June 19, 1997

Resistor packs eliminate temperature drift

Jim Hansen, Kollsman Inc, Merrimack, NHFinding low-temperature-coefficient resistors for 10-bit or more accurate analog applications is often a problem in the laboratory. Even if you have a good collection of precision, low-temperature-coefficient resistors in stock, chances are they won't track across even room temperatures. A pair of 50-ppm resistors--clearly better than most common laboratory stock--can exhibit a worst-case drift of about 0.5 bits/°C in a 12-bit system.

Many analog designs call for ratiometric resistor combinations in which the precise ratio is more important than the actual resistor values in the network. One way to nearly eliminate resistor-induced temperature drift in such an application is to use a humble resistor pack to obtain a set of resistors whose values, although not precise, track better over temperature than many common amplifiers or sensors.

In a given lot of single-in-line resistor packs, one package probably tracks the next about as well as any batch of equivalent resistors. But within one resistor package, the individual resistors track each other over temperature because they are made at the same time with the same resistor material. The trick to exploiting this characteristic is to find the best combination of resistors in the pack to make the desired ratio.

You can make a variety of useful voltage dividers and gain- setting ratios from a single "common terminal" resistor pack. (These units typically have five to nine equal-value resistors tied to a common terminal.) For example, you can make a precise 5-to-1 voltage divider by feeding the input to one resistor, grounding four of the remaining resistors, and taking the output from the common terminal (Figure 1a). You can make temperature-stable op-amp gain-setting resistors in the same way by connecting the resistor pack's common terminal to the op-amp feedback (inverting) input, and tying one or more resistors to the output and ground to set the gain (Figure 1b).

A simple spreadsheet program finds the best combination of resistors for the 5-to-1 voltage-divider network. (Click here to download the file from DI-SIG, #2049.) Of 12 random five-resistor packages from stock, 10 had at least one--and, usually, two--"no-trimming" combinations yielding 12-bit, 0.0244% or better ratios. Worst-case combinations are generally 0.2 to 0.3%, which is good enough for 8-bit applications but nothing to write home about. A similar program (available on request from the author at jmhansen@worldnet.att.net) evaluates nine-resistor packages for an 8-to-1 ratio with slightly less accurate no-trim results.

The spreadsheet also calculates the required trim, typically less than 5 or 6 ohms and often 1 ohm and less. Because the trimming values are low, you can use nearly any quality resistor without harming the overall temperature coefficient, or if your system needs two or more such dividers, you may be able to reduce trimming requirements by using packages with complementary errors.

Use common sense when measuring resistor packs for this application. Pick a quality 5 or better digit meter to read resistances and measure all the resistors in a package at once and at the same temperature. And don't handle the packs with your fingers before measuring them because, like any other resistor, these resistors drift with temperature.

Finally, test your resistor pack. Put a fixed power supply across the terminals of your network and measure the voltage ratio with a DVM while applying a heat gun to the resistor pack. Some units, particularly older hybrid ones, may exhibit drift. If so, use modern thin-film modules, and you should get excellent results. (DI #2049)

| EDN Access | Feedback | Table of Contents |