1. Separate bias-current error from input-voltage offset
A high-resistance source changes the importance of input current. ADI draws an analogy between noise and DC offsets: an output error attributed to input offset voltage may instead be dominated by bias current acting through high input impedance. The current and voltage descriptions represent different mechanisms even though both can appear as an unwanted output voltage.
Input offset voltage is already a voltage error at the amplifier input. Bias current needs the surrounding impedance context to express its voltage consequence. A low offset-voltage entry therefore cannot, by itself, establish that the source-to-amplifier DC interface has a small error. The source resistance belongs beside the current-related description rather than being treated as unrelated background.
Keep this operating-error question separate from random noise. Bias current and current-noise specifications are not interchangeable, and reducing a static error does not establish a complete spectral-noise result. A focused comparison first distinguishes the DC contributions and then considers the noise mechanisms under their own conditions. This avoids choosing an input technology from one attractive voltage number or prescribing a compensation network before the actual source interface has been assessed.
2. Compare voltage noise with current-noise conversion
AN-940 describes input-referred voltage noise as a voltage source and current noise as current sources associated with the differential inputs. Source resistance converts a current-noise contribution into voltage. As that resistance becomes important, comparing only the voltage-noise entry removes part of the mechanism that determines the combined input-referred noise.
This explains why the lowest voltage-noise candidate need not answer the high-resistance requirement. The current-noise contribution depends on the impedance through which it acts, whereas the amplifier voltage-noise description is already expressed as voltage. Their relative importance can change with the source context, even when neither amplifier specification has changed.
Retain both contributions when comparing candidates at like conditions. AN-940 discusses differences among amplifier technologies, but those observations do not establish a universal bipolar or FET winner for an unidentified source. The useful selection question is which input-noise mechanisms matter with the actual resistance basis. It is distinct from a numerical crossover calculation, a complex-impedance analysis or the design of a transimpedance stage.
3. Include source and feedback resistance noise over the intended band
The source resistance itself is another noise contributor. ADI identifies Johnson noise from resistance at the noninverting input and notes that feedback-resistor noise can be significant in high-resistance circuits. Consequently, replacing the amplifier does not remove every noise source associated with the input and feedback network. The circuit contributes noise as well as converting amplifier current noise into voltage.
Those resistance-related effects do not scale in the same way. In the note's simplified model, Johnson noise increases differently with resistance from the voltage produced by amplifier current noise. This explains why the dominant contribution can change as source resistance changes, rather than remaining permanently assigned to whichever specification was largest in an initial comparison.
The frequency and bandwidth basis remains important when interpreting those quantities. A spectral-noise description and a total over a band are different forms of information. Keep the source and feedback contributions identifiable before combining them, rather than treating all resistance as one unexplained noise penalty. This preserves the model's technical meaning without producing an integrated-noise number, introducing a balancing resistor or transferring a device example into a component recommendation.
4. Carry the input-referred budget through the actual noise gain
Input-referred noise provides a common point at which contributions can be discussed, but the output question still needs the circuit's noise gain. AN-940 explains that the combined referred-to-input description is carried to the output through that gain. An input-noise comparison without its gain context therefore does not yet describe the corresponding output-noise requirement.
Noise gain is not always identical to signal gain. ADI explicitly distinguishes cases where the signal transfer and the transfer applied to input-referred noise differ. A single gain label can conceal which relationship is being used. This is particularly relevant when a comparison combines amplifier input specifications with an output objective supplied for the surrounding circuit.
Keep the reference point, frequency or bandwidth basis, and applicable gain definition together. The same source-resistance discussion can then be followed from individual mechanisms to an intelligible output requirement. The purpose is to compare like quantities, not to infer stability or perform a noise integration from a general application note. A satisfactory voltage-noise figure still needs the current, resistance and noise-gain context that gives it meaning.
Sources and further reading
ADI AN-940: input-referred voltage/current noise, summing noise sources, noise gain and source-resistance selection prose. Simplified resistive model only; no numerical crossover, noise integration, compensation/TIA recipe or amplifier ranking is supplied.