Microwave Journal

Using Off-Chip Passive Components to Maximize GaN Performance & Reduce Cost

October 13, 2021

Decades of research and development dedicated to GaN RF power semiconductor technology has led to an increasing supply of affordable RF power devices with impressive performance. GaN semiconductors have reduced material capacitance and enhanced electron mobility, resulting in remarkably low conduction losses, considerably faster switching times and higher frequency-temperature and frequency-voltage characteristics than silicon technologies. Extensive lab testing conducted by numerous sources consistently demonstrates these performance advantages over competing technologies, which has hastened the deployment of GaN power devices in numerous applications. Now, design engineers worldwide are harnessing these compact, low loss and fast switching semiconductors to develop smaller, lighter and more reliable systems that extend the capabilities of solid-state RF power design.

With the many benefits of GaN come a set of challenges with new circuit designs. For example, passive components at the output of a GaN device can reduce the output power of the active component. Even if the passive components don’t introduce excessive loss, they can degrade GaN’s ability to operate at maximum performance. Some of these design challenges can be overcome by using high performance passive components, such as advanced capacitors and surface-mount technology (SMT) heat pipes. This symbiotic relationship between passive components and GaN RF power devices makes the capacitor selection process critical.

This article addresses several high performance passive component technologies that pair well with GaN to provide impedance matching, bias filtering, DC blocking and thermal control, helping GaN power devices operate optimally.


Single-layer capacitors (SLC) are comprised of a single ceramic dielectric layer with terminations for conductive epoxy attachment and wire bonding. Providing good performance through 40 GHz, they can be used in internal and external configurations—for example, playing an integral role in maximizing the power transfer of a GaN power amplifier as part of the impedance matching networks. When placed inside a device package, SLCs can be elements of a matching network between the lead frame and gate of the transistor, helping provide a broadband impedance match at the input of the device. Used outside the package, SLCs can be used for impedance matching, DC blocking and broadband bypassing. SLCs can be configured as single, dual or multiple SLC arrays to minimize component count.

Most of the electrical characteristics of SLCs are determined by the ceramic dielectric used for their construction. The two most common dielectrics are SiO2 and C0G (NP0) EIA Class I temperature-compensating ceramic. Both have high temperature stability (0 ±30 ppm/°C), ideal for impedance matching where temperatures are high and thermal stability is critical. A new type of dielectric, the grain boundary barrier layer (GBBL) material system, has demonstrated noteworthy performance as a replacement for the general-purpose Z5U and Y5V ceramic dielectrics where bulk capacitance is a concern. A typical GBBL dielectric exhibits X7S temperature characteristics with better temperature stability compared to the Z5U and Y5V dielectrics (see Figure 1).

Figure 1

Figure 1 Temperature stability of Z5U, Y5V and X7S dielectrics. The GBBL material system has X7S temperature characteristics.

Figure 2

Figure 2 Bordered (a) and non-bordered (b) SLCs. Excess conductive epoxy can climb up the sidewall of a non-bordered SLC and cause a short. Source: TJ Green Associates, LLC.1


SLC terminations are typically comprised of sputtered TiW/Au or TiW/Ni/Au. This combination of sputtered materials yields thin, high-quality termination surfaces with excellent adhesion, essential for conductive epoxy attachment and wire bonding, particularly with high-power RF devices subject to severe temperature cycling. The terminations can be bordered, which means the metallization does not extend to the capacitor edges, or non-bordered, which means it does (see Figure 2). Bordered SLCs minimize the chance of conductive epoxy climbing the sidewalls to touch the top plate and short the capacitor. Ideally, epoxy fillets should flow about halfway up the side of an SLC, but since MIL-SPEC requirements don’t specifically specify conductive epoxy fillet height on die edges, using bordered SLCs can avoid the possible negative outcome from overflow.1 Non-bordered SLCs are typically used in source bypass configurations because they reduce the bond length between the top termination and the active device.

Figure 3

Figure 3 Gate (a) and drain (b) bias filter banks. Source: Qorvo.2


GaN power devices require a stable bias voltage for optimum operation. Since bulk capacitor banks are stable with voltage and temperature and have good aging characteristics, they are often used to filter bias line noise and provide a fast source of charge to supply the high current changes (Δi/Δt) in a power amplifier (see Figure 3). The transient response of these filter capacitor banks is determined by the combination of high capacitance and high frequency response. Bulk capacitors suitable for voltage bias banks include high-CV multilayered ceramic capacitors (MLCC) and tantalum, tantalum-polymer, aluminum and aluminum-polymer electrolytic capacitors (see Figure 4).

Figure 4

Figure 4 Capacitance stability, DC bias and temperature performance for tantalum polymer (a), high-CV MLCC (b) and aluminum electrolytic (c) capacitors.

Figure 5

Figure 5 Tantalum capacitor packaging options.

Table 1

While high-CV MLCCs can achieve the capacitance ranges required for many bias networks, they don’t provide stable capacitance values across operating conditions such as temperature, time and DC bias.3 For example, the capacitance stability of a 100 μF X5R MLCC can vary from 100 μF at 25°C to approximately 85 μF at -55°C and 80 μF at 125°C. High-CV MLCCs also suffer from DC bias voltage effects that can significantly reduce the capacitance value present in the circuit. For example, the capacitance value of a Class II MLCC can decrease by 35 to 65 percent at the fully rated DC current. Additionally, low voltage AC ripple current can further reduce the capacitance of high-CV MLCCs by another 5 percent, and aging can reduce the capacitance by another 2 to 5 percent per decade. Depending on the operating conditions and the MLCC chosen, all these losses combined with the temperature coefficient can reduce the total expected capacitance of high-CV MLCCs by approximately 80 percent.

The remaining bulk capacitor options suitable for bias filtering include traditional and polymer versions of tantalum and aluminum electrolytic capacitors. Although this article focuses on tantalum bulk capacitors for bias filtering, aluminum capacitor technology and performance are also improving. Tantalum capacitors have size, weight and stability advantages over traditional aluminum electrolytic capacitors. For example, tantalum capacitors have an average capacitance of 0.6 μF/mm3 compared to miniature aluminum electrolytic capacitors, which have an average capacitance of 0.1 μF/mm3. Tantalum-polymer capacitors exhibit approximately one-eighth the equivalent series resistance of traditional tantalum capacitors, meaning a current capacity approximately 8× of traditional tantalum capacitors. Advances in tantalum-polymer capacitor technologies have also extended the voltage rating of miniature SMT capacitors to 125 V. While traditional tantalum capacitors require 50 percent derating, tantalum-polymer capacitors rated up to 16 V only require 10 percent derating for polymer substrate devices and 20 percent derating for those rated for greater than 16 V operation.

Tantalum and tantalum-polymer capacitors are available in multiple case sizes with reduced height profiles relative to aluminum electrolytic capacitors and with novel lead frame packages that have dramatically lower inductances than aluminum electrolytic capacitors (see Table 1 and Figure 5). This enables bulk capacitor designs with better fit. As such, both tantalum and tantalum-polymer capacitors are highly competitive with aluminum electrolytics and well suited for use in GaN power amplifier designs, despite their derating.


Another passive component requirement for GaN power amplifiers is DC blocking. DC blocking in circuits operating at higher frequencies and wide bandwidths requires stable, low loss capacitors that can be easily configured to the circuit. Three unique capacitor technologies are worth considering: ultra-broadband capacitors (UBC), metal-insulator-metal (MIM) capacitors and metal oxide semiconductor (MOS) capacitors (see Figure 6). While there are additional options, these three types have proven to be practical for DC blocking.

UBCs — UBCs have a multilayered ceramic dielectric form factor that is compatible with standard printed circuit board manufacturing, including fully automated, high speed pick-and-place processing. They are available in 0201 and 0402 package sizes to match transmission lines, respectively, rated for 10 and 100 nF and they have ultra-low insertion loss, flat frequency response and excellent return loss from 16 kHz to approximately 70 GHz (see Figure 7). UBCs are an optimal passive component for DC blocking, DC coupling, bypass and feedback circuits in GaN power amplifiers.

MIM Capacitors — MIM capacitors are small, have low loss and can be used to compensate for the inductance effects of wire bond attachments, making them useful for DC blocking in high frequency transmit and receive stages. MIM capacitors use quartz, alumina or glass substrates to minimize loss and have a transmission line wire bond pad with backside ground to extend the frequency performance and reduce loss. Copper traces maximize conductivity, and front and backside gold metallization are compatible with high integrity epoxy, gold wire or ribbon attachment. They have 60 ppm/°C temperature stability with capacitance values from 0.3 to 15 pF and up to 100 V operating voltage. Custom capacitors can be designed, using a 50 to 100 pF/mm2 capacitance to area ratio.

MOS Capacitors — MOS capacitors are SLCs with SiO2 dielectrics and are small, temperature-stable capacitors with high Q, high breakdown voltage and low leakage. Manufactured with copper terminations in standard or custom patterns, MOS capacitors can be as thin as 127 μm for integration in 2.5D and 3D multi-chip modules, enabling higher frequency and lower power designs—potentially eliminating the wire bonds to the capacitor to reduce series inductance and extend the frequency response. Other termination options are gold or aluminum metallization on the top side and no metallization silicon, gold on bare silicon or chrome-gold on the bottom. Standard MOS capacitors are compatible with epoxy and solder die attachment and gold or aluminum wire bonding. Standard sizes and capacitance values range from 0.010 to 0.070 in.² and 1.0 to 1,000 pF, respectively.

Figure 6

Figure 6 UBC (a), MIM (b) and MOS (c) capacitors.

Figure 7

Figure 7 Typical |S21| (a) and |S11| (b) of the 550L series UBCs.



Figure 8

Figure 8 IR thermal scan of a 100-W GaN PA MMIC without heat pipes (a) compared to the PA using three heat pipes (b). Q-Bridge SMT heat pipe (c).

A signature benefit of GaN power devices is their ability to deliver high power in a small package. This can generate a massive amount of heat in a small area, requiring thermal control methods to remove, spread and couple the heat from the active device to get maximum performance and ensure reliability.

The thermal challenges of GaN can benefit from novel solutions like miniature SMT heat pipes.4 Miniature SMT heat pipes are a cost-effective solution for providing additional heat flow from the pins of an active device. Unlike traditional heat pipes, SMT heat pipes deliver high thermal conductivity with reduced parasitic capacitance, higher insulation resistance and high breakdown voltage. The small size of the SMT heat pipe supports the requirements for small size, weight and power (SWaP) designs. Although performance depends on the case size, the typical parasitic capacitance for the standard 0402, 0603 and 0805 EIA sizes is just 0.04 to 0.13 pF and the typical thermal conductivity is from 40 to greater than 500 mW/°C.

Miniature SMT heat pipes are available with several terminations: Sn/Ni/Pt, Ag/Pt and non-magnetic Ag. Custom SMT heat pipes have been designed to optimize the heat flow around MMICs, other high-power devices have been developed in case sizes spanning 0302 to 3737.

To illustrate the effectiveness of the SMT heat pipe, AVX engineers performed infrared (IR) measurements on an ultraminiature, high performance GaN power amplifier MMIC rated for 100 W output power. The MMIC was operated with a CW signal for 20 seconds at the amplifier’s center frequency, measuring the thermal rise with no heat pipes and using three Q-Bridge SMT heat pipes. Using the heat pipes, the IR measurement showed a 38°C decrease in MMIC temperature without any design modifications (see Figure 8), from approximately 80°C with no heat pipes to 42°C using the heat pipes. Neither the MMIC nor the larger power amplifier assembly were designed to include the Q-Bridge SMT heat pipes, i.e., the test was run with an existing design. Arguably, greater heat reduction could have been demonstrated if the power amplifier had been designed to include these SMT heat pipes.


GaN RF power devices have been proven through extensive lab tests and products, hastening their deployment for defense and commercial applications, in turn driving prices down and increasing accessibility. To fully realize GaN’s impressive performance—reduced capacitance, higher electron mobility, low conduction losses, faster switching and higher frequency-temperature and frequency-voltage characteristics than silicon—engineers must pair GaN devices with high performance passive components. Passive components are important circuit elements for impedance matching, bias filtering, DC blocking and thermal management. Without paying attention to the effects of these passive components, designers risk limiting the technology from reaching the full capabilities of its performance.

The development of new material systems, such as the GBBL ceramic dielectrics for SLCs and conductive polymers for tantalum bulk capacitors, are enabling designers to mitigate these potential performance limitations. New capacitor sizes and package styles are enabling smaller layouts to support SWaP requirements and improve electrical performance. New devices like SMT heat pipes are expanding the solutions available for removing, spreading and coupling the heat from the higher power density of GaN—improving both performance and reliability.

As GaN device technology continues to evolve and deliver improved performance, new and improved passive component technologies will be engineered to support these impressively compact, low loss and fast switching semiconductors, leading to the next generation of smaller, lighter and more reliable RF power amplifiers.


  1. T. J. Green, “Microwave Packaging Technology.” TJ Green Associates, LLC Virtual Training Course, June 13–15, 2021, www.tjgreenllc.com/virtual-training/.
  2. Qorvo, “Application Note: GaN Bias Circuit Design Guidelines,” October 2019, www.qorvo.com/products/d/da006903.
  3. T. Zednicek, “High CV MLCC DC Bias and Ageing Capacitance Loss Explained,” European Passive Components Institute, November 2019, epci.eu/high-cv-mlcc-dc-bias-and-ageing-capacitance-loss-explained.
  4. R. Demcko, “The Impact of High Temperature Exposure on Capacitor Reliability and Performance,” AVX SEMI-THERM® Thermal Technologies Workshop, November 2020, semi-therm.org/wp-content/uploads/2020/11/TTW2020-Final-Program-DRAFT-20201103.pdf.