TDR Painter

Paint an impedance profile, drop in parasitics, and watch return loss, insertion loss and ILD respond. Each cell of the profile is a short lossy line, and the S-parameters come from cascading them.

TDR Canvas the Z(x) you paint and the TDR from S11

Brush
60 mil
Drop in
Drag chips along the strip under the plot
Painted profile Z(x) TDR an instrument would show (from S11)

Return loss dB(S11), dB(S22)

S11 (Tx side)S22 (Rx side, dashed)

Insertion loss dB(S21) and its fit

S21Fit a0 + a1√f + a2·f + a4·f²

Where the power goes |S21|² + |S11|² + absorbed = 1

TransmittedReflectedAbsorbed in line and R

Insertion loss deviation loss minus its fit

ILD (extra loss points up)
How the model works
  • The painted profile is 500 equal cells. Each cell is a uniform line of its painted impedance; lumped elements sit on cell boundaries. S-parameters come from the ABCD cascade, so they are exact for this structure.
  • R′ and L′ scale with Z and C′ with 1/Z, as if width alone changed the impedance. Loss per inch is the same everywhere, and only the reflections change with what you paint.
  • Dielectric: ε(f) = ε∞ + Δε·ln((f₂+jf)/(f₁+jf)) / ln(f₂/f₁), anchored to Dk and Df at f_ref. Conductor: R_dc + (1+j)·K·√f.
  • TDR: S11 times a Kaiser half-window, inverse FFT, trapezoid-rule integration, then Z = Z_Tx(1+ρ)/(1−ρ). The window sets rise time and ringing; the orange trace is what a TDR would show, the indigo line is the truth.
  • With loss on, the orange trace sits a little above the painted line and creeps upward along it: series resistance (DC and skin effect, with its internal inductance) adds up along the line and raises the impedance the TDR sees. Tick Lossless conductor and the climb disappears; the dielectric alone leaves a much smaller droop.
  • Labels on short painted bumps use the TDR area method: L = 2Z₀∫ρ dt, C = −(2/Z₀)∫ρ dt over round-trip time, with the area computed from the painted profile (exact on a lossless line). Long or large excursions show average impedance and length instead.
  • ILD is the insertion loss, −dB(S21), minus its fitted value a0 + a1√f + a2·f + a4·f² (f in GHz). The fit weights each point by |S21| and runs from 0.05 GHz to f_max, as in OIF-CEI-04.0; the coefficients are not bounded. Positive ILD means more loss than the fit.
  • Privacy: the page makes no network requests and keeps nothing, not even in browser storage. Save writes a .tdrp file to your computer (a ZIP holding project.json, with a CRC-32 that Open verifies) and Open reads it back; Export JSON writes the same project as readable text. The standalone build adds a Content-Security-Policy that blocks every outgoing connection, so the browser itself enforces this.
  • Attenuator: an ideal resistive T pad with image impedances Zin (Tx side) and Zout (Rx side). With K = 10^(dB/10), the shunt resistor is 2√(K·Zin·Zout)/(K−1) and each series resistor Z(K+1)/(K−1) minus it. When Zin ≠ Zout the loss can't go below the minimum-loss pad, 20·log₁₀(√r + √(r−1)) with r the impedance ratio (Johnson and Graham, Advanced Black Magic, Table 5.6; Reference Data for Engineers, ch. 11).
  • Nothing radiates and nothing couples. Power that is neither reflected nor transmitted is absorbed in the line, in series R or in an attenuator.
  • Experimental impedance range (gear menu): 1 mΩ is a short and 1 MΩ an open, for the ports, the profile, stubs and series R. The S-parameters stay referenced to Tx and Rx, so with Rx shorted S21 is tiny by definition; the TDR is the useful view. A painted 1 mΩ section keeps the same loss per inch as any other cell. Stubs can be open or shorted at the far end: a 1 mΩ shorted stub is a short to ground at its position, DC included, while a 1 mΩ open stub is a nanofarad capacitor that is open at DC.