Calculators / 09
PCB Channel Loss Budget Calculator
A PCB trace loses signal two ways: in the copper, where skin effect and roughness raise the resistance, and in the laminate, where the field does work on the dielectric. The copper term grows as √f and the dielectric term as f, so at high enough frequency the laminate always wins — and which one dominates your channel decides whether a wider trace or a better laminate is the fix.
Reading the numbers
- The dielectric loss is 2.3·f·Df·√Dk dB per inch with f in GHz: halving Df halves it.
- A wider trace lowers only the conductor term. It is a real fix only while the conductor term is the larger one.
- The margin is the budget minus the loss; a negative margin means the channel needs a better laminate, smoother copper, a shorter run or more equalisation.
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 conductor term is first-order: it puts the current on one face of the strip and treats the return path as lossless. In stripline the current splits across both faces, which halves the strip's resistance, while the planes add some back; the combination lands within roughly ±30% of a field solve. Dk and Df are taken as constant with frequency, and roughness uses the Hammerstad–Bekkadal multiplier on the RMS roughness.
How it is checked. Zero Df gives zero dielectric loss; dielectric loss is linear in f and in length; smooth-copper conductor loss is exactly √f; and the conductor term agrees with the published rule of thumb 36·√f / (w·Z0) dB/in to within 5%.
Sources. D. M. Pozar, Microwave Engineering, 4th ed., Wiley, 2012, §2.7; E. Bogatin, Signal and Power Integrity — Simplified, 3rd ed., Pearson, 2018, ch. 9.
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 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.