Driving a component spec down to zero is an ongoing engineering challenge, but one that has esthetic and customer benefits
In the past few months, we've seen a slew of zero-parameter analog components from leading vendors, including a zero-drift instrumentation amplifier (Texas Instruments), an auto- zero comparator (Fairchild), a zero-offset op amp (Microchip Technology) and a zero-IF downconverter (Analog Devices). Even if these components are not perfectly at "zero," they
come very, very close (it's that asymptotic factor). Even the elusive zero-power IC is getting closer and closer for some basic analog components.
What's going on here? First, there's the natural engineering imperative to strive for perfection. For many analog parameters, such as drift, leakage, bias current, or offset, zero is the ideal specification we seek. True, engineers always must balance attainable performance with cost and many other factors, but there's still an esthetic impulse toward achieving perfection, if only in one or two dimensions.
True, engineers sometimes do things largely because they can and want to, yet this has led to some on-the-fringe advances that are later accepted as must-have. Improvements in IC process technology and associated device modeling at the deep-physics level help make these achievements possible, of course.
There's a more mundane rationale for pushing toward zero, as well. ICs manufacturing is a business, and vendors know that, in many cases, their customers will pay anywhere from a modest to a substantial premium for a product whose key spec is zero, or close to it.
For example, a sensor-signal amplifier with near-zero drift over time and temperature can eliminate the need for factory trim procedure, which usually also requires adding an electronic or mechanical trimming potentiometer to the design and the bill of materials, as well as a factory calibration step. Even more important, it may eliminate entirely the need for additional calibration in the field, which is always awkward for users and often inaccurate.
But even a brilliant zero design won't work if you can't test it effectively, and that's the real trick in a successful zero-spec product. It's hard enough to test for this performance in a low-volume bench set-up; it's really hard to do in a higher-volume production environment. As test engineers like to remind us, but with mixed success, the test expertise is the real secret behind a successful, high-performance product, especially one with zero as a specification's value. p