Analog Angle Blog

Active signal correction: When is the gain worth the pain?

Deciding how to improve performance is a standard design challenge, especially when analog sensor and signal channel are involved. Three commonly used approaches include:

  • Use of better components for lower drift, tolerance, bias current, or noise as well as higher bandwidth and more drive.
  • Employing circuit topologies such as a bridge or differential design, which cancel out some or most of the errors, or controlling some error-inducing factors such as temperature.
  • Calibrating the sensor or circuit and using the results to improve—usually linearize—the signal in an analog scheme or as a correction factor during data reduction.

These techniques, especially the first two, were widely employed by the late Jim Williams and Bob Pease in their designs and associated articles. For an excellent example, see Jim’s 1976 classic EDN article “This 30-ppm scale proves that analog designs aren’t dead yet“; though it is nearly 50 years old, it still has many important lessons (Figure 1).

Figure 1 The first page of Jim Williams’ EDN article on design and error minimization for a high-precision, ultra-stable weigh scale makes it clear that it achieves 30-ppm absolute-accuracy (0.02%) performance and never needs adjustment in the field.

There’s yet another way to improve performance: adding some sort of active cancellation circuitry to negate the offending behavior. This is not new, of course, as it has been used for years in applications as diverse as noise-cancelling headphone or active damping for vibration-minimization in sensitive optical tables (Figure 2).

Figure 2 Active vibration control systems for optical tables use a support spring and cancellation actuator in parallel or in series to attenuate the remnants of larger vibrations cancelled by static isolation arrangements. Source: TMC/AMETEK Ultra Precision Technologies

The use of active approaches and associated circuitry for improving performance have increased in recent years, now encompassing situations which were previously handled by passive approaches. A recent Texas Instruments blog “How to reduce EMI and shrink power-supply size with an integrated active EMI filter” showed how to do so for the noise of a switching supply. Note that this task is simplified since the critical attribute of this noise—its frequency—is known.

Of course, in many cases, the attributes of the noise and other error sources are not known, but it’s still worth trying to attenuate using active schemes. Doing so may sound like a good idea—and it can be—but implementation of dynamic compensation and correction schemes is complicated and has its many subtleties. As with all closed-loop situations, some feedback-loop types and parameters such as time constants result in only marginal improvement at best and even performance degradation at worst.

Furthermore, there are issues such as loop saturation to analyze and deal with, so the solution does not blow up, so to speak. Finally, adding the sensor and other closed-loop components has a cost and space impact, although these may be less severe than using sophisticated mechanical vibration-damping schemes.

I think we’ll be seeing more of such advanced approaches, including dynamic closed-loop compensation as the cost—in terms of dollars and power—and size of the needed hardware come down while they significantly enhance overall performance. Still, it’s not an approach to be implemented without careful analysis of what it requires, what it may yield, and what may go wrong.

It’s my general sense that these sensor-based active compensation schemes are especially effective and viable when the challenge is to take care of gross problems. This is the case for acoustic noise-cancelling headphones and systems, for example, where the goal is not to totally eliminate outside noise, which can’t be done and isn’t necessary to do at all, but only to attenuate it by perhaps 20 to 40 dB.

In contrast, if you are using a closed-loop scheme to go from very, very good performance to truly stellar performance, squeezing out that last drop may become a challenge. The improvement scheme may bring with it many downsides as well, some of which you can anticipate and some you may not be able to foresee. In other words, getting to 80% or 90% using cancellation is one thing, while going from 98% to 99.99% is far more challenging and prone to unwanted surprises, and at greater cost.

What’s your sense of the practicality and desirability of active schemes to take performance to the next level in high-performance designs? Have you tried it, or at least considered it?

Related Content

0 comments on “Active signal correction: When is the gain worth the pain?

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.