What is the formula for bandwidth from rise time?
BW = 0.35 ÷ tr for a 10 to 90 % rise time, so a 1 ns edge corresponds to 350 MHz. For a 20 to 80 % rise time use 0.22 ÷ tr. Both describe a single-pole or roughly Gaussian response.
A faster edge uses more of the spectrum, and this rise time to bandwidth calculator shows how much. Convert in either direction, find the knee frequency, then see how much oscilloscope bandwidth it takes to measure the edge and how long a trace can get before it needs termination.
160 ps/in is Philips AN246’s nominal 50 Ω stripline. Multiply ps/mm by 25.4.
Optional: replaces the delay above with √εr ÷ c.
Optional: compares the edge with the 5 × clock scope rule.
Bandwidth (−3 dB)
350 MHz
0.35 ÷ rise time
0.35 assumes a single-pole or roughly Gaussian response. A maximally flat response is closer to 0.4 ÷ bandwidth (Keysight).
Critical length under each published rule
| Rule | Critical length | Meaning |
|---|---|---|
| tr ÷ 2 | 79.4 mm (3.13 in) | Round-trip delay equals the rise time (Philips AN246). The loosest common limit; Johnson warns of severe ringing on unterminated nets at this length. |
| tr ÷ 3 | 52.9 mm (2.08 in) | Upper end of the range in use (Johnson). |
| tr ÷ 6 | 26.5 mm (1.04 in) | Johnson’s rule of thumb for simulating a net. |
| tr ÷ 10 | 15.9 mm (0.625 in) | Lower end of the range: a stiff driver and a large capacitive load (Johnson). |
Use BW = 0.35 ÷ tr for a 10 to 90 % rise time. A 1 ns edge corresponds to 350 MHz. The knee frequency, 0.5 ÷ tr, is 500 MHz, and that is the figure to use when choosing a scope. On a 160 ps/in line the edge is 159 mm (6.25 in) long, and Howard Johnson’s rule of thumb for simulating a net is anything over one sixth of that, 26.5 mm (1.04 in).
Rise time and bandwidth are two views of the same low-pass response. The knee frequency describes the signal rather than the system, and the critical length compares the edge with the delay of the line.
These assume a single-pole or roughly Gaussian response. Keysight puts a maximally flat response at about 0.4 ÷ BW. The knee frequency formula comes from Johnson and Graham:
The edge occupies a length of line set by the delay per unit length. Critical length is a fraction of that:
BW = 0.35 ÷ 1 ns = 350 MHz and fknee = 0.5 ÷ 1 ns = 500 MHz. The edge is 159 mm (6.25 in) long. Johnson’s one-sixth rule gives 26.5 mm (1.04 in), and AN246’s round-trip rule gives 79.4 mm (3.13 in).
fknee = 0.5 ÷ 500 ps = 1 GHz. For 20 % timing error a 1 GHz scope is enough. For 3 %, a Gaussian scope needs 1.9 × 1 GHz = 1.9 GHz, and a maximally flat one needs 1.4 GHz.
A lightly loaded ABT output has a nominal 0.9 ns rise time, about 389 MHz. On a 160 ps/in, 50 Ω stripline, tr ÷ (2 × tpd) = 71.4 mm (2.81 in), which AN246 rounds to 2.8 in. One sixth of the edge length is shorter, 23.8 mm (0.938 in).
10 to 90 % rise times on a line with 160 ps/in of delay.
| Rise time | Bandwidth | Knee | Edge length | tr ÷ 2 | tr ÷ 6 |
|---|---|---|---|---|---|
| 50 ps | 7 GHz | 10 GHz | 7.94 mm (0.313 in) | 3.97 mm (0.156 in) | 1.32 mm (0.0521 in) |
| 100 ps | 3.5 GHz | 5 GHz | 15.9 mm (0.625 in) | 7.94 mm (0.313 in) | 2.65 mm (0.104 in) |
| 200 ps | 1.75 GHz | 2.5 GHz | 31.8 mm (1.25 in) | 15.9 mm (0.625 in) | 5.29 mm (0.208 in) |
| 500 ps | 700 MHz | 1 GHz | 79.4 mm (3.13 in) | 39.7 mm (1.56 in) | 13.2 mm (0.521 in) |
| 1 ns | 350 MHz | 500 MHz | 159 mm (6.25 in) | 79.4 mm (3.13 in) | 26.5 mm (1.04 in) |
| 2 ns | 175 MHz | 250 MHz | 318 mm (12.5 in) | 159 mm (6.25 in) | 52.9 mm (2.08 in) |
| 5 ns | 70 MHz | 100 MHz | 794 mm (31.3 in) | 397 mm (15.6 in) | 132 mm (5.21 in) |
| 10 ns | 35 MHz | 50 MHz | 1.59 m (62.5 in) | 794 mm (31.3 in) | 265 mm (10.4 in) |
| 20 ns | 17.5 MHz | 25 MHz | 3.18 m (125 in) | 1.59 m (62.5 in) | 529 mm (20.8 in) |
Keysight’s multipliers on the knee frequency, with the result for a 1 ns edge (knee 500 MHz).
| Rise-time error | Gaussian scope | Maximally flat scope | Gaussian, 1 ns edge | Maximally flat, 1 ns edge |
|---|---|---|---|---|
| 20 % | 1.0 × fknee | 1.0 × fknee | 500 MHz | 500 MHz |
| 10 % | 1.3 × fknee | 1.2 × fknee | 650 MHz | 600 MHz |
| 3 % | 1.9 × fknee | 1.4 × fknee | 950 MHz | 700 MHz |
Use the fastest edge your driver can produce, not the one its typical load gives. AN246 works from the faster of the rise and fall times, and a part from a newer process can be much faster than its data sheet’s maximum suggests. Critical length is not a hard edge: the published conventions span a factor of five, and the right one depends on the driver impedance, the load and whether you can wait for ringing to settle. A stiff driver into a large capacitive load is the worst case, where Johnson reports termination being needed on lines as short as a tenth of the edge length, and he warns that a line delay of half the rise time can already give severe overshoot and ringing on an unterminated net. Microstrip has part of its field in air, so its effective εr is below the laminate value and it runs faster than stripline. When in doubt, simulate the net with the driver and receiver models you intend to use.
BW = 0.35 ÷ tr for a 10 to 90 % rise time, so a 1 ns edge corresponds to 350 MHz. For a 20 to 80 % rise time use 0.22 ÷ tr. Both describe a single-pole or roughly Gaussian response.
Divide the constant by the bandwidth: tr = 0.35 ÷ BW for a 10 to 90 % rise time, or 0.22 ÷ BW for 20 to 80 %. A 100 MHz bandwidth gives 3.5 ns. This assumes a single-pole or roughly Gaussian response.
A 10 to 90 % edge has a bandwidth of 0.35 ÷ tr and a knee frequency of 0.5 ÷ tr, so a 1 ns edge has 350 MHz of bandwidth and a 500 MHz knee.
A single-pole step response takes τ × ln 9 to go from 10 to 90 %, and its bandwidth is 1 ÷ (2πτ). Multiplying them gives ln 9 ÷ 2π = 0.3497, rounded to 0.35. Keysight gives about 0.4 for scopes with a maximally flat response.
The frequency above which the spectrum of an edge no longer has much effect on its shape: 0.5 ÷ tr for a 10 to 90 % edge and 0.4 ÷ tr for a 20 to 80 % edge. The idea comes from Johnson and Graham. A 1 ns edge has a knee of 500 MHz.
Multiply the knee frequency by 1.0 for 20 % timing error, or by 1.9 (Gaussian scope) or 1.4 (maximally flat scope) for 3 %. A 500 ps edge needs 1 GHz to 1.9 GHz. Keysight says most scopes at 1 GHz and below are roughly Gaussian.
There is no single limit. Philips AN246 uses the point where round-trip delay equals the rise time (length = tr ÷ 2tpd). Howard Johnson reports that people terminate at between 1/10 and 1/3 of the edge length and simulates anything over about 1/6. For a 1 ns edge at 160 ps/in those limits run from 15.9 mm (0.625 in) to 79.4 mm (3.13 in).
No, the edge rate does. A 10 MHz clock with a 500 ps edge has the same knee frequency as a 100 MHz clock with that edge. Keysight’s 5 × clock rule for scope bandwidth holds for edges that are slow compared with the clock period, and falls short for fast edges on slow clocks.
About 3.5 ns, from 0.35 ÷ bandwidth, for a scope with a Gaussian response. That is the fastest edge the instrument can display, not the fastest it can measure accurately.
See all free electronics tools →
ProtoFlow is free AI PCB design software: describe a circuit and get a schematic with real parts, PCB layout with an autorouter, SPICE simulation and Gerber files. Native KiCad, Altium and EAGLE import and export. Free on macOS, Windows and Linux.
Free to start. No card required.