Silicon oscillator timing is better than RC

One might say that the 17th century was a barbaric and brutal age, and it was true as far as medical technology was concerned. For barbers were surgeons, and they were the ones who solved any problem, no matter how serious it was. Their solution was often surgical, and they did it without antiseptics and anesthesia. Barbers were experts in removing anything that was a hindrance to one's health: hair, teeth, appendages, vital fluids, and so on. Fortunately for us, the barber's scope of practice has shrunk to a more modest level, and now it is specialized medical professionals who solve the "really serious problems." The long history of barbers as surgeons is a good example of how technology progresses. Granted, in the world of electronics, life cycles are sometimes very short, and we don't have time to get enamored with any particular technology. However, this is not the case with the venerable RC circuit, which has been a popular timing component for decades. You only need to look at the widespread use of the 555 timer, which relies on the RC circuit and was invented nearly 40 years ago. Even in recent years, new versions of the 555 timer have continued to appear on the market. And it's not just the 555 timer, there are countless integrated devices that rely on RC circuits for timing, because RC circuits have always been the simplest, most flexible and most programmable choice. However, no matter how the RC circuit is implemented, the use of RC circuits has always been accompanied by many limitations. Now, with the emergence of a new class of timing devices based on silicon oscillator technology, the above situation has changed.

Perhaps the simplest and most common electronic circuit is a resistor and capacitor in series connected to ground. As shown in Figure 1, when a voltage is applied to the resistor, the voltage across the capacitor responds exponentially: V _C = V _R (1-e ^-t/ ), and when the resistor is connected to ground, the voltage across the capacitor responds in a similar but inverse exponential: V _C = V _INITIAL (e ^-t/ ). This simple and predictable timing response makes this circuit an ideal solution for filtering noise, slowing down fast signal edges, protecting device inputs, avoiding race conditions, and solving countless other timing problems. Even without adding resistors or capacitors to the circuit, this circuit often exists in practice due to resistance in the traces or wires.

Figure 1:

With the addition of a few components, an RC circuit with predictable charge and discharge characteristics can be used as an electronic timing component. A nice feature of the RC circuit is that it can be used to set timing sequences with both monostable and astable responses, as shown in Figure 2. In the monostable mode of operation, the flip-flop opens the switch, the capacitor charges, and when the capacitor reaches 2V, the comparator resets the output. Monostable operation is necessary to implement asynchronous timing such as starting or stopping a sequence or delaying an event. In the astable mode of operation, the feedback signal continuously changes the direction of the capacitor's charge and discharge, keeping the capacitor voltage within a fixed range (between 1V and 2V in this case). As a result, a continuous pulse train, or oscillator, can be generated as long as the circuit remains powered.

Figure 2: A simplified 555 circuit illustrating how to use an RC circuit

Programmability of frequency and time are key features of RC circuits and depend on whether the engineer can find the right combination of resistors and capacitors. There are many different sizes and types of capacitors, but there is a trade-off between accuracy, size and cost. The best tolerance is 1-2% with NP0/COG type capacitors. But because NP0/COG capacitors are very expensive when the capacity exceeds 1uF, designers may compromise with capacitors with a tolerance of 5% or worse, which is a typical value for other types of capacitors. For very small capacitance values, designers should be aware that stray capacitance or gate capacitance can cause errors. For example, in Figure 2, only a few pF of capacitance at the comparator input will cause a 1% error. In addition to these problems, there may be other capacitor error sources, such as ESR, temperature coefficient and leakage current. Faced with all these capacitor issues, it seems a good idea to integrate RC capacitors in semiconductor chips. However, because accurate semiconductor-based capacitors require a large chip area (even with very small capacitance values) and require a lot of fine-tuning, this is an expensive solution. Because of the limited range of RC value choices and the relatively high cost, this is not a common choice when implementing RC circuits, so the headaches of external capacitors may never go away.

Faced with the practical limitations of capacitors, the choice of resistors becomes more critical, but resistors are also subject to some limitations. If the resistance of the RC circuit is very small, there will be power consumption consequences because a lot of power is wasted in the resistor. For example, the RC circuit in Figure 2 draws more than 1mA of peak current, and in the astable operation (oscillator) mode, the two external resistors themselves draw an average current of 450uA1 ^. On the other hand, surface leakage and input bias currents limit the maximum resistance value. In the presence of several nanoamps of stray or bias currents, resistors greater than 10MΩ will produce appreciable errors due to these currents.

Assuming suitable resistors and capacitors are available, there is still another significant error source in the RC circuit due to the nonlinearity of the charge and discharge response curves. Any comparator threshold error is amplified by more than 2.5 times in the timing response. For example, a ±2% comparator threshold error produces a timing error of approximately ±5.4% ² . For the astable mode of operation, this problem manifests itself not only in the form of frequency error, but also in duty cycle error. Figure 3 illustrates the effect of this error source. Note that the inherent error of the exponential response curve is eliminated by the silicon oscillator, which has a linear response curve.

Figure 3: RC circuit error caused by comparator threshold variation

One of the advantages of programmability is the ability to implement voltage controlled modulation (VCO), pulse width modulation (PWM), pulse duration modulation, and other types of dynamic time or frequency modulation. Many applications require this capability: tone generation, heater control, motor control, pulse generation, etc. The need is evident in the countless web pages, books, articles, and short essays devoted to implementing such applications with 555-type or other RC circuits. Whether it is a 555 timer or other circuit, implementing timing with an RC circuit requires adjusting the comparator threshold voltage or the RC response curve. With comparator threshold adjustment come a variety of errors, as described above. The simplest method of implementing control requires the use of a potentiometer or variable capacitor to adjust the RC time constant. In fact, most implementations require a lot of additional circuitry (such as a closed-loop feedback network) to compensate for the many error sources.

In summary, as each generation of electronic products continues to demand higher accuracy, power, and size, the inherent limitations of RC circuits become more and more apparent. RC circuits generally do not operate at frequencies above 1MHz, are inaccurate, consume a lot of power, and can be more expensive than it first appears, especially when significant effort is put into enhancing functionality or improving performance.

Traditional alternatives to timing circuits require fixed frequency crystal oscillators. Adding programmability or asynchronous functionality requires additional circuitry. Whether adding programmability or asynchronous functionality with discrete components or by programming a microcontroller, simple timing applications end up being complex and inflexible. For many applications, this is definitely not an attractive option. With the advent of silicon oscillators, RC-based timers have a real competitor.

Silicon oscillators are completely solid-state devices that utilize a frequency-locked servo loop controlled by current. The frequency control current can be set with a single resistor, and their programmable architecture also allows monostable operation3 ^. Replacing RC circuits with silicon oscillators as the basic “timing engine” provides a simpler, more flexible, and more programmable timing method.

Because silicon oscillators do not rely on mechanical resonant components and are manufactured using standard CMOS technology, they are highly resistant to shock, vibration and wear. This also means that silicon oscillators can integrate other features such as frequency programmability, multi-phase outputs, spread spectrum frequency modulation and intelligent startup circuits. Silicon oscillators can operate over a frequency range of 10kHz to 170MHz, with built-in voltage dividers, and this frequency range can be extended to very low frequencies4 ^. The accuracy and power consumption performance of silicon oscillators are excellent. For example, Linear Technology's LTC6906 has an accuracy of better than 99% from 10kHz to 1MHz while drawing less than 80uA of current.

Figure 4: Silicon oscillator “engine”

Linear Technology Corporation recently introduced a new series of timing devices based on silicon oscillators, the TimerBloxTM series. These devices contain intelligent circuits that provide 5 basic timing functions: voltage-controlled oscillators (VCOs), very low-frequency clocks for long-duration timing, pulse-width modulated oscillators (PWMs), single-pulse generators, and delay components. As a series, these devices cover an operating range of 29uHz (9.5 hours) to 2MHz, with a typical frequency or time accuracy of 98% or higher. Each TimerBlox device provides 8 separate operating frequency/time ranges that can be selected with a simple resistor divider. Once the range is selected, the user can set the exact frequency or time with a single resistor from 25kΩ to 800kΩ. This architecture allows the TimerBlox series of devices to cover a wide operating range while ensuring that the effect of the resistor size on the overall accuracy or power consumption is negligible (the stray current and power consumption generated by setting the current are negligible). The TimerBlox family of devices operates over a supply voltage range of 2.25V to 5.5V, with a supply current range of 60uA to 250uA, and the TimerBlox devices offer fast, first-cycle accurate startup. The 20mA output source and sink capability allows direct drive of opto-isolators and transformers for electrical isolation. The TimerBlox family of devices is available in SOT23 and tiny DFN packages, and all devices operate over a temperature range of -40°C to 125°C. All of these features are designed to ensure that the TimerBlox family of devices offers the simplest, most flexible, and most programmable timing solution. With these devices, designers are no longer faced with the unforgiving choice of using RC circuits to implement timers or provide clocks. The role of the RC circuit can now be scaled back to a more modest level, and people are no longer expected to use RC circuits to deal with "really serious timing problems."

Figure 5: Typical application using TimerBlox device

1 A resistor current of 450uA is quite large when you consider that the CMOS version of the 555 timer draws less than 100uA of supply current.
2 It is worth noting that the comparator threshold error of a typical 555 timer is > ±5%.
3 Some Linear Technology silicon oscillators include built-in current setting control, eliminating the need for external setting resistors.
4 Linear Technology's LTC6991 provides a very low frequency clock that can operate at a 9.5 hour timed interval.