Reference Clock Design for Handheld Communication Equipment

In communication handheld devices, the contribution of the reference clock is like the role of the heart in the human body. The slightest difference will lead to system function disorder. The reason why it is defined as a reference is that whether such products can work properly depends entirely on the accuracy of the clock; and once there is an error, the error will intervene in the working frequency band of the application end as the frequency multiplication increases, affecting both the digital and analog parts of the baseband and the up-conversion and down-conversion of the RF. Usually the central frequency point used by the reference clock is 10MHz ~ 30MHz, and most of them use 13MHz, 20MHz, or 26MHz, especially when the RF and baseband share the reference clock, 13MHz and 26MHz are the most common.

Why use 13MHz and 26MHz as the reference clock? Objectively speaking, for RF, the frequency resources of handheld device traffic communication are currently mainly concentrated in the frequency range centered on 1GHz and 2GHz. From the perspective of anti-interference and harmonic suppression, it is required that the frequency multiplication of the reference clock should fall as little as possible in the intermediate frequency and high frequency range involved in these frequency bands. Another objective reason is that the physical properties of the quartz crystal used to generate the frequency determine the selection range of the reference clock. In addition, subjectively, these frequency points are close to the fuzzy range of high frequency and low frequency, which makes it relatively easy to generate other frequencies based on the reference clock frequency. It is worth mentioning that the use of this range of frequencies as the reference clock is a compromise from the perspective of power consumption.

Quartz crystal oscillator circuit

Physical properties of quartz crystal

The importance of the reference clock was mentioned earlier. To generate such a precise frequency, the selection of its basic components is crucial. After more than a century of industrial experience, quartz crystal has finally become the most ideal vibration source device. Quartz crystal was first used in watches. Its main component is SiO2, which is grown from single crystals. The lattice is neatly arranged and is a good piezoelectric material. At present, the use of artificial single crystal quartz crystals is more than 3,000 tons per year, and its use scale is second only to silicon. Quartz crystal is mainly used because its physical properties meet the requirements of reference clock, and quartz crystal is currently the only material with the following properties:

· Piezoelectric effect;
· Stress compensation and zero temperature coefficient cutting;
· Low loss and high quality factor Q;
· Simple manufacturing process, hard but not brittle, insensitive to conditions other than fluoride and high alkalinity;
· Abundant reserves in nature, can be grown into polycrystalline, and purity is easy to control.

Piezoelectric effect

The piezoelectric effect was discovered by Jacques and Pierre Curie in 1880 and is one of the most important physical effects applied to sensing and control disciplines. When external stress is applied, some crystals with special structures can generate a voltage difference. Conversely, under the action of an external electric field, the crystal can produce elastic deformation. The piezoelectric effect is an important physical process that converts stress into an electrical signal, as shown in Figure 1.

Figure 1. Principle of piezoelectric effect

The rate of change of vibration is the frequency we care about, which is determined by the orientation, size and shape of the source crystal cut and the degree of polishing. The final adjustment and determination of the center frequency is achieved by plating a layer of atomic-level gold on the surface of the crystal to achieve vibration stability.

Equivalent model of quartz crystal oscillation circuit

The physical constants of the quartz crystal determine C0, C1 (Motional Capacitor), L1 (Mass) and R1 (Bulk Loss) shown in the equivalent circuit Figures 2 and 3, where the other part of C0 comes from the electrode, fixture and guide wire.

Figure 2 Equivalent physical model of a quartz crystal, where C0 includes the capacitance effects of the holder and lead connections

Figure 3 Equivalent circuit of a quartz crystal resonator

Reference clock circuit design
The main issues that need to be considered when designing a circuit are: how to make the circuit oscillate, how to maintain the oscillation, and how to compensate for errors caused by environmental changes.

Figure 4. Series connection of oscillator circuit

Figure 5. Parallel connection of oscillator circuits

Figure 6. The process of converting a model into a circuit.

Figure 7. Frequency response of crystal near the resonance point

The piezoelectric effect makes it possible to generate oscillation. However, it is not difficult to find from the conversion model of Figure 6 that the oscillation generated by one excitation will soon disappear through the attenuation of impedance Zl, which requires the addition of a negative resistor "-Zl" to offset the consumption in the circuit or to continuously provide energy to the oscillator, that is, the amplifier shown in the two circuit forms of Figures 4 and 5. Because the oscillation circuit design of the reference clock always tries to use the self-excited oscillation of the circuit, as long as a positive feedback circuit is used, the circuit oscillation can be maintained at a specific frequency point fA, and the expected theoretical effect is shown in Figure 7. In

order to solve the temperature drift of the oscillator, it is necessary to introduce a temperature sensor. Here, the temperature compensation circuit is mainly composed of the characteristic that thermistor changes its resistance value with temperature. By changing the resonance point of the RC circuit, the entire circuit is adjusted to work at the desired frequency center, but the frequency adjustment direction of the RC circuit must be opposite to the temperature drift trend of the oscillator.

The simulation model and parameters of the circuit are listed in detail in Figure 8, and the functions of each part are shown in the figure. The initialization values ​​of AFC and Ref_Cal (reference clock calibration) in the circuit are used to determine the default values ​​of the crystal start-up frequency center. The buffer between the RF output (RF_Out) and the baseband output (BB_Out) is mainly used to adjust the level of the baseband output and isolate the mutual interference between the RF and BB sides. The temperature compensated TCXVCO is the integrated circuit implementation of this model, that is, the module design of the reference clock.

Figure 8 Reference clock simulation circuit and module description

It is worth mentioning that the module setting sequence in the circuit is not limited to that shown in Figure 8. The positions of the varactor, temperature compensation circuit and crystal itself can be adjusted based on experience. For example, the varactor can be designed on the other side of the crystal as part of the feedback circuit to form different application circuits.

Test and calibration of reference clock

The parameters that need to be tested for the reference clock mainly include: stabilization time, harmonic amplitude, waveform duty cycle, frequency drift relationship with temperature, frequency relationship with system output power change, AFC automatic frequency control linearity, controller DAC response to frequency error and reference clock module power consumption.

Stabilization time refers to the time taken from the crystal to start oscillating until a certain range of stable output, that is, to reach the specified ppm accuracy range according to the design index requirements. It is a key parameter to measure the success of the circuit design. In addition to the communication tester, MDA (Modulation Domain Analyzer) is also used in the test, such as Agilent53310A. When collecting frequency signals, a high-impedance test head (Probe) must be selected. The test of harmonic amplitude should focus on verifying the frequency points that fall within the receiving and transmitting frequency bands and the frequency synthesizer band. The relationship between frequency drift with temperature and the relationship between frequency and system output power change are mainly used to analyze frequency stability. Although the linearity of AFC automatic frequency control and the response of controller DAC to frequency error test the same physical quantity, they have different focuses. The former is the response speed to frequency changes, and the latter indicates the ability to follow and calibrate the frequency. The power consumption of the reference clock module is also becoming more and more the design focus of low-power devices. In addition to reducing the power consumption of the reference clock module itself, it is also necessary to use a relatively low clock to replace the reference clock when the system enters the power saving mode, so that the system can have the opportunity to flexibly switch the reference clock according to the system message processing volume, so as to achieve the purpose of energy saving.

Issues related to the reference clock

The radiation interference of the reference clock mainly affects the performance of the RF. As we all know, there is no fixed formula for dealing with RF problems, and it depends largely on experience accumulation. As the core module of the RF part, the reference clock has many problems that are inextricably linked to it. The most common problems can be summarized as: frequency error, network synchronization error, training sequence loss, phase error, reduced sensitivity caused by phase noise and frequency error, and harmonic interference.

The main and most fundamental way to solve such problems related to the reference clock is to do a good job of wiring the reference clock on the printed circuit board PCB. If conditions permit, it is best to design a separate shielding cover for this part of the circuit.

Wiring skills for reference clock circuits

With the improvement of chip integration, the workload of adjusting circuit performance by adjusting the values ​​of discrete devices in the circuit has been greatly reduced. Instead, it is replaced by meticulous circuit wiring work. The initiative of RF engineers often depends on the quality of wiring. Before wiring, you should carefully consider the placement of the device, actively communicate with the structural engineer, and various attempts and discussions under the conditions allowed by the organization are essential. When wiring,

all routing lines should be prioritized from a global perspective, and the reference clock line should be given priority first. The route from the reference clock output to the device pin should be as short as possible. For long-distance clock lines, double the line spacing can be used as a protective ground when necessary. There should be no traces in the same layer or adjacent layers that are parallel to the clock line in close proximity. In particular, the position between the reference clock line and the power line of the power amplifier and the logic control power line of the RF unit should be properly handled.

Careful consideration should be given to laying the ground at the location of the crystal, and a compromise should be made between parasitic capacitance and heat dissipation effect for the design of the reference clock separation device and the reference clock module. Practice has shown that parasitic capacitance in the reference clock line should be avoided and eliminated as much as possible; the most taboo in the ground wiring of the reference clock line is to interconnect with the ground of the shielded room or the ground of the phase-locked loop before reaching the main ground, and to avoid the existence of island-type ground as much as possible. If possible, a single point to the main ground method should be preferred.

Production calibration

From the perspective of the values ​​that need to be calibrated, the production calibration of the reference clock only needs to meet the frequency error requirements. However, the influence of temperature on this part of the circuit must be considered during the production calibration process. According to the arrangement of the production process, a temperature compensation coefficient needs to be added to the calibration algorithm, and the statistical selection of this coefficient must cover the effect of the change in the temperature recovery time enjoyed after reflow soldering to the start of calibration in actual production operations.