Written by Senior Applications Engineer, Colby Hobart
Insertion loss as it pertains to this discussion is defined as the loss of power of a radio frequency (RF) signal as it propagates through a system. Insertion loss is generally an undesired effect of a necessary transmission and all efforts are made to minimize this effect. It is measured in deciBels, or dB, which is a logarithmic ratio of power at the destination port relative to power at the incident port and as such, contains no unit of power, voltage, or current. Insertion loss is commonly known as loss, S21, or Smn.
There are several reasons why a critical goal of the RF designer is to minimize system or component insertion loss. A few of these reasons are discussed below:
1- Power dissipation as heat
Some of the forms of insertion loss (dielectric loss and ohmic, metallic loss) are manifested in the power loss converted to heat in the system. This heat, when excessive, can cause several different issues in the system or component where the introduction occurs. For instance, a more intricate heatsink may need to be integrated into the system, driving up cost, or increasing the footprint of the device. If it is not possible to manage the heat in this way, the excessive heat can cause expansion of system parts, changing their electrical response, such as increasing phase or changing impedance causing excess reflections. Beyond this, uncontrolled heat can cause permanent system failure, changing the physical attributes of the components in the system beyond the point where they will return when cooled.
2- Need for higher power amplifiers
Using a RADAR system as an example, there is a power budget, which is partially driven by the insertion loss of the system. On the transmit end, a high power amplifier drives an array of antennas through a beamforming network and some other components, such as an isolator. A certain amount of power must radiate out of each of the antennas in the array in order for the overall gain of the array to be enough that the beam can reach the target and reflect enough power back into the system to discriminate the size and shape of the object being studied. Insertion loss in the beamforming network, the isolators, and the impedance transformers and other transmission lines all take away from the realized gain of the combined beam of the array. If the gain is not high enough to meet the specification, then the system will need a higher power amplifier specified, which will take more system power to operate and a higher degree of cooling.
3- More elements needed in a phased array
Alternatively, this same problem could be solved by adding additional elements to the array, which will increase system size, beamformer complexity, and the number of amplifiers used to drive the system. All of these effects are cost drivers and some may prevent the system from meeting size or weight specifications. Higher power amplifiers with higher system insertion loss will also ultimately reduce system signal-to-noise ratio that could violate the spec in an additional way, or cause the system to require additional expensive high-power filters to reduce the noise levels.
Insertion loss, which is a measure of how much power traverses a system relative to how much power is used to excite the system, is measured between two points, or ports of a system. There are four contributing factors to reduction in power sensed at port 2 relative to that injected at port 1.
1- Dielectric Loss
Since the electrical field of transmission lines is directed from a signal conductor to a ground through a dielectric, that dielectric material will absorb some signal power. The percentage of power absorbed per unit length is related to the dielectric loss tangent of the material, known as Df, or tan δ. For materials that are considered to be low loss, the dielectric loss will account for no more than 20% of the total insertion loss of a circuit. Losses in these cases will be dominated by metallic, or ohmic losses.
2- Radiation Loss
In an open system, not fully bounded by ground planes or ground vias, there is the opportunity for some power to be lost as radiation to the outside world, to an adjacent circuit, or to an open cavity in the system. This form of insertion loss is known as radiation loss and is typically a sign of something amiss in the design. It can be remedied, typically, by changing the design to have additional grounding in the areas where the radiation is occuring. A good example of an open-style transmission line is a microstrip line. An uncovered microstrip will have a small amount of radiation loss, but using a grounded cover over the microstrip line can contain this radiation so that it is not a component of the insertion loss metric.
3- Reflection Loss
All microwave systems have reflection, known as return loss, S11, or VSWR. Return loss is almost always undesired and the focus of the designer is to minimize it. It is due to system geometries that cause unintended impedance changes creating an incident reflection of power at the point of the impedance blip. Some of these geometries are caused by imperfect fabrication techniques, others by an oversight in design, and still others are caused by an unavoidable geometry critical to the design. An example of this is a transition from a stripline circuit to a via to move the signal in the z-axis. Both transmission lines are TEM (transverse electromagnetic), but the direction of propagation changes at a 90 degree angle causing some reflection. Furthermore, the via will typically have a stub on the ends due to practical fabrication.
4- Conductor (Ohmic) Loss
The final component of insertion loss and the most significant in low-loss material systems is that of the conductor, or ohmic losses. This component of insertion loss is based on the conductivity of the signal trace, the surface roughness of the trace at the points incident to the electric field, and the line definition at the trace edges. Each of those parameters will have a significant impact on the total insertion loss of the system.
The most common transmission lines on PCBs are made from copper traces. Copper is highly conductive, flexible, and strong. In RF circuits, the most common copper weights are ½ oz and 1 oz copper. This specification refers to an amount of copper per square foot. ½ oz copper equates to about 0.0007” thickness and 1 oz copper is about 0.0014”
The roughness of the copper in a transmission line and its ground will impact the amount of insertion loss in a device or system. Rougher copper makes a trace electrically longer, meaning it increases the phase of the network. Since the phase is greater and the transmission line already has a set loss per unit length based on the parameters discussed above, plus its impedance and its trace width, a longer phase will induce greater insertion loss. Surface roughness is generally specified by a metric that describes the size of the peaks and valleys. It is generally reported as Ra (roughness average), RZ (tallest peak – lowest valley), or Rq (RMS value of the sample of peaks and valleys). While these metrics do a good job of describing the magnitude of the roughness, they do not cover the periodicity of this roughness, which will play a major role in the final phase length and overall insertion loss.
For this reason, the standard metrics are good for comparing runs of the same copper type, or differences in processing of foils of the same copper type, but may fall short when comparing copper foils that were created using two different methods, such as rolled copper vs. electrodeposited. The importance of copper roughness is related to the linewidth of the signal trace. The narrower the linewidth is, the more pronounced the impact of copper roughness on phase length. So, assuming a 50-ohm characteristic impedance, a thinner substrate will require a narrower line width. For example, on a 0.0073” thick substrate Rogers Corp’s RO4350B, a standard ED copper would produce 1.7 times the insertion loss per inch that their LoPro copper would for the same 50-ohm microstrip run. This would be of enormous impact to the microwave design engineer. Looking back to some of the tradeoffs discussed earlier, this could mean 42% fewer elements in a phased array, or the same reduction in power to the LNAs. However, if the same experiment is run on 0.0173” thick laminate of the same type, the higher profile copper is 1.35 times as lossy as the LoPro, so the advantage, while still substantial, is cut in half.
There are two main types of copper used in PCB formation – electrodeposited and roller. Electrodeposited copper is formed by dissolving bulk copper in an acid and plating it onto a drum. Its grain structure is quite random looking and has tight periodicity. In contrast, rolled copper is formed by passing a large slug of copper through progressively smaller mechanical rollers, stretching it out until it is the desired thickness. The grain structure of this copper will then be fairly tight periodicity in one axis and extremely long, stretched out peaks and valleys in the other. Because of this, rolled copper acts, electrically, almost identical to the theoretical models of perfectly smooth copper. While electrodeposited copper does sound inferior to rolled copper in this comparison, it does have its merits. Its grain structure allows for better adhesion to the laminates. It also can be used with different nickel oxide resistive coatings that can be used to create etched resistors directly on the PCB. There have also been some strides made in recent years on electrodeposited copper to drive down insertion loss by making the Ra surface roughness extremely small, sub-1 micron. By making the copper extremely smooth, its effect on insertion loss is getting very close to that of rolled copper. Rolled copper is also very expensive and only available on a restricted selection of laminates to which it can achieve acceptable adhesion.
Copper roughness is driven by more than just the way in which it is formed. After the copper supplier forms the copper foils, they are adhered in a heated press to a dielectric laminate system at the laminate supply house. The foil often needs an adhesion promotion processing step before this lamination occurs, to increase the roughness to a level needed for good adhesion and to pass the laminate supplier’s peel strength step. There is also a topology to the laminate itself, which, if rigid enough, may imprint on the copper being adhered. Beyond this, at the PCB shop there are a number of processes in the fabrication of the PCB which may increase the surface roughness of the copper. Many board processes require cleaning after, which may include a scrub that can add roughness to the copper. Some steps may require a sanding operation, which can do the same. The board shop may also require an adhesion promoter to traces before the different PCB cores are laminated together with a prepreg or other adhesive. Conversely, there could be plating steps at the PCB shop, which could smooth out the copper on the exposed side to some level.
Fortify is working now on characterizing a series of structures with different native print roughnesses to understand the impact on insertion loss. Since the copper in this case is additive, it will take on the roughness of the 3D printed substrate. This substrate will have roughnesses on the order of stock copper foils, but will have different periodicities, based on the print parameters used. For this reason, it is unlikely that insertion loss prediction models, such as Hammersted, or Hall-Hurray, used today for stock copper foils, will accurately predict the expected loss of these 3D structures. Follow Fortify on LinkedIn and subscribe to our website to stay tuned for the updates on these experiments and case studies.
It has been shown that it is of top importance to microwave component and system engineers to design for low insertion loss. This can translate to cost, weight, and power savings in a system, which can directly drive if a system is an adequate solution to the problem at hand. While a number of aspects can affect insertion loss, such as reflections, radiations, and substrate loss factor, a critical piece of this is the surface roughness of both the signal conductor and ground copper of a transmission line. The difference between rough and smooth copper can be nearly a 2X factor in overall insertion loss, which could mean a system would need almost full redundancy to make up for the extra loss due to rough copper. Rough copper can come from how it is formed, treated at the laminate supplier, or processed in the PCB shop, so it is important to monitor the roughness and processes at all three of these stages. It is important to choose a copper that will be smooth enough after all processing to achieve the system insertion loss spec, while meeting the goals of cost, adhesion, and current capacity for the system. Additionally, while typically adequate for traditional copper foils, reporting only the roughness value as a function of peak and valley height can fall short in predicting insertion loss of a transmission line. The periodicity of the roughness is also a significant factor that becomes more obvious with new manufacturing methods, such as 3D printed substrates. Fortify is working to show this significance on a series of additive transmission lines, whose results will be reported in a subsequent blog.