How do I measure the group delay flatness of a filter using a VNA?
Filter Group Delay Measurement
Group delay flatness determines: pulse distortion (a filter with non-flat group delay distorts the shape of a transmitted pulse, spreading it in time and reducing the peak amplitude), EVM (group delay variation across the signal bandwidth causes ISI and constellation spreading), and radar range resolution (group delay ripple degrades the range sidelobe level of a compressed radar pulse).
Group Delay Specifications
- Communication filters: Group delay flatness less than 10-50 ns over the channel bandwidth for basestation filters. Less than 1-5 ns for point-to-point microwave radio filters
- Radar filters: Group delay flatness less than 1-5 ns over the pulse bandwidth for low-sidelobe pulse compression
- Test equipment filters: Less than 0.1-1 ns for instrument-grade filters
Flatness: Δτ = τ_max - τ_min [within passband]
Aperture: Δf_aperture = N × Δf_step (N = smoothing points)
Recommended aperture: 1-5% of passband BW
For 100 MHz passband: aperture ≈ 1-5 MHz
Frequently Asked Questions
What VNA settings give the best results?
For accurate group delay measurement: number of frequency points: 1001-4001 (more points = better frequency resolution, but: more points increase the measurement time). IF bandwidth: 100 Hz-1 kHz (lower = less noise but slower). Aperture: 1-5% of the filter's passband (adjust to balance noise and resolution). Average: 4-16 averages to reduce random noise. Power level: set the VNA output power to the filter's linear operating range (typically -10 to 0 dBm for passive filters). An excessive number of averages and very narrow IF bandwidth can make the measurement take minutes; for most filters, 1001 points, 1 kHz IF BW, and 4 averages is a good starting point.
Why is the aperture setting important?
The aperture determines how the VNA computes the derivative of the phase: the VNA calculates: tau(f_n) = -(phase(f_n+k) - phase(f_n-k)) / (2 × k × delta_f), where k is the aperture in number of points. Small aperture (k=1): the derivative is computed between adjacent points. This gives the highest frequency resolution but: the group delay is very noisy (noise on the phase measurement is amplified by differentiation). Large aperture (k=10-50): the derivative is smoothed over a wider frequency range. Lower noise but: fine features (group delay ripple at narrow frequency spacing) are averaged out. The optimal aperture: for measuring broad group delay variation (the overall shape): use a large aperture. For measuring fine group delay ripple: use a small aperture with averaging to reduce noise.
What if the group delay is not flat enough?
If the filter's group delay exceeds the specification: redesign the filter (if possible): group delay flatness trades off with other filter parameters (selectivity, rejection, insertion loss). A filter designed for minimum group delay variation (equiripple or Bessel response) will have lower selectivity than a Chebyshev or elliptic filter. Add an equalizer: a group delay equalizer (all-pass filter) can be cascaded with the filter to flatten the group delay without affecting the amplitude response. The equalizer is designed to have a group delay that is complementary to the filter's group delay variation. Accept and compensate digitally: in digital systems, the channel equalizer (CTLE/DFE or OFDM channel estimation) can compensate for moderate group delay variation. This shifts the burden from the filter to the digital signal processing.