Hewlett-Packard Company introduced the first in a new family of low-noisbe pseudomorphic high-electron mobility transistors (PHEMT), which use HP's latest-generation gallium arsenide (GaAs) fabrication process.

The ATF-34143 features 0.5dB noise figure, +14dBm third-order input intercept point and 17.5dB gain at 2GHz, 4V, 60mA. The new transistor is optimized for 0.9GHz to 2.5GHz cellular PCS base station low-noise amplifiers (LNAs). The FET (field-effect transistor) also can be used in many other applications in the 450MHz to 10GHz frequency range. This ultra-low-noise transistor has a gate width of 800 microns and is supplied in the 2mm x 2mm SOT-343 surface mount plastic package, providing thermal resistance (theta jc) of 25 degrees C/W.

HP's new fabrication process is optimized for low noise and superior uniformity in 0.9GHz to 2.5GHz applications. The result is devices with exceptionally tight RF and dc parameter distribution. As an example, the 6-sigma spread (sample of 450 parts from nine wafers) for NF at 2GHz, 4V, 60mA is only 0.2dB; gain variation is less than 1dB. The 0.5 micron PHEMTs are fabricated using precision optical stepper-projection lithography on molecular beam epitaxial (MBE) material. The material is grown in a state-of-the-art multiwafer MBE system using proprietary targeting and characterization techniques, yielding extremely uniform and reproducible material. Multilayer photoresist is used to generate a mushroom gold gate with very low gate resistance, and selective dry etching is used to produce devices with very tight distribution of dc and RF characteristics. The devices have been extensively life tested and demonstrate excellent reliability.

Demo boards, application notes, models and characterization data at 2V and 4V are available on the HP Web site. Once at the site, visitors should jump to the HPRFhelp section for product selection information.

Just when I thought SiGe was going to be the only answer to really low-noise RF performance HP shows that GaAs processes can be bettered to match. The 0.5 dB noise figure at 2 GHz is superb for a device with a 17.5-dB gain and the overall uniformity is quite amazing. Guaranteeing such small spreads of noise and gain variations it will be possible to make front-end designs that are completely fixed which will not require separate testing. As always I have to be concerned about HP's interest in this end of the business long term but this process technology cannot be ignored by PCS base station designers. Additionally, we have been conditioned to expect parts for the base station side of things to be inherently more expensive -- that is not the case here.