The very name input offset voltage indicates that it has been referred to the op amp input. This is done with all of the error sources because the actual output created by any error source depends on the closed-loop gain (A_CL) of the circuit, as seen from the error source. Thus, V_OS must be multiplied by the non-inverting circuit's A_CL to be referenced to the output.

Table 5-1 shows the V_OS portion of the OPA2991 data sheet. Each V_OS specification has a column on the data sheet for the minimum, typical, and maximum values to be listed. There are three major V_OS specifications that may be provided for an op amp: V_OS at 25°C, V_OS full range, and drift over temperature (ΔV_OS/ΔT). In the past, devices fell under specific operating temperature ranges with letter designators as shown in Table 5-2. However, most new devices have a temperature range from –40°C to +125°C regardless of application. Some devices are also marked with letters such as A or B. These grades indicate the accuracy of the part—the better the quality (grade), the lower the DC errors. This is not always the case, so always check the data sheet for the actual V_OS specifications.

The first entry is for V_OS under static temperature conditions, where the maximum and typical values are listed for a temperature of 25°C. This specification is listed on virtually all op amp data sheets and is expressed in millivolts (mV) or microvolts (μV). It is possible that natural variations or future changes in the process used to create a device result in a V_OS that is much different from the typical value. Every device is tested at the factory to ensure that it does not exceed the maximum specified value before it is shipped to the customer, so a robust design may require a designer to look at maximum instead of typical values.

On a data sheet, V_OS full range is usually only provided with a maximum rating. This specification lists the worst possible V_OS that can be encountered over a specified temperature range. Occasionally, the full range is specified for a temperature range less than the maximum operating temperature range, so strict attention must be paid to the conditions.

The drift of a device with temperature is indicated by ΔV_OS/ΔT. This is an average that is calculated using the ends of the specified temperature range as shown in Equation 13. For example, V_OS for many parts is measured at –40°C (T_A1) and 125°C (T_A2) and the results are calculated and expressed as the number of microvolts of increase in V_OS per degree Celsius of temperature change (μV/°C). Drift is usually given as a typical value on data sheets. Note that only maximum values are production tested by TI, so look at the full range data or measure drift to get a more complete understanding of the device's drift.

Some devices have multiple slopes in their drift plot, so Equation 14 is used. This is useful when there are different slopes in different regions, since a calculation over the whole range would show only a small drift when there could be larger drifts which cancel.

Figure 5-2 shows V_OS measured from –40°C to 125°C (a typical temperature range) for several devices and compares the V_OS drift behavior. The slope of each line indicates the magnitude of drift, or ΔV_OS/ΔT, for that part. The steeper the line, the greater the ΔV_OS/ΔT. The OPA2991 is shown to have less drift over temperature as compared to the LM358B. A designer should consider the full drift range behavior if temperature variance is expected in the application.

Figure 5-3 and Figure 5-4 is an example of the relationship between V_OS and the input common mode voltage (V_CM) using TLV9161, OPAx991, and TLV07xH. Since V_OS is normally tested at one value of V_CM, such graphs are useful in inferring behavior that is not specified in the data sheet. They provide the designer with an understanding of the behavior of the device over a wide range of values.