Circuit performance may start with the choice of printed-circuit-board (PCB) material, but achieving a desired level of circuit performance can also have a great deal to do with how circuits are fabricated on a chosen PCB material. Attendees to the MicroApps sessions at the recent IEEE 2015 International Microwave Symposium (Phoenix, AZ, www.ims2015.org), in particular, the Rogers Corp. MicroApps session, “PCB Fabrication Influences on Microwave Performance,” learned how such factors as circuit-board thickness and choice of transmission-line metal can impact the final performance of both active and passive circuits fabricated on a particular PCB material.
Insertion loss is usually an important parameter for most high-frequency circuits, especially where signal power is limited, and for RF/microwave printed circuits, insertion loss is highly dependent upon the choice of PCB material thickness. To demonstrate, three high-frequency 50 ohm circuits were made and modeled from the same circuit-board material, but at three different thicknesses. The material was RO4835™ circuit laminate from Rogers Corp. and the thicknesses were 6.6, 10.0, and 30.0 mils. RO4835 rigid thermoset laminate material has a dielectric constant of 3.48 at 10 GHz through the z-axis (thickness) of the material, controlled to a tight tolerance of ±0.05. This circuit material exhibits thermal conductivity of 0.69 W/m/°K and features excellent dimensional stability in the x-y plane. But what some designers may not realize when using this material is that the choice of thickness does matter, especially regarding insertion-loss performance.
All three circuits were modeled, fabricated, and measured, to compare simulated and measured performance levels for the different PCB thicknesses. Modeling was performed with the aid of the MWI-2014 simulation software from Rogers Corp., using Hammerstad and Jensen closed-form equations. Simulations were compared with measurements performed on a wideband vector network analyzer (VNA), a model E8346C from Agilent Technologies (now Keysight Technologies), capable of performing broadband S-parameter measurements from 10 MHz to 50 GHz. The swept-frequency simulated and measured results for each different circuit thickness agree quite closely across a modeled/measured frequency range of DC to 20 GHz. It is the ways in which the total losses for each thickness of PCB material break down in the simulations that are quite different.
The measured and modeled swept-loss plots showed total insertion loss. The modeled total insertion loss, however, is further broken down and compared in terms of dielectric and conductive circuit losses for each thickness of circuit-board material. Any fabricated circuit can be evaluated in terms of its insertion-loss components, which include the dielectric loss of the circuit material, the conductor loss of the circuit traces, the leakage loss of the PCBs, and the radiation loss of the circuit traces. In this presentation, two of the four insertion-loss components, dielectric loss and conductor loss, were examined for the three different PCB thicknesses to better understand how circuit thickness played a role in this important high-frequency circuit performance parameter.
The dielectric losses for the three thicknesses of PCB material are quite close in value, increasing steadily with frequency and with an overall increase in modeled dielectric loss as a function of frequency. But these differences are very slight—almost negligible when comparing the modeled dielectric losses for the 6.6- and 10.0-mil circuit materials. The largest differences in simulated dielectric losses occurred between the thinnest and thickest circuit materials, with about 0.2 dB/in. dielectric loss at 15 GHz for the 6.6-mil-thick RO4835 material compared to about 0.25 dB/in. dielectric loss at 15 GHz for the 30-mil thickness of the same circuit material.
Simulated conductor losses, on the other hand, were not quite as similar for the three thicknesses of PCB material, with this loss component of PCB insertion loss increasing steadily as thinner circuit laminates are used. For the thickest of the three RO4835 PCB materials, the modeled conductor losses remained under 0.1 dB/in. of transmission line through about 10 GHz and only slightly above 0.1 dB/in. of transmission line through 20 GHz. In comparison, for the thinnest of the three RO4835 PCB materials, the modeled conductor losses were slightly less than 0.4 dB/in. at 10 GHz, climbing to about 0.6 dB/in. at 20 GHz. Conductor losses for transmission lines on the 10-mil-thick RO4835 PCB material were just about midway between the conductor loss values for the thinnest and thickest of the circuit laminates that were modeled.
As a further examination of how fabrication choices can affect PCB performance, microstrip edge-coupled bandpass filter circuits fabricated on RO4835 circuit material were compared for circuit laminates with bare copper transmission lines and for circuit laminates with copper transmission lines having solder mask protection. Solder mask is a polymer layer added to PCB copper traces to protect against the effects of oxidation and to prevent solder bridges from forming between closely spaced circuit traces. Both filter circuits were fabricated on 20-mil-thick RO4835 circuit laminates, identical except for the solder mask. Adding the solder mask provides reliable long-term protection against the deleterious effects of oxidation, but it also results in some tradeoffs, such as additional transmission-line insertion loss. For example, the modeled insertion loss for the filter circuit with bare copper transmission lines was about 0.25 dB/in. at 10 GHz, climbing to about 0.50 dB/in. at 20 GHz. In comparison, the modeled insertion loss for the filter circuit with copper transmission lines covered with solder mask was slightly more than 0.30 dB/in. at 10 GHz, rising to about 0.60 dB/in. at 20 GHz.
Perhaps even more significant, especially in the design process for such a filter, the choice of using or not using solder mask on the PCB material made a difference in the location of the filter center frequency and quality factor (Q), with the center frequency slightly lower for the filter using PCB material with solder mask. S-parameter measurements on the microwave VNA revealed a center frequency of 2.9499 GHz and a Q of 8.7993 for the filter with solder mask, and a center frequency of 3.0144 GHz and a Q of 8.9627 for the filter with bare copper conductors. The bandwidths for the two filters were almost identical, at 335.25 MHz for the filter with solder mask and 336.32 MHz for the filter with bare copper transmission lines
Of course, these are just a few examples using RO4835 circuit laminates, but they point to the importance of carefully considering PCB fabrication approaches and material parameters before starting a design. As shown, such parameters as thickness and use of solder mask can affect performance and perhaps make the difference between achieving circuit performance results in just one design iteration versus having to perform multiple design iterations. For those interested in further information on these comparisons, using additional RO4000® series materials, copies of the MicroApps presentation are available for free download from Rogers Corporation’s Technology Support Hub at www.rogerscorp.com/techub.
Do you have a design or fabrication question? Rogers Corporation’s experts are available to help. Log in to the Rogers Technology Support Hub and “Ask an Engineer” today.