Silicon-Germanium: The superior semiconductor technology for solid-state TV tuners

Because of the ubiquity of television in the modern world, practically everyone is aware of the rapid and stunning advances in TV technology over the last decade. During The last 10 years, common television sizes have evolved from 27-to-32 inches in the 90s to 37-inch and larger (much larger). At the same time, the once-universal Cathode Ray Tube (CRT) has been dying a fairly quick but silent death. The dim, murky pictures seen on analog big screen TVs a decade ago have been completely supplanted by bright, saturated color displayed in photographic-like detail on LCD, Plasma, and DLP screens of huge dimensions.

What is less apparent to the consumer is the tremendous challenge that has faced designers of solid-state tuners. The majority of televisions produced today still use "CAN-type" tuners, little different than those used when transistors replaced vacuum tubes in TV design in the late 1960s. That, however, is on the precipice of change.

The Solid-State Tuner Challenge
The CAN tuner, so called because it is housed in metal enclosures to eliminate crosstalk and stray radiation, is as mature as any technology can get. Manufacturing processes are standard and efficient. Size and signal capability is consistent from manufacturer to manufactureras are the shortcomings.

CAN tuners are large and bulky. The required tunable and fixed coils in the tuned circuits require discrete transistors, concomitant with higher voltage and current demands than those of most other digital and analog circuitry in the television set. Worst of all, the coils require tuning in the manufacturing process that is both time consuming and expensive and are prone to losing alignment over time. Relying on discrete components, CAN tuners' parameters are not very well controlled and vary significantly from unit to unit. Typically, CANs can handle only one standard from one region of the world, requiring many different parts be stocked and to cover today's global market.

Clearly, though, CAN tuners have jealously held on to advantages over available solid-state designs that have ensured a long life. On one end of the spectrum, low-end CAN tuners offer mediocre quality but optimized cost; while, on the other end, high-end devices offer good performance at a heightened complexity and cost, so covering the complete spectrum of applications. Enclosing all RF components in a standalone unit facilitates the design of the final product.

Early silicon designs could not compete with the superb signal-to-noise (SNR) capability and high dynamic range of high-end CANs and were not competitive enough on price at the low-end. With the advent of digital TV, however, hybrid CANs have proven difficult to design in, are complex, and more expensive than traditional solutions. Meanwhile, silicon tuners have proven its ability handle multiple standards with a single chip. Still, to supersede CAN tuners, solid-state tuners must compete in both signal-handling ability and cost.

Television Signal Fundamentals
Television signals have tremendous signal-strength variability. Not only does the signal strength vary from channel to channel based on the distance from the transmitter (for direct reception) or distance from the last cable repeater (for CATV systems) but the signal strength also varies in the time domain due to environmental factors, number of receivers on the cable and other factors. The tuner must be able to accommodate this signal strength variability and still produce a solid, linear output to the television circuitry.

Further, a television tuner must tune over an enormous bandwidth. Unlike, say, a wireless router that tunes a few MHz on either side of its center frequency, a television tuner must be able to tune over nearly 1 GHz of bandwidth.

Clearly, solid-state tuners have advantages over CAN tuners. They are much smaller and require significantly less power. The small size means that multiple tuners can be housed in even less space than a CAN tuner, allowing picture-in-picture and multiple program recording. More importantly, they don't require tuning and can work with any transmission standard.

So, what is the holdup? The answer to that question is in the semiconductor technology itself. Early tuners have attempted to use a CMOS process, which has well served the digital realm, but has problems with the very heart of the TV tuner: dynamic range and SNR. In addition, voltage-frequency tradeoffs have not allowed CMOS to operate from the lower voltages desired in the television circuit.

New Process Technology Solves the Problem
The answer to the problem lies in the physics of the semiconductor itself. As has been demonstrated in microprocessor technology, shrinking the geometry of the CMOS transistor allows for greater speed. The most modern CMOS geometry has allowed impressive switching speed. But, like every physical process, there is a limit. Realization of the smaller geometry requires, in part, thinning of the oxide layer of the transistor. The gate of an MOS transistor comprises a metal layer separated by a very thin oxide layer (Figure 1).

Figure 1: The Gate is insulated from the Source and Drain by a thin oxide layer

Increased speed has led to the gate oxide being as small as 1.2 nm, which is just five atoms thick. Physics dictates that the thinner the layer of oxide, the lower maximum voltage the transistor can tolerate without tunneling leakage and, ultimately, failure. Thus as speeds have increased due to decreased geometry, operating voltage has decreased. The fastest CMOS chips operate at supply voltage close to one volt.

Therein lies the problem, when the application is a television tuner, where SNR is king, a high SNR first requires a large enough signal to produce the desired SNR. At 1.8 volts or below, there just isn't enough signal to produce the requisite SNR. To operate at a high enough voltage to produce acceptable SNR, CMOS speed is not great enough to allow high-speed linearization techniques required to produce stable, constant, and noise-free output signals. It isn't a question of "waiting for a better CMOS;" the physical constraints do not allow for the requisite speed with acceptable SNR, dynamic range and linearity. The answer lies in a different process.