PCB News - How can we design a perfect PCB board?

Everyone knows that PCB Layout is to turn a circuit diagram into an actual PCB circuit board, but this is not a simple process. Many people in foreign countries call PCB design art. It is not difficult to lay a PCB board, but it must be done well. It is not an easy task to use a PCB to realize its functions perfectly.

The two major difficulties in the field of microelectronics are the processing of high-frequency signals and weak signals. In this regard, the level of PCB production is particularly important. The same principle design, the same components, and PCBs produced by different people have different results., So how can we make a good PCB board? Based on our past experience, I would like to talk about my views on the following aspects:

One: clear design goals

Receiving a design task, we must first clarify its design goals, whether it is an ordinary PCB board, a high-frequency PCB board, a small signal processing PCB board or a PCB board with both high frequency and small signal processing. If it is an ordinary PCB board, As long as the layout and wiring are reasonable and tidy, and the mechanical dimensions are accurate, if there are medium load lines and long lines, certain measures must be used to reduce the load, and the long line must be strengthened to drive, and the focus is to prevent long line reflections. When there are signal lines exceeding 40MHz on the board, special considerations should be made to these signal lines, such as crosstalk between lines. If the frequency is higher, there is a stricter limit on the length of the wiring. According to the network theory of distributed parameters, the interaction between the high-speed circuit and its wiring is a decisive factor and cannot be ignored in system design. As the gate transmission speed increases, the opposition on the signal lines will increase accordingly, and the crosstalk between adjacent signal lines will increase proportionally. Generally, the power consumption and heat dissipation of high-speed circuits are also very large, so high-speed PCBs are being made. Enough attention should be paid.

When there are millivolt or even microvolt-level weak signals on the board, these signal lines need special attention. Small signals are too weak and are very susceptible to interference from other strong signals. Shielding measures are often necessary, otherwise they will Greatly reduce the signal-to-noise ratio. As a result, the useful signal is submerged by noise and cannot be extracted effectively.

The commissioning of the board should also be considered in the design stage. The physical location of the test point, the isolation of the test point and other factors cannot be ignored, because some small signals and high-frequency signals cannot be directly added to the probe for measurement.

In addition, other relevant factors should be considered, such as the number of layers of the board, the package shape of the components used, and the mechanical strength of the board. Before making a PCB board, you must have a good idea of the design goals for the design.

two. Understand the layout and routing requirements of the functions of the components used

We know that some special components have special requirements in the layout and routing, such as the analog signal amplifier used by LOTI and APH. The analog signal amplifier requires a stable power supply and small ripple. Keep the analog small signal part as far away from the power device as possible. On the OTI board, the small signal amplifying part is also specially equipped with a shielding cover to shield the stray electromagnetic interference. The GLINK chip used on the NTOI board uses ECL technology, which consumes a lot of power and generates heat. Special consideration must be given to the heat dissipation problem in the layout. If natural heat dissipation is used, the GLINK chip must be placed in a place with relatively smooth air circulation., And the heat radiated can not have a big impact on other chips. If the board is equipped with speakers or other high-power devices, it may cause serious pollution to the power supply. This point should also be paid enough attention.

Three. Component layout considerations

The first factor that must be considered in the layout of components is electrical performance. Put closely connected components together as much as possible, especially for some high-speed lines, make them as short as possible when laying out power signals and small signal devices. To be separated. On the premise of meeting the circuit performance, the components must be placed neatly and beautifully, and easy to test. The mechanical size of the board and the location of the socket must also be carefully considered.

The grounding and the transmission delay time on the interconnection line in the high-speed system are also the first factors to be considered in the system design. The transmission time on the signal line has a great influence on the overall system speed, especially for high-speed ECL circuits. Although the integrated circuit block itself is very fast, it is due to the use of ordinary interconnect lines on the backplane (the length of each 30cm line is about The delay amount of 2ns) increases the delay time, which can greatly reduce the system speed. Synchronous working parts such as shift registers and synchronous counters are best placed on the same plug-in board, because the clocks on different plug-in boards The signal transmission delay time is not equal, which may cause the shift register to produce a major error. If it cannot be placed on one board, the length of the clock line from the common clock source to each plug-in board must be equal where synchronization is the key.

Fourth, the consideration of wiring

With the completion of the design of OTNI and the star optical fiber network, there will be more boards with high-speed signal lines above 100MHz that need to be designed in the future. Some basic concepts of high-speed lines will be introduced here.

1. Transmission line

Any "long" signal path on the printed circuit board can be regarded as a kind of transmission line. If the transmission delay time of the line is much shorter than the signal rise time, the main reflections produced during the signal rise period will be submerged. Overshoot, recoil and ringing are no longer present. For most of the current MOS circuits, since the ratio of rise time to line transmission delay time is much larger, the trace can be as long as meters without signal distortion. For faster logic circuits, especially ultra-high-speed ECL.

For integrated circuits, due to the increase in edge speed, if no other measures are taken, the length of the trace must be greatly shortened to maintain the integrity of the signal.

There are two ways to make high-speed circuits work on relatively long lines without serious waveform distortion. TTL adopts Schottky diode clamping method for fast falling edges, so that the overshoot is clamped to a diode voltage drop lower than the ground potential. At the level of “H”, this reduces the amplitude of the backlash. The slower rising edge allows overshoot, but it is attenuated by the relatively high output impedance (50～80Ω) of the circuit in the “H” state. . In addition, due to the greater immunity of the level "H" state, the kickback problem is not very prominent. For HCT series devices, if the Schottky diode clamp and series resistance termination method are combined, it will improve The effect will be more obvious.

When there is fan-out along the signal line, the TTL shaping method introduced above appears to be somewhat inadequate at a higher bit rate and a faster edge rate. Because there are reflected waves in the line, they will tend to be synthesized at a high bit rate, causing serious signal distortion and reduced anti-interference ability. Therefore, in order to solve the reflection problem, another method is usually used in the ECL system: the line impedance matching method. In this way, the reflection can be controlled and the integrity of the signal can be guaranteed.

Strictly speaking, for conventional TTL and CMOS devices with slower edge speeds, transmission lines are not very necessary. For high-speed ECL devices with faster edge speeds, transmission lines are not always needed. But when using transmission lines, they have the advantages of predicting the connection delay and controlling reflection and oscillation through impedance matching. 1

There are five basic factors in deciding whether to use a transmission line. They are: (1) the edge rate of the system signal, (2) the connection distance (3) the capacitive load (how much fan out), (4) the resistive load (the line termination method); (5) allowable The percentage of backlash and overshoot (the degree of reduction in AC immunity).

2. Several types of transmission lines

(1) Coaxial cable and twisted pair: They are often used in the connection between systems. The characteristic impedance of coaxial cable is usually 50Ω and 75Ω, and twisted pair is usually 110Ω.

(2) Microstrip line on the printed circuit board

The microstrip line is a strip conductor (signal line). A dielectric is used to isolate it from the ground plane. If the thickness, width, and distance between the line and the ground plane are controllable, its characteristic impedance can also be controlled. The characteristic impedance Z0 of the microstrip line is:

In the formula: [Er is the relative permittivity of the printed board dielectric material

6 is the thickness of the dielectric layer

W is the width of the line

t is the thickness of the line

The transmission delay time of a microstrip line per unit length depends only on the dielectric constant and has nothing to do with the line width or spacing.

(3) Strip line in printed board

A stripline is a copper strip line placed in the middle of a dielectric between two conductive planes. If the thickness and width of the line, the dielectric constant of the medium, and the distance between the two conductive planes are controllable, then the characteristic impedance of the line is also controllable. The characteristic impedance B of the strip line is:

Where: b is the distance between two ground boards

W is the width of the line

t is the thickness of the line

Similarly, the transmission delay time of a strip line per unit length has nothing to do with the width or spacing of the line; it only depends on the relative permittivity of the medium used.

3. Unterminated transmission line

If the line delay time is much shorter than the signal rise time, the transmission line can be used without series termination or parallel termination. If a non-terminated wire has a round-trip delay (the time it takes for a signal to travel on the transmission line once) than pulse The rise time of the signal is short, so the kickback caused by non-termination is about 15% of the logical swing. The maximum opening route length is approximately:

Lmax＜tr/2tpd

Where: tr is the rise time

tpd is the transmission delay time per unit line length

4. Terminate the transmission line

At the receiving end of a line, a resistance equal to the characteristic impedance of the line is used to terminate, then the transmission line is called a parallel terminal connection. It is mainly used to obtain the best electrical performance, including driving distributed loads.

Sometimes in order to save power consumption, a 104 capacitor is connected in series to the terminating resistor to form an AC termination circuit, which can effectively reduce DC loss.

A resistor is connected in series between the driver and the transmission line, and the terminal of the line is no longer connected to the termination resistor. This termination method is called series termination. The overshoot and ringing on the longer line can be controlled by series damping or series termination technology. Series damping is achieved by using a small resistance (generally 10 to 75Ω) connected in series with the output of the drive gate. This damping method is suitable Used in conjunction with lines whose characteristic impedance is controlled (such as backplane wiring, circuit boards without ground planes, and most winding wires, etc.).

In series termination, the sum of the value of the series resistance and the output impedance of the circuit (driving gate) is equal to the characteristic impedance of the transmission line. The series connection end wiring has the disadvantages that it can only use the lumped load at the terminal and the transmission delay time is longer. However, this This can be overcome by using redundant serially terminated transmission lines.

5. Comparison of several termination methods

Both parallel connection and series connection have their own advantages. Which one to use or both depends on the designer's preference and system requirements. The main advantage of the parallel connection is that the system is fast and the signal is transmitted on the line without distortion. The load on the long line will neither affect the transmission delay time of the drive gate driving the long line nor the signal edge speed thereof, but will increase the transmission delay time of the signal along the long line. When driving a large fan-out, the load can be distributed along the short branch line, instead of the terminal where the load must be lumped on the line as in the series termination.

The series termination method makes the circuit have the ability to drive several parallel load lines. The delay time increment caused by the capacitive load of the series terminal wiring is about twice as large as the corresponding parallel terminal wiring, and the short circuit is caused by the capacitive load. The speed slows down and the drive gate delay time increases. However, the crosstalk of the series connection is smaller than that of the parallel connection. The main reason is that the amplitude of the signal transmitted along the series connection is only one-half of the logic swing. The switching current is only half of the switching current of the parallel termination, and the signal energy is small and the crosstalk is small.

Two PCB board wiring technology

Whether to choose a double-sided board or a multi-layer board when making a PCB depends on the highest operating frequency, the complexity of the circuit system, and the requirements for assembly density. It is best to choose a multilayer board when the clock frequency exceeds 200MHZ. If the operating frequency exceeds 350MHz, it is best to choose a printed circuit board with PTFE as the dielectric layer, because its high-frequency attenuation is smaller, the parasitic capacitance is smaller, and the transmission speed is faster. Large and low power consumption, the following principles are required for the wiring of the printed circuit board

(1) When designing signal transmission lines, avoid sharp turns to prevent reflections caused by sudden changes in the characteristic impedance of the transmission line. Try to design a uniform arc line with a certain size.

(2) Keep as much space as possible between all parallel signal lines to reduce crosstalk. If there are two signal wires that are close together, it is best to run a ground wire between the two wires, which can play a shielding role.

The width of the printed line can be calculated according to the above-mentioned characteristic impedance calculation formula of the microstrip line and the strip line. The characteristic impedance of the microstrip line on the printed circuit board is generally between 50 and 120Ω. To get a large characteristic impedance, the line width must be very narrow. But very thin lines are not easy to make. Considering various factors, it is generally appropriate to choose an impedance value of about 68Ω, because the characteristic impedance of 68Ω can achieve the best balance between delay time and power consumption. A 50Ω transmission line will consume more power; of course, a larger impedance can reduce the power consumption, but it will increase the transmission delay time. The negative line capacitance will increase the transmission delay time and decrease the characteristic impedance. However, the intrinsic capacitance per unit length of the line segment with very low characteristic impedance is relatively large, so the transmission delay time and characteristic impedance are less affected by the load capacitance. An important feature of a properly terminated transmission line is that the short branch line should have no effect on the line delay time. When Z0 is 50Ω. The length of the branch stub must be limited to 2.5 cm or less. In order to avoid loud ringing.

(4) If there is a small signal amplifier on the board, the weak signal line before amplification should be far away from the strong signal line, and the trace should be as short as possible, and if possible, use a ground wire to shield it.

(5) If there are high-current devices on the printed circuit board, such as relays, indicator lights, speakers, etc., their ground wires should be separated to reduce noise on the ground wire. The ground wires of these high-current devices should Connect to an independent ground bus on the plug-in board and backplane, and these independent ground wires should also be connected to the ground point of the entire system.

(6) For double-sided boards (or four-layer lines in six-layer boards). The lines on both sides of the circuit board should be perpendicular to each other to prevent crosstalk caused by mutual induction.