Skip to content

Rise Time to Bandwidth Calculator

Convert digital rise time to signal bandwidth. Get the knee frequency, 3 dB bandwidth, IPC-2251 spectral bandwidth, and the highest significant harmonic of your clock.

Single-pole ln 9/(2π·tᵣ) bandwidth · Johnson–Graham knee 0.5/tᵣAPI slug rise-time-bandwidth

Sign in to saveSave

Linked values

Type in any field — the others follow.
ExampleDecimals
Optional — used to report how many harmonics fall below the knee frequency; 0 leaves it out

0 MHz or more

ps

1 ps to 100 ns

ps

1 ps to 100 ns

MHz

3.5 MHz to 350 GHz

GHz

5 MHz to 500 GHz

MHz

3.18 MHz to 318 GHz

×f_clk

0.05 to 5 000 ×f_clk

Common
RISE TIME10 ps20 ps50 ps100 ps200 ps500 ps1 ns2 ns5 ns10 ns500.000 ps3 DB BANDWIDTH20 GHz10 GHz5 GHz2 GHz1 GHz500 MHz200 MHz100 MHz50 MHz699.398 MHzKNEE FREQUENCY50 GHz20 GHz10 GHz5 GHz2 GHz1 GHz500 MHz200 MHz100 MHz50 MHz1.000 GHzHIGHEST SIGNIFICANT HARMONIC, ×F_CLK300100301053110.000

Reference values

Select a row to load it
Rise TimeRise Time (10–90% equivalent)3 dB BandwidthKnee FrequencySpectral Bandwidth (1/πt_r)Highest Significant Harmonic
ps20.000ps17.485GHz25.000GHz15.915GHz250.000×f_clk
ps50.000ps6.994GHz10.000GHz6.366GHz100.000×f_clk
ps100.000ps3.497GHz5.000GHz3.183GHz50.000×f_clk
ps200.000ps1.748GHz2.500GHz1.592GHz25.000×f_clk
ps500.000ps699.398MHz1.000GHz636.620MHz10.000×f_clk
ns1.000ns349.699MHz500.000MHz318.310MHz5.000×f_clk
ns2.000ns174.850MHz250.000MHz159.155MHz2.500×f_clk
ns5.000ns69.940MHz100.000MHz63.662MHz1.000×f_clk

Formula

H. Johnson & M. Graham, "High-Speed Digital Design: A Handbook of Black Magic", Prentice Hall 1993, §1.3; IPC-2251
fknee=0.5tr,f3dB=ln⁡92πtr≈0.35trf_{knee} = \frac{0.5}{t_r},\qquad f_{3dB} = \frac{\ln 9}{2\pi t_r} \approx \frac{0.35}{t_r}
SymbolQuantityYour value
trRise time, 10–90% (s)500 ps
fkneeKnee frequency (Hz)1 GHz
f3dB3 dB bandwidth (Hz)699.4 MHz
fclkClock frequency (Hz)100 MHz

Assumptions

5 items
  1. 01The edge is the step response of a single pole, so the 3 dB bandwidth is ln(9)/(2π·t_r) = 0.3497/t_r for a 10–90 % rise time; a channel with more than one pole has a somewhat different constant, which is not modelled.
  2. 02The knee frequency is 0.5/t_r (Johnson and Graham) and the spectral bandwidth 1/(π·t_r); all three frequencies use the 10–90 % rise time.
  3. 03A 20–80 % rise time is converted to 10–90 % by multiplying by ln(9)/ln(4) = 1.585, the single-pole ratio; it is the largest ratio in common use, so the least conservative for sizing an interconnect.
  4. 04The highest significant harmonic is the knee frequency divided by the clock frequency, not rounded to a whole harmonic, and a clock of 0 leaves it at 0. Warnings flag a knee below the third harmonic, a 10–90 % rise time over 40 % of the clock period and a knee above 20 GHz.
  5. 05The rise time must be above zero, from 1 ps; the loss and dispersion of the interconnect, which slow the edge further along a trace, are not included.

Worked example

Scope reading: a 300 ps 20–80 % edge on a 125 MHz clock

An oscilloscope measures a 300 ps rise time, 20–80 %, on a 125 MHz clock line. Starting from that reading, find the bandwidth the edge occupies and how many clock harmonics it carries.

StepExpressionResult
10–90 % equivalent300 ps × ln(9)/ln(4)475.5 ps
Knee frequency1 / (2 × 475.5 ps)1.052 GHz
Half-power bandwidthln(9)/(2π) / 475.5 ps735.5 MHz
Highest significant harmonic1.052 GHz / 125 MHz8.412 ×f_clk

How it works

Digital designers talk in nanoseconds; interconnect designers talk in gigahertz. The bridge between them is the rise time, and there are two different constants for crossing it — used for two different purposes, and routinely confused.

The 0.35 constant is exact

For a single-pole response, the step output crosses 10% at t=−τln⁡(0.9)t = -\tau\ln(0.9) and 90% at t=−τln⁡(0.1)t = -\tau\ln(0.1). Subtracting gives tr=τln⁡9t_r = \tau\ln 9. With the 3 dB bandwidth at f3dB=1/(2πτ)f_{3dB} = 1/(2\pi\tau), the product falls out:

f3dB tr=ln⁡92π=0.34965…f_{3dB}\, t_r = \frac{\ln 9}{2\pi} = 0.34965\ldots

So 0.35 is not a rule of thumb — it is ln⁡9/2π\ln 9 / 2\pi rounded to two figures. It answers the question what analogue bandwidth does my amplifier or scope need to reproduce this edge?

The knee frequency is a design bound

Johnson and Graham's knee frequency answers a different question: how far up in frequency does my interconnect have to keep behaving?

fknee=0.5trf_{knee} = \frac{0.5}{t_r}

Above the knee, a digital edge carries so little energy that channel behaviour there stops mattering. It is deliberately conservative, and it is the number you should use when deciding how far to trust a via model, a connector S-parameter file, or a trace loss estimate.

The two differ by roughly 43%, which is why quoting the wrong one leads to either an over-specified amplifier or an under-modelled channel.

Rise time conventions

Datasheets quote either 10–90% or 20–80% transitions, and the two are not interchangeable. For a single pole, the ratio is ln⁡9/ln⁡4=1.585\ln 9/\ln 4 = 1.585 — a 20–80% figure must be multiplied by 1.585 before feeding it into either bandwidth relation. Modern SerDes datasheets increasingly use 20–80% because it avoids the noisy tails of the waveform, so check which convention you are reading.

What sets the rise time

It is the driver, not the clock. A 25 MHz oscillator with a modern CMOS output stage can have a 500 ps edge, which puts its knee at 1 GHz — the interconnect must behave to 1 GHz even though the fundamental is 25 MHz. This is the single most common misjudgement in board design: slow clock, fast edges, and a channel modelled only to the fundamental.

Practical tips

  • Always design to the fastest edge the driver can produce, not the typical. Fast-corner silicon at light load sets the emissions.
  • Use series termination or a deliberately slew-limited driver where speed is not needed — slowing an edge is the cheapest EMC fix available.
  • Pick a scope with at least three times the signal's 3 dB bandwidth so the instrument contributes under about 5% to the measured rise time.
  • When a datasheet omits the convention, assume 10-90% for older logic families and check carefully for SerDes and DDR parts, which often use 20-80%.
  • Remember rise times add in quadrature through a cascade: total is the square root of the sum of squares of each stage.
  • Translate the knee frequency into a channel requirement early — it tells you whether an S-parameter model or a connector spec covers your band before you commit to the part.

Common mistakes

  • Designing the channel to the clock frequency instead of the knee frequency. A slow clock with fast edges still demands a wideband interconnect — this is the classic reason a 25 MHz design fails EMC.
  • Mixing the 0.35 and 0.5 constants. Use 0.35 divided by rise time for amplifier and instrument bandwidth; use 0.5 divided by rise time for how far the interconnect model must remain valid.
  • Feeding a 20-80% datasheet number into a 10-90% formula. That under-predicts rise time by 37% and over-predicts bandwidth by the same factor.
  • Forgetting that a measurement instrument adds its own rise time in quadrature. A 500 ps edge measured on a 500 ps scope reads about 707 ps.
  • Assuming the specified typical rise time is what you will get. Lightly loaded fast-corner silicon can be two or three times quicker than the datasheet typical, and that is the corner that sets your EMC result.

Frequently asked questions

5 questions
Where does the 0.35 constant come from?

It is exactly ln(9) divided by 2 pi, which is 0.34965, derived from the single-pole step response. The 10% and 90% crossings occur at fixed multiples of the time constant, giving a rise time of tau times ln(9), and with the 3 dB bandwidth at one over 2 pi tau the product is fixed. It is not empirical.

Should I use 0.35 or 0.5 divided by rise time?

Both, for different jobs. The 0.35 form gives the 3 dB bandwidth — what an amplifier, receiver, or oscilloscope needs to reproduce the edge. The 0.5 form gives the knee frequency — how far up the interconnect model has to stay valid. Design the channel to the knee and the electronics to the 3 dB figure.

How do I convert a 20-80% rise time?

Multiply by ln(9)/ln(4) = 1.585 to get the 10-90% equivalent, then apply the bandwidth relations. Going the other way, multiply by 0.631. Note that 1.585 is the single-pole value and the largest factor in circulation — a Gaussian edge gives about 1.52 and a linear ramp 1.33 — so for a real driver it slightly over-states the rise time and therefore under-states the knee frequency. Design to the shorter estimate where that margin matters.

Why does my 25 MHz clock cause EMC problems?

Because emissions follow the edge rate, not the clock rate. A 25 MHz clock with a 500 ps edge has a 1 GHz knee, so it radiates meaningfully up to 1 GHz. Slowing the edge with series resistance is usually the fastest fix.

How much scope bandwidth do I need?

At least three times the signal's 3 dB bandwidth to keep the instrument's contribution under about 5%. Rise times add in quadrature, so a scope matching the signal's rise time inflates the reading by 41%.

Shop Components

As an Amazon Associate we earn from qualifying purchases.

  • Op-Amp IC AssortmentGeneral-purpose and precision operational amplifiersAmazon →
  • Function GeneratorDDS function generator for signal and filter testingAmazon →