How does the CMRR of difference amplifier ICs change over time?

Hi Simeon,

The summary answer, is that unless we have specifically measured a part for this data, we don't have absolute certainty. We do, however, have a number of features which give us a significant amount of confidence in the product: unlike discretes, the thin-film resistors are effectively identical due to the processing on the wafer, we also have relative-trimming to get the parts to spec, which means the actual deviations from resistor-to-resistor is small. In this case, the parts track each one another very well, which is how we're able to obtain performance that is otherwise very difficult to get with discrete solutions (even with very expensive precision parts).

The specific measurements which embodies these sorts of measurements, are called "Long-term Drift." At ADI, we do actually perform Long-term drift on a few select parts, and the number of parts which are being measured this way is growing. Typically, however, these are the higher precision parts (references, ultra-low VOS amplifiers, etc) so the lower cost parts like the AD8276 would not normally receive this treatment (due to test cost). All is not lost however, the Industry standard test for observation is 1000-hours of use. From many precision references (not just from ADI), to amplifiers and even actual bench-top equipment, much of the real error accumulation happens over the first 1000 hours of life, after which the majority of errors settle out and the drift takes on a small, but non-zero error as a function of time.

The positive take-away here, is that the worst of the errors will settle themselves out in roughly the first month of service life (if always powered). The device should still be within specifications at that time, and from there on the total error over the lifetime should be negligible.

Please let me know if I can clarify anything further!

Cheers,

-David

Hi Simeon,

The summary answer, is that unless we have specifically measured a part for this data, we don't have absolute certainty. We do, however, have a number of features which give us a significant amount of confidence in the product: unlike discretes, the thin-film resistors are effectively identical due to the processing on the wafer, we also have relative-trimming to get the parts to spec, which means the actual deviations from resistor-to-resistor is small. In this case, the parts track each one another very well, which is how we're able to obtain performance that is otherwise very difficult to get with discrete solutions (even with very expensive precision parts).

The specific measurements which embodies these sorts of measurements, are called "Long-term Drift." At ADI, we do actually perform Long-term drift on a few select parts, and the number of parts which are being measured this way is growing. Typically, however, these are the higher precision parts (references, ultra-low VOS amplifiers, etc) so the lower cost parts like the AD8276 would not normally receive this treatment (due to test cost). All is not lost however, the Industry standard test for observation is 1000-hours of use. From many precision references (not just from ADI), to amplifiers and even actual bench-top equipment, much of the real error accumulation happens over the first 1000 hours of life, after which the majority of errors settle out and the drift takes on a small, but non-zero error as a function of time.

The positive take-away here, is that the worst of the errors will settle themselves out in roughly the first month of service life (if always powered). The device should still be within specifications at that time, and from there on the total error over the lifetime should be negligible.

Please let me know if I can clarify anything further!

Cheers,

-David