Figure 1

Figure 1 Symmetric Doherty PA delivering 2 W average in Band 42. The design uses two TQP0103 GaN transistors, each in a 3 × 4 mm plastic QFN package.

The demands on the wireless network infrastructure are expected to continue to grow at a substantial, if not alarming, pace. The infrastructure for wireless communication is constantly adapting to address these demands. Network operators are perpetually striving for continuous improvements in cost, capacity, coverage, quality of experience, reliability, functionality, interoperability, spectral efficiency, mobility and more acronyms and words ending in ‘ibility’ than one cares to mention.

This is not to imply that the demands are trivial and unfounded. Cisco has reported mobile traffic grew 69 percent in 2014 and reached 2.5 exabytes per month. To state this data transfer rate differently, the mobile traffic volume in 2014 was nearly 30 times the entire globe’s internet traffic in 2000. By 2019, monthly data rates are forecast to grow 10 times the current rate to 24 exabytes per month. The increase in data traffic will be driven both by traditional mobile users and the anticipated growth of various data connections, the so-named Internet of Things (IoT). The demand for additional network capacity appears endless, and even as LTE continues to rollout in areas of the world for the first time, network operators are already planning to address the next major evolution of the mobile network – 5G.

5G is the fifth generation of the mobile infrastructure network; it will encompass many network advancements and promises. As previous generations provided, 5G is expected to continue to increase spectral efficiency, support more users, provide higher data rates and improve the user’s experience. Exactly how network operators will provide a 5G experience is yet to be known; however, it is clear that all network operators share a common need to satisfy the growing network demand with more bandwidth. The use of additional frequency spectrum is a major requirement and focus of next generation network systems. Multiple research and development programs are active in all areas of the spectrum: low frequency TV white space, unlicensed bands at 2.4 and 5 GHz, existing point-to-point and point-to-multipoint bands at 15, 28 to 30, 50, 60 and 71 to 86 GHz (E-Band) and 3.5 GHz. Each band has advantages and disadvantages, and it is likely that next-generation networks will include solutions in all of these bands as the heterogeneous network continually widens.

Figure 2

Figure 2 Symmetric Doherty PA gain and efficiency as a function of output power, using a single carrier, 64 DPCH, 10 dB PAR WCDMA test signal at 3.5 GHz.

ATTRACTIVENESS OF 3.5 GHz

3.5 GHz is a band that offers an immediate solution to the growing spectral needs, and it does not require network operators to wait for a 5G solution. Providing solutions at 3.5 GHz offers hardware manufacturers a design platform that is very similar to existing traditional cellular bands, compared to higher frequency alternatives.

3.5 GHz provides 200 MHz of spectrum, from 3.4 to 3.6 GHz, that is available in most of the world and has been recognized as a potential global harmonized band for TDD. Japan is at the forefront of using 3.5 GHz for mobile infrastructure, and recent announcements reporting field trials in China are expanding the footprint. Europe has long allocated the band for fixed broadband; mobile infrastructure is expected to dominate future use. The U.S. has a more difficult challenge to harmonize the band with the rest of the world since portions are used for radar. However, the FCC recently opened 100 MHz for commercial use, the newly dubbed “innovation band.” Given these spectrum allocations, 3.5 GHz will play a key role in future network expansion, with the potential for both carrier aggregation and stand-alone operation.

These new installments are expected to continue with a common theme, i.e., networks will become denser. Increasing network density utilizes a layered approach for coverage, meaning installing multiple layers of access to improve capacity in high traffic areas. Depending on whether the coverage is indoor or outdoor, base stations are developed to provide various power levels, which often differ by manufacturer:

  • Femtocell, less than 0.25 W,
  • Picocell, 0.25 to 0.5 W,
  • Microcell, 1 to 5 W,
  • Metrocells, 5 to 10 W, and
  • Traditional macrocells, greater than 10 W.

Varying power levels allows maximum flexibility for the operators to create smaller, denser, higher capacity coverage areas within the network. Initially thought to be an ideal candidate for only small cells, 3.5 GHz is expected to see deployment at all power levels to offer a fully layered solution for network operators.

DESIGN CONSIDERATIONS

In response to the growing demand for outdoor solutions at 3.5 GHz, Qorvo is developing new GaN amplifier products and Doherty power amplifier reference designs targeting 1, 2 and 20 W average output power at the antenna reference plane. Future developments will target 10 W metrocell and 40 W macrocell solutions. GaN is ideally suited for the band due to excellent gain, high power density and high efficiency, providing significant performance improvements over competing technologies. These designs utilize Qorvo’s 0.25 µm gate length GaN on SiC process on 100 mm wafers. Process options allow for 28 to 32 V operation for small cells and 48 V for macrocell applications. 100 mm GaN costs have dropped significantly in the last few years, and a planned transition to 150 mm wafers will further reduce cost.

Figure 3

Figure 3 ACPR vs. output power below P3dB for an asymmetric Doherty PA and 10 MHz, 6.5 dB PAR WCDMA signal (a) and 20 MHz, 6.5 dB PAR LTE signal (b).

0.25 µm GaN provides higher gain and higher frequency operation than GaN processes with longer gate lengths (e.g., 0.5 µm). To maintain high efficiency of the power amplifier line-up, including driver and pre-driver stages, the gain of the final Doherty amplifier – a premium at 3.5 GHz – needs to be as high as possible. GaN’s high power density results in lower drain-source capacitance, compared to GaAs or silicon, which enables higher bandwidths. The low Cds and high inherent impedance of the device allow for internal package matching networks that are suitable for high video bandwidth applications. The video bandwidth for the 3.4 to 3.6 GHz band needs to be high, since 100 MHz signal bandwidth is planned and 200 MHz is being discussed.

At 3.5 GHz, the insertion loss between the power amplifier (PA) and antenna – which includes the circulator, board losses and filtering – is estimated to be 2 dB. Thus, 20 W average radiated power at the antenna will require 32 W at the Doherty PA reference plane. The required peak power of the Doherty PA is a function of the modulated carrier peak-to-average ratio. For the downlink LTE signals of macrocell base stations, these are typically on the order of 7 dB with crest factor reduction. Further headroom of 1 dB for digital predistortion (DPD) is added to the PA specification to compensate for performance over temperature and device-to-device variability. Thus, 20 W average power at the antenna will require 200 W of peak power, or 8 dB above 32 W.

DOHERTY PA REFERENCE DESIGNS

To show the performance available from GaN, a symmetric Doherty PA reference design for Band 42 (3.4 to 3.6 GHz) was developed. It delivers 2 W at the antenna, using TQP0103 GaN transistors for both the carrier and peaking amplifiers (see Figure 1). The PA delivers more than 20 W peak power with greater than 44 percent efficiency at 8 dB back-off. The gain and efficiency as a function of the output power are shown in Figure 2.

An asymmetric Doherty PA reference design for 1 W average output at the antenna uses a TQP0102 GaN transistor for the carrier amplifier and the TQP0103 for the peaking amplifier.  The power ratio of the peaking amplifier to the carrier amplifier is 2:1. The asymmetric Doherty will have higher efficiency than the symmetric Doherty at 8 dB back-off. This reference amplifier design will achieve greater than 50 percent efficiency at the amplifier reference plane. At the same back-off power, this boosts efficiency more than 6 points above a symmetric Doherty.

Designing an asymmetric Doherty requires extra caution to ensure the AM-AM and AM-PM responses of the Doherty are smooth and monotonic, necessary to work with DPD. The gain and phase responses must be monotonic through the transition when the peaking amplifier switches from off to on and load modulates the carrier amplifier to peak power. Achieving the appropriate gain and phase responses is challenging, since two different devices are used, each with different gain and phase responses. Further, the modes of operation are also different, with the carrier amplifier operated in Class AB and the peaking amplifier biased to Class C. Better than -60 dBc ACPR has been demonstrated with asymmetric GaN Doherty designs with 10 MHz signal bandwidth using third party DPD systems (see Figure 3a). With a 20 MHz LTE signal, the DPD corrects to better than -57 dBc ACPR over a wide range of back-off power levels (see Figure 3b). The range extends from deep back-off, where only the carrier amplifier is active, through the transition of the peaking amplifier turning on and load modulating the carrier amplifier. The DPD will improve the linearity as long as the average power plus the peak-to-average ratio of the modulated carrier is less than the Doherty amplifier saturated power. The DPD system cannot compensate for nonlinearities at powers exceeding the saturated power capability of the Doherty amplifier. ACPR degrades rapidly when this is attempted, to a near vertical slope. Figure 3 shows the case when the power of a 6.5 dB PAR is driven to less than 6.5 dB back-off.

Figure 4

Figure 4 Load-pull output power, efficiency and gain contours for a QDP3600 carrier amplifier transistor used in an asymmetric, 20 W Doherty amplifier.

The asymmetric Doherty architecture is also being used for a 20 W average power (at the antenna) reference design. This amplifier uses the QDP3600 as the carrier amplifier and a T1G4012036-FS as the peaking amplifier, both operating at 48 V. The loads are designed to achieve carrier and peaking amplifier powers of 70 W and 140 W at peak power, netting a combined P3dB of 200 W at the Doherty PA output. The design is based on load-pull measurements of the transistors. The carrier amplifier load-pull contours at 3.5 GHz are shown in Figure 4. The red contours show the peak power (P3dB) capability of the device. The black circle is a 3:1 VSWR contour around the maximum power and shows carrier amplifier efficiency match conditions that the device can achieve in a Doherty configuration. The green contours show the gain, and the blue contours show the drain efficiency at the 45 dBm targeted average power of the Doherty amplifier. Following the 3:1 VSWR circle around to the left side shows that 60 percent drain efficiency and 19.6 dB of gain can be achieved at the carrier amplifier efficiency match. When the peaking amplifier turns on, it will load modulate the carrier amplifier to the center of the 3:1 VSWR circle and be at a load condition with 48.9 dBm P3dB. With the peaking amplifier matched to a 2:1 power ratio, the expected design performance is 14 dB gain and 55 percent efficiency at 45 dBm average output power.

This efficiency is substantially higher than what can be achieved with silicon LDMOS, which is the incumbent PA technology for the cellular infrastructure market. The gap between GaN and LDMOS increases with increasing frequency. Based on data sheet specifications, the latest LDMOS devices in development at 3.5 GHz show expected Doherty efficiencies of 37.5 percent.

Conclusion

There is much discussion and excitement around the future 5G network. While the 3.5 GHz band has yet to reach the potential of a globally harmonized band, there are many opportunities for 3.5 GHz to fulfill the immediate need for mobile infrastructure bandwidth, well in advance of a 5G deployment. Plans for developing the new spectrum are growing rapidly for both small cell and macrocell deployments, with GaN well suited to address the needs for high power, high efficiency and wide video bandwidth.