I always said that selling RF power LDMOS is a lot like selling sports cars: Whoever has the best performing product is the one that gets the sell. In cars, that would be all about the top speed, acceleration, road handling and the like. With RF power devices, it is other key performance parameters such as DC-to-RF conversion efficiency, gain, instantaneous bandwidth capability and maybe a touch about high terminal impedances that wins the socket. But all that is changing now. Other key performance parameters are taking the lead in determining the ultimate RF power device.
There are some very hard and fast theoretical limits as to just how high the efficiency of a standard Class B/AB power amplifier can be. Pi/4 or 78.5% is the maximum limit, due to the 180 degree conduction angle attribute and no matter what , you are not going to get a better number than that. Modern day LDMOS devices, with their extremely low Rds(on) and low parasitic Drain-Source capacitances, are very close to this maximum theoretical limit now so there is not much improvement in RF performance that one can achieve with changes in the standard metrics of Ft or shifted I-V curves. No, future design wins will rely on a different set of RF performance criteria that are not easily measured and not limited to the confines of standard RF performance parameters.
Since the vast majority of all current cellular basestation designs are DPD Doherty, the previously listed Class B/AB limitations are obsolete, as the devices now are required to operating under both class AB and class C bias conditions. Class C, in theory, can be as high as 90% efficient. That, coupled with the ever improving error correction capability of the digital predistortion systems, allows the devices to operate at a much lower back-off ratio (i.e. closer to saturation). Some key HV8 design goals include reduced phase distortion, improved gain flatness and increased VBW capability which are essential in enhancing the broadband DPD performance of the Doherty circuit. Improved DPD performance has a direct link to system efficiency, since better DPD correction means the Doherty circuit can be pushed even closer to saturation to realize higher levels of amplifier efficiency while still meeting the linearity requirements.
Freescale’s RF division has released 14 advanced HV8 devices over the course of the past 6 months covering most popular cellular frequency bands. They have been winning significant sockets in Doherty DPD applications. Clearly, HV8 has been living up to its design expectations. One can expect about the same number of additional HV8 devices to be released over the upcoming 6 month period, along with greater expansion in the RFIC and GaAs MMIC market space.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.