A frequency response measurement is a description of a circuit and its test conditions together, never of the circuit alone. Change the source impedance, the load, or the drive level, and the curve moves. A plot published without those conditions is an advertisement, not data, and a sweep you run yourself is only reproducible if you write them down.
What a response curve is actually recording
The basic procedure is simple: apply a signal of known amplitude at one frequency, measure the output, compute the ratio, and repeat across a range. Plotting that ratio against frequency gives the magnitude response, and measuring the timing difference between input and output gives the phase response. The complication is that most circuits are not ideal voltage sources driving ideal loads. The generator has output impedance, the circuit under test has input impedance that varies with frequency, and the load you attach interacts with the output stage. Each of those forms a divider, and each divider is frequency dependent. The curve you record therefore includes the test fixture. Measuring the input at the circuit’s own terminals rather than at the generator removes one of those errors immediately.
Source and load impedance move the result
A filter designed to be driven from a low impedance and terminated in a specific load will not show its intended corner frequency if either condition is violated. Driving a passive filter from a generator with significant output impedance shifts the corner and changes the damping. Loading an amplifier more heavily than the design intends reduces high-frequency response as the output stage runs out of current. The same is true in reverse: measuring with an unrealistically light load flatters the result. Decide which loading represents the real use case and test with it, and if the circuit is expected to operate across a range of loads, sweep at the extremes of that range rather than only in the comfortable middle. Two curves that bracket reality are worth more than one optimistic curve.
Drive level exposes nonlinearity
Frequency response as usually plotted assumes the circuit is linear, meaning the ratio does not depend on amplitude. Real circuits stop being linear as they approach their limits, and some are noticeably nonlinear well before. Sweeping at a level high enough to clip produces a curve that looks compressed at the frequencies where gain is highest. Sweeping too low buries the output in noise and produces a ragged curve that appears to roll off early. Running the same sweep at two or three different drive levels is the standard check: if the normalized curves overlay, the circuit is behaving linearly across that range. If they separate, the response is amplitude dependent and a single curve cannot represent it honestly. Record the drive level on every plot.
Sweep method and settling time
Whether you step frequency point by point or sweep continuously, the circuit needs time to reach steady state at each frequency before the measurement is taken. Sharp filters and high-Q resonances settle slowly, and sweeping faster than they settle drags peaks sideways and flattens them. If a resonance appears at a different frequency when you sweep upward than when you sweep downward, the sweep is too fast, not the circuit unstable. Automated analyzers handle this with a settling delay you can set. Point-by-point measurement with a generator and a meter is slower but immune to the problem if you simply wait. Logarithmic frequency spacing is usually the right choice, because it spreads points evenly across decades rather than crowding them at the top.
Making the sweep reproducible later
A response curve is most valuable when compared against another one, which means the second measurement must be made the same way. Record the generator amplitude and its source impedance, the load used, the measurement points on the circuit, the probe or input configuration, the frequency range and spacing, the settling time, and the ambient temperature if the circuit is at all temperature sensitive. Save the raw numbers, not just an image, so a later comparison can be plotted on the same axes. If the circuit passes through a connector or a cable in normal use, decide whether those are part of the device under test and state that decision. Two engineers producing different curves for the same board almost always differ in one of these conditions rather than in the board.
What to check before you act
- Measure the input at the circuit’s terminals, not at the generator output.
- Sweep with the load the circuit will actually drive.
- Repeat at a second drive level to confirm linear behavior.
- Record source impedance, load, level and spacing with every curve.
Common questions
Why does my measured corner frequency differ from the design value?
Source and load impedance are the usual causes, followed by component tolerance. Check what impedances the design assumed before concluding the circuit is wrong.
Do I need a network analyzer?
Not for audio and low-frequency work, where a generator and a voltmeter or scope will do. A dedicated analyzer mainly buys speed, dynamic range and phase accuracy.
The practical takeaway
Sweep the circuit under the conditions it will actually face, verify linearity at more than one level, give sharp responses time to settle, and record every condition alongside the numbers. A curve with its test conditions attached can be compared and defended; one without them cannot be reproduced, even by you.
References and further reading
- FCC: background on radio frequency signals and measurement
- Department of Energy: electronics and equipment energy performance
- NFPA: electrical safety resources and code background
Numerical scenarios are illustrative unless identified otherwise. Follow the exact product instructions; component ratings and local installation requirements can differ.
