How do I calculate the required filter order to achieve a specific rejection at a given frequency offset?
Filter Order Determination
Determining the correct filter order is the first step in filter design. Too few resonators means insufficient rejection; too many means excessive insertion loss, size, cost, and group delay distortion. The mathematical relationships between filter order, passband specification, and stopband rejection are well-established for each filter type.
The frequency selectivity parameter Ωs = fs/fc (for lowpass) or Ωs = Δfs/Δfp (for bandpass, ratio of stopband to passband bandwidth) captures the steepness requirement. Small values of Ωs (stopband close to passband) require high filter orders. Large values (stopband far from passband) require lower orders.
For bandpass filters, the required order depends on fractional bandwidth. A narrowband filter (1% bandwidth) needs more resonators than a wideband filter (20% bandwidth) for the same relative rejection, because the same absolute frequency offset corresponds to a larger Ωs in the narrowband case.
n ≥ log(10^(As/10)-1) / (2·log(Ωs))
Chebyshev order:
n ≥ cosh⁻¹(√((10^(As/10)-1)/(10^(Ap/10)-1))) / cosh⁻¹(Ωs)
Example: As=60 dB, Ap=0.5 dB, Ωs=2:
Butterworth: n ≥ 10
Chebyshev: n ≥ 5
Elliptic: n ≥ 4
Frequently Asked Questions
Does higher order always mean better?
Higher order provides steeper rolloff but increases insertion loss (each resonator adds loss proportional to 1/Qu), group delay (more resonators means more delay and more delay variation), physical size, and cost. The optimal order is the minimum that meets the rejection specification.
How does fractional bandwidth affect order?
Narrower fractional bandwidth requires more resonators for the same relative rejection. A 1% bandwidth filter at X-band (100 MHz at 10 GHz) requires approximately twice the order of a 10% bandwidth filter for the same shape factor because the frequency selectivity parameter Ωs is much larger relative to the passband.
Can I verify the order with simulation?
Yes. After calculating the theoretical order, simulate the filter response using Genesys, HFSS, or a filter design tool to verify that the rejection meets specification with realistic component Q values. The finite Q of real resonators degrades rejection compared to the ideal (infinite Q) mathematical prediction.