Published Nov 20, 2026 · Last updated Nov 20, 2026 · 5 min · IndieRF

Mismatch Uncertainty in a Cascade: +1.26 to −1.18 dB

The antenna port is two thirds of the gain ripple.

A 915 MHz whip at its printed 2:1 VSWR, a PSA4-5043+ LNA and a CBP-915C+ filter into a second PSA4-5043+ land between +1.26 dB and −1.18 dB of the 50 Ω gain. Independent phases give +0.88 / −0.81 dB RSS; a sheet's mismatch loss says a fixed 1.45 dB.

Four ports from three data sheets

The Linx sheet prints "VSWR ≤ 2.0 typical" for the ANT-916-CW-HWR, which this note takes at 2.0:1, 9.54 dB of return loss. The PSA4-5043+ prints typical return loss only, 11.4 dB in and 13.9 dB out on its 1.0 GHz row at 5 V, the nearest row to 915 MHz. On the 0.5 GHz row (10.5 dB in, 15.0 dB out) the worst case grows to +1.36 / −1.26 dB, so the frequency between the rows is worth about 0.1 dB. The CBP-915C+ prints one pass-band VSWR, 1.24:1 typical and 2.10:1 maximum, and this note applies it to both ports: 19.40 dB typical. The load is the second PSA4-5043+'s 11.4 dB input.

Where the ripple comes from

Two stages with good reverse isolation cascade exactly to

S21=S21A S21B1−ΓoutΓinS_{21} = \frac{S_{21A}\,S_{21B}}{1 - \Gamma_{out}\Gamma_{in}}

so with x=∣Γout∣ ∣Γin∣x = |\Gamma_{out}|\,|\Gamma_{in}| and the phase unknown, the gain moves between −20log⁡10(1+x)-20\log_{10}(1 + x) and −20log⁡10(1−x)-20\log_{10}(1 - x). In phase the denominator shrinks and the gain rises: 10 dB of return loss against 10 dB (x = 0.1) runs from +0.92 dB to −0.83 dB. The form quoted on mismatch-uncertainty calculators, 20log⁡10(1±x)20\log_{10}(1 \pm x), is the right pair of magnitudes for the denominator itself, so it puts +0.83 dB on top and −0.92 dB underneath, with the signs crossed for gain.

InterfaceReturn loss, out against in (dB)xHigh (dB)Low (dB)
Whip to LNA9.54 against 11.40.0897+0.82−0.75
LNA to filter13.9 against 19.400.0216+0.19−0.19
Filter to stage 219.40 against 11.40.0288+0.25−0.25

Cascade adds the highs and the lows for the worst case, which the chain reaches when all three products line up, and takes the root-sum-square for independent phases. The whip interface is 65% of the worst case, so the antenna match is the cheap place to spend: a whip at 1.5:1 takes the chain to +0.92 / −0.89 dB.

Rule of thumb: one interface ripples about ±8.7·x dB, so 6 dB more return loss on either side halves it. Two 2:1 ports give x = 0.111 and an estimate of 0.97 dB against an exact +1.02 / −0.92 dB, and the high side pulls further away from the estimate as x grows. It also assumes the blocks between ports pass signal one way only, which a filter does not.

What a spreadsheet carries

A level-plan sheet carries one of two numbers here. One is nothing, 0 dB. The other is a mismatch-loss column, 10log⁡10(1−∣Γ∣2)10\log_{10}(1 - |\Gamma|^2) per port, which sums to 1.45 dB over the six ports, 0.51 dB of it the 2:1 whip. The printed gain and insertion loss were measured between 50 Ω ports, so each part's own share of that is already in them, and a fixed loss has no phase in it. Neither sheet can say the second stage may see the signal 1.26 dB hot or 1.18 dB cold, and that window is the level tolerance an AGC set point or a power reading at that plane inherits.

The bound is on gain. The first LNA's noise figure also moves with the antenna match, through noise parameters the PSA4-5043+ sheet leaves out.

The filter sends reflections back

The budget rows treat every block as one-way, which suits the two amplifiers. The CBP-915C+ is passive and reciprocal, so the second LNA's reflection crosses it twice and reaches the first LNA's output 2.2 dB down: a fourth path with x = 0.042, worth +0.37 / −0.36 dB on its own. Cascading the filter as reciprocal at 1.10 dB each way, with every phase free, the chain reaches +1.66 / −1.52 dB, about 0.4 dB past the budget's high and 0.3 dB past its low. More loss in the passive block shrinks that path by twice its loss in dB, and the gap shrinks with it.

The same chain with other parts

Each row is the gain ripple at the second stage's input, in dB. Worst case and RSS are Cascade's Mismatch bound rows. The sum of 20log⁡10(1±x)20\log_{10}(1 \pm x) is the calculator form, high and low as it labels them. The last column takes the filter as reciprocal with 1.10 dB of loss.

CaseWorst caseRSSΣ 20·log10(1 ± x)Filter two-way
As printed, typical+1.26 / −1.18+0.88 / −0.81+1.18 / −1.26+1.66 / −1.52
CBP-915C+ at its 2.10:1 maximum+2.33 / −2.14+1.36 / −1.24+2.14 / −2.33+2.78 / −2.45
PSA4-5043+ 0.5 GHz row (10.5 in, 15.0 out)+1.36 / −1.26+0.97 / −0.88+1.26 / −1.36+1.75 / −1.60
Antenna at 1.5:1+0.92 / −0.89+0.58 / −0.55+0.89 / −0.92+1.32 / −1.23
Every port 10 dB+2.75 / −2.48+1.59 / −1.43+2.48 / −2.75+3.62 / −3.02

The filter's 2.10:1 limit nearly doubles the worst case, so the filter's VSWR is the line worth holding at incoming inspection.

References

  • Linx Technologies, ANT-916-CW-HWR-ccc Data Sheet, revised 8/1/2017, Electrical Specifications: VSWR ≤ 2.0 typical, 916 MHz centre, on a 10.16 cm reference ground plane.
  • Mini-Circuits, PSA4-5043+ Low Noise, High IP3 Monolithic Amplifier, data sheet Rev. D (ECO-026573), page 2, Electrical Specifications at +25 °C, VD = +5.0 V: input and output return loss at 0.5 and 1.0 GHz.
  • Mini-Circuits, CBP-915C+ Surface Mount Bandpass Filter, data sheet Rev. B (M174392), page 2, Electrical Specifications at 25 °C, pass band 902.5 to 927.5 MHz: insertion loss and VSWR.

Questions? Contact

Discussion

No comments yet. Start the thread with a measurement, a correction, or a worked example.