Robots are transitioning from exhibition floors to factory workshops. The torso of a humanoid robot is packed with components such as millimeter-wave (mmWave) radar, LiDAR, Wi-Fi 6/6E modules, EtherCAT buses, and servo drives.
A complete unit typically contains 20 to 40 PCBs, each serving a specific function; any board handling radio frequency (RF) signals or high-speed differential signals is generally classified as a “high-frequency” board.
Many teams developing high-frequency PCBs for robots for the first time fall into the same trap: assuming that because the signal frequency is relatively low, they can use standard FR-4 manufacturing processes. The result? Antenna gain falls 3dB short of simulations, communication range is slashed by half, and joint drive boards suffer batch failures during aging tests. The root cause often lies not in the frequency value itself, but in factors that might seem unrelated to high-frequency performance—such as batch-to-batch variations in the dielectric constant (Dk) of the substrate, “smear” (resin residue) on via walls, or uneven dielectric thickness.
Below, we will walk through the manufacturing process, covering four critical stages—material selection, via drilling, impedance control, and surface finish—while highlighting key parameters and common failure points.
I. Sources of High-Frequency Signals in Robotic Systems
High-frequency signals in robotic systems primarily originate from four types of modules. Sensing modules include mmWave radar (commonly operating at 24GHz, 60GHz, or 77GHz for obstacle avoidance and short-range proximity sensing), as well as the high-speed interconnects and clock links found within solid-state LiDAR. Communication modules encompass Wi-Fi 6/6E, 5G modules, UWB (Ultra-Wideband) positioning systems, and video transmission links.
Motion control components—such as EtherCAT, Industrial Ethernet, and high-speed encoder feedback—also fall into this category. Furthermore, the high $di/dt$ switching in servo drives can generate broadband noise, making the entire board highly sensitive regarding electromagnetic compatibility (EMC) and signal return paths. Main control and vision systems utilize interfaces like MIPI (for cameras), high-speed differential pairs, PCIe, and memory buses; while their nominal frequencies may not be extremely high, their signal edges are exceptionally fast, resulting in an equivalent bandwidth that can reach the GHz range.
The requirements for PCBs in robotics differ from those in smartphones or servers: they must fit into compact structural spaces, withstand repeated flexing at robotic joints, and operate across a wide temperature range. A single unit contains between 20 and 40 circuit boards. Based on mass production plans for 2026, the delivery of 100,000 units will translate into millions of PCBs. This dictates that the boards used in these robots must be small in size, highly diverse in type, frequently involve mixed-material lamination, and demand exceptional dynamic reliability.
II. Material Selection: Base it on Frequency Bands and Processability—Don’t Just Focus on Nominal Dk
High-frequency PCB materials generally fall into three categories. The first consists of thermoset materials—hydrocarbon resins filled with ceramics—exemplified by the Rogers RO4000 series. The RO4003C has a Dk of approximately 3.38 and a Df of 0.0027, while the RO4350B has a Dk of about 3.48 and a Df of 0.0037. Their processing workflow is similar to that of standard FR-4, making them the most widely used option.
The second category comprises ceramic-filled PTFE materials, represented by the Rogers RO3000 series. The RO3003 has a Dk of about 3.00 and a Df of 0.0013, while the RO3035 has a Dk of approximately 3.50 and a Df of 0.0015. The third category consists of pure PTFE-based systems, such as the RT/duroid 5880 and Taconic TLY-5, both of which feature a Dk of 2.20 and a Df of 0.0009. The 5880 utilizes non-woven glass fiber, offering excellent dielectric uniformity—ideal for phased-array applications that require extremely high phase consistency—whereas the TLY-5 employs woven glass fiber, providing superior mechanical strength and dimensional stability.
Selection by frequency band: For wireless and RF modules operating below 6 GHz, medium-Dk materials like the RO4003C or RO4350B are sufficient. For radar and millimeter-wave links in the 24 GHz to 28 GHz range, low-loss materials such as the RO3003 are necessary to keep insertion loss and phase error under control. For frequencies above 77 GHz, one should opt for low-Df PTFE-based systems or the RO3035.
There is a common misconception here. Many engineers operate under the assumption that a lower dielectric constant (Dk) is always better. While this may hold true for prototypes, it can cause complications in mass-produced antenna layers. PTFE-based materials, for instance, have low mechanical strength; during the multi-layer lamination process, the dielectric material can flow, compromising the thickness uniformity of the antenna layer—which ultimately manifests as batch-to-batch shifts in resonant frequency. For mass-production projects, a more reliable choice for antenna layers is a hydrocarbon-based material with a Dk of approximately 3.0 and a dissipation factor (Df) between 0.0015 and 0.002. While this involves sacrificing a small amount of theoretical insertion loss, it ensures greater manufacturing stability and batch-to-batch consistency.
For projects under significant cost pressure, “mixed-material” (hybrid) stack-ups are commonly used: Rogers materials are employed for RF layers, while FR-4 is used for power and control layers. A typical 4-layer configuration might feature an RO4350B top layer for the RF front-end, two middle FR-4 layers for power and digital control, and an RO4350B bottom layer for the RF output—a setup that can substantially reduce costs.
When selecting materials, consider asking three additional questions: Are the Dk and Df values measured or merely nominal? Can the batch-to-batch Dk variation be controlled within ±0.05? And can the supplier provide data on the coefficient of thermal expansion (CTE) and moisture absorption rate? While some domestically produced high-frequency substrates are now viable for cost-sensitive applications, imported materials remain the standard for critical millimeter-wave links. Furthermore, any change in material necessitates a complete re-evaluation of impedance and antenna tuning. It is also worth noting that a high glass transition temperature (Tg) does not equate to high-frequency performance; Tg relates solely to thermal resistance and has no direct bearing on Dk or Df.
III. Drilling and Via Metallization: The Process Steps with the Highest Scrap Rates
Among all the manufacturing stages for high-frequency PCBs, drilling and via metallization are the processes where the highest number of defects—and thus the highest scrap rates—occur. Drill bit lifespan varies significantly depending on the material: for standard FR-4, a single drill bit can perform over 1,000 hits, whereas for ceramic-filled materials like the RO4000 series, the lifespan drops to 300–500 hits—ceramic particles are two to three times more abrasive to cemented carbide than glass fibers. For pure PTFE-based systems, the lifespan is approximately 600–1,000 hits; the primary failure mode here is “smearing,” where the PTFE softens due to frictional heat generated by the drill bit and deposits onto the hole walls.
There are several established “rules of thumb” for drilling parameters: when processing RO4000 materials, the entry (plunge) rate should be reduced by 20%–30% compared to FR-4. For PTFE, the approach is different: increase the feed rate, decrease the retraction speed, and use high spindle speeds (over 80,000 RPM) for small-diameter holes. When drilling PTFE, use an aluminum entry sheet (cap sheet) to conduct heat away from the hole opening. Limit the number of panels in a stack: one to two sheets for PTFE, two to three for RO4000, and three to four for FR-4. Maintain a pressure foot force of approximately 40 psi.
Mixed-stack (hybrid) constructions present the greatest challenge, as a single drilling pass must penetrate both high-frequency materials and FR-4; in such cases, parameters must be a compromise, weighted toward the requirements of the high-frequency material. For blind and buried vias, laser drilling is preferred to avoid mechanical stress.
Once drilling is complete, the holes cannot be immediately copper-plated. PTFE has very low surface energy—much like a non-stick pan—making it impossible for electroless copper to adhere. Traditional potassium permanganate desmearing is largely ineffective for PTFE; instead, plasma treatment is required. A gas mixture of CF4 and O2 is used to etch away any smearing on the hole walls and roughen the surface. If this step is not performed correctly, defects may not be apparent initially, but issues such as hole-wall pull-away, “pink ring” (separation at the interface), or even fractures in the copper plating can emerge after reflow soldering or thermal cycling.
The thickness of the copper plating within the holes is another area where quality standards are sometimes compromised. For high-frequency and high-speed circuit boards, the copper plating in the holes is generally required to be at least 18 μm thick, with an average thickness of 20 μm. Vias are inherently points of impedance discontinuity. Insufficient copper plating thickness within a via not only compromises mechanical reliability but, at high frequencies, directly manifests as increased insertion loss and degraded return loss.
For inspection, it is recommended to perform metallographic cross-section analysis on 2–3 vias per board, focusing on four key aspects: the presence of gray, smudged deposits on the via walls; white spots indicating where ceramic inclusions have been dislodged; “nail-heading” (lifting) of the copper foil at the via entry/exit points; and the existence of micro-cracks along the via walls. Overlooking any of these defects creates potential failure risks when the board is subjected to mechanical stress (such as joint movement) or thermal cycling.
IV. Impedance Control: From Stack-up Calculation to Final Product Measurement
Impedance is determined by four parameters: trace width (W), dielectric thickness (H), dielectric constant (Dk), and copper thickness (T). High-speed differential pairs on robotic main control boards and communication boards typically require a tolerance of ±5% (e.g., a 100Ω differential impedance must fall between 95Ω and 105Ω). For RF sections, a 50Ω single-ended impedance often carries a tolerance of ±5% to ±10%. A common industry rule of thumb is that a 10% deviation in trace width results in an impedance shift of 5–8Ω—a margin sufficient to cause signal reflection and eye-diagram closure in the GHz frequency range.
During the design phase, use tools like Polar SI9000 or electromagnetic field simulation to calculate the stack-up, incorporating a 10%–15% margin to account for variations in raw material batches and manufacturing processes. Critical high-frequency traces should be kept as short, straight, and uniform as possible.
On the manufacturing side, control is maintained through three key processes: etching is managed via LDI (Laser Direct Imaging) combined with vacuum etching, keeping trace width tolerances within ±0.02mm (or ±0.01mm for more stringent requirements); lamination controls dielectric thickness to within ±7%–±10% (or as tightly as ±0.025mm for demanding applications); and layer-to-layer registration is maintained with an alignment error of less than ±18μm.
Deformation of the panel (array) is frequently overlooked. For large-format PCB panels, uneven heat distribution and mechanical stress during lamination, etching, and reflow soldering can cause local impedance deviations of up to ±8Ω—exceeding the standard ±5% tolerance. Mitigation strategies include: ensuring symmetrical copper distribution across the panel layout; maintaining a minimum distance of 3mm between high-speed impedance-controlled traces and V-cut grooves or perforated break-away tabs (stamp holes); utilizing low-flow prepregs (PP) for hybrid-laminate constructions; and allowing the panels to stabilize at a constant temperature after etching and lamination before proceeding to the panel-separation (depanelization) stage.
Rigorous verification is essential. Impedance test coupons should be incorporated along the panel edges, with actual impedance values derived using Time-Domain Reflectometry (TDR) and Vector Network Analyzers (VNA). For mass production, sampling should be conducted across various zones, including the center, edges, and four corners of the panel.
Robotic systems also employ specialized boards, such as those for joint actuation and power management. These boards carry high currents and often utilize copper weights of 2oz or greater. Because heavy copper is prone to significant side-etching—which increases the deviation between the designed and actual trace widths—impedance compensation must be calculated separately. For 2oz copper, minimum trace width and spacing should be relaxed to 6/6 mil; adhering to the 4/4 mil design rules used for 1oz copper would likely result in open circuits or excessively thin traces.
V. Lamination and Hybrid Lamination: Thickness Uniformity and Warpage
The lamination process for high-frequency boards cannot simply follow standard FR-4 profiles. PTFE-based materials have higher melting points, requiring elevated lamination temperatures; however, they also exhibit greater fluidity in their molten state compared to standard FR-4 prepregs. Excessive pressure can cause the dielectric layer to thin and the resin to flow out, leading to impedance drift. In practice, a multi-stage temperature and pressure profile is employed: initial heating at low pressure to soften the material, followed by a gradual increase in pressure to ensure proper filling, and finally, a holding phase for curing. For hybrid-laminate interfaces, low-flow prepregs should be selected to mitigate interlayer stress caused by differences in the Coefficient of Thermal Expansion (CTE) between the materials. Following lamination, the boards should undergo a stabilization phase at a constant temperature, allowing them to cool slowly and release internal stress; otherwise, they are prone to warpage during depanelization or component assembly.
Warpage is a common issue in complex hybrid-laminate boards. There is a significant disparity in the Z-axis coefficient of thermal expansion (CTE) between PTFE and FR-4; even slight deviations in lamination parameters can lead to warping, or in severe cases, delamination. Industry standards typically require that warping does not exceed 0.5 mm per 100 mm of length. The risk increases with the layer count—for 8- to 12-layer millimeter-wave radar boards, if the antenna-layer material is too thin or lamination is poorly controlled, micro-vias are prone to fracturing during temperature cycling tests. Furthermore, control boards located near robotic joints must also endure repeated bending and vibration.
VI. Surface Finish: Selecting Based on Frequency and Reliability
High-frequency currents tend to flow along the surface of a conductor; the higher the frequency, the shallower the skin depth. For copper, the skin depth is approximately 2.1 μm at 1 GHz, 0.66 μm at 10 GHz, 0.40 μm at 28 GHz, and only about 0.24 μm at 77 GHz. Consequently, the thin layer of metal used for surface treatment plays a direct role in the conduction of high-frequency signals. The electrical resistivities of different metals vary significantly. Using a reference unit of 1×10⁻⁸ Ω·m, the values are: silver 1.59, copper 1.72, gold 2.44, nickel 6.99, and tin 11.0. Silver has an even lower resistivity than copper, while the resistivity of nickel is more than four times that of copper.
The trade-offs between common surface finish processes are clear. Electroless Nickel Immersion Gold (ENIG)—consisting of a 3–5 μm nickel layer topped with a 0.05–0.1 μm gold layer—offers excellent flatness and is the most widely used finish. However, due to the high resistivity of nickel, it increases signal loss by approximately 19% compared to bare copper at 10 GHz, and introduces an additional insertion loss of 0.06–0.15 dB/cm at 28 GHz. Additionally, there is a risk of “black pad” defects, which necessitates keeping the phosphorus content in the nickel layer strictly controlled between 7% and 9%. Immersion silver (ImAg) provides a silver layer 0.15 to 0.4 μm thick with a resistivity even lower than that of copper. At 28 GHz, its insertion loss is approximately 0.5 dB/100 mm lower than that of immersion gold (ENIG), making it the preferred choice for millimeter-wave applications. The trade-off is that silver is prone to sulfidation (tarnishing/blackening), necessitating vacuum packaging. OSP (Organic Solderability Preservative) consists of an organic film 0.2 to 0.5 μm thick; it contains no metal layer and is the most cost-effective option, though it has limited resistance to multiple reflow cycles—typically supporting no more than two.
HASL (Hot Air Solder Leveling) must be avoided entirely; its uneven surface disrupts the path of skin-effect currents, rendering it unsuitable for any high-frequency PCB. In terms of frequency selection: for applications above 20 GHz or in the millimeter-wave band, immersion silver or OSP are preferred. For the 10–24 GHz range, immersion gold or ENEPIG (Electroless Nickel Electroless Palladium Immersion Gold) are suitable. For frequencies below 10 GHz, where long-term storage may be required, ENIG offers the best overall balance.
For robotics products, additional real-world constraints apply: boards often undergo multiple reflow cycles, and the finished units frequently operate in environments characterized by high humidity, dust, and vibration. Consequently, the corrosion resistance and shelf life of the surface finish must be carefully considered. OSP should not be exposed to the environment for more than 7 days, and immersion silver should be processed within 24 hours of opening its vacuum packaging. If such details are not explicitly noted in the procurement specifications, they can easily lead to widespread “cold” (weak/defective) solder joints during mass production.
VII. Mass Production Consistency: The Process Challenge from Engineering Prototype to 100,000 Units
Specifications met during the prototype phase are not guaranteed to be maintained during mass production; the greatest risk for high-frequency PCBs is batch-to-batch variation. Incoming material inspection is the first critical checkpoint: the dielectric constant (Dk) and dissipation factor (Df) of every batch of high-frequency laminate must be measured upon arrival. If the measured values deviate from the nominal specifications, the material should not be used, as batch-to-batch variation in the laminate is the primary cause of impedance drift.
On the manufacturing side, process capability is key. For high-reliability sectors like automotive electronics and industrial control, a Cpk (Process Capability Index) of at least 1.33 is typically required. By integrating AOI (Automated Optical Inspection) and SPC (Statistical Process Control), some in the industry have successfully reduced scrap rates from 2.1% to 0.7%. Robotics-related PCBs are characterized by a high variety of product types and small batch sizes. In this context, flexible manufacturing and proactive DFM (Design for Manufacturing) reviews are far more critical than raw production capacity. Technical requirements—such as layer stack-up, impedance, via specifications, and surface finishes—must be explicitly documented in the manufacturing files; simply mentioning the use of “high-Tg” material in a chat group is insufficient. Furthermore, the certification cycle for high-performance materials can extend up to 18 months, while lead times for key manufacturing equipment can exceed 24 months.
Conclusion
KingTop Technology specializes in PCB manufacturing and processing, serving the robotics and high-frequency electronics sectors. The core challenge in manufacturing high-frequency PCBs for robotics lies not merely in achieving fine-line circuitry, but in ensuring that the impedance requirements defined during the design phase are faithfully replicated in the final product—through precise control of material selection, drilling, etching, lamination, and surface finishing. Any lapse in these processes will manifest in the signal integrity, rather than in the physical appearance of the board.
For robotics hardware teams, we offer the following recommendations: select materials based on both frequency range and manufacturability—prioritizing stability over extreme performance parameters for antenna layers. Request impedance coupons and TDR (Time-Domain Reflectometry) reports as early as the prototyping stage, rather than waiting until mass production. Finally, ensure that all specific requirements are formally documented in the manufacturing notes so that the factory can verify them, rather than relying on informal verbal communication.
