High-performance applications such as third-generation wireless basestations and measurement and medical equipment demand faster response times from high-speed, high-resolution data converters and the amplifiers used to drive them. Companies are meeting this demand with advanced processes that enable bipolar transistors with faster transition frequencies, lower noise and better linearity, as well as precision resistors and more linear capacitors.
One such method is BiCom-III, a silicon germanium process developed for ultrahigh-precision analog ICs. It is a dielectrically isolated silicon-based procedure with germanium added to the base region, which greatly increases carrier mobility and makes for extremely fast transient times. The process produces truly complementary bipolar npn and pnp transistors with transition frequency of 15 to 20 GHz and maximum frequency of 40 to 60 GHz. Complementary transistors allow Class AB (push-pull) amplifier stages that support the design of high-speed, high-performance analog circuits.
Other technology improvements that support high-speed op amp design include metal-insulator-metal (MIM) capacitors with very low voltage coefficients, resistor matching to within small percentages (0.1 percent) and dielectric isolation for the transistor junctions. Sometimes called silicon-on-insulator, or SOI, this feature reduces parasitic capacitance and produces very high transistor current gain (beta) and early-voltage product for increased loop gain.
Designing ideal op amp
To understand how a process supports op amp design, it is helpful to understand the architecture and the goals of the design. All op amps have he same basic architecture: an input stage, a high-impedance node and an output stage. The purpose of the input stage is to acquire the differential input signal and turn the differential voltage applied to the input into a differential current that is then routed to the high-impedance node. The current from the input is converted to a voltage at the high-impedance node. If the op amp is single-ended output, the voltage developed is single-ended. If the op amp is differential output, the voltage developed is differential. The output stage is then used to buffer the voltage from the high-impedance node to drive the output.
The goal is to attain the highest possible gain while reducing error sources and maintaining stability. It is also generally desired to have very high input impedance and very low output impedance. In other words, make an "ideal op amp."
Many designers use transistor current gain (beta) times the early voltage (VA) as a figure of merit when talking about bipolar transistors. For the input stage, having high-beta transistors allows a lower input bias current, which in turn means higher input impedance. For the output stage, the higher the beta, the less the output stage loads the high-impedance node, resulting in increased amplifier gain.
Early voltage is a measure of the collector impedance of a transistor. The impedance of the high-impedance node is directly related to the early voltage of the transistors used. Higher VA = higher impedance = higher amplifier gain.
High beta x VA results in a better op amp because it increases loop gain, and loop gain reduces errors in the amplifier. It also lowers distortion and input offset. In BiCom-III, the beta x VA product equals 50,000 for npn's and 20,000 for pnp's, which is much higher than other complementary bipolar processes.
During fabrication, a silicon oxide (glass) insulating trench is formed around the transistors, which isolates them from the surrounding structures. Junction-isolated processes use reverse-biased pn junctions to isolate the transistors. Two advantages arise from dielectric isolation: Stray capacitance to other devices and the substrate is reduced, and the stray capacitance has a very low and linear voltage coefficient.
The speed of a transistor is determined by a number of things, one of the most important being stray capacitance. With less current "siphoned off" as frequency increases, the transistor works better at high frequency.
Variation in capacitance with changing voltage leads to nonlinearity, which causes distortion. Junction-isolated processes fall prey to this effect, which typically manifests itself in op amps that have better distortion characteristics in inverting gains than in noninverting gains. This is because the input common-mode voltage is fixed for inverting gains, while the common-mode voltage follows the input signal for noninverting gains.
Capacitance is purposely added at the high-impedance node for dominant-pole compensation. The lower the voltage coefficient of the added capacitance, the less distortion is generated. This is important at the high-impedance node since, generally, the highest voltages seen in the op amp are developed here.
BiCom-III MIM capacitors have a typical voltage coefficient (lin) of -6 ppm/V, which is extremely low and on par with the best passive components available.
Resistor matching is important in setting gain, matching current sources and reducing input offset voltage; the BiCom-III process has intrinsic matching of 0.1 percent without any trimming. Temperature coefficients are also lower than those of most commercially available resistors with 25ppm/ degrees C (lin) for the thin-film resistors and -6 ppm/ degrees C (lin) for the poly resistors.
An example of the matching performance is found in the THS4302 and THS4303. They are fixed-gain amplifiers (5 V/V and 10 V/V) with absolute gain accuracy of 0.1 percent from -40 degrees C to 85 degrees C. Applications include wideband signal processing; wireless transceivers; IF amplifiers; A/D converter preamplifiers; D/A converter output buffers; test, measurement and instrumentation; and medical and industrial imaging.
In addition, a 5-V CMOS can be integrated into the BiCom-III process flow to support BiCMOS mixed-signal products such as high-speed A/D converters.
James Karki is a member of the group technical staff and manager of high-speed amplifier strategic marketing at Texas Instruments Inc. (Dallas).