Part Number: INA241B
Hello,
We're designing with the INA241B and had a quesiton about using an opamp to buffer that input.
It doesn't seem like this is necessary, but i found another post discussing better accuracy with the opamp.
Could you please commnet on the best configuration & errors we could anticipate?
Thank you, Keith
Hi Keith,
Thank you for your post.
The short answer: you don't need a buffer on the signal inputs (IN+/IN-), but you almost certainly do need one on the REF pins if you're using a resistor divider to set a custom reference voltage. That's likely what the post you found was referring to.
Signal Inputs (IN+/IN-): No Buffer Needed
The INA241B is designed to connect directly to your shunt resistor [1]. Its 120 dB minimum CMRR [1] handles the high common-mode voltage natively, and no buffering is required or recommended on the sensing inputs. One important caution: avoid adding RC filters at the inputs. Component mismatches from tolerances and temperature drift can degrade CMRR from ~100 dB down to 40–60 dB. If you need filtering, place it at the output instead [3].
REF Pins: Buffer Required for Resistor Dividers
This is where the accuracy problem lives. The REF pins are not high-impedance inputs — they draw approximately 35 µA of bias current through an internal ~606 kΩ resistance tied to an internal bias voltage [2]. If you're driving them from a resistor divider, that current will load down the divider and pull your reference voltage away from the intended value.
This isn't theoretical. A real-world case confirmed a 13 mV drop on a 0.613 V reference (Thevenin impedance of ~8.33 kΩ) — a ~2% error before the INA241B even processes a signal [2]. TI explicitly states that using an unbuffered resistor divider for single-ended output is not recommended because internal impedance shifts adversely affect device performance specifications [4].
The fix: Use an op-amp configured as a voltage follower between your resistor divider and the REF1/REF2 pins, exactly as shown in Figure 7-9 of the INA241B datasheet [1]. The op-amp's low output impedance absorbs the REF pin bias current without disturbing the reference voltage.
When you don't need a buffer: If your REF pins connect directly to VS, GND, or a dedicated low-impedance precision reference IC, no buffer is needed — those sources can sink the 35 µA without issue [5].
Error Budget to Anticipate
Even with a properly buffered reference, account for the INA241B's inherent error sources:
|
Error Source |
Typical |
Maximum |
|
Gain error (25°C) |
±0.02% |
±0.1% |
|
Gain error (full temp) |
— |
±0.15% |
|
Gain error drift |
±0.2 ppm/°C |
±5 ppm/°C |
|
Input offset voltage |
±25 µV |
±150 µV |
|
Offset drift |
±100 nV/°C |
±500 nV/°C |
Source: INA241B datasheet [1]
Without a buffered reference, REF pin loading adds a fixed offset error on top of all of the above — and that offset scales with the Thevenin impedance of your divider, making it difficult to calibrate out.
Recommended Configuration Summary
|
Scenario |
Buffer Needed? |
|
REF pins → VS or GND |
|
|
REF pins → precision reference IC |
|
|
REF pins → resistor divider (single-ended output) |
|
|
IN+/IN- → shunt resistor |
|
|
Input RC filter |
|
A few things would help sharpen this further: Are you measuring unidirectional or bidirectional current (this determines whether you need a mid-supply reference at all)? What's your target reference voltage and supply rail? And do you have a precision reference IC available, or are you planning to use a divider? If you're going the divider route, the op-amp selection matters too — you'll want rail-to-rail output capability and low offset to avoid introducing new errors at the reference node.
Citations
I hope this helps,
Joe
Hello,
Thank you for the information. Would you agree that we can use the Internal REF1 and REF2 buffer without adding an opamp if 1.65V works in our system and the accuracy of 0.15%.
Thank you, Keith
