9 standards for RF in PCB design

2) For a mixed-signal PCB, the RF section and the analog section should be kept away from the digital-to-digital section (this distance is usually above 2 cm, at least 1 cm is guaranteed), and the grounding of the digital section should be separated from the RF section. It is strictly forbidden to use the switching power supply to directly supply power to the RF part. The main reason is that the ripple of the switching power supply modulates the signal of the RF section. This modulation tends to severely damage the RF signal, leading to fatal results. Normally, the output of the switching power supply can pass through a large choke, and a π filter, followed by a linearly regulated low-noise LDO (Micrel's MIC5207, MIC5265 series, for high-voltage, high-power RF circuits, It is conceivable to use a LM1085, LM1083, etc.) to obtain a power supply to the RF circuit.

3) In the RF PCB, the components should be arranged closely to ensure the shortest connection between the components. For the ADF4360-7 circuit, the distance between the VCO inductor on the pin-9 and pin-10 pins and the ADF4360 chip should be as short as possible to ensure that the series inductance of the distribution between the inductor and the chip is minimized. For the ground (GND) pin of each RF device on the board, including the resistor, capacitor, inductor and ground (GND) pin, the hole and ground should be made as close as possible to the pin (second Layer) connected.

4) Use surface mount devices whenever possible when selecting components to operate in high frequency environments. This is because the surface mount components are generally small in size and the components are short. This minimizes the effects of additional parameters due to component leads and internal traces. Especially for discrete resistors, capacitors, and inductive components, using a smaller package (0603\0402) is very helpful in improving the stability and consistency of the circuit;

5) Active devices operating in high-frequency environments often have more than one power supply pin. At this time, be sure to set a separate decoupling capacitor near the pin of each power supply (about 1mm). The capacitance is 100nF. about. In the case where board space allows, it is recommended to use two decoupling capacitors per pin with capacitance values of 1nF and 100nF respectively. Ceramic capacitors made of X5R or X7R are generally used. For the same RF active device, different power pins may supply different functional parts of the device (chip), and each functional part of the chip may operate at a different frequency. For example, the ADF4360 has three power pins that supply the on-chip VCO, PFD, and digital sections. These three parts implement completely different functions and work at different frequencies. Once the low-frequency noise in the digital portion is transmitted to the VCO section through the power trace, the VCO output frequency may be modulated by this noise, causing spurs that are difficult to eliminate. To prevent this from happening, the power supply pins of each functional part of the active RF device must be connected together via an inductive magnetic bead (around 10uH) in addition to a separate decoupling capacitor. This design is very advantageous for the isolation performance of the active mixers LO-RF, LO-IF including LO buffer amplification and RF buffer amplification.

6) For the feeding and feeding of RF signals on the PCB, special RF coaxial connectors must be used. The most commonly used one is the SMA type connector. For SMA connectors, they are divided into in-line and microstrip. For signals with frequencies below 3 GHz, and the power of the signal is not large, and we do not care about the weak insertion loss, we can use the in-line SMA connector. If the frequency of the signal is further increased, we need to carefully select the RF connection cable and the RF connector. At this point, the in-line SMA connector may cause a relatively large signal insertion loss due to its structure (mainly turning). At this point, a microstrip SMA connector of good quality (the key is the PTFE insulator material used in the connector) can be used to solve the problem. Similarly, if your frequency is not high, but you are looking for indicators such as insertion loss and power, you can also consider the microstrip SMA connector. In addition, the small RF connector also has SMB, SMC and other models. For SMB connectors, this type of connector generally only supports signal transmission below 2 GHz, and the snap structure of the SMB connector will appear in high vibration occasions. "Flash off" situation. So be careful when choosing an SMB connector. Most RF connectors have 500 insertion and removal limits. Inserting and removing too often may permanently damage the connector, so do not screw the RF connector when debugging the RF circuit. Since the part of the SMB's PCB holder is a pin-type structure (male), the connector loss of the frequent plug-and-pull soldering on one end of the PCB is relatively small, which reduces the difficulty of maintenance, so in this case the SMB connector is also a kind. A good choice. In addition, for those places where space is extremely demanding, there are also miniature connectors such as GDR. For those digital signals such as high-frequency clocks, low-jitter clocks, high-speed serial signals, such as high-frequency clocks, low-frequency, small-signal, precision DC, etc., or other digital signals, SMA can be used as a feed-through connector. .

7) When designing the RF PCB, the width of the trace of the RF signal is strictly regulated. When designing, according to the thickness and dielectric constant of the PCB, it is necessary to strictly calculate and simulate the impedance of the trace at the corresponding frequency to ensure that it is 50 ohms (the standard of CATV is 75 ohms). However, it is not always that we all need strict impedance matching. In some cases, a small impedance mismatch may not matter (for example, 40 ohms to 60 ohms); and even if your simulation of the board is based on Under the ideal circumstances, when the actual production is given to the PCB factory, the process used by the manufacturer will cause the actual impedance of the board and the simulation results to differ by a thousand miles. Therefore, for the problem of impedance matching of small-signal RF PCB, my suggestion is: Step-1: Properly communicate with the PCB factory to obtain the width range of the 50-ohm trace of the board corresponding to the thickness and the number of layers used; Step-2: Select a suitable width within this width to apply uniformly to all 50 ohm RF signal lines; Step-3: When the PCB is delivered for production, note all lines of this width on the Script for a 50 ohm impedance match. At this point, you don't need to point out a lot of lines that need to be impedance matched. (For PCB manufacturers, they will make an impedance strip in the situation of the PCB extension you designed, in the factory. Test the impedance of a sample trace of a corresponding width on an impedance strip to roughly determine the impedance of the same width trace on the board. Finally, the impedance strip is cut and recycled by the PCB factory and will not be seen by you. At different frequencies, the impedance of the same width line will be slightly different, but the difference is generally within 10%. Of course, you can also write a very complicated impedance setting script that allows the board factory to fine-tune the width of the working lines at different frequencies according to their process so that the impedance is strictly set to 50 ohms, and then ask the PCB factory for each The root line is filtered. This results in a logarithm increase in cost and a large amount of scrap rate; and after such PCB mounting is completed, the resistance variation can still be caused by the solder distribution and the RF component itself. Such a situation is extremely rare, because even with sophisticated RF test and measurement instruments, the error caused by the weak mismatch of the RF small signal trace impedance (within 5%) can be easily corrected by software; In terms of the communication machine, it is even more unnecessary to care about the 5% difference. But I want to emphasize that the RF traces of the LNA (low noise amplifier) and PA (amplifier) are very sensitive, but fortunately, whether it is LNA circuit or PA circuit, routing The frequency must be the same, and the number of traces is small (no more than two nodes for input and output). At this time, I suggest that in sensitive cases, LNA and PA should be used separately, using high-quality RF-specific PCB sheets (Rogers/Arlon/Taconics) with uniform dielectric constant distribution, and no solder resist oil is used in the RF signal line. Call green oil) to avoid impedance drift caused by solder masking; and require PCB board manufacturers to provide impedance test reports. Since the signal power of the input portion of the LNA circuit itself is already very small (-150 dBm or less), the insertion loss caused by the impedance mismatch further reduces the valuable signal strength; for the PA circuit, since it operates at a very high power, The insertion loss caused by the impedance mismatch can consume a lot of energy (compared, the insertion loss is 1dB: the difference between the 10dBm signal attenuation is 9dBm and the 50dBm attenuation is 49dBm. Oh, the latter can generate 20W of heat. In some power kW of PA, 1dB insertion loss may bring the effect of fire and splash.

8) For those RF microstrip circuits that are simulated on the PCB, such as ADS, HFSS and other simulation tools, especially those directional couplers, filters (PA narrowband filters), microstrip resonators (such as you Design VCO), impedance matching network, etc., it is necessary to communicate well with the PCB factory, using plates with strict thickness and dielectric constant and the same indicators used in the simulation. The best solution is to find the corresponding board for the microwave PCB board agent and then commission the PCB factory for processing.

9) In the RF circuit, we often use the crystal oscillator as the frequency standard. This crystal oscillator may be TCXO, OCXO or ordinary crystal oscillator. For such a crystal circuit, be sure to keep away from the digital part and use a special low-noise power supply system. More importantly, the crystal oscillator may have a frequency drift with changes in ambient temperature. For TCXO and OCXO, this will still happen, but to a lesser extent. In particular, small-sized crystal oscillators with patches are very sensitive to ambient temperatures. For such a situation, we can add a metal cover on the crystal circuit (not directly in contact with the crystal package) to reduce the sudden change of the ambient temperature and cause the frequency of the crystal to drift. Of course, this will lead to an increase in size and cost.