FS Technology's detailed introduction to high-frequency PCBA design
The demand for high-speed and accurate data transmission makes high-frequency data cables become the mainstream of the times, such as USB4, HDMI, Thunderbolt, and DisplayPort. The signal transmission is carried out through the old version of HDMI 2.1TMDS, which can reach up to 18Gbps, and can transmit high-definition images of 3840x2160p and 4K resolution. With the innovation of high-frequency PCBA technology, the bandwidth using the latest FRL mode can be increased to 48 Gbps. If we use compression technology on this basis, it becomes easy to transmit 10K resolution images; and by using the USB4 new version you can speed up to 40 Gbps. Regardless of the type of high-frequency wiring, there will be PCB thermal noise or other types of PCB noise. How to solve PCBA noise has become the focus of manufacturers or designers. Here the question arises which value of frequency is considered high frequency? 400MHz, 10GHz, 5 GHz, or 1Ghz.
The concept of high frequency line
Reduce high frequency reflection wavelengths
The definition of high frequency can be judged not only by frequency, but also by whether reflection occurs. When electromagnetic waves are transmitted on a medium (air or circuit board), the following three situations will occur:
- Continuous medium: no reflection phenomenon occurs;
- Discontinuous medium: if reflection occurs, the frequency is regarded as a low frequency on this medium;
- Discontinuous medium: If reflection occurs, the frequency is considered high frequency on this medium.
So we can conclude that Reflection depends on the wavelength.
The transmission speed of electromagnetic waves in air or a vacuum is equal to the speed of light:
While in the case of any medium-speed electromagnetic waves is:
Through the use of this expression, we can find the transmission rate of electromagnetic in case of medium such as high frequency PCBA board:
If we relate the wave velocity with the electromagnetic wave equation when it passes through a medium. As a result, we have a new expression:
From above equation, we can conclude that if the wavelength is larger than the length of the transmission path there will be no reflection.
So in results to minimize the reflection we can increase the wavelength of the electromagnetic signal to larger than possible, normally 4 times the trace length is employed. One conclusion we get from the above formula is that by lowering the frequency and lowering the relative permittivity of the board, the wavelength can also be made longer. But in reality, due to the transmission requirements, the data transmission rate is not possible to be reduced at will, and the medium of the PCB cannot be 1.
There is another option that shortens the length of Trace, letting its length be less than the transmitted wavelength. But FS Technology believes that this approach has limitations. When the transmission rate reaches a certain height or the position between the high-frequency PCBA components is limited, the trace length of the high-frequency PCBA needs to be increased by at least 5 to 10 cm.
High Frequency Traces and Impedance
When we study the high-frequency traces of the PCB board there is a question arises about the impedance design of the trace, but there is nothing expressed about this in the above data. Above we discuss that in case of no reflection there are no impedance issues for Trace, so there is no need for impedance design.
As mentioned in the previous paragraph, as the demand for high-frequency transmission increases, blindly shortening the length of the trace is very weak. For projects that require ultra-high transmission rates, this method cannot be used at all. So lengthened Trace will unavoidably show reflection. The theory based on the electromagnetic waves says that there is another technique that exists, and has no reflection: Trace impedance = load impedance = internal impedance.
This theory is related to the discussion that we have at the start that if we have a continuous medium, there will be no reflection. In a simple way, trace, load, and internal impedance are the same.
High frequency PCBA signal reflection and misjudgment
Whether it is a PCBA manufacturer or a customer who purchases PCBA services, we do not want reflections to appear. Reflection denotes the energy portion that is not used or transmitted and moves back to its origin like a transmitter. The main point we want to express above is that when there is no reflection, no matter what the frequency is at this time, the trace will not have impedance problems, that is to say, we do not need to do any impedance design.
But in real happening that reflected energy is again reflected from the source due to the phenomena of superposition of waves. If waves are reflected back two times to receiving points will be superimposed with another signal that makes the error signal. Let’s assume we have a digital signal that is 1 at starting point will be zero at receiving point and the signal which was zero will be one that results in Miscalculation or misjudgment of the signal.
Precautions for Manufacturing High Frequency PCBA
Till now we have got the idea of problems caused by the reflection we will have to choose the right board to make high-frequency PCB. Normally two types of PCB boards are used Rogers and FR4.
The relative permittivity is the key point that affects the quality of the signal:
There is a relation between frequency and relative permittivity and variation of frequency also changes the relative frequency below formula explains this factor:
Here we can see that change in relative permittivity according to a frequency transmission rate of electromagnetic wave on the board and wavelength also changes. It is a dispersion phenomenon given by the electromagnetic wave theory and this medium is known as dispersion.
Detailed overview of high frequency PCBA dispersion problems
The digital signal is transmitted in square waves that are either 0 or 1. If we observe the changing frequency of a square wave, it will be reflected in the formation of countless sin waves of different frequencies. In simple words, it consists of fundamental waves and 2 odd harmonics superimposing.
If three different frequency waves are passed through PCB traces due to different relative permittivity, there will be a transmission rate difference. In an ideal case, the 3 frequencies have the same speed, we set the speed value to X, and as a result, we will get a complete square wave signal at the receiving end of the signal. But in the reality, it can’t happen, the three frequencies have their own rates, either fast or slow. If we blindly increase the trace length, we will receive electromagnetic waves that are decomposed into three different frequencies at different times.
A similar high signal reaches at different time intervals and amplitude also reduces that will make error for judgment at receiving point and results in bit errors.
Or, if there are multiple signals of different rates in our equipment since the high-speed signal will surpass the low-speed signal, this will cause the high-frequency PCBA to receive a disordered signal at the receiving end, making it impossible for us to judge its accuracy.
As we discussed earlier that although we have designed the Trace impedance (single-end 50 Ω or differential 100 Ω), the trace can move for a larger time without causing the reflection but when it meets with the dispersive medium it also restricts the length of the trace. So for impedance design, we can opt to make the trace as short as possible.
High Frequency Material Comparison: FR4 vs Rogers
Through using the understanding of relative permittivity and dispersion phenomena now make a comparison between FR4 and Rogers materials. The below figure gives the idea of the relative permittivity of FR4 and Rogers with frequency.
From the above figure, we can see that ∈ r (f) of FR4 changes highly according to frequency, while there is no charge for Rogers. Suppose that we are using the frequency range of transmission rate as PRBS31, so the use of FR4 must have serious attention to dispersion issues. For the reduction of dispersion issues, traces must be small as possible.
You may think the solution is not to use FR4 but use Rogers PCB and problems solved. It is correct but rogers prices are high, you must also consider this too. The reduction of prices and how to maintain the quality of the signal to fulfill the communication requirements. It always comes to mind engineers that make layouts for making balance.
At last, we have not discussed the ∈ r (f) effect on the trace impedance. Basically, it will affect, but to avoid too messy analysis and explanation, the impedance design issue is idealized.
Analysis of Large Waveguide Structure of Three High Frequency PCBAs
In the above discussion, we explain the relationship between the high frequency and PCB, like the selection of accurate material and price comparison. Now we will discuss the high-frequency PCB design layout.
When we thought about how to design the Layout of PCB? The first question that comes to mind is about the simulation. The use of simulation software is easy and saves extra time and is expensive, but how we can formulate simulation parameters? First, you must have an idea of starting the layout design.
Before high frequency PCB board fabrication, the first problem we face is choosing a suitable waveguide structure. The three most basic transmission structures are described below:
- Coplanar Waveguide
The architecture and design are simple. For similar high frequency PCBA substrates, there are many parameters that cannot be varied due to simple structure, such as line width substrate thickness and line thickness.
With the increment in the frequency, there is another issue for Microstrip that is air value above the trace is ∈ eff = 1, and the substrate below is ∈ eff ≠ 1, the asymmetric pattern will be for an upper and lower field that results in electromagnetic wave field pattern to be asymmetric. This asymmetric behavior affects the quality of signal transmission.
The structure is complicated and ∈ eff of the upper and lower substrate can be close to the substrate, therefore the pattern of the electromagnetic wave is complete. Also, stripline high frequency traces not shown a long-range crosstalk effect among one another.
There is signal attenuation and reflection problems that will come due to redundant via segment since internal layers are connected with outer parts through the use of vias. To minimize this issue use back drilling or blind vias, but this increases the costs.
For the creation of a waveguide, both sides of the trace are covered through a copper guard like the Microstrip. Then the Microstrip, coplanar waveguide has the ability to change more physical parameters for example distance between trace and ground and distance between via and trace existing on the ground.
A coplanar waveguide is flanked through GND that minimizes the effect of crosstalk. There is bottleneck exists in the design, if there are high frequency and high-density traces used in circuits like PCIe the structure will not have enough space to apply copper ground on both sides of the traces.