(Keynote) Perspective on the Development of RF Compound Semiconductors from GaAs to GaN and Beyond

Monday, 2 October 2017: 09:10
Chesapeake A (Gaylord National Resort and Convention Center)
P. Saunier (Retired)
In this presentation, we are reviewing the present RF compound-semiconductors technologies, their applications, and their latest development (devices and MMICs). We will give a perspective on future development and also outline some of the latest achievements of the competing Si devices and MMICs.

The largest production volume is obviously that of GaAs-based HBT for cell-phone amplifiers with annual production probably in the 5- to 6 billion units. We will look at the potential device improvement as the phone operating frequencies are continuously increasing.

GaAs HEMTs and MMICs continue to be used for RF microwave and millimeter-wave driver amplifiers and lowest noise amplifiers.

InP based HBT’s hold the record for Ft and Fmax now approaching the TeraHertz. State-of-the-art MMIC performances have been demonstrated at mmWave; at 86GHz a MMIC achieved 358mW Pout (25.5dBm) with 10.2dB gain and 23.6% PAE.

 GaN has matured rapidly and has entered the commercial world. Due to the initial high cost, the first applications were military (radars, communications and counter-measures), but the two biggest markets under development are for the base-stations (L- and S-band) and the upcoming LEO satellite constellations (Ku-, K- and Ka-band).

Linearity is one of the key requirements for base-stations and excellent performances have been achieved with GaN Doherty amplifiers. Hybrid amplifiers have been the norm so far and we will show a few examples. But MMIC solutions are rapidly being developed for higher frequencies.

We will give more details on a few key GaN MMICs and devices which are benchmarks in the field:

A 2-stage MMIC with >300W and 55% efficiency at L-band was recently demonstrated with a released 0.5 micron GaN-on-SiC process. A packaged 100 W MMIC at X-band is on a product list. Recently, a 40W MMIC at Ka-band was published with 32- to 37% PAE across the 26- to 30GHz band. Greater than 1W MMICs at W-band are commercially available with efficiencies above 20%. New Low-Noise MMICs are being designed using 0.1micron gate-length for NF of 2.0- to 2.6dB NF from 18- to 52GHz.

Self-aligned devices with Ft/Fmax of 454GHz/444GHz (Fmax of 600GHz was shown with lower associated Ft) were demonstrated 3 years ago and their development continues for millimeter-wave amplifiers. They are fabricated using advanced “silicon-like” processes, such as 20- to 40nm self-aligned gates and material re-growth (for ohmic contacts). This type of device also opens the door for very low voltage applications; we will give examples of very good performances (power and low-noise) down to 1- to 5V bias at millimeter-wave.

 A recently published N-polar device showed a record 6.7W/mm with 14.4% PAE at 94GHz. The same technology recently showed a record 34.2% PAE at 86GHz with 2.5W/mm power density. This technology is very promising certainly an alternative/competitor to the conventional Ga-face GaN.

Small gate-length GaN devices are now used for Low-Noise applications, for survivability and inherent overdrive protection. They could avoid the Noise Figure penalty of having a protection-limiter in front of the LNA

GaN-on-Diamond is receiving renewed attention as Thermal Management is critical for upcoming communication satellites. Although the technology is difficult, the advantages are multiple: for the same layout (as with GaN-on-SiC) the junction temperature can be 40- to 600C lower. For the same junction temperature, the power per unit area can be 3- to 4 X higher. For the same layout and power density, the PAE will be higher (due to lower junction temperature). In the same way, for the same layout and junction temperature, the output power will be higher.

Ga2O3 is being investigated as a potential new wide-bandgap semiconductor; preliminary results have been published. One of the key advantage could be cost. The main drawbacks are the very poor thermal conductivity of the substrate and, right now, the low maximum current available. Other recent achievements show that the technology could be very promising for power device/switches.

We will benchmark the development of Si-based devices (SiGe and CMOS) and MMICs, particularly at mm-wave, for phased arrays. Tremendous progress is continuously being made. Recently, SiGe HBT showed fT/fmax/BVCEO = 505 GHz/720 GHz/1.6 V. Si is benefiting from low-cost (for high volumes), tremendous integration capability with 8- to 12 levels of metal/interconnect, thick metal levels for integration of inductors, (output) combiners and patch antenna, and integration with the necessary functions such as antenna calibration.