Two identical boards, the same sensor, the same firmware — and readings that differ by 3% between units. In a large share of the cases we have debugged, the cause is not the sensor or the ADC. It is the op-amp offset error that nobody put into the error budget, and the temperature drift that turns a small offset into a large one. This article walks through the underlying logic: what offset and drift actually are, how to convert them into a system error, and how that decides whether a cheaper amplifier is a safe substitution.
Two separate errors, not one
The datasheet gives you two numbers, and they behave completely differently.
- Input offset voltage (VOS) is the differential voltage you would have to apply at the inputs to make the output zero. It is a fixed DC error at a given temperature, specified at 25 °C, and it is trimmed at wafer level or cancelled by a chopper loop.
- Offset voltage drift (dVOS/dT) is the rate at which that error moves with temperature, in µV/°C. It is not trimmed, because it is a property of the matching between devices on the die.
This distinction is the whole point. A general-purpose amplifier with 2 mV of offset and 10 µV/°C of drift contributes only 2 mV at room temperature — but 0.85 mV of extra error at 105 °C. A zero-drift part with 10 µV of offset and 0.05 µV/°C contributes 10 µV across the entire industrial range.
The error budget, written out
The input-referred error of a real stage is roughly:
Verr ≈ VOS + (dVOS/dT × ΔT) + (IB × Rsource) + Vnoise
Four terms, and they trade against each other. Note the third one: input bias current flowing through the source impedance is an error too. This is why a low-offset bipolar amplifier can be worse than a higher-offset CMOS part in a high-impedance bridge circuit, and why a low-bias-current part can be worse in a low-impedance shunt circuit. You cannot pick an amplifier from a single number.
Worked example: a load-cell front end
Take a bridge load cell with 2 mV/V sensitivity, a 5 V excitation, so 10 mV at full scale, amplified by 100 for a 1 V span. Now compare two amplifiers at 105 °C, an 80 °C swing from the 25 °C trim point.
| Amplifier class | Typical VOS | Typical drift | Extra error at 105 C | Referred to 10 mV full scale |
|---|---|---|---|---|
| General purpose | 2 mV | 10 µV/°C | 0.80 mV | 8.0% of full scale |
| Precision | 150 µV | 1 µV/°C | 0.08 mV | 0.8% of full scale |
| Zero-drift | 10 µV | 0.05 µV/°C | 0.004 mV | 0.04% of full scale |
Same circuit, same layout, same sensor — an error difference of two orders of magnitude. This is the number that decides whether a substitution is safe, and it is absent from most "equivalent part" conversations.
How this drives the substitution decision
The temptation when a part is short or expensive is to look at the pinout, confirm it matches, and swap in whatever has the same GBW. On offset-critical stages that is exactly the wrong test. Three questions settle it:
- Is the error budget temperature-limited or trim-limited? If the design is trimmed at test and runs in a stable enclosure, initial offset dominates and a cheaper part may pass. If it runs over −40 to +85 °C without a trim, drift dominates and the cheaper part will fail in the field, not on the bench.
- What is the noise gain? High-gain stages multiply input error directly. A gain-of-1 follower can tolerate an offset the same amplifier would never survive at gain 200.
- What does the qualification actually cost? Re-qualifying an instrumentation front end is not a single bench test: it is a temperature sweep, a repeatability run, and often a customer-facing accuracy claim to re-issue. That cost normally dwarfs the per-unit saving on a ¥1-class part, which is why we usually advise leaving precision stages alone and taking the substitution saving where it is genuinely free.
Parts we hold for precision and zero-drift work
| Model | Class | Typical position | Link |
|---|---|---|---|
| COS2333SRA | Zero-drift, 5 V, RRIO | Sensor conditioning, transmitters, battery instruments | COS2333SRA |
| COS8551TRA | Zero-drift, single | Precision DC gain stages, thermocouple and bridge front ends | COS8551TRA |
| COS2388SRA | Zero-drift, dual | Two-channel precision acquisition | COS2388SRA |
| COSOP07SR | 36 V precision, trimmed | Classic low-offset DC instrumentation, 4-20 mA loops | COSOP07SR |
Browse the full range in amplifiers and data conversion. Tell us your full-scale signal, the temperature range, the noise gain and whether a production trim is available, and we will quote the parts that fit that budget rather than the closest part number.
我们作为独立分销商保证所供元器件为原装正品，并提供日期码与检验选项，助您在量大量前完成验证。
价格随采购数量与行情流动，所示价格仅供参考。欢迎询价获取实时报价，量大从优。
About us & trademarks: JTDZ Tech (Shenzhen Jinteng Electronics Technology Co., Ltd.) is an independent distributor of electronic components. We are not an authorised distributor or representative of the manufacturers named in this article. Manufacturer and part names are used for indicative purposes only, to help identify the components discussed and the parts they are commonly compared with; all trademarks remain the property of their respective owners. Always verify specifications against the manufacturer's datasheet before design-in or purchase.
