Silicon oscillators can offer advantages over crystals

Paul Boughton

Silicon oscillators can give significant advantages to the designer. Greg Zimmer

Is high temperature, high vibration, or high acceleration challenging your circuit design? Is power or size a constraint? Lead-times? Cost? If so, you are probably already aware of the limitations imposed by crystal oscillators or ceramic resonators.
However there is a new alternative: silicon oscillators.

At the heart of most clock circuits are crystals or ceramic resonators. These tried and true components boast excellent accuracy and stability, but their downside lies in their mechanical operation. As such, they are subject to wear-out, and physical impact can induce errors in the output frequency and phase. Vibration and temperature extremes can damage them and because they rely on a tuned circuit with a matched driver, they don’t always start up as planned or oscillate at the intended frequency.

One alternative to resonator-based clocks is RC-based clocks, which rely on the time-constant set by a resistor and capacitor. These are simple and more adjustable than fixed frequency crystal or ceramic devices, but poor accuracy and stability limit their usefulness. A new class of silicon oscillators combines excellent accuracy and linearity, reliability and low power in a small footprint.

Just as vacuum tubes, relays and core memory were transformed by solid-state technology, silicon-based oscillators are displacing traditional crystal and ceramic-based oscillators. Silicon oscillators do not rely on mechanical resonating elements and are not plagued by the poor performance of RC-based circuits. Standard silicon fabrication and assembly means inherent immunity to shock, vibration and wear-out.

Start up is consistent and fast. These devices operate with a single 2.7V to 6V power supply and operate over a temperature range of -40°C to 125°C. Linear Technology’s silicon oscillators includes fixed frequency, resistor programmable and serial programmable devices, and covers the frequency range from 1kHz to 170MHz.

Fixed frequency

Fixed frequency oscillators are simple, reliable, and robust devices. No trim components are required and these devices exhibit a maximum frequency error of ±1percent at 25°C with stability of 20ppm/°C. Outstanding jitter, rise and fall time, and duty cycle provide an exceptionally ‘clean’ square-wave. Typical start up time is 100µsec and an enable pin is provided for glitch-free control over the output. A tri-state divider input allows for division of the master clock frequency (+1, +2, +4). The LTC6905-133, can generate an output frequency of 133MHz, 67MHz or 33.3MHz.

A single external resistor (RSET) and a tri-state divider input are used to set the frequency of the resistor programmable oscillators. RSET is chosen by a simple formula. Although this circuit looks simple, an internal feedback loop works to maintain a precise relationship between RSET and the output frequency. It does this with a typical temperature coefficient of 20ppm/°C and stability over the supply voltage range of 0.5percent/V. Using a 0.1percent resistor typically provides better than 0.6percent accuracy from 0° to 70°C.

Since one resistor sets any frequency in the oscillators operating range, non-standard values, such as those used in switched-capacitor filters can easily be generated. Resistor programmability allows for adjustments late in the design cycle or during production calibration as a final trimmer. Resistor programmability also offers unique functionality since the frequency is controlled via any method that sources current into the SET pin. This allows for voltage-controlled or current-controlled operation suitable for use in instrumentation.

Serial programmable oscillators offer similar functionality, but instead of using a resistor, these devices are programmed via an SPI or I2C-compatible interface. Frequency is set ‘on-the-fly’ via a 10-bitDAC, with four additional bits to set the desired range, resulting in an output span of 1kHz to 68MHz with 0.1percent resolution. These devices automatically start-up at 1kHz, at which point, the logic can reset its own frequency as needed.

For power reduction, consider the advantages of replacing a fixed frequency, power-hungry crystal device with a silicon oscillator. Frequency programmability provides the possibility for dynamic power reduction. Via serial interface or simply by toggling the oscillator divider pins, reducing the oscillator frequency will minimise power consumption across all of the clocked devices. Using dual-edge clock devices, where both clock edges are used, can also reduce power. With dual-edge clocking, the clock rate is reduced by 50percent. Since silicon oscillators provide outstanding rise and fall time and duty cycle of 50percent ±5percent, the rail-to-rail square wave output is ideal for dual-edge clocking.

Using these devices for just power reduction, however, ignores their full potential. The following examples illustrate a few novel uses for silicon oscillators.

First, a micropower time interval generator with large dynamic range and excellent stability. The circuit consists of an LTC6906 silicon oscillator, a counter and a dual flip-flop. The LTC6906 silicon oscillator has an extremely low current draw (18uA Max at 100kHz) and a fast start up, enabling a standard CMOS gate to awaken the circuit via the power supply pin. The oscillator’s frequency and the counter’s modulo set the output time interval pulse width of this circuit. As shown, the interval is programmable from 16µsecs to 1.6 seconds, and additional counters can extend this range. 100ppm stability is achievable over temperature by combining 50ppm resistors with the 50ppm stability of the LTC6906. By powering-down the clock when not needed, the current draw is reduced to the quiescent current of the CMOS flip flop and counter.

Wide range, linear VCO

Frequency can be set with Linear Technology’s resistor programmable oscillators via any method that sources current into the SET pin. This is particularly useful for the LTC6905, which has the wide frequency range of 17MHz to 170MHz. The RSET resister establishes a constant current into the SET pin, and the current through RCNTRL will subtract from this current to change the frequency. Thus, increasing VCNTRL increases the output frequency.

Replacing RSET with a resistive-based sensor, such as a thermistor, allows for direct conversion of a sensor output into frequency. By conversion to a digital signal (the CMOS clock output), measurements can be taken remotely and transmitted digitally. Using an optoisolator will electrically isolate the sensor and avoid ground loops. Adding series and parallel resistors for specific thermistors and temperature ranges can improve linearity. Because the resistor programmable oscillators have a small footprint, wide operating range, and low drift, they can be used in this application directly at the point of measurement with minimal design impact.

Greg Zimmer with Linear Technologies, Milpitas, Ca, USA.