How to Solve the Frequency Deviation of Crystal Oscillator? Disassemble the 7 Core Factors That Affect the Stability of Crystal Oscillators

Among all electronic components, the crystal oscillator is probably the most inconsistent one - it looks like a small block from the outside and does not require debugging when soldered onto a board, but its output frequency is inaccurate, which directly determines whether the entire system can operate normally. Many engineers encounter frequency deviation problems and often only blame the poor quality of the crystal oscillator, without understanding that every step from chip cutting to circuit design is quietly affecting the final output frequency. To understand how frequency is determined, we need to break down seven core influencing factors from two dimensions: innate attributes and acquired environment. Each link contains pitfalls that are easy to step on.

1. Chip thickness: innate background color of frequency

Crystal oscillators oscillate by the piezoelectric effect of quartz crystals, and the fundamental determinant of frequency is the physical thickness of the chip. This rule is as simple as drumming: the thinner the skin of the drum, the higher the tone produced; The thicker the drum skin, the lower the pitch. Crystal oscillators follow the same logic: the oscillation frequency is inversely proportional to the thickness of the chip. There is a very precise correspondence in the industry: a 40MHz ordinary crystal oscillator requires a chip thickness of 41.75 microns, which is about half the diameter of a human hair; If you want to make a 100MHz crystal oscillator, you have to grind the chip to 16.7 microns, which is thinner than plastic wrap. At this thickness, the chip is as brittle as a cicada's wing, breaking easily with cutting, polishing, and packaging, resulting in extremely low yield and naturally soaring costs. This is also the core reason why crystal oscillators below 40MHz are generally cheaper on the market, while high-frequency crystal oscillators have doubled in price.

2. Cutting angle: direction code hidden in the crystal

Quartz is an anisotropic crystal, and the frequency characteristics of crystals cut from the same quartz ore at different angles vary greatly. The most mainstream cutting method currently is AT cutting, and the temperature frequency curve of the chip cut at this angle has a very gentle inflection point near 25 ℃ room temperature, which covers the operating temperature range of the vast majority of electronic devices. The frequency drift caused by temperature changes is minimal, so more than 90% of civilian crystal oscillators use AT cutting. Another common type of BT cutting frequency temperature curve has a higher inflection point and is more suitable for high-temperature working scenarios, but in ordinary scenarios, the temperature drift is much larger than that of AT cutting, and it is now being used less and less. In addition, there is also SC cutting designed specifically for high-precision scenarios, which has better temperature stability, but the cutting process is more complex and costly, and is only used in high-end constant temperature crystal oscillators. Many times, the stability difference between two crystal oscillators with the same nominal frequency and size is several orders of magnitude, which is essentially due to the difference in cutting angle.

3. Oscillation mode: fundamental frequency or overtone, different paths result in different outcomes

What if you want a higher frequency when the thickness of the chip has been ground to the limit of the process? Engineers have come up with the idea of overtone oscillation: utilizing the harmonic characteristics of quartz crystals to oscillate the chip on odd multiples of the fundamental frequency, without the need to grind thinner chips to achieve higher frequencies. For example, a chip with a fundamental frequency of 20MHz can achieve 100MHz output with five overtones, perfectly avoiding the process difficulty of grinding thin wafers. But the overtone crystal oscillator cannot be used directly to produce the desired frequency. It must be paired with an inductor capacitor frequency selection network in the circuit to screen out the overtone frequency, otherwise the crystal oscillator will only oscillate stably at the fundamental frequency. This is also why many beginners cannot obtain the correct frequency with high-frequency overtone crystal oscillators, mostly due to the lack of a frequency selection network. Generally speaking, using a fundamental frequency crystal oscillator below 40MHz is sufficient, with low cost and simple circuit; 40MHz and above are mostly dominated by overtone crystal oscillators, which is already the industry's default rule.

4. Working temperature: external disturbance with the highest frequency

Even if the chip size and cutting angle are perfect, the biggest frequency disturbance source in actual use is still temperature. Quartz crystals undergo thermal expansion and contraction with temperature changes, and their size and frequency naturally change accordingly. This is an unavoidable physical law. The core difference between different levels of crystal oscillators lies in their ability to cope with temperature drift: ordinary civilian passive crystal oscillators have no compensation, and the temperature drift range is ± 10ppm~± 30ppm. It is not a problem to place them in a constant temperature room, but they will exceed the allowable range when exposed to large temperature differences outdoors; The temperature compensated crystal oscillator (TCXO) is equipped with temperature sensing and compensation circuits, which can adjust the frequency in real time according to temperature changes, compressing the temperature drift to ± 0.1ppm~± 2.5ppm, which is sufficient to cope with large temperature difference scenarios such as vehicle mounted and outdoor base stations; The top-level constant temperature crystal oscillator (OCXO) directly places the crystal in a constant temperature bath and maintains it at the most stable inflection point temperature of frequency. The temperature drift can reach several hundred ppb level, which is only used for extreme precision requirements such as Beidou timing and 5G core base stations.

5. Load capacitance: The final accuracy is determined between millimeters

Almost all crystal oscillator data manuals will indicate the nominal load capacitance, and many engineers feel that this is just a random parameter. In fact, when calibrating the frequency of the crystal oscillator at the factory, it is calibrated under the nominal load capacitance. If the total load capacitance of the actual circuit does not match the nominal value, the output frequency will definitely deviate: if the total load is greater than the nominal value, the frequency will shift downwards; If the total load is smaller than the nominal value, the frequency will shift upwards, and a difference of 2-3 pF can cause several ppm deviation. What's even more frustrating is that the total load capacitance of the actual circuit not only includes the two external matching capacitors, but also the stray capacitance of PCB wiring and the parasitic capacitance of chip pins. These stray capacitances often add up to 3-5 pF and are easily overlooked. Many people directly calculate the external capacitance according to the parameters given in the manual when designing circuits, but the total capacitance is too large and the frequency deviates from the specifications without finding the reason. To achieve accurate frequency, it is necessary to include stray capacitance, which is the easiest pit for many beginners to step into.

6. Drive level: If the force is not applied correctly, the frequency will float

A crystal oscillator requires an external circuit to provide driving in order to start oscillating, and the magnitude of the driving level directly affects frequency stability. If the driving level is too low, the crystal oscillator vibration amplitude is not enough, and it cannot start oscillating at all, or the frequency is unstable after starting oscillating; If the driving level is too high and the vibration amplitude of the chip is too strong, not only will the temperature of the chip rise and cause frequency drift, but long-term high-intensity vibration will also cause hidden damage to the chip, and even directly crack and scrap it. Just like pushing a swing, if the force is too weak, it won't lift, and if the force is too strong, it will directly break the rope. Only with appropriate force can it swing steadily. Different specifications of crystal oscillators have recommended driving level ranges, generally between tens of microwatts and a few milliwatts. When designing circuits, it is necessary to adjust the driving resistors according to the manual requirements, and not use just one resistor to make up the numbers. Many inexplicable frequency drift problems are caused by incorrect driving levels.

7. Long term aging: slow drift left by time

Even if all conditions are perfect, the frequency of the crystal oscillator will gradually drift over time, which is the impact of aging. During the long-term operation of a crystal oscillator, the stress inside the chip will gradually release, and the packaging material will also undergo small deformations. These changes are cumulative and irreversible, causing the frequency to drift slowly year by year. The industry uses aging rate to measure this change, usually in ppm/year. Generally speaking, the aging drift of a crystal oscillator is the largest in the first year, around ± 1ppm~± 3ppm, and the drift speed will gradually slow down afterwards. High quality and low aging rate crystal oscillators can control drift within 0.5ppm per year, and constant temperature crystal oscillators can even achieve ppb levels. For ordinary consumer electronics, this drift of three to five years has no impact at all; But for devices such as communication base stations and smart meters that have been unattended for more than ten years, the aging rate is too high and the frequency will deviate from the range in a few years. The maintenance cost will increase several times. Therefore, for long-term equipment, low aging rate crystal oscillators must be prioritized.

Overall, the frequency of a crystal oscillator is never fixed from the moment it leaves the factory. The innate size, cutting, and oscillation mode determine the basis of frequency, while the temperature, load, driving, and aging that occur later in life constantly change the final output frequency. Only by thoroughly understanding the impact of each link can we avoid pitfalls in the selection and design process, ensure the stable output of the crystal oscillator at the required frequency, and lay a reliable timing foundation for the entire system.