Measuring Filters with a VNA
A VNA is the ideal instrument for characterizing filters. A spectrum analyzer with a tracking generator can show the amplitude-versus-frequency response of a filter, but it cannot tell you the filter's input impedance, phase response, or group delay. The VNA measures all of these simultaneously in a single two-port sweep, giving you a complete picture of the filter's behavior. Building a low-pass harmonic filter, verifying a bandpass filter for a monoband antenna, or characterizing a commercial helical filter — the VNA does all of it in seconds.
Two-Port Filter Measurement Setup
Measuring a filter requires a two-port measurement. The test setup is straightforward:
- Port 1 connects to the filter input (signal injection)
- Port 2 connects to the filter output (transmission measurement)
- Both ports must be properly calibrated — perform SOLT calibration (short, open, load on Port 1; through between ports) before the measurement
- The calibration reference plane must be at the filter connectors, not at the VNA ports — use any cables you will also use for measurement during calibration
After calibration, connect the filter between Port 1 and Port 2. The VNA will sweep the frequency range and simultaneously display S11 (input reflection, upper or lower trace as set) and S21 (transmission, the other trace). You can display both traces simultaneously with different vertical scales.
NanoVNA with two SMA cables, Port 1 connected to filter input and Port 2 to filter output, with the filter component between them"
width="1536" height="1024"
fetchpriority="high"
class="hrb-figure-img">
Two-port VNA filter measurement. Port 1 drives the filter input; Port 2 measures the filter output. S21 (transmission) shows insertion loss and frequency response. S11 (reflection) shows how well the filter input is matched to 50 Ω.
View LargerReading the S21 Display
S21 displayed in log-magnitude format gives a direct picture of the filter's frequency response. The vertical axis is in dB; 0 dB means perfect transmission (all signal passes through), and negative values represent loss. Key things to read from the S21 display:
Passband insertion loss: The value of S21 at the intended operating frequency (or within the passband). Ideally close to 0 dB. A well-designed 7-element LC bandpass filter might show −0.5 to −1.5 dB insertion loss. Higher insertion loss means more signal is lost to resistive losses in the filter components. Exceeding 3 dB in the passband suggests the filter uses lossy components or is significantly out of tune.
Passband ripple: Some filter designs (Chebyshev, elliptic) deliberately allow the passband response to ripple up and down by a fixed amount (commonly 0.5 dB or 1 dB). The VNA shows this as periodic peaks and dips in the passband S21 trace. Butterworth (maximally flat) designs have no passband ripple. If a filter designed for no ripple shows ripple, it indicates component value errors or detuning.
3 dB frequencies: The points where S21 falls 3 dB below its maximum passband value. These define the filter's −3 dB bandwidth. Use the VNA's markers to find these frequencies precisely.
Stopband attenuation: The S21 value at frequencies outside the passband. Read from the trace at the specific frequency you care about — for a 40m low-pass harmonic filter, read S21 at 14 MHz (second harmonic) and 21 MHz (third harmonic).
Measuring Bandwidth and Shape Factor
Use the VNA's marker system to measure bandwidth precisely. Most VNA software (including the NanoVNA companion software) allows placing markers and reading the frequency and S21 value at each marker position.
3 dB bandwidth: Place a marker at the S21 peak (maximum passband value). Note the S21 value — for example, S21 = −0.8 dB at the peak. The 3 dB bandwidth extends from the lower frequency where S21 = −0.8 − 3 = −3.8 dB to the upper frequency where S21 = −3.8 dB. Use two markers to find these frequencies and compute their difference.
Shape factor: The ratio of the filter's 60 dB bandwidth to its 3 dB bandwidth. A shape factor close to 1 indicates a very sharp, steep filter skirt. Practical LC filters achieve shape factors of 2 to 10 depending on the number of elements. Quartz crystal filters achieve shape factors less than 2.
A 5-element crystal filter centered at 7.150 MHz shows:
- −3 dB bandwidth: 7.148 to 7.152 MHz = 4 kHz
- −60 dB bandwidth: 7.144 to 7.156 MHz = 12 kHz
Shape factor = 12 kHz / 4 kHz = 3.0
A shape factor of 3.0 is excellent for a crystal filter. This means the filter is relatively steep-sided — only 3× wider at −60 dB than at −3 dB, providing good adjacent-channel rejection without excessive passband narrowing.
Measuring Stopband Rejection
Place a marker at the stopband frequency of interest and read S21. Note that the VNA's dynamic range limits how deep a stopband you can measure. A NanoVNA with 70 dB of dynamic range cannot accurately measure a filter that attenuates more than about 65–70 dB — the measurement hits the noise floor of the VNA and shows a false attenuation floor rather than the actual filter attenuation.
To verify that a stopband measurement is valid and not limited by VNA dynamic range:
- Insert a known, precise attenuator pad (say, 10 dB) in the measurement path and observe whether the stopband S21 decreases by exactly 10 dB. If it does, you are measuring the filter; if not, you have hit the dynamic range floor.
- Compare the stopband S21 level to the VNA's noise floor, measured by connecting Port 1 to Port 2 through a large (40–50 dB) attenuator. Any S21 reading at or above this noise floor is real attenuation.
What S11 Tells You About a Filter
A well-designed 50 Ω filter should have low S11 (good impedance match) within the passband. High S11 in the passband indicates the filter input is not well-matched to 50 Ω, which can cause interaction with the source (transmitter or preamplifier). The input VSWR of a filter in the passband should typically be below 1.5:1 (S11 below −14 dB) for amateur radio applications.
In the stopband, S11 approaches 0 dB (nearly all input power reflects). This is expected and correct behavior — a filter in its stopband is a high-impedance or low-impedance reflective load, not a matched load. Do not be alarmed by high S11 in the stopband; it is the intended behavior.
If S11 in the passband is high (S11 > −10 dB), the most common causes are:
- Filter designed for a different impedance (e.g., a 75 Ω filter in a 50 Ω system)
- Component values significantly out of specification (especially inductors wound by hand)
- Ferrite core saturating due to excessive test signal level (use low VNA power levels)
- Unbalanced ground connections in the filter's physical layout
Measurement of Different Filter Types
| Filter Type | Expected S21 Shape | Key Measurement Points | Ham Radio Application |
|---|---|---|---|
| Low-Pass | Flat from DC to cutoff; rolls off above cutoff | −3 dB cutoff frequency; S21 at 2× and 3× cutoff for harmonic rejection | Harmonic filter on TX output |
| High-Pass | Rolls off below cutoff; flat above | −3 dB cutoff; S21 at 0.5× cutoff to verify rejection | AM broadcast rejection for HF RX |
| Bandpass | Peak at center frequency; rolls off both sides | Center frequency, −3 dB bandwidth, stopband rejection at ±1 octave | Monoband TX/RX filter, crystal filter for SSB IF |
| Band-Stop (Notch) | Deep dip at notch frequency; near-flat elsewhere | Notch frequency, notch depth, bandwidth of notch at −3 dB and −20 dB | Interference notch, broadcast rejection |
Full Example: 40m Bandpass Filter Verification
A Butterworth 7-element bandpass filter is designed for the 40m band (7.0–7.3 MHz). After construction, a VNA measures the following with the filter connected between Port 1 and Port 2 (calibration performed at the filter connectors):
- Set frequency range: 1 MHz to 30 MHz (wide enough to see the passband and key harmonics)
- Calibrate: SOL on Port 1, Through between ports
- Connect filter: Port 1 to filter input, Port 2 to filter output
- Set two traces: Trace 1 = S21 log magnitude, Trace 2 = S11 log magnitude
- Place markers at: 7.0, 7.15, 7.3, 14.0, 21.2, 28.0 MHz
Results:
| Frequency | S21 | S11 | Interpretation |
|---|---|---|---|
| 7.00 MHz | −1.8 dB | −11 dB | Near passband lower edge — slight increase in S11 |
| 7.15 MHz | −0.9 dB | −22 dB | Passband center — excellent match, low insertion loss |
| 7.30 MHz | −1.6 dB | −12 dB | Near passband upper edge — symmetrical behavior |
| 14.00 MHz | −38 dB | ≈0 dB | Second harmonic — 38 dB rejection (exceeds FCC requirement) |
| 21.20 MHz | −52 dB | ≈0 dB | Third harmonic — 52 dB rejection |
| 28.00 MHz | −61 dB | ≈0 dB | Fourth harmonic — approaching VNA dynamic range limit |
Conclusion: The filter passes with 0.9 dB passband insertion loss and exceeds FCC spurious emission requirements across all measured harmonics. The S11 of −22 dB at the passband center indicates excellent 50 Ω matching.
Frequently Asked Questions
Why does my filter show higher insertion loss than the design specification?
The most common cause is inductor Q (quality factor) lower than assumed in the design. The Q of a hand-wound coil on a ferrite core depends strongly on the core material, the winding technique, and the frequency. If your design assumed Q = 100 and your actual inductors have Q = 50 due to core loss or wire resistance, insertion loss will double. Other causes include resistive losses in the filter's copper traces or component leads, and inductors wound with too few turns (reducing Q). Measure the inductors on an RLC bridge or Q meter before building to verify Q at the operating frequency.
My S21 measurement shows periodic ripple even though the filter is a Butterworth design. Why?
The ripple is most likely caused by standing waves in the test cables between the VNA and the filter, not by the filter itself. If the test cables are not perfectly terminated (or if the filter's S11 is not exactly −∞ dB), multiple reflections between the VNA ports and the filter cause a standing-wave pattern in the transmission measurement. The solution is to ensure the VNA is properly calibrated, the filter is connected at the calibration reference plane, and the test cables have no damaged or dirty connectors. A small ripple of ±0.2 dB at all frequencies is usually acceptable; larger ripple indicates a calibration or connection problem.
Test Your Knowledge
Answer the questions below to check your understanding. Every answer can be found in the lesson above.