The verdict is in: more designers will need to get comfortable working at microwave and mmWave frequencies. As more devices become mobile-capable and must work with 5G in the future, and as autonomous vehicles, UAVs, advanced IoT products, and radar-enabled robots become more commonplace in daily life, designers will need to understand how to create analog layouts that ensure signal integrity at mmWave and higher frequencies.
5G has barely rolled out in the US, and researchers in Europe are already looking 6G. The pace of evolution is truly breathtaking. The telecom industry figured out long ago that you can avoid many signal integrity problems by switching to fiber; now we have data rates in the 10’s to 100’s of Gbps in multiple lanes, and ICs are still catching up to these capabilities. RF designers working at 40 GHz and above will have to contend with signal rise times (~20 ps, or an equivalent knee frequency of 50 GHz) only seen in data center and optical networking environments. Here at NWES, we want to give innovators the information they need to prepare their new products for the future, and this means ensuring PCB designers and our prospective clients have a solid understanding of RF PCB layout techniques.
RF PCBs are a different beast than a typical digital PCB. The signal integrity problems that can arise above WiFi frequencies (i.e., 10’s GHz and higher) occur due to fundamental physical mechanisms in the substrate material. Most RF designers have worked up to the 5 GHz range in the past, but this is set to change going forward, and new products will push analog devices to their limits.
Personally, I come from an optics background, so I’ve always been more of an analog guy rather than a digital designer. Many PCBs will eventually start to look a lot more like optical devices in that we will inevitably move to electronic-photonic integrated circuits (EPICs) and PCBs, and ultimately fully photonic PCBs. In the meantime, more designers will need to become comfortable with faster data rates, which are accompanied by faster rise times and higher frequencies.
In this post, I want to focus on two important areas of RF PCB design and layout: routing and signal integrity. Obviously, the two are related, but there are many more areas to explore. I’ll be producing a series of posts on these and other topics in the coming weeks. For now, let’s focus on trace geometry and losses on mmWave interconnects as these two aspects of your RF PCB layout will be major determinants of signal integrity.
Your trace geometry and layer stack will have a major effect on signal integrity as you route high frequency analog signals throughout your board. Newer designers will likely default to a microstrip geometry, possibly with guard vias to provide isolation between different circuit blocks and to suppress radiated emissions. The question of which trace geometry to use is about more than isolation; it is also about avoiding losses due to dielectric absorption, roughness losses, and scattering from your PCB substrate.
The alternative to using microstrip transmission lines is to use stripline transmission lines. Routing striplines on a dedicated interior layer provides isolation from the surface layer, although cavity resonances and radiated EMI from substrate edges can still be a problem. A better transmission line geometry is grounded coplanar waveguide (GCPW) routing. This is much better for routing on a surface layer than microstrip traces or stripline traces. This geometry provides much lower losses and stronger isolation compared to these other two routing schemes.
Fig. 1: Example edge-coupled GCPW structure for mmWave frequencies. Thanks to user someonr on StackExchange.
Transmission lines should be routed as controlled impedance lines to ensure consistent impedance throughout the interconnect. The level of damping seen by transients in your transmission lines should be maximized. This will help reduce interference between interconnects due to parasitic coupling between signals.
This need to reduce ringing is critical for mmWave signals that use ASK, PSK, QAM, QPSK, or any other signal modulation scheme that uses sudden changes in phase or amplitude to encode digital data. FSK and other frequency modulation schemes are sufficiently smooth that any transient response will be weak.
In order to reduce losses, ensure isolation, and properly size your conductors, you’ll need to choose an appropriate trace geometry for use in your RF PCB layout. I would argue for using GCPW routing on the surface layer of your PCB and symmetric stripline routing in an interior layer. However, you should preferentially route on the surface layer as GCPWs will provide high isolation (typically -45 dB), lower losses, and lower dispersion than a microstrip or stripline trace with the same impedance.
If you look at data on insertion loss in different trace configurations on FR4 (see Fig. 2), you’ll find that microstrips traces losses tend to increase at a faster rate than losses in GPCWs at frequencies beyond ~25 GHz. Even if you use guard vias on these microstrips, you’ll find that losses in microstrips become higher than losses in GPCWs beyond ~40 GHz. This makes GPCWs ideal for routing at mmWave and higher frequency bands.
Fig. 2: Insertion loss in microstrips, isolated microstrips, and grounded coplanar waveguides. Thanks to John Coonrod from Rogers Corp for compiling this data.
In addition to dielectric absorption in the substrate, the progressively higher losses in microstrip and stripline traces is due to copper roughness combined with the skin effect. At progressively higher frequencies, more current becomes confined near the edge of a conductor. Rough etching on copper surfaces creates surface roughness, which is a well-known source of DC losses. The roughness between the deposited copper foil and the substrate in microstrips and striplines tends to have the highest roughness, thus it will incur the highest resistive losses.
This should illustrate the advantage of GCPWs: the skin effect still occurs, but the ground planes surrounding GCPWs draws the currents to the horizontal boundaries of the waveguide (which borders air), where the surface roughness is very low. No matter which routing geometry you use, you should also use a substrate with low loss tangent at the relevant frequency. GCPWs are a better choice for FR4, while microstrip and stripline will operate better on Rogers or Isola laminates at mmWave frequencies.
A recent simulation study provides and excellent comparison of insertion and return loss spectra for grounded and ungrounded coplanar waveguide interconnects on thin polyerm substrates. Their results show excellent insertion less (less than 15 dB) up to ~80 GHz, making these interconnects an excellent choice for innovative designs on unique substrates.
RF PCB layout for mmWave frequency devices takes the right design tools and analysis experience. If you’re looking for a knowledgable firm that offers cutting-edge PCB design services and technology research services for innovative electronics companies, contact NWES for a consultation.