy and stability, phase noise, jitter, load and temperature variation, and cost, along with how best to apply oscillators for design success.
Oscillators are electronic circuits that generate periodic waveforms at a desired frequency. A functional block diagram of a generic oscillator contains an amplifier and a feedback path with a frequency selective feedback network (Figure 1). Oscillation can be initiated and sustained if the loop gain is equal to or greater than unity at the desired oscillation frequency, while at the same time the phase shift about the loop is equal to a multiple of 2p radians. This is a positive feedback condition.
The frequency dependent network can be an inductor-capacitor (LC) or resistor-capacitor (RC) network, but precision oscillators usually employ a resonator. The choice of the resonator type is one of the specifications to be dealt with as each has its own strengths and weaknesses.
Figure 1: Functional diagram of a basic oscillator consisting of an amplifier with a frequency selective network or resonator in a positive feedback configuration. (Image source: Digi-Key Electronics)
Commonly employed resonators are quartz crystal, surface acoustic wave (SAW) filters, or microelectromechanical systems (MEMS).
When an oscillator like this is first powered on, the only signal in the circuit is noise. The element of noise at a frequency that satisfies the gain and phase states for oscillation is circulated around the circuit loop with increasing amplitude due to the circuit’s positive feedback. The signal amplitude increases until it is limited by the amplifier characteristics or by an external automatic gain control (AGC) unit. The waveshape of the oscillator output can be controlled at this point, with the common waveshape choices being sinusoidal, clipped sinusoidal, or logic (“0” or “1”) outputs. If logic outputs are chosen, then a logic family (HCMOS, TTL, ECL, LVDS…) must also be selected.
Sinusoidal outputs are primarily used in carrier and local oscillator signal generation in communications related applications where spectral purity is a key concern. The sine waveshape has significant power only at the fundamental frequency and little or no power at harmonic frequencies.
The key specification for oscillators is frequency stability which defines how well the oscillator maintains its frequency. A related specification is aging which specifies the drift in the oscillator’s frequency over a long interval, generally a year. As the speed of applications increase, the short-term variation in the oscillator’s phase becomes an important issue. This short-term variation in phase is described as the oscillator’s phase noise. The phase noise is a frequency domain specification. The equivalent time domain specification is phase jitter or time interval error.
The feedback network in the basic oscillator can be any of several resonant structures. The most common is the quartz crystal. Quartz crystal resonators use the piezoelectric effect. A small voltage applied across a crystal causes it to deform, and a force applied to the crystal produces an electric charge. This series of electromechanical interchange forms the basis of a very stable oscillator. This effect produces oscillations at specific frequencies related to the type of crystal, the geometric orientation at which the crystal is cut, and its dimensions.
Crystals are held between two electrodes which form the input and output of the crystal resonator. Under these conditions, the crystal acts like a highly selective LC circuit (Figure 2). Observe that the crystal in its holder is represented by a series RLC circuit, which represents its series resonant frequency dominated by the model components LS and CS. The parallel connected capacitor represents the capacitance of the holder and associated wiring. The parallel capacitance CP reacts with the series inductance LS resulting in a parallel resonant frequency. In operation, the series resonance dominates the resonator operation. Crystal fundamental frequencies range from the kilohertz (kHz) range to about 200 megahertz (MHz).
Figure 2: The equivalent circuit model for a quartz crystal. Model component LS and CS determine the series resonant frequency, while LS, CS, and CP determine the parallel resonance. (Image source: Digi-Key Electronics)
Another common resonator is the surface acoustic wave (SAW) device (Figure 3).
Figure 3: A SAW filter/resonator uses interdigital transducers mounted on a piezoelectric substrate to generate surface acoustic waves across the gap between transducers, producing a frequency dependent response at the output. (Image source: Digi-Key Electronics)
The SAW filter is a frequency selective device which uses a surface acoustic wave propagated along the surface of an elastic substrate. SAW’s are generated and detected using interdigital transducers (IDT) formed by conductive paths on the substrate as shown in the figure. SAW filters/resonators operate with a frequency range from 10 MHz to 2 gigahertz (GHz). The frequency is dependent on the dimension of the IDT elements and the characteristics of the substrate material. The circuit models for a SAW device are like those of a quartz crystal. SAW resonators can be manufactured economically using photolithography in small packages at low cost. These oscillators are referred to as SAW oscillators or SOs.
The final resonator technology to be discussed in this article is based on microelectromechanical systems (MEMS). MEMS use standard semiconductor manufacturing processes to produce miniature mechanical elements. The sizes of these devices can vary from microns to millimeters. Resonators which are akin to high frequency tuning forks are designed to vibrate under electrostatic excitation. These resonators’ die structures are combined with a programmable oscillator/controller IC (Figure 4).
Figure 4: A MEMS oscillator module combines a MEMS mechanical structure with an oscillator/controller IC in a single package. (Image source: SiTime)
The oscillator/driver excites the MEMS structure and feeds its output to a fractional-N phase locked loop (PLL) which multiplies the output frequency of the MEMS device by a programmable factor ‘N’. The one-time-programmable (OTP) memory stores the module configuration parameters. Temperature compensation is achieved by adjusting the output frequency within the PLL. The PLL can also be programmed, giving the oscillator a digitally controlled frequency output.
The greatest advantage of the MEMS oscillator is its immunity to mechanical shock and vibration. This is an important factor in mobile applications like cell phones, cameras, and watches.
Oscillator circuit types
The circuit topology of modular oscillators has been developing over many decades and there are currently many available technologies. In almost every case the circuit improvements have been made to improve the accuracy and stability of the oscillator’s output frequency. Examples seen in the previous paragraph included the non-quartz-based SAW and MEMS oscillators. The techniques applied to quartz oscillators can also be applied to any type of oscillator. These oscillators are all rated to operate into a 15 picofarad (pF) load capacitance. Variations in load capacitance affect the operating frequency.
The basis of comparison for these topologies is the bare quartz crystal oscillator (XO) (Figure 5). This example is implemented using logic gates and includes a varactor diode to permit tuning. These simple oscillators exhibit frequency stability on the order of 20 to 100 parts per million (ppm).
Figure 5: A basic crystal oscillator implemented using logic inverters includes a provision for voltage control via a varactor diode in series with the quartz crystal. (Image source: Digi-Key Electronics)
The Abracon ASV-10.000MHZ-LCS-T is a surface mount crystal clock oscillator. It has a digital output with HCMOS logic level. Oscillators of this type have the chief advantage of low cost. Its frequency stability is ±50 ppm, but other devices in this family of oscillators have stability specifications from 20 to 100 ppm. The primary source of frequency drift is temperature change. Another source is crystal aging or frequency change over time. Aging rates are proportional to the basic stability. In the case of this oscillator, the aging rate is ±5 ppm per year. XO’s match general purpose applications which do not require high frequency stability. Such applications include a clock source for a microprocessor.
The temperature compensated crystal oscillator or ‘TCXO’ adds circuit elements to compensate for temperature related variations of the quartz resonator and amplifier (Figure 6).
Figure 6: The quartz resonator and amplifier are sensitive to temperature so the TCXO adds a temperature sensor and a temperature compensation network to correct frequency drift. (Image source: Digi-Key Electronics)
A temperature sensor such as a thermistor is used to develop a correction voltage, which is applied through a suitable network to a voltage variable varactor diode in series with the crystal to control the frequency. This works by changing the capacitive loading of the quartz crystal. A twenty times improvement, or more, in frequency stability can be achieved with temperature compensation.
The Abracon ASTX-H12-10.000MHZ-T is a typical TCXO with an HCMOS output level and a frequency stability specification of ±2 ppm. The cost is about three times that of a basic XO.
Another approach to temperature stabilization is to enclose the oscillator module in a temperature controlled oven (Figure 7). This topology is called the oven controlled crystal oscillator (OCXO).
Figure 7: The OCXO stabilizes the temperature of the oscillator by enclosing it in an oven with a temperature set to match the temperature where the crystal’s frequency vs. temperature curve has a slope of zero. (Image source: Digi-Key Electronics)
The crystal oscillator is enclosed in a temperature controlled oven. The oven temperature is set to a value where the crystal’s frequency vs. temperature curve has zero slope so that small changes in temperature result in little or no change in the oscillator’s frequency. The OCXO can improve the oscillator stability by over a thousand times. Oscillators like this are needed in applications requiring precise timing such as navigation systems or high-speed serial data communications.
The Connor-Winfield DOC050F-010.0M is an OCXO with LVCMOS output levels. It has a specified frequency stability of ±0.05 ppm. This improved performance comes with a higher power consumption due to the oven, larger size, and cost (about 30 to 40 times that of a XO) relative to the basic crystal oscillator.
The MEMS oscillator discussed previously is an example of the digitally controlled oscillator (DCXO).
The SiTime SIT3907AC-23-18NH-12.000000X is a MEMS-based DCXO with an LVCMOS logic output and a 10 ppm frequency stability. It features the ability to program a change in frequency using its internal PLL with “pull” ranges of ±25 to ±1600 ppm.
A microcomputer-controlled crystal oscillator (MCXO) features frequency stability equal to that of the OCXO with a smaller package at lower power requirements. MCXOs stabilize their output frequencies using either of two methods. The first is to have the source oscillator operate at a frequency higher than the desired output and use pulse deletion to achieve the desired output frequency. The second method is to operate the internal source oscillator at slightly below the desired output frequency, and add a correction frequency generated by an internal direct digital synthesizer (DDS) to the source output frequency.
The IQD Frequency Products LFMCXO064078BULK is an MCXO that is HCMOS compatible with a frequency stability of 0.05 ppm. The product family includes oscillators at key fixed frequencies between 10 and 50 MHz. Its physical volume is only 88 mm3 and it requires only 10 milliamperes (mA) at 3.3 volts, for a total power consumption of 33 milliwatts (mW).
Some applications require that the frequency of an oscillator be adjusted. This can be done either digitally or via analog control. Analog control is accomplished using a voltage-controlled crystal oscillator (VCXO). Figure 5 showed how an oscillator can be tuned by applying a voltage to a varactor diode in series with the resonator and shifting its frequency by changing the load capacitance. This is the underlying principle of the VCXO.
The Integrated Device Technology Inc.'s XLH53V010.000000I is an example of a VCXO supplying HCMOS output levels and a frequency stability of ±50 ppm. The pull range of a VCXO indicates the maximum frequency offset that can be achieved by varying control voltage. This oscillator has a pull range of ±50 ppm. For the nominal output frequency of 10 MHz the pull range is ±500 Hz.
The SAW oscillator described in the section on resonators is another low-cost oscillator characterized by high reliability. The EPSON XG-1000CA 100.0000M-EBL3 is an example of an SO. These devices are used in fixed frequency applications such as remote control transmitters. They offer good stability and jitter specifications, but the greatest benefit is reliability.
Matching oscillators to applications
In general, applications using oscillators as a precision time base require devices with better frequency stability. As such, GPS-related applications are well matched with OCXO or MCXO-based oscillators. Where isolation from shock and vibration is a requirement, an SO oscillator is the best match for the application. Clocking high-speed serial interfaces requires low timing jitter. Cost is a factor in all designs and generally varies with the degree of frequency stability being offered. Other factors such as size or power requirements are device dependent based on the technology used. These may require engineering trade-offs. A comparison of the key specifications of the oscillators discussed in this article to help focus on their individual features and benefits is shown in Table 1.
- Calculated estimate from phase noise
- Start-up /steady state
Table 1: Typical parameters under which to compare various oscillators. Each is chosen based on the design requirements and other factors such as cost and availability at time of design. (Table source: Digi-Key Electronics)
The oscillators in the table are ordered by frequency stability. Note that specific output frequencies were used in the article, but that all these oscillators offer a range of output frequencies within each model series.
A good understanding of oscillator construction and operation goes a long way toward helping designers zero-in on the right one for their application’s requirements. As always, the choice of the oscillator to use for a design project will involve engineering trade-offs involving cost, power, space, stability, and accuracy, but the variety of oscillators now available minimizes those tradeoffs with off-the-shelf solutions.