Current-Feedback Amplifier Networks: Preserve Feedback Resistance When Changing Gain

System boundary

A noninverting current-feedback operational amplifier with external feedback resistor Rf and gain resistor Rg. This is not a generic photodiode transimpedance circuit or a voltage-feedback amplifier with a constant gain-bandwidth assumption.

Integration interfaces

Integration risks

Distinguish amplifier architecture from a current-to-voltage application

A voltage-feedback amplifier can be wired as a transimpedance receiver, converting input current into output voltage. That application does not make the underlying device a current-feedback amplifier. Confirm the active part's architecture from its documentation before applying resistor-selection rules. Similar circuit symbols and the same nominal closed-loop gain are not sufficient identification.

In the simplified current-feedback architecture, the inverting terminal is the low-impedance output of an input buffer and an error current drives an internal transimpedance. The external feedback resistor determines how the output returns that current. This difference explains why an absolute resistance can control dynamics even when the familiar noninverting gain expression remains approximately one plus Rf/Rg.

Separate the gain numerator from the feedback denominator

For an ideal unity input buffer with negligible output resistance, let Z(s) be the amplifier's open-loop transimpedance in ohms. Current balance gives closed-loop voltage gain G divided by one plus Rf/Z(s), where G equals one plus Rf/Rg. Rf is therefore the feedback transresistance that competes with Z(s) in the loop; G sets the nominal signal gain.

With this ideal buffer and fixed Rf, changing Rg changes G without changing that loop denominator. This is the limited sense in which bandwidth can remain approximately independent of voltage gain. It is not a promise that every real device has identical response at every gain, and it does not authorize arbitrary Rf values. The selected amplifier's higher-order dynamics and operating limits still apply.

Check a same-gain substitution that changes the loop by a decade

Assume an approved circuit uses Rf equal to 1 kilohm and Rg equal to 1 kilohm for nominal gain two. Replacing both with 10 kilohms preserves that ratio and ideal gain. However, the feedback term Rf/Z(s) becomes ten times larger at every frequency in the simplified model. The closed-loop response is therefore not unchanged merely because a DC ratio test passes.

For a single-pole internal transimpedance Z(s) = Z0 divided by one plus s tau, algebra gives a closed-loop pole rate of one plus Z0/Rf, divided by tau. This illustrative model predicts a lower pole frequency as Rf rises. Real amplifiers have further poles and device-specific limits, so this expression is useful for understanding the substitution, not selecting a production bandwidth from an invented internal model.

Include the input buffer when comparing gain settings

Let ri be the nonzero resistance in the input-buffer model. The effective feedback transresistance becomes Rf plus ri times G. Thus even if Rf is held fixed, a higher gain increases the second term and can alter the loop. Both terms have units of ohms, which is a useful check against accidentally adding a dimensionless voltage gain to a resistance.

For assumed Rf of 1 kilohm and ri of 25 ohms, gain two gives effective feedback transresistance 1050 ohms. Gain ten gives 1250 ohms. If both external resistors are scaled tenfold at gain two, the corresponding value becomes 10050 ohms, not 1050. These numbers illustrate the model's sensitivity; ri and an acceptable feedback value must come from the actual amplifier and operating condition.

Do not treat a smaller feedback resistor as a free speed upgrade

Reducing feedback resistance can move the loop crossover toward frequencies where additional internal poles contribute phase shift. The apparent gain in bandwidth may be accompanied by peaking, overshoot or instability. A unity-gain connection therefore must not automatically short output to the inverting input as one might do with an appropriate voltage-feedback follower.

Use the specified feedback arrangement for the selected device, gain, load and supply. Adding a feedback capacitor also changes impedance with frequency and is not a universal remedy copied from another amplifier family. If an alternate compensation network is needed, it requires its own loop analysis and validation. Do not remove the absolute feedback-resistance requirement from the drawing to simplify procurement.

Preserve the physical feedback path as well as nominal values

The resistor layout has parasitic capacitance and inductance, and the output load may include a cable or receiver capacitance. A remotely routed feedback path does not necessarily behave like a compact reference circuit. Identify which pads and interconnect lengths are included in the reviewed network, and avoid treating the ceramic substrate as an ideal zero-parasitic carrier.

A switched-gain design must also preserve a valid feedback path during transitions. Evaluate the actual switch impedances and intermediate states rather than only the two settled settings. If the network is laser trimmed, state whether Rf may move or whether adjustment is restricted to another element. Achieving the final gain by moving the wrong resistor can alter dynamic behavior while satisfying a static electrical target.

Verify dynamic behavior at every supported gain and load

Keep the exact resistor values, amplifier supply, load and board configuration with the test record. Compare small-signal frequency response and an appropriate step response at each supported setting. Look for peaking, overshoot, ringing and settling to the application's actual error band, rather than reporting rise time alone as the performance criterion.

Check relevant output levels as well as small signals because slew behavior, current limits and thermal settling can affect large transitions. Measurement probes and fixtures can add capacitance to the feedback or output nodes; use a method suitable for the bandwidth and retain its loading model. A clean waveform under one light load does not establish all cable or receiver configurations.

Order the gain network with an absolute-value constraint

The drawing package should name the amplifier architecture and identify Rf and Rg unambiguously. Specify absolute feedback resistance, permitted trim range, ratio, relevant parasitics and the operating configuration used to approve them. If there are multiple gain settings, list the actual resistor pair for each rather than a gain column alone.

ChipSimple can review a custom thick-film network against the supplied circuit requirements. The analog owner establishes the active-loop design, and the system owner validates the connected response. A future resistor-value or amplifier substitution should trigger that review even when nominal voltage gain remains identical, preserving the distinction between ratio accuracy and a stable, usable wideband signal path.