OpAmp offset trimming can be done using the following methods (fig. 5- 14):
1. Size adjustment of the input transistors by connecting/disconnecting the sources of fractional transistors (fig. 5-14a) to the sources of main devices. Laser cutting or fusing of poly links performs the source
disconnection. As the input voltage has the rail-to-rail span, the commutation in packaged units should be done by NMOS/PMOS switching pairs.
2. Size adjustment of the input devices by switching the drains of fractional transistors (fig. 5-14b). Due to the residual link resistance, laser cutting is not used. The drains of disconnected fractions should connect to the supply rail to avoid parasitic effects. The commutation switches using only one type of transistors are usually sufficient as the drain potentials are usually near one of the supply rails.
3. Adjustment of the current mirror degeneration resistors (fig. 5-14c).
This method used to be the most popular technique for laser offset trimming.
Resistor network trimming can also be applied using metal or poly fuses and zener zapping. CMOS switches here are not optimal due to their relatively large, supply- and temperature-dependent turn-on resistor Ron.
4. Adjustment of the load current sources of the input pair’s drains by switching the sources of fractional current sink transistors (fig. 5-14d). This can be done by fusing metal or poly links or by zener zapping.
5. Same as 4., only by switching the drains of fractional current sink transistors (fig. 5-14e). It can be done using MOS switches with one type of transistors.
6. Addition of a trimmable current source to one of the current sinks (fig.
5-14f).
The offset voltage temperature drift may be as important to the customer as the offset voltage itself. It was mentioned that the temperature drift of the bipolar amplifier is proportional to the offset. Trimming offset reduces the temperature drift as well. Drift values of 0.1-0.2 uV/oC are realistic for high precision trimmed bipolar amplifiers.
Figure 5-13. Offset trimming techniques
Trimming of both the offset and offset temperature drift has been realized for JFET input amplifiers as well. The procedure includes two temperature offset measurements, calculations and trimming of the offset-controlling
variables (resistors or currents) [55]. This procedure can only be used in the wafer probe test (i.e., with unpackaged units) because it requires the knowledge of the OpAmp ID. On a wafer, this ID is defined by the OpAmp location. The offset measurements are done for all amplifiers on the wafer at the first temperature, then the temperature of the test environment changes, (heated up or cooled down) and the measurements are repeated. The trimming is performed using the data of the first measurement as a reference.
An automated test of packaged parts at an arbitrary temperature is technically realistic. However, tracking the identity of a packaged unit up to the final test is unfeasible unless the ID number is programmed into each unit. Storing this ID number would at least double the amount of required on-chip ROM. It would also require the addition of a communication port to the OpAmp in order to read this ID back into test system. As a result, keeping the packaged amplifier ID is feasible only if the OpAmp is a part of the system already having flash memory and a communication port.
The temperature drift of a MOS differential pair is not correlated with the offset while operating in either weak or strong inversion. Temperature drifts on the order of 10 uV/oC are not uncommon. In a correctly designed OpAmp, without drastic changes of input stage or biasing operation mode with temperature, this drift is relatively constant. A significant curvature in the offset versus temperature dependency may appear only at 110-120 oC.
This curvature is attributed to the junction leakage currents rapidly growing at high temperatures.
Measurement results of the offset versus temperature (normalized to zero at –55oC) dependencies for a set of OpAmps are shown in fig. 5-14.
Voff uV
-1400 -1000 -600 -200 200 600
-50 -40 -30 -20 -10 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 1
TEMP
Figure 5-14. Offset versus temperature for the typical CMOS OpAmp
A recently introduced method [15, 62] allows one to implement independent trimming of the offset and first-order temperature drift without amplifier identification. As one can see from fig. 5-14, the dependence of offset versus temperature is approximately linear. This method relies on this fact, and uses two temperature test results to eliminate this main linear component. This drift trimming can be done in large production batches.
After testing and trimming at the first temperature, rejection of bad units, etc., the whole batch proceeds to the second step in any unit order. No amplifier ID, except the data on what trims have been done to which batch, is required.
Using this method, drifts in the range of 0.5-1 uV/oC (box method) for CMOS OpAmps can be achieved. Further reduction of the drift is limited by the curvature in the offset versus temperature dependence.