With the rate at which CMOS process technology has been expanding on all fronts during the last 20 years, older bipolar technology seems to be on its way out. CMOS devices are everywhere—in analog, digital, and even the RF domain. Some CMOS proponents believe it will completely displace bipolar in the analog sector and dominate the RF region in the near future.
But that is unlikely. With confidence, bipolar supporters have been refurbishing and polishing the technology for the last two decades. And, a recent spurt of activity suggests that it is very much alive and certainly healthy and growing.
Just as complementary MOS infused new life and energy into nMOS in the late 1970s, adding complementary pnp transistors to npns has given conventional bipolar a shot in the arm. Subsequently, the complementary-bipolar (CB) process also has dramatically changed to withstand the challenges of modern applications as well as gear up for future needs. Newer materials like silicon germanium (SiGe) further promise to help bipolars maintain their lead in performance and adequately serve applications down the road.
In the last 20 years, the CB process has seen a phenomenal reduction in chip area and power while simultaneously boosting transistor speeds to greater heights. Traditionally, slower pnp transistors in particular have significantly improved in terms of transition-frequency performance.
Also, IC developers have lately migrated from the junction-isolation techniques of the early days to more sophisticated trench or dielectric isolation to substantially cut parasitic capacitance. This in turn enhances amplifier output-drive capability. The result is very low quiescent current, high-output drive, faster speed, and lower distortion. Cost has been checked by going to smaller features and keeping the die size smaller.
For Instance, Analog Devices' CB technology progression table indicates that the supplier's 12-V XFCB-1 technology, introduced in 1992, achieved an fT of 2.5 GHz for pnp and 4 GHz for npn transistors. With enhancements over the years, the same technology, now labeled XFCB-12 and introduced last year, offers 3.5 GHz for pnps and 7 GHz for npns.
Lowering the breakdown voltage to 4.5 V and refining the process has catapulted pnp performance to 12 GHz and npn performance to 20 GHz. Besides scaling voltages, feature sizes have been shrunk from about 6.0 µm to 1.5 µm and below, while power consumption has been drastically reduced. Additionally, the road-map shows that the CB evolution continues, with better performance to come.
Recently, National Semiconductor unwrapped its latest achievement on this front by introducing a new family of high-speed amplifiers that can operate down to 2.7 V. National claims that its VP10 CB process can provide a high-speed op amp with a 1-V minimum supply voltage and a gain-bandwidth product of 100 MHz. That's an accomplishment indeed!
Other CB backers include Texas Instruments' Burr-Brown division and Elantec Semiconductor. With broadband communications and other high-bandwidth applications on the rise, both are busy cranking out high-performance line drivers, video amplifiers, and other high-end op amps and analog circuits that are beyond the reach of CMOS.
No doubt, CMOS will continue to improve and further encroach on bipolar territory. But there will always be applications whose requirements will be far more stringent. And, CB or some other form of bipolar will be there to serve the needs of analog designers.