Wi-Fi product RF circuit debugging experience and problem analysis-EEWORLD

Collect

This document is a summary of some RF debugging experiences in the Wi-Fi product development process by a front-line RF engineer of Shenghua Communications . It records and describes the process of encountering and solving problems in actual project development.

1 Introduction

This document summarizes my experience in Radio Frequency debugging (hereinafter referred to as debugging) over the past year and a half, and records the process of problems I encountered and solved in actual project development. Now I want to use this document to share these experiences with you. If this document can help you in your work, it will be my greatest honor.

Personally, I feel that the debugging process uses the simplest basic knowledge. If you have a very deep (note, very deep) understanding of the basic knowledge of RF, I believe that all bugs will be easy to solve. Similarly, this document will also describe the most easy-to-understand debugging techniques in the most easy-to-understand language.

In this article, I try to avoid writing some empty theoretical knowledge, but the content of Chapter 2 is an exception. The content of "Passive Components at Microwave Frequencies" is taken from my unfinished "long-winded" General RF Circuit Design of Wi-Fi Products (Second Edition).

I believe this document has more than one error. If anyone finds one, I hope you can point it out so that we can make progress together.

2 Passive devices at microwave frequencies

In this chapter, we mainly explain passive components at microwave frequencies. A simple question: a 1K resistor has a resistance of 1K in DC conditions, and may still be 1K in a loop with a frequency of 10MHz, but what about at 10GHz? Will its resistance still be 1K? The answer is no. At microwave frequencies, we need to look at passive components in a different way.

2.1. Conductors at microwave frequencies

Conductors at microwave frequencies can exist in many forms, such as microstrip lines, strip lines, coaxial cables, component pins, and so on.

2.1.1. Skin effect

At low frequencies, the current inside the wire is uniform, but at microwave frequencies, a strong magnetic field is generated inside the wire, which forces electrons to gather at the edge of the conductor, so that the current flows only on the surface of the wire. This phenomenon is called the skin effect. The skin effect causes the resistance of the wire to increase. What is the result? When the signal is transmitted along the conductor, the attenuation will be very serious.

In actual high-frequency applications, such as the induction coil of a radio, in order to reduce the signal attenuation caused by the skin effect, multiple strands of wire are usually wound side by side instead of a single strand of wire.

We usually use skin depth to describe the skin effect. Skin depth is the combined effect of frequency and the wire itself, which we will not discuss in depth here.

2.1.2. Linear inductance

We know that a magnetic field is generated around a wire with current flowing through it. If the current in the wire is an alternating current, the magnetic field strength will also change with the change of the current. Therefore, a voltage that prevents the current from changing will be generated at both ends of the wire. This phenomenon is called self-inductance. In other words, the wire at microwave frequency will show the characteristics of inductance, which is called linear inductance. You may think that linear inductance is very small and can be ignored, but we will see in the following content that as the frequency increases, linear inductance becomes more and more important.

The concept of inductance is very important because at microwave frequencies, any wire (or conductor) will exhibit certain inductance characteristics, even the pins of resistors and capacitors are no exception.

2.2. Resistance at microwave frequencies

Basically, resistance is the property of a material that blocks the flow of electric current. The relationship between resistance, current and voltage is given in Ohm's law. However, at microwave frequencies, we cannot use Ohm's law to simply describe resistance. At this time, the characteristics of resistance have changed greatly.

2.2.1. Equivalent circuit of resistor

The equivalent circuit of a resistor is shown in Figure 2-1. R is the resistance of the resistor itself under DC conditions, L is the pin of the resistor, and C varies depending on the resistor structure. It is easy to imagine that the same resistor will have different resistance values at different frequencies. Think about it, when we design Wi-Fi products, we rarely use direct-insertion components (except large-capacity electrolytic capacitors). On the one hand, it is to reduce the volume, and on the other hand, and more importantly, to reduce the inductance caused by the component pins.

Figure 2-1 Equivalent circuit of resistor

Figure 2-2 qualitatively gives the relationship between the resistance value and frequency.

page

Figure 2-2 Resistance value and frequency

Let's try to analyze why resistors have such characteristics. When the frequency is 0 (corresponding to a DC signal), the resistance value presented by the resistor is its own resistance value; when the frequency increases, the resistance value presented by the resistor is its own resistance value plus the inductive reactance presented by the inductor; when the frequency increases further, the resistance value of the resistor itself plus the inductive reactance of the inductor is already quite large, so the resistance value presented by the resistor is the capacitive reactance of the parallel capacitor, and the higher the frequency, the smaller the capacitive reactance.

2.3. Capacitance at microwave frequencies

Capacitors are widely used components in radio frequency circuits, such as bypass capacitors, interstage coupling, resonant circuits, filters, etc. Like resistors, the capacitive reactance characteristics of capacitors will also change greatly at microwave frequencies.

2.3.1. Equivalent circuit of capacitor

We know that the material of the capacitor determines the characteristic parameters of the capacitor. The equivalent circuit of the capacitor is shown in Figure 2-3. C is the capacitance of the capacitor itself, Rp is the parallel insulation resistance, Rs is the heat loss of the capacitor, and L is the inductance of the capacitor pin.

Figure 2-3 Equivalent circuit of capacitor

Regarding capacitors, I would like to introduce several parameters that you may not pay attention to when selecting materials.

Figure 2-4 qualitatively shows the reactance characteristics of capacitors at different frequencies. The vertical axis in the figure is insertion loss, which is the loss caused by the addition of capacitors.

Figure 2-4 Reactance characteristics of capacitors at different frequencies

Obviously, before the turning point, the capacitor shows the characteristics of a capacitor, and after the turning point, the capacitor shows the characteristics of an inductor. Generally speaking, a large-capacitance capacitor will show more inductive characteristics than a small-capacitance capacitor. Therefore, at a frequency of 250MHz, a 0.1uF bypass capacitor is not necessarily better than a 100pF capacitor. In other words, the classic formula for capacitive reactance is

It seems to indicate that when the frequency is constant, the larger the capacitance, the smaller the capacitive reactance. However, at microwave frequencies, the conclusion is the opposite. At microwave frequencies, a 0.1uF capacitor will exhibit a greater impedance than a 100pF capacitor. This is why we need to connect a large-capacity electrolytic capacitor and a small-capacity capacitor in parallel at both ends when designing a power supply circuit. These small-capacity capacitors are used to eliminate high-frequency noise signals.

2.3.2. Capacitor Capacitance and Temperature Characteristics

When selecting materials in the CIS library, we always find that the capacitor has a parameter of X7R or X5R, NPO, etc. I have searched for relevant information, translated it, and written it in this section.

This type of parameter describes the type of dielectric material used in the capacitor, temperature characteristics, and error parameters. Different values also correspond to a certain range of capacitance. Specifically, they are:

X7R is commonly used for capacitors with a capacity of 3300pF~0.33uF. This type of capacitor is suitable for filtering, coupling and other occasions. The dielectric constant is relatively large. When the temperature changes from 0°C to 70°C, the change in capacitance is ±15%;

Y5P and Y5V are commonly used for capacitors with a capacity of 150pF~2nF. They have a wide temperature range. As the temperature changes, the capacitance changes by ±10% or +22%/-82%.

For the relationship between other codes and temperature characteristics, please refer to Table 2-1. For example, X5R means that the normal operating temperature of the capacitor is -55°C~+85°C, and the corresponding capacitance change is ±15%.

Table 2-1 Capacitor temperature and capacity error coding

page

2.4.1. Equivalent circuit of an inductor

It is not difficult to imagine that the wire itself has a certain resistance, and there is a certain capacitance between adjacent coils, so we get the equivalent circuit of the inductor as shown in Figure 2-5. Among them, Rs is the resistance of the wire, L is the inductance value of the inductor itself, and C is the equivalent capacitance. The inductance-frequency curve of the inductor is quite similar to the impedance-frequency curve of the resistor, which is directly related to their similar equivalent circuits. Readers can analyze the frequency characteristic curve of the inductor by themselves.

Figure 2-5 Equivalent circuit of inductor

2.4.2. Q value of inductor

The ratio of the inductive reactance of an inductor to the series resistance Rs is called the Q value of the inductor, that is, Q=X/Rs. Similar to capacitance, the larger the Q value, the better the quality of the inductor. If the inductor is an ideal inductor, then the Q value should be infinite, but in reality there is no ideal inductor, so an inductor with infinite Q value does not exist.

At low frequencies, the Q value of the inductor is very large, because at this time Rs is only the DC resistance of the wire, which is a very small value. When the frequency increases, the inductive reactance X of the inductor will increase, so the Q value of the inductor will increase with the increase of frequency (the skin effect is not obvious at this time); however, when the frequency increases to a certain extent, the skin effect cannot be ignored. At this time, the series resistance Rs will increase with the increase of frequency, and the series capacitance C will also begin to play a role, causing the Q value to decrease with the increase of frequency. Figure 2-6 shows the relationship between the Q value and frequency of an inductor of a certain company.

Figure 2-6 The relationship between the Q value and frequency of a company's inductor

In order to maximize the Q value of the inductor, we can usually use the following methods when making the inductor:

Using a larger diameter wire can reduce the DC resistance of the inductor;

Pulling the coils of the inductor apart can reduce the distributed capacitance between the coils;

Increase the permeability of the inductor, which is usually achieved using a magnetic core, such as a ferrite core.

In fact, the handmade production of inductors is a required course for RF engineers, but this part is relatively complicated and will not be discussed in this article. Interested readers can refer to relevant literature.

3 RF Debug Experience Sharing

3.1. A wireless AP 2.4GHz Chain0 has no output power

During a test of a wireless AP (dual-band high-power 11n wireless AP), I suddenly heard a clear and pleasant cracking sound, and then saw a wisp of green smoke slowly rising from the board (unfortunately I couldn't see where it was), and the surroundings were quickly filled with an unpleasant burning smell, and the power on the VSA (Vector Signal Analyzer) also dropped below 0dBm. Anyone with a little experience can come to the conclusion that "something is burned."

There is no output power. It is conceivable that a component in the Tx circuit must be damaged, but which one is it?

First, we used visual inspection (the so-called visual inspection is to directly observe the appearance of components with your eyes to check whether there are signs of cracks or burning), but nothing was found.

Then the "point measurement method" is used. At this time, you may ask: "What is the point measurement method?" The point measurement method is to use a probe or a probe rod to directly detect the signal status of the point to be tested. It is often used for time domain signal detection, such as an oscilloscope. However, due to the high operating frequency of Wi-Fi products, signal detection is generally performed in the frequency domain, and the point measurement method is rarely used for detection.

Practice has proved that the spot test method is a fast and effective means to determine where the RF problem is.

When talking about the point measurement method, we have to talk about the production of a simple probe. Take an SMA Cable (as shown in Figure 3-1), remove the SMA connector at one end (do not remove both ends), and strip off the shielding layer of 1~2cm to expose the core wire. In this way, an ordinary SMA Cable is transformed into a point measurement probe, which becomes a detection tool and a good assistant for RF engineers.

3.2. Output power is too large

Phenomenon: The output power is extremely large, the constellation diagram is blurred and cannot be demodulated. page

This is a slightly complicated question.

We know that Atheros solutions all have an output power control part, that is, to make the target power consistent with the actual power value. How is this achieved? We took out the 2.4GHz PA circuit of AP96 for research, as shown in Figure 3-2.

Figure 3-2 2.4GHz PA circuit

In Figure 3-2, U27 and its peripheral circuits form a power amplifier, which is sent to the subsequent circuits through C208 and R263. PC1 in the figure is a printed directional coupler, and the voltage of its 3 and 4 pins increases with the increase of output power. L18, L19, D1, C217 and R248 form a half-wave rectifier circuit, which converts the voltage sensed by the directional coupler into a DC signal and sends it to the Transceiver for detection, that is, the AR9223_PDET_0 network. In this way, the Transceiver can know the current output power at any time. The relationship between power and voltage value is established during the Calibrate process.

After the board has been calibrated and EEPROM has been loaded, we use ART to continue Tx. At this time, the board will send out signals according to the Target Power we set, and the Transceiver will increase its output power until it is consistent with the corresponding voltage value (AR9223_PDET_0) recorded during the Calibrate process.

At this time, we return to the original question: "The output power is too high, the constellation diagram is blurred, and it cannot be demodulated." What's going on? The Transceiver must not be able to get the correct voltage value, so it can only increase its output power until the PA output power reaches saturation. Check L19, L18, D1, C217, and R248 and find that D1 is open. Replace a new diode and it will return to normal.

It should be pointed out that the use of directional couplers for output power control is a method unique to Atheros, and Broadcom and Ralink have not yet adopted this method. In addition, the PA itself generally has a built-in power detection unit, which is output through a pin, usually called V_DET.

3.3. A wireless network card has severe static heating

Phenomenon: After a wireless network card is powered on, without any operation, the four PAs generate a lot of heat. The surface temperature of the PA is very high and very hot.

The first judgment is that the PA is not in a real "static" state, they are working secretly! So, how to verify it? Take the demo board of PA (SKY65137-11) and use the Power Supply to power it so as to observe its current consumption. When powered on, it is found that the current consumption is almost zero, and there is no heating phenomenon, which is different from the situation of the wireless network card. Studying the datasheet of SKY65137-11, a key pin PA_EN caught my attention. This pin is the enable pin of PA. When powered on, pull this pin up to 3.3V, and find that the current consumed by 5V increases sharply, and a lot of heat is emitted, and the surface temperature of PA rises immediately. Disconnect PA_EN from 3.3V, and the current consumed by 5V decreases. At this time, touch the PA_EN pin with your hand and find that the current consumed by 5V is jumping. This shows that the weak electrical signal sensed by the human body is enough to put the PA in the "Enable" state. At the same time, it shows that PA_EN is a very sensitive pin, and a very weak signal is enough to trigger it.

Analyze the SKY65137-11 unit circuit of the wireless network card, as shown in Figure 3-3 (excluding Level Shift).

Figure 3-3 SKY65137-11 unit circuit

It is easy to find that the PA_EN pin of SKY65137-11 is directly connected to the control pin of AR9220 through a Level Shift circuit. In this way, a slight disturbance of the control pin of AR9220 can trigger the PA, which will cause the PA to heat up in static conditions.

Solution: Use a 10K resistor to pull down the PA_EN pin to ground so that the PA is normally in the off state.

Through the above method, the heating problem of PA is solved page

3.4. Calibrate of a wireless network card is inaccurate

Phenomenon: After the wireless network card is calibrated, the actual output power is inconsistent with the Target Power.

First, after investigation, it was determined that it was not a problem with the settings of Cable Loss and ART. The RF part of this wireless network card was designed by us independently, and there are too many uncertain factors, so we will not conduct an in-depth analysis here. As discussed in 3.2, Atheros' solution achieves output power stability by detecting the voltage value corresponding to the output power of the PA; in static conditions, if the PA has no output power, the corresponding voltage value is zero. Through detection, it was found that the V-Detect output of SKY65135-21 (2.4GHz PA) in static conditions is not zero, but a voltage value of a few tenths of a volt. This may be caused by the problem of the PA itself, and it is for this reason that the Calibrate of this wireless network card is inaccurate. We all know the unidirectional conductive characteristics of the diode. In order to prevent the 2.4GHz and 5GH bands of the wireless network card from affecting each other during the Calibrate process, they can be separated by a diode. In the subsequent versions of this wireless network card, we adopted this method, which can solve the problem of inaccurate Calibrate.

3.5. A wireless AP has no output signal

Phenomenon: ART runs normally, but when observed with VSA, there is no output signal.

Recall the content explained in 3.1, we mentioned the point test method. Personally, I think that the point test method is the fastest way to solve problems like this. When using ART for Continue Tx, use the probe to detect the Transceiver output, PA input, PA output, low-pass filter output, T/R Switch input and T/R Switch output in turn. Generally speaking, it is sufficient to detect these points.

According to the above method, we detect each point of the Tx loop in turn (taking 2.4GHz link 0 as an example), as shown in Figure 3-4.

Figure 3-4 2.4GHz 0 link detection point

In the actual detection process, it was found that there was a signal at the input of the T/R Switch, that is, there was a normal RF signal at C379, but there was no signal at the output of the T/R Switch. After consulting the Datasheet of the T/R Switch uPG2179, it was found that the control signal at this time was inconsistent with the expected one. For details, readers can refer to the uPG2179 Datasheet and the reference design of AR9280 (the Transceiver of this project).

div id=yued>