Stop blindly pursuing advanced mmWave PCB: your design may not need it

I participated in a vehicle radar project team last year. At first, we were obsessed with specifying a certain expensive imported base material. Later, we changed our mind and spent time to make the simulation model of the antenna array very detailed, even taking into account the impact of the connector pads. Then we found a domestic processing factory that we have cooperated with for many years. Although they have not used such high-end materials, they have a thorough understanding of our design intentions and are very patient in adjusting the process parameters. The cost of the final board dropped a lot but the performance test results exceeded expectations.

This experience makes me think that by around 2025, this industry may undergo a very important change: everyone will no longer blindly pursue “top materials” but will pay more attention to “system-level design and manufacturing collaboration capabilities.” Whoever can seamlessly integrate the three links of design, material selection, and processing technology can truly make stable and reliable high-frequency products.

Many discussions now focus on the technical specifications of the 77GHz or higher frequency band, which is of course important, but I think the focus of competition in the next stage will shift to how to achieve large-scale, consistent high-quality manufacturing. After all, no matter how good the laboratory sample is, it will lose its commercial value if it cannot be stably produced in batches. For those who actually make products, the actual significance of a one percentage point improvement in yield may be much greater than a reduction in the loss tangent value of a certain material by a few tenths.

So my point is don’t be led by those gorgeous parameters and material models.
Calm down and do a solid job in basic skills. Understand what your circuit needs. Understand what your factory can do well. On this basis, choose the most suitable rather than the most expensive material route. This may be a more practical way to deal with future high-frequency PCB challenges.

I recently chatted with some friends who make high-frequency circuits and found that when everyone mentioned millimeter wave design, their first reaction was to look for Rogers plates. This is of course true. Materials such as Rogers 4350B do have a good reputation in the industry, but I feel that this line of thinking is a bit biased, as if things cannot be made well without it. In fact, many projects focus on the highest-end materials from the very beginning, but in the end the cost cannot be realized at all.

My experience is that what really determines whether your circuit can run is often not the accuracy of the nominal Dk value of the board itself, but the ability to control the loss of your entire link. The bulk of the loss here is often not the medium itself, but the conductor loss. Many people spend a lot of money to buy low-loss board rolls. However, in the PCB processing factory, due to the mismatch of copper foil processing technology or wrong lamination parameters, the insertion loss of the final board is measured, which is far from expected.

This has to mention a key thing: the surface roughness of the copper foil. For millimeter wave frequency bands, such as 77GHz, signals are transmitted in a very thin layer on the surface of the conductor. This phenomenon is called the “skin effect.” At this time, if the surface of the copper foil is rugged, the signal will be like climbing over mountains, and the energy loss will be very large. No matter how good the low-loss substrate you choose is, it will be useless. Therefore, more and more high-frequency PCB designs now begin to specify HVLP (ultra-low profile) copper foil, or even rolled copper. Although the cost of rolled copper is much higher than that of commonly used electrolytic copper foil, its surface is extremely smooth and flat, and its conductor loss advantage at high frequencies is very obvious.

Of course, I am not saying that media materials are not important. The Df value, which is the loss tangent, must be as low as possible. But in actual projects, you have to calculate the general ledger. For a Sub-6GHz 5G antenna board, if it is extremely cost-sensitive and the performance requirements are not the limit, sometimes it is not impossible to use some improved FR-4 materials. Its Df may be between 0.008 and 0.012. Although it is not as good as the Rogers series, which is often a few tenths, but as long as your traces are not long and the design is appropriate, it can fully meet the link budget of many consumer-grade products. Blindly pursuing the ultimate low Df will cause your BOM cost to skyrocket.

The most pragmatic approach I’ve seen is mixed pressure design. This is actually a very smart idea: only spend money where you really need it. For example, a 5G small base station motherboard integrates radio frequency front-end, digital processing and power management. You can use a small piece of Rogers 5880 or similar high-frequency materials to make the radio frequency antenna and key millimeter-wave feeder parts, while the digital circuit and power supply parts use ordinary FR-4. They are made together through the lamination process of multi-layer boards.
This not only ensures the low-loss characteristics of the core radio frequency channel, but also controls the overall cost from being too exaggerated. After all, the entire board is made of Rogers materials. It is very common for the cost to triple or five times, which is simply unrealistic for products that require large-scale commercial use.

So my opinion is that when designing high-frequency PCB or mmWave PCB, don’t just focus on the parameters in the material manual. It is a system engineering: you have to comprehensively consider your frequency, trace length, tolerance for loss, process capabilities of the processing plant, and the most realistic project budget.

From the design end, it is necessary to think about how to use the appropriate material combination, match the correct copper foil type and processing requirements, to achieve the best balance between performance and cost. This is where the real test of engineering ability is far more complex and valuable than simply specifying a “best” material.

Every time I see someone talking about millimeter wave PCB design in such a mysterious way, I want to laugh. It seems like you have to use some exorbitantly expensive materials to do it, but it’s actually not that complicated. After working on many projects, I found that what really affects performance is often some basic choices, such as boards.

Many people stare at Rogers boards as soon as they get started, thinking that the more expensive ones are better. In fact, ordinary FR-4 can also be used very well in many situations. The key lies in how you deal with the problem of its dielectric constant changing with frequency. I have done a comparison test. Using a suitable FR-4 laminate design in a certain frequency band, the performance is not much worse than those high-end boards.

When it comes to lamination, I find that many people have a too rigid understanding of the middle layers of L2, L3, and L4. Why do we have to follow what is in the textbook? Sometimes placing the power plane on L3 and the second ground plane on L4 can better control the impedance. I have tried adding a very thin dielectric layer between L2 and the top layer to reduce loss, and the effect is much better than simply pursuing surface roughness.

The most easily overlooked aspect in high-frequency PCB design is processing accuracy. It doesn’t matter how beautiful the drawings are, it’s all useless if the factory can’t make it or the deviation is too big. After working with several factories for a long time, I learned that there are only so many tolerances they can stably control, and you have to take them into consideration during the design stage.

There are also so-called “advanced materials”, such as certain specially treated copper foils, which are highly publicized. But after actual testing, I found that in many cases the performance improvement is not obvious and the cost doubles. It would be more practical to spend this budget on optimizing wiring and reducing vias.

I have seen many engineers become obsessed with simulation software, but forget that the actual circuit board is a three-dimensional physical structure. The coupling between microstrip lines and the parasitic effects caused by vias are difficult to simulate accurately. So my experience is that doing more physical tests and adjusting based on actual measured data is better than anything else.

Ultimately, millimeter wave design is a constant trade-off process—finding a balance between cost, performance, and manufacturability. There is no absolutely right solution, only the option that best suits the needs of the current project.

I always feel that many people have misunderstandings about high-frequency circuit board design.
It seems that all problems can be solved by choosing the right materials. In reality? Materials are only the foundation of foundations.

I’ve seen too many projects get stuck on mmWave antenna layout. Those dense arrays cannot be placed casually. You have to consider the interaction between them – it’s more complicated than many people think. Sometimes two antennas are too close and interfere with each other; sometimes they are too far apart and affect overall performance. The balance point is difficult to find. For example, in the 28GHz frequency band, a small adjustment in the array spacing may cause the beam direction to shift by several degrees, which requires a combination of electromagnetic simulation and multiple physical iterations for optimization.

When it comes to material selection, it’s even more interesting. Many people ask which Rogers board should be used as soon as they come up? In fact, this question itself is somewhat problematic. Because the characteristics of different models of Rogers PCB are quite different – some are suitable for this frequency band but not that frequency band; some have stable dielectric constant but less than ideal thermal expansion coefficient; some are just the opposite. For example, RO4350B performs well in the Ku band, but its thermal expansion coefficient is quite different from that of copper foil, so it needs to be carefully evaluated in application scenarios with drastic temperature changes.

I encountered this problem in a project I was working on recently: I paid too much attention to the stability of the Dk value when selecting materials. As a result, I ignored the problem of CTE matching. Later, the board showed slight deformation during the temperature cycle test. Although it was not visible to the naked eye, the impact on the millimeter wave signal was real. This deformation changed the characteristic impedance of the transmission line, causing the insertion loss to increase by nearly 1.5dB near the 77GHz frequency point, and the material pairing had to be re-evaluated.

The manufacturing process of high-frequency circuit boards is also an easily underestimated link. Do you think you can just design it and hand it over to the factory? That is not necessarily the case, especially when it comes to the mixing process, it is easy for problems to occur at the junction of different materials. If the signal is not processed well, reflection or loss will occur at that location. For example, when FR-4 is laminated with a high-frequency dielectric layer, control of the lamination pressure and temperature profile is crucial, otherwise micro-air gaps may form at the interface, becoming a stealth killer of signal integrity.

I suggest that everyone should consider manufacturing feasibility in the design stage and make the drawings very beautiful. If the factory cannot make it or the performance is not up to standard, then your work will be in vain. It is best to communicate with process engineers in advance to understand the actual capabilities of the production line for minimum line width, hole spacing, and inter-layer alignment accuracy, and use these as design constraints.

Another point that I think is very important but is often overlooked is that in the testing process, many teams spend a lot of time on design but are hasty in testing and verification. This is very dangerous because many problems in the millimeter wave frequency band cannot be completely predicted through theoretical calculations, and there will always be some unexpected situations in actual testing. Building a reliable test environment has its own challenges, such as probe calibration and multipath effects in shielded rooms, which will affect measurement authenticity.
For example, once we designed a mmWave PCB that performed very well in simulation, but during actual measurement we found that the standing wave ratio in a certain frequency band was particularly high. After a long investigation, we found that the location of a certain ground via was not right. Although it was only a few tenths of a millimeter off, it was enough to have an impact on high-frequency signals. This via happens to be located in the sensitive area of ​​the microstrip line fringe field, and its deviation destroys the expected return path and induces resonance.

So now I have developed a habit: try to leave room for design and don’t tighten the parameters too much. Leave some room for manufacturing errors and actual environmental changes. This way, the product will be more reliable in actual applications. After all, what we make is not a laboratory sample but a product that can work stably in various environments. For example, impedance tolerances are relaxed to ±10% and adjustable pad options are designed for critical matching networks.

High-frequency circuit design is like this. There is always a gap between theory and practice, and what we have to do is to continuously narrow this gap. The experience accumulated in this process is the most valuable asset. It is more real than the formulas in any textbook, because these are insights gained from actual projects rather than empty talk on paper.

When I first started getting into millimeter wave circuit board design, I always thought it was very mysterious. When everyone mentions high-frequency signal transmission, they immediately bring up various complicated formulas and high-end-sounding materials, such as Rogers sheets. But to be honest, my experience tells me that what really determines the quality of a board is often not the expensive materials used, but whether the most basic processing techniques are in place. Take etching as an example. Many people think this is a mature process, but in the millimeter-wave frequency band, the challenges it brings are completely different dimensions.

Think about it, once the signal frequency reaches millimeter waves, the micron-level error is enough to make a huge difference in the performance of the entire circuit. Slight changes in line width, such as if you think it is just a few microns difference, will actually cause the characteristic impedance to completely deviate from the design value. I have seen some antennas and transmission line models that were beautifully drawn on the design drafts, and the simulation results were perfect. However, as soon as the test signal was put on the board, the signal was completely attenuated. This is often where the problem lies.

So now I pay special attention to the actual process capabilities of the processing plant rather than the top materials used in their brochures. A manufacturer that can stably control the precision etching process is much more reliable than a manufacturer that only sells high-end plates. For example, they need to be able to accurately manage the concentration, temperature and spray pressure of the etching solution, because these factors directly affect the uniformity of line width and verticality of sidewalls. A trace with sloping sidewalls will have unpredictable line width and impedance characteristics that deviate from the design.

Of course, material selection is not unimportant. Plates like Rogers that are optimized for high frequencies can indeed provide a more stable dielectric constant and help ensure signal integrity, but this does not mean that everything will be fine with it. The batch stability, hygroscopicity and resin flow control during lamination of the board also affect the dielectric constant distribution of the final product.
If the lamination process is poor, resulting in uneven thickness of the dielectric layer, the performance of the actual circuit board will be greatly compromised, no matter how good the parameters of the board itself are.

In one of my recent projects, we tried a mixed-material solution. Instead of using all expensive special boards, we made targeted material selection on key signal layers and strict processing control. In the end, the cost was well controlled and the performance was fully up to standard. Specifically, we use high-performance materials on the surface layer that carries the millimeter-wave antenna and front-end transmission lines, while using more cost-effective FR-4 materials in the inner digital control and power supply parts. This solution requires the factory to have superb craftsmanship when laminating different materials to ensure alignment and bonding between layers and avoid delamination under thermal stress.

This reminds me of another key point: the application of laser technology is becoming more and more common in high-frequency PCB manufacturing, especially when making tiny blind holes and fine lines. The precision of ultraviolet laser drilling is difficult to match with traditional mechanical methods. It can achieve very clean hole walls, which is crucial for high-frequency signal transmission because any burrs or irregularities will cause unnecessary signal reflection and loss. In addition, laser direct imaging technology is used for pattern transfer and can achieve higher alignment accuracy and finer line width than traditional film exposure. This is almost a necessary process for complex gradient lines or coupling structures on millimeter wave circuits.

After all, designing and manufacturing high-frequency circuit boards is a highly collaborative process. You can’t just throw the drawings to the factory and expect them to make perfect products. You must have a deep understanding of their process limits, such as what precision their etching lines can stably achieve and how small holes their laser equipment can make, and then adjust your design parameters based on these actual processing capabilities. For example, if the factory’s etching process best controls the uniformity of 10 mil line widths, then the design should try to avoid using 8 mil or finer line widths that are close to the limit of its capabilities, or evaluate the impact of process fluctuations in advance through simulation.

The armchair theoretical school and the hands-down craftsman school must sit together to actually make things and do them well. Designers need to explain the electrical significance of each critical dimension to process engineers, and process engineers need to provide feedback on the accuracy range that can be achieved repeatedly in actual production. This kind of early and in-depth communication is often much more efficient than repeatedly modifying and verifying problems after problems are discovered during testing, and can also avoid many costly reworks caused by “taking things for granted”.

I’ve seen too many people overcomplicate millimeter wave and high frequency circuit design. Everyone gets nervous when they mention mmWave PCB or high Frequency PCB. In fact, many times the problem is not that mysterious.

Many people are confused about what material to use Rogers or other boards. Of course, Rogers PCB is indeed good in some scenarios. The material properties are stable, but this does not mean that it is the only solution, and not all projects have to use such expensive materials.