The 5G wireless network has a special requirement for the board material working on the 5G circuit at the millimeter wave frequency because it covers a wide frequency band. This paper discusses the effect of the surface roughness of the inner wall of the metallized via for transmitting signals between the top copper foil and the bottom copper foil of the PCB material on the final RF performance of the material.

The fifth-generation wireless network is hailed as one of the most important technological achievements in modern communication. The 5G technology uses both signal frequencies below 6 GHz and millimeter-wave frequencies for short-range backhaul and high-speed data links. Circuits in such a wide frequency range require special board materials, and Rogers' RO4730G3TM board material is the choice of many circuit design engineers because of its excellent performance from RF to millimeter wave frequencies. However, one difference between the presence of such laminate materials and conventional circuit materials is that the material uses hollow microspheres as a dielectric fill material, a difference that has caused some circuit designer concerns.

The texture of the wall surface of the metallized vias of all circuits has different subtle differences, even when comparing the roughness of the surface of the hole wall of the same board. Since the drilling process involves a number of factors, the surface of the wall of the metallized via may vary from hole to hole. In materials with microsphere fillers, the drill bit may or may not affect the microsphere filler, resulting in a difference. When the drill bit impacts and breaks the hollow sphere, the copper plating of the via will grow along the contour of the broken sphere, and the surface of the pore wall will no longer be smooth and flat. Figure 1 shows how the presence of microsphere filler in a circuit board affects the increase in surface roughness caused by the formation of metallized vias in the circuit material. It is natural to question whether this roughness can adversely affect the electrical performance or reliability of a circuit compared to conventional circuit materials that have a smoother surface with metallized vias.

With the increasing demand for high frequency circuit materials in a wide frequency range in 5G wireless networks, it is very meaningful to know whether the surface roughness of metallized vias in the board material with hollow microsphere fillers has an impact on circuit performance, because There is no such filler in conventional circuit board materials. Through a series of studies, a 20.7 mil thick RO4730G3TM circuit board material with glass reinforcement and microsphere filler from Rogers and a 20 mil thick RO3003G2 material without glass reinforcement and a smaller and non-hollow filler were compared. Whether the difference in the wall of the hole will have an impact. In order to test whether the surface roughness of the hole wall has an effect, we have developed a number of different test circuits to compare the metallized vias on the board in the 5G wide frequency range.

The test circuit is based on a microstrip transmission line structure with a via in the middle of the circuit for conductor and signal transitions from the top copper layer of the dielectric substrate material to the bottom copper layer. The length of the test circuit is basically about 2 inches. We also used other high frequency transmission line technologies as a reference to evaluate the effect of metallized via wall surface roughness, including 8-inch and 2-inch long microstrip circuits without signal vias, and 8-inch and 2" long grounded coplanar waveguide (GCPW) circuit without through holes. To ensure consistency in measurement, the test used the same two 2.4 mm coaxial connectors for testing all circuits. And the test connector is always connected to the test port of the VNA in the same way to maintain phase consistency.

Designers accustomed to studying microscopic images of printed circuit boards (PCBs) as shown in Figure 1 may be concerned about the effects of the roughness of the metallized vias, especially at the high frequency of 5G circuits. In general, for conventional high frequency circuit materials that do not use microsphere filling, a rough hole wall surface may mean that some problems occur during the manufacturing process and may affect the reliability of the via holes. However, for the hollow microsphere filled circuit material, it is normal to form a rough metallized via, which does not mean that its performance is poor. To demonstrate that the rough metallization vias in this circuit material do not affect via reliability and electrical performance, we have new materials (rougher metallized vias) and more traditional circuit materials (smooth metallization) Through-hole studies are conducted to eliminate any doubts about the use of this material for 5G wireless network circuit design and any other circuit that applies to the millimeter-wave frequency range.

Designers accustomed to studying microscopic images of printed circuit boards (PCBs) as shown in Figure 1 may be concerned about the effects of the roughness of the metallized vias, especially at the high frequency of 5G circuits. In general, for conventional high frequency circuit materials that do not use microsphere filling, a rough hole wall surface may mean that some problems occur during the manufacturing process and may affect the reliability of the via holes. However, for the hollow microsphere filled circuit material, it is normal to form a rough metallized via, which does not mean that its performance is poor. To demonstrate that the rough metallization vias in this circuit material do not affect via reliability and electrical performance, we have new materials (rougher metallized vias) and more traditional circuit materials (smooth metallization) Through-hole studies are conducted to eliminate any doubts about the use of this material for 5G wireless network circuit design and any other circuit that applies to the millimeter-wave frequency range.

In fact, in a number of research tests on such circuit board materials and their microsphere fillers, we used two materials with different metallized via wall features to study the surface roughness variation of metallized via walls. Comparison of various effects brought about by RF performance. The research test is based on a specially designed microstrip transmission line circuit with microstrip lines on the top and bottom layers. The intermediate medium is a dielectric material that connects the top-to-bottom microstrip lines through metallized vias. These tests are designed to provide a very meaningful data reference for 5G applications, so the test circuit has good RF performance in the 100 MHz to 40 GHz range.

The dielectric constants (Dk, or εr) of the two materials used in this study were very close, with values around 3. Both materials are also made of materials of the same thickness, both 20 mils. The main difference between the two is that one of them can make a metallized via with a smooth surface of the hole wall, while the surface of the other fabricated metalized via hole is rough. The material that can be used to form the smooth metallized via wall surface is Rogers' RO3003G2TM circuit board material, while the RO4730G3TM circuit board with glass reinforcement and hollow microsphere filler produces a rough surface of the metalized via wall.

The difference in texture of the circuit metallized via wall surface is generally considered a problem in circuit fabrication, not a material issue. However, some material properties can optimize the surface of the metallized via wall, including the type of circuit material filler, filler size, glass reinforcement, and resin type. As RO4730G3TM circuit board and its hollow microsphere filler (rough metallized via wall surface), the RO3003G2TM circuit board material is comparatively glass-free and the filler particles are very small. Assuming that both use the best PCB processing method, the latter will have a very smooth metallized via wall surface. As shown in Figure 2, it is a very smooth metallized via wall that can be formed by the RO3003G2TM board.

Figure 2. The micrograph shows a smooth surfaced metallized via wall formed in a 20 mil thick RO3003G2 circuit material.

For the two circuit materials of the same thickness, the difference in surface roughness of the metallized vias of the two materials shown in Figures 1 and 2 is very obvious. Observing the two figures may raise the question of whether the higher surface roughness of the metallized vias means that there is a problem with its RF performance. For test circuits, the microstrip transmission line circuit is an effective method to compare the effects of smooth and rough metallized via wall surfaces on RF performance, because microstrip lines are processed during manufacturing compared to other high frequency transmission line structures. Some changes have little effect on RF performance.

In order to provide meaningful results for metallized vias in different circuit materials at 40 GHz, we have put a lot of effort into optimizing these microstrip circuits. One of them is the signal transition from the RF test connector to the PCB microstrip line is a big design challenge. In general, it is difficult to obtain better echo characteristics on the signal transition of a microstrip transmission line on a 20 mil thick circuit board, especially a transmission line having a frequency above 25 GHz. For wideband microstrip circuits, return loss of less than 15 dB or better is generally considered acceptable.

Through-hole transition is another important factor to consider, especially at millimeter-wave frequencies where it is difficult to achieve low-loss transitions from one layer to another. In general, it is difficult to achieve good performance of microstrip line via transitions above 20 GHz on 20 mil thick circuit materials. However, considering the above difficulties, the microstrip test circuit of this study is designed to achieve good results when the frequency reaches 40 GHz, as shown in Figure 3.

Figure 3. These circuits are used to evaluate the effect of the surface roughness of the metallized via walls on the RF performance at high frequencies. The left picture shows the standard microstrip transmission line and the right side is the microstrip line with metallized vias. Circuit.

The "standard" microstrip line circuit shown on the left side of Figure 3 is a microstrip circuit that implements signal transition conversion through a grounded coplanar waveguide (GCPW) structure. The body of the circuit consists of a microstrip transmission line that is used at the end of the circuit for a coaxial (2.4 mm) connector to microstrip transition (Southwest Microwave Model #1492-04A-5). The circuit on the right side of Figure 3 is the top and bottom circuit of the test circuit used in this study. They are loosely coupled grounded coplanar waveguides with metallized vias in the middle that provide a transitional connection from the top layer to the underlying circuitry. The test circuit is 2 inches in length and the loosely coupled grounded coplanar waveguide transmission line circuit will have very similar RF performance as the microstrip transmission line circuit. Loose coupling has good performance at higher frequencies and is ideal for testing at 40 GHz.

Figure 4. This is an example of S-parameters for metallized vias with different circuit surface textures from network analyzer tests, including frequency and time domains, respectively.

Figure 4 is a graph of the results of the frequency domain and time domain measured by a vector network analyzer. The two markers of the return loss (S11 and S22) in the lower right corner of the figure represent the return loss values at different frequencies. Mark 2 is located at 40.7 GHz and is the highest frequency at which the test circuit has good return loss. The impedance of the reflected wave S22 is shown in the upper right corner of the figure, and the impedance of the reflected wave S11 is shown in the lower left corner of the figure. As indicated by the mark of S11, the impedance values in the via conversion, marks 1, 2 and 3, the circuit has an impedance of approximately 48 Ω. A small impedance change can be observed in the via transition region with an impedance variation of less than 2 Ω, with little effect on the RF performance of the circuit. From these test results, the circuit can be considered to have a good through-hole transition from the top-to-bottom signal, while it also has good insertion loss performance up to 40 GHz (as shown in the upper left hand corner). Many identically designed circuits are fabricated on the same large PCB to better understand the changes caused by normal material variations and PCB manufacturing processes leading to changes in RF performance. We also processed two large PCB boards (Board 1 and Board 2) with multiple test circuits on it, and the two large boards came from the same and larger area of the same circuit material.

The larger slab has a raw material size of 24 x 18 inches and is cut into two 12 x 18 inch slabs that maintain material consistency on both 12 x 18 circuits due to the same large slab. In the fabrication of the microstrip line test circuits of the two selected 20mil RO3003G2 and 20.7mil RO4730G3 materials, the same circuit fabrication process and process were used to reduce the impact of processing.

Comparison of test results

Through the research and testing of circuit materials, a large amount of test data is obtained, including the insertion loss, return loss, impedance, group delay and phase angle (as shown in Figure 4). Through-through measurements are used as a means of determining the effect of metallization vias on circuit performance. The impedance of the circuit is also measured, but it is not considered to be the best indicator of the effect of metallized vias on RF performance. The impedance of the microstrip line circuit (or loosely coupled grounded coplanar waveguide) is in turn affected by parameters such as dielectric thickness, conductor width, copper thickness variation, and dielectric Dk. The impedance in the metallized via transition region will be more affected by these variables than the effect of the metallized via wall surface. For the above reasons, although impedance data was collected, the impedance was not used to judge the effect of the metallized via wall surface on RF performance.

The phase angle of S21 is a measure of the RF variation of the circuit that is used as a change in the surface of the metallized via hole wall because the surface roughness of the conductor along the microstrip transmission line will affect the phase angle 1, 2 of the signal through the transmission line. The through measurement is sensitive to the RF signal path with the converted via. To verify the accuracy and repeatability of the test, a repeatability study conducted on one of the test circuits found that the standard deviation of the S21 phase angle measured at 39 GHz was less than ±1.2 degrees. The S21 phase angle we used in the test is the unwrapped phase angle of S21, which is the sum of the absolute values of the phase angles of -180 to +180 degrees. What makes more sense with this approach is the increased resolution, because even for frequencies up to 39 GHz in 5G applications, the resolution of non-expanded phase changes is less sensitive. However, for a 2-inch long microstrip transmission line on a Dk board material of approximately 3, the unfolded phase angle range at 39 GHz can reach thousands of degrees, so test circuits and measurement schemes provide suitable phase resolution.

Although the data collected in the metallized via wall study is broad, some results can still be shared here. For example, Figure 5 shows data for six different circuits of the same design fabricated on the same board and compared to a microstrip transmission line without a via transition as a reference. Figure 5 also shows the data for six different circuits of the same design made on the second board (the two boards were originally cut from the same piece of 24 x 18 material). The test results are based on a 20 mil RO3003G2 with a smooth metallized via wall surface.

Figure 5. The phase angle measurement of the S21 deployment is a 2-inch long microstrip transmission line circuit with metallized vias. The board material is a 20 mil thick RO3003G2 which provides a very smooth metallized via wall surface.

The circuit ID in Figure 5 can show which 12 x 18 inch slab the circuit is from, and the circuit ID number on the board. For example, P1 C4 comes from board 1 and the circuit number is 4. The circuits are spaced apart from each other and evenly on a 12 x 18 inch panel to maintain consistency. Some changes are expected in advance because they are very sensitive to differences in phase angle. Some of the changes are due to the PCB manufacturing process, not the roughness of the metallized via walls, including variations in conductor width, changes in copper plating thickness, and variations in drill quality. In addition, the gap around the metallized vias may vary somewhat due to the normal manufacturing tolerances of the PCB. Similarly, small material changes on each board, such as small changes in Dk values, can also cause phase changes. Considering the test values shown in Figure 5, the standard deviation of the phase data repeatability at 39 GHz is less than ± 1.2 degrees, which is very good.

Although not a factor in the measurement, the D47 tolerance of the RO4730G3TM circuit material is considered to be very good performance. However, at higher frequencies, even a slight Dk change is sometimes a significant effect. For example, at 39 GHz, a Dk shift of 0.05 would result in a phase angle change of approximately 15.3 degrees. For a tolerance of ±0.05 or a total Dk offset of 0.10, the phase angle at 39 GHz may move up to 30.6 degrees due to variations in circuit material Dk. This value has a good reference when considering the number of phase angle changes in Figure 5. However, since the board of circuit material evaluated as these metallization vias are all from the same original slab, the phase angle variation in this study due to Dk variations will be small. Figure 6 provides a comparison of circuits with smooth metallized via walls (repeated test data from RO3003G2TM of Figure 5) and circuitry with rough metallized via walls (RO4730G3TM).

Figure 6. Comparing the phase angle difference statistics of microstrip transmission line circuits fabricated on different boards at three critical 5G frequencies. The data on the left is the result of a smooth metallized via wall surface circuit, and the data on the right is the result of a rough metallized via wall surface.

As mentioned earlier, during the research process, we have minimized the effects of material changes, such as plates 1 and 2 taken from the same large plate to ensure the smallest difference in material Dk. Therefore, the change in phase angle and any difference in appearance are mainly affected by the circuit manufacturing process. When analyzing the results of the circuit of the same board, the difference in phase angle at this time comes from the least impact of PCB manufacturing and material changes, because the same board is processed completely simultaneously. Because of this, studying multiple circuits on the same board provides a good understanding of the quality of metallized vias in microstrip lines. The PCB fabrication process may also result in a rougher metallized via wall surface than would be expected. As shown in Figure 6, there is a certain change in the S21 unfolding phase angle on each board, but this change is not significant when comparing the phase changes of the circuits on two different materials.

Figure 7. Surface characteristics of the metallized via wall (rougher) from the top to bottom line of the RO4730G3 material and phase measurements at 3 millimeter wave frequencies.

Obviously, by observing the photomicrograph, the surface walls of the metallized vias used to connect the top layer to the bottom line may exhibit a large difference. For example, Figure 2 shows an ID of P1/C1 as a circuit metallization via made on a 20 mil thick RO3003G2 material with a very smooth metallized via wall. Figure 7 shows the appearance of a circuitized via of P2/C6, which is a via on a 20.7 mil thick RO4730G3 board material. The surface of the metallized via wall on this material is relatively rough. From the appearance alone, there may be some concerns as to whether the surface roughness of the metalized via hole wall will affect the RF performance. But as the previous studies have shown, the difference between the rough and smooth metallized via sidewalls is only surface, at least for these test circuits at 40 GHz, there is no need to worry about them for RF/microwave. / Millimeter wave performance will have a performance impact.

参考文献

1. J. W. Reynolds, P. A. LaFrance, J. C. Rautio, and  A.F. Horn III, “Effect of conductor profile on the insertion loss, propagation constant, and dispersion in thin high frequency transmission lines,” DesignCon 2010.
2. A. F. Horn III, P. A. LaFrance, C. J. Caisse, J. P. Coonrod, and B. B. Fitts, “Effect of conductor profile structure on propagation in transmission lines,” DesignCon 2016.