100nF

Termination resistor calculator: series, parallel, Thevenin and AC

A driver of 45 Ω on a 50 Ω line needs about 6.80 Ω in series; a 50 Ω line terminated to 1.3 V from 3.3 V needs 130 Ω up and 82.0 Ω down. This calculator sizes series, parallel, Thevenin and AC (RC) termination for a single-ended line from TI's application notes, picks the stocked values, shows the reflection that is left on each edge, and draws the receiver's step response with and without it.

3.3 Vunterminatedseries termination01T3T5T7Ttime, one-way delays T
Fig 1 — the receiver after a 3.3 V step, by the SNLA027 lattice: series termination against the same line unterminated, which overshoots by 7 % and rings.
Series resistor R_S = Z0 − Z_out
6.50 Ω exact; fit 6.80 Ω (E24)
Source reflection, rising edge
+0.018 (Z_out 45.0 Ω + R_S)
Source reflection, falling edge
-0.012 (Z_out 42.0 Ω + R_S)
Where it goes
at the driver pins; the receiver end stays open

How this is calculated

Standard: TI SNLA027B (AN-807), Equations 14 and 16; TI SCAA080A §2 and Table 1; TI SNAA377 §3.1 Equations 1–4 and §4.1.1; TI SNLA034B (AN-903) §4 and §5 Equation 1

ρL=RL−R0R0+RL,ρS=RS−R0RS+R0\rho_L = \frac{R_L - R_0}{R_0 + R_L}, \qquad \rho_S = \frac{R_S - R_0}{R_S + R_0}
SNLA027 Equations 14 and 16 (p. 6), the voltage reflection coefficients of the load and the source. For series termination R_S is the driver impedance plus the series resistor; for the far-end schemes R_L is the termination.
RS=Z0−12(ZPU+ZPD)R_S = Z_0 - \tfrac{1}{2}\left(Z_{PU} + Z_{PD}\right)
SCAA080 §2: "The sum of the driver impedance Zout and RS should be close to the transmission line impedance Z0"; SNAA377 §4.1.1: Z_O = R_O + R_S. The mean of the pull-up and pull-down impedances is this page's choice; SCAA080 asks for "a good matching ... for PU and PD". Floored at zero.
RT=Z0 (1+k),0≤k≤0.1,P=VOH2RT DR_T = Z_0\,(1 + k), \quad 0 \le k \le 0.1, \qquad P = \frac{V_{OH}^2}{R_T}\, D
SNLA034 §4: R_T "is typically chosen to be equal to ZO. When over-termination is used RT is typically chosen to be up to 10% larger than ZO." The power, with D the fraction of time high, is derived.
Z0=RTOP RBOTTOMRTOP+RBOTTOM,VTERM=VCCRBOTTOMRTOP+RBOTTOM  ⇒  RTOP=Z0VCCVTERM,  RBOTTOM=Z0VCCVCC−VTERMZ_0 = \frac{R_{TOP}\, R_{BOTTOM}}{R_{TOP} + R_{BOTTOM}}, \quad V_{TERM} = V_{CC} \frac{R_{BOTTOM}}{R_{TOP} + R_{BOTTOM}} \;\Rightarrow\; R_{TOP} = \frac{Z_0 V_{CC}}{V_{TERM}}, \; R_{BOTTOM} = \frac{Z_0 V_{CC}}{V_{CC} - V_{TERM}}
SNAA377 Equations 1 and 2 (§3.1), solved for the two resistors. The note's Equations 3 and 4 check 130 Ω and 82 Ω from 3.3 V as "≈ 50Ω" and "≈ 1.3V". The standing power V_CC²/(R_TOP + R_BOTTOM) is derived.
CT≤2 tdZ0,RTCT≤0.1 TUIC_T \le \frac{2\, t_d}{Z_0}, \qquad R_T C_T \le 0.1\, T_{UI}
SNLA034 Equation 1 (§5), with t_d the one-way delay: "CT ≤ (Cable round trip delay) / ZO", and "the resulting RC time constant should be less than or equal to 10% of the unit interval (TUI)".
vrx(n)=∑k=0nVZ0ZS+Z0 (1+ρL) (ρLρS)kv_{rx}(n) = \sum_{k=0}^{n} \frac{V Z_0}{Z_S + Z_0}\,(1 + \rho_L)\,(\rho_L \rho_S)^k
The lattice of SNLA027 summed at the receiver: the n-th arrival, at (2n + 1) one-way delays. It converges to the DC divider V·R_L/(Z_S + R_L). Used for the figure.

Assumptions

What it computes

A line needs terminating when its delay is long against the edge — the critical-length calculator decides that — and then one of four arrangements absorbs the wave that would otherwise bounce between the ends. TI's AN-807 (SNLA027) gives the quantity all of them work on, the voltage reflection coefficient, as Equation 14 for the load, ρL = (RL − R0)/(R0 + RL), and Equation 16 for the source. A matched end has ρ = 0 and sends nothing back. The calculator sizes the resistor for the scheme you choose, reports what is still reflected with the value you can buy, and draws the receiver's step response by the same note's lattice method.

Series termination puts one resistor at the driver. SCAA080: "The sum of the driver impedance Zout and RS should be close to the transmission line impedance Z0", and SNAA377 writes the same rule as ZO = RO + RS. The wave leaves at half amplitude, doubles at the open receiver, and is absorbed when its reflection returns to a matched source. Drivers rarely pull up and down with the same impedance, so the calculator uses the mean of the two and reports each edge's residual reflection; SCAA080 asks for "a good matching to the transmission line impedance for PU and PD" without saying how.

The far-end schemes match the receiver instead. Parallel termination is a resistor equal to Z0 to ground, or, per SNLA034, "up to 10% larger". Thevenin termination splits it into a pair whose parallel combination is Z0 and whose ratio sets a DC level, SNAA377's Equations 1 and 2. AC termination puts a capacitor in series with the resistor so no DC flows, sized by SNLA034's Equation 1, CT ≤ round-trip delay ÷ Z0.

Worked example: a clock buffer on a 50 Ω trace

The calculator's defaults: the CDCx706/x906 at 3.3 V on a 50 Ω line, with the pull-up and pull-down impedances SCAA080 measures in Table 1.

Driver, CDCx706 at 3.3 V   pull-up 45 Ω, pull-down 42 Ω (SCAA080 Table 1)
R_S = Z0 − mean(Z_out)    50 − 43.5                          = 6.50 Ω
Nearest E24                                                   = 6.80 Ω
ρS, rising    (45 + 6.8 − 50)/(45 + 6.8 + 50)                 = +0.018
ρS, falling   (42 + 6.8 − 50)/(42 + 6.8 + 50)                 = -0.012
Table 1's 5 Ω, rising edge  (45 + 5 − 50)/(45 + 5 + 50)       = +0.000

Both edges come within 1.8 % of a match with the stocked 6.80 Ω. SCAA080's own proposal for this row is 5 Ω, which leaves +0.000 on the rising edge: a difference of no practical weight on this line.

SCAA080's table against Z0 − Zout

The note proposes a series resistor for four line impedances at two supplies. Set beside the rule the same note states, and beside the worst residual reflection each choice leaves:

Z0, supplyPU / PDTable 1 RSWorst |ρ|, Table 1Z0 − meanWorst |ρ|, calculator
35 Ω, 3.3 V49 / 45 Ω0 Ω0.1670 (driver above Z0)0.167
35 Ω, 2.5 V62 / 67 Ω0 Ω0.3140 (driver above Z0)0.314
50 Ω, 3.3 V45 / 42 Ω5 Ω0.0316.50 Ω0.018
50 Ω, 2.5 V57 / 58 Ω0 Ω0.0740 (driver above Z0)0.074
75 Ω, 3.3 V43 / 38 Ω10 Ω0.22034.5 Ω0.026
75 Ω, 2.5 V51 / 49 Ω0 Ω0.21025.0 Ω0.014
100 Ω, 3.3 V41 / 35 Ω12 Ω0.36162.0 Ω0.015
100 Ω, 2.5 V49 / 44 Ω5 Ω0.34253.5 Ω0.024

Where Z0 − Zout is a few ohms or less, the table and the rule agree. On the 75 Ω and 100 Ω lines they part: for 100 Ω at 3.3 V the rule asks for 62.0 Ω and the table proposes 12 Ω, leaving a reflection of 0.36. The note calls Table 1 "a basic value" and recommends signal-integrity simulation "with IBIS models ... to optimize the value of the series resistor by including the receiver and board parasitics", which may be where its smaller values come from; the page does not reproduce that simulation and follows the stated rule. Where the driver alone exceeds Z0, as on the 35 Ω line, no series resistor helps, and both give 0 Ω.

Thevenin pair: SNAA377's 130 Ω and 82 Ω

SNAA377 terminates a 50 Ω LVPECL line at VCC − 2 V from 3.3 V. Equations 1 and 2 solved exactly give 127 Ω and 82.5 Ω; the nearest E24 values are the note's own 130 Ω and 82 Ω, which its Equations 3 and 4 check as "≈ 50Ω" and "≈ 1.3V". The calculator gives 50.3 Ω and 1.28 V, and one more number the note leaves out: the pair draws 51.4 mW from the supply whether the line is switching or not.

AC termination: SNLA034's cable

SNLA034's example is 100 ft of 100 Ω cable at 1.7 ns/ft. "CT ≤ (100 ft × 2 × 1.7 ns/ft)/100Ω or ≤ 3,400 pF", and with the RC time constant held to 10 % of the unit interval "the maximum switching rate therefore should be less than 300 kHz". The calculator gives 3.40 nF, an RC of 340 ns and 294 kHz. It is a termination for slow control lines, which is how the note itself classes it.

What an unterminated line does

The figure's grey staircase is the same line with no termination. A strong driver makes it worse: 20 Ω into an open 50 Ω line launches 2.36 V, which doubles to 4.71 V at the receiver, then falls to 2.69 V one round trip later as the driver's mismatch inverts the returning wave, and only settles towards 3.3 V after several round trips. Change the driver impedances to see it.

Where the model stops

Further reading