In electronic devices, some applications require highly stable AC signals, making precise frequency control crucial. Traditionally, LC oscillators have been widely used to generate such signals. However, LC oscillators have poor stability and are easily affected by factors such as ambient temperature and power supply variations, leading to frequency drift. To address this issue, a type of oscillator based on quartz crystals, namely crystal oscillators, has been developed, which can generate highly stable signals. But strangely, despite the advanced state of modern chip manufacturing technology, we rarely see crystal oscillators built into chips. So, why aren't crystal oscillators built into chips?
Practicality and cost
First, for decades, chip manufacturing processes may not have allowed for the direct integration of crystal oscillators into the chip itself. However, with technological advancements, modern chip manufacturing processes now enable the embedding of crystal oscillators. So, why not do it?
On the one hand, some applications do not require a highly stable signal frequency. For these applications, a built-in crystal oscillator increases costs without offering much practical benefit. On the other hand, even if an application requires a highly stable signal frequency, an external crystal oscillator can be provided by being packaged next to the chip or connected to the chip.
Material limitations and cost considerations
The primary material of a chip is silicon, while a crystal oscillator uses a quartz crystal. These two materials cannot be directly bonded together, but this can be achieved through encapsulation. However, this increases cost and complexity because the encapsulation process requires additional steps and materials.
Furthermore, quartz crystals are relatively expensive to manufacture. If crystal oscillators were directly integrated into each chip, regardless of the required precision and stability, it would increase the chip's cost. In contrast, treating the crystal oscillator as a separate component allows for selection of the appropriate oscillator based on specific needs, making it more economical.
Frequency flexibility and circuit design
Integrating a crystal oscillator into a chip limits its frequency. Once integrated, the frequency is fixed and cannot be changed to provide different frequencies. However, placing the crystal oscillator externally allows for the free selection of different frequencies as needed. This is crucial for applications such as systems that require operation at different frequencies or systems with adjustable frequencies.
Furthermore, treating the crystal oscillator as a separate component simplifies chip circuit design and layout. Chip design requires consideration of crystal-related wiring and pins to ensure signal transmission quality and stability. Treating the crystal oscillator as a separate component allows for more flexible chip design and a more compact layout.
Although technology has made it possible to integrate crystal oscillators into chips, in practice, they are still rarely built-in due to trade-offs in practicality, cost, material limitations, and frequency flexibility. Instead of integrating crystal oscillators, modern chips tend to offer a wide range of clock and frequency control options to meet the needs of different applications. Using an external crystal oscillator allows us to select the appropriate frequency as needed, maintaining system stability and flexibility. Therefore, the fact that crystal oscillators are not built into chips does not indicate a lack of technology, but rather a choice based on practical needs and cost-effectiveness.
Reason 1: In the past, chip manufacturing processes may not have allowed for integrating crystal oscillators into the chip itself, but this is possible now. This issue is mainly determined by practicality and cost.
Reason 2: Chips and crystal oscillators are made of different materials. Chips (integrated circuits) are made of silicon, while crystals are made of quartz (silicon dioxide). They cannot be made together, but they can be packaged together, which is already possible, but the cost is relatively high.
Reason 3: Once a crystal oscillator is packaged inside a chip, its frequency is fixed, and changing the frequency is virtually impossible. With an external crystal oscillator, however, the frequency can be freely changed to provide different frequencies to the chip. Some might argue that since the chip has an internal PLL, the crystal frequency doesn't matter; you can simply use the PLL to multiply or divide the frequency. However, this brings us back to the cost issue. If a 100MHz crystal oscillator is integrated into a chip, but I only need 10MHz, why would I buy a chip with an integrated 100MHz crystal oscillator? It's expensive and wasteful.
The "on-chip clock" we usually talk about doesn't actually mean there's no crystal oscillator on the chip, but rather an RC oscillation circuit.
As shown in the diagram, the system clock can be supplied in three ways: HSI, HSE, and PLL. If the internal clock is used as the system clock, its multiplication factor cannot reach 72MHz; the maximum multiplication factor is 8MHz/2*16 = 64MHz.
