Calculators / 06
Via Stub Resonance Calculator
The unused length of a via below the layer a signal leaves on is an open-circuited stub. At the frequency where it is a quarter wavelength long it looks like a short to the line, and energy that should pass through it is sent back instead. The fix is to keep that frequency well above everything the channel carries — by backdrilling, or by choosing layers that leave no stub.
Reading the numbers
- The notch falls as the stub grows: halve the length and it doubles.
- Loss starts to build well below the notch. Read |S21| at Nyquist to see what the stub already costs where the budget is quoted.
- The margin k is yours to set. Nothing in a specification fixes it; it depends on how much of the notch's shoulder the channel can afford.
Every value is in the address bar as you change it, so a link reproduces the calculation exactly.
Type values with SI prefixes — 100n, 2.2p, 16G — or
drag the sliders; a bare number keeps the prefix already shown.
The formulas
Go deeper — assumptions, how it is checked, and sources
Assumptions. The stub is lossless and has the line's impedance, so the notch is infinitely deep. A real stub has loss and a different impedance, which make the notch finite and shift its shoulder; the frequency is the part to trust. The effective Dk of a via barrel is not simply the laminate's, because the field around it is not uniform.
How it is checked. |S21| is 0 dB at DC and at twice the notch frequency, where the stub is a half wave; it vanishes at the notch; and the notch is exactly where the stub is a quarter wavelength.
Source. D. M. Pozar, Microwave Engineering, 4th ed., Wiley, 2012, §2.3 and §5.2.
Laminate presets. Each preset is copied from its manufacturer’s own document for one named construction, and every value is typical, not guaranteed: Shengyi S1141 technical data sheet (2021), a standard-grade FR-4 measured on a 1.6 mm 8×7628 board; Isola 370HR Dk/Df tables, Rev. C (2020), core 1×2116 at 51 % resin; Panasonic MEGTRON 4 R-5725, data sheet 22040132 (2022), 2116 at 56 %; MEGTRON 6 R-5775(N), 22040130 (2022), 2116 at 56 %; MEGTRON 7 R-5785(GE) and (GN), 22040128 (2022), 2116 at 53 % and 55 %; and Rogers RO4350B at its design Dk of 3.66, not the 3.48 process Dk. Dk and Df change with frequency, construction and test method, so the preset reads them at the Nyquist frequency you set. It interpolates only between points measured by one method, and outside them it uses the nearest tabulated point and says so. Two cores of the same product can differ by more than two products do, so sign off with your fabricator’s stackup values. The sources are rows C-27 to C-33 of the claim ledger.
Related
Sources
- Isola — 370HR Dk/Df tables C-27
- Panasonic — MEGTRON 4 data sheet C-28
- Panasonic — MEGTRON 6 R-5775(N) data sheet C-29
- Panasonic — MEGTRON 7 data sheet C-30
- Rogers — RO4350B laminates C-32
- Shengyi — S1141 technical data sheet C-33
Rows marked with a claim id are tracked in the claim ledger, which records what each source can and cannot establish.