A bare transistor stage is a moody thing. Its gain depends on beta, on temperature, on the exact bias point, and on manufacturing tolerances that shift from one device to the next even within the same batch. Wrapping that stage in negative feedback, most commonly by inserting a resistor between the emitter and ground or between the source and ground, trades away some of that raw gain in exchange for a circuit whose behavior depends far more on a resistor value than on the transistor sitting next to it. This trade sits at the center of almost every practical amplifier design, and it plays out differently depending on whether the feedback wraps a single stage, a whole multistage amplifier, or an entire integrated circuit built around a current feedback topology instead of a conventional voltage feedback one.
Emitter Degeneration Feeds A Fraction Of The Output Signal Back Into The Input To Tame Transistor Variability
A common emitter or common source stage without any degeneration has a transfer characteristic that follows an exponential curve for a bipolar transistor or a square law curve for a MOSFET, which means the output is a strongly nonlinear function of the input except for signals small enough that the curve looks locally straight. Inserting a resistor in the emitter or source leg changes this picture directly, because any increase in output current now raises the voltage across that resistor, which subtracts from the effective base to emitter or gate to source drive and pushes the current back down. This is negative feedback in its most literal form, current flowing through the output leg is sampled as a voltage and fed back to oppose the very signal that produced it. The practical payoff shows up in two places at once, distortion drops because the transfer curve straightens out over a wider signal range, and the gain becomes far less sensitive to the beta or transconductance variation that would otherwise make every unit behave slightly differently off the production line.
The Feedback Factor Beta Sets How Much Gain A Degenerated Stage Trades Away For Predictable Behavior
The standard feedback model treats the transistor as a forward gain block A and the degeneration network as a feedback factor B, with the closed loop gain given by A divided by the quantity one plus the loop gain AB. When the loop gain AB grows much larger than one, the closed loop gain collapses to approximately the reciprocal of B alone, meaning the transistor's own gain drops almost entirely out of the equation and the circuit behavior is set by the feedback network instead. For a common emitter stage with an emitter resistor Re and a collector resistor Rc, this asymptotic gain works out to approximately minus alpha times Rc divided by Re, a ratio of two resistors rather than a transistor parameter, which is precisely why degenerated stages show up wherever a designer needs gain that survives temperature swings and part to part variation without hand selecting transistors. With no degeneration at all, Re equals zero, the feedback factor collapses to zero, and the stage reverts to a pure forward amplifier with a gain of minus Rc times the transconductance, fully exposed to every variation the bare transistor brings with it.
Negative Feedback Stretches Bandwidth By The Same Factor It Shrinks Gain Across A Constant Gain Bandwidth Product
The reason engineers accept a lower gain from a degenerated stage is that the lost gain does not simply vanish, it reappears as bandwidth. A voltage feedback amplifier's gain times its bandwidth stays close to a constant value, the gain bandwidth product, so pulling the closed loop gain down by a factor of ten through heavier feedback pushes the closed loop bandwidth up by roughly that same factor of ten. This relationship is why a single operational amplifier with a gain bandwidth product of a few megahertz can be configured as a high gain, narrow bandwidth stage for one part of a signal chain and a low gain, wide bandwidth buffer for another part, using nothing more than a change in the feedback network around it. The same principle explains why an emitter degenerated common emitter stage can trade its bare, high, unpredictable gain for a smaller but far more stable gain that also happens to extend usefully further up in frequency before rolling off, since the dominant pole set by the transistor's internal capacitances shifts outward as the loop gain that used to amplify signals at that pole gets reduced.
Multistage Amplifiers With Negative Feedback Risk Instability Once Accumulated Phase Shift Approaches A Full Cycle
A single degenerated stage rarely causes stability trouble because it has only one dominant pole contributing meaningful phase shift inside the useful frequency range. A multistage amplifier is a different problem entirely, because the total time delay through the chain equals the sum of the delays through every individual stage, and each of those delays adds its own phase lag that grows with frequency. Once the accumulated phase shift around the feedback loop approaches 180 degrees at a frequency where the loop gain is still equal to or greater than one, the negative feedback that was supposed to stabilize the circuit turns into positive feedback instead, and the amplifier either rings badly on transient signals or breaks into sustained oscillation. This is exactly why three stage transistor amplifiers built for high gain with heavy overall negative feedback need careful attention to where each stage's pole falls, since simply cascading gain stages and wrapping the whole chain in one feedback loop, without managing the individual pole locations, is a well known route to an amplifier that looks fine on paper and oscillates on the bench. Designers manage this risk with dominant pole compensation, which deliberately makes one stage's rolloff happen at a low enough frequency that the loop gain drops below one before the phase shift from the other stages can accumulate to a dangerous level, trading away some bandwidth in exchange for a guaranteed margin against instability.
Current Feedback Amplifiers Break The Fixed Gain Bandwidth Rule By Routing Feedback Through A Low Impedance Node
A conventional voltage feedback amplifier senses the output voltage and feeds a fraction of it back to a high impedance input node, which is exactly the arrangement that locks gain and bandwidth into a fixed product. A current feedback amplifier restructures the input stage so the inverting input is a low impedance node fed by a unity gain buffer, and the small error current flowing into that node, rather than an error voltage, gets mirrored onto a high impedance node where it develops the output signal through the amplifier's transimpedance gain. Because the feedback resistor connects to this low impedance node instead of a high impedance one, the gain setting resistors do not enter the loop gain expression the same way they do in a voltage feedback design, and the closed loop bandwidth stays relatively constant across a wide range of closed loop gains instead of shrinking as gain increases. A commercial current feedback part built for video and RF signal chains illustrates the effect directly, holding roughly 550 megahertz of small signal bandwidth at unity gain along with a slew rate near 940 volts per microsecond and gain flatness within 0.1 decibel out to about 65 megahertz, numbers that a comparable voltage feedback design would struggle to hold once configured for anything beyond unity gain.
The Feedback Resistor In A Current Feedback Amplifier Controls Stability In A Way Closed Loop Gain Never Does
The flip side of decoupling bandwidth from gain is that current feedback amplifiers get their stability set by a different knob entirely, the value of the feedback resistor itself rather than the closed loop gain a designer might naturally reach for first. Because the feedback resistor sits inside the loop gain expression at the low impedance input node, lowering it pushes bandwidth higher but also erodes phase margin, while raising it well above the manufacturer's recommended value tends to degrade frequency response and settling behavior rather than improving stability the way a designer might expect from voltage feedback intuition. A typical stability analysis for this kind of amplifier walks through the transimpedance curve on a logarithmic scale and checks where the closed loop response crosses zero decibels relative to the phase shift of the open loop transimpedance at that frequency, with a worked example landing around 120 degrees of phase shift at the crossover point, equivalent to 60 degrees of phase margin, a healthy figure that keeps step response free of excessive overshoot. This is precisely why current feedback amplifier datasheets specify a recommended feedback resistor value for each closed loop gain rather than leaving it to the designer's judgment, and why substituting an arbitrary resistor value chosen purely to hit a target gain, the way a designer might casually do with a voltage feedback part, is one of the most common ways to turn a current feedback design unstable.
Several practical consequences follow directly from how differently the two feedback topologies behave, and they matter every time a designer chooses between them:
- voltage feedback amplifiers trade gain for bandwidth along a fixed product, so a high gain configuration always costs bandwidth in a predictable and calculable way;
- current feedback amplifiers hold bandwidth roughly constant across gain settings, which suits video, RF, and high speed data acquisition chains where gain changes without a bandwidth penalty;
- current feedback stability depends on the feedback resistor value rather than the closed loop gain, so datasheets specify a recommended resistor for every gain rather than leaving the choice open;
- multistage voltage feedback designs need dominant pole compensation to keep accumulated phase shift from turning negative feedback into oscillation at high loop gain;
- degenerated single stage transistor amplifiers gain predictability and linearity from the same feedback mechanism, at the cost of raw gain that the feedback network absorbs.
Choosing Between Voltage Feedback And Current Feedback Topologies Depends On What The Application Actually Demands
A designer building a precision instrumentation front end, where DC accuracy and low input bias current matter more than raw speed, generally reaches for a voltage feedback topology, since its high input impedance and its predictable, well behaved response to gain changes make error analysis straightforward even though a comparison of DC precision reveals a real trade-off. An operational amplifier with an open loop gain of fifty thousand delivers roughly 0.002 percent DC error from the feedback loop alone, while a typical current feedback part with a transimpedance around six megohms and a feedback resistor near one kilohm lands closer to 0.02 percent, a full order of magnitude worse on that specific metric even though it wins decisively on bandwidth and slew rate. A designer building a video distribution amplifier or a high speed data acquisition front end, where gain often needs to change from unity to ten without redesigning the frequency response around it, gains far more from a current feedback part's nearly gain independent bandwidth than they lose from that DC precision gap, particularly once AC coupling or downstream calibration removes most of the DC error from the signal path anyway. Neither topology is universally better, and the underlying physics that makes one excel at DC precision is the same physics that makes the other excel at maintaining bandwidth across a range of gains, so the choice comes down to which of those two properties the application actually needs more.
A worked numeric comparison makes the trade-off concrete rather than abstract. Suppose a designer needs a stage with a closed loop gain of twenty, built first around a voltage feedback amplifier with a gain bandwidth product of 10 megahertz. The closed loop bandwidth for that configuration works out to roughly 10 megahertz divided by 20, or about 500 kilohertz, a hard ceiling the designer cannot exceed without either raising the gain bandwidth product of the chosen device or accepting a lower closed loop gain. Building the same gain of twenty around a current feedback amplifier instead changes the calculation entirely, since the bandwidth depends primarily on the feedback resistor rather than on the gain setting resistor ratio, so a part rated for several hundred megahertz at unity gain might still deliver well over one hundred megahertz at a gain of twenty once the feedback resistor is adjusted according to the manufacturer's gain versus resistor table, a bandwidth improvement of roughly two orders of magnitude over the voltage feedback alternative purely because the topology decouples gain from bandwidth. This gap is exactly why current feedback parts dominate applications like cable driving, video multiplexing, and high speed analog to digital converter buffering, where the required gain is rarely fixed at unity and a voltage feedback part's shrinking bandwidth at higher gain would otherwise force the designer into a much faster, more expensive, and higher power device just to claw back the bandwidth that current feedback provides at no extra cost in the topology itself.