Oscilloscope Bandwidth and Rise Time: How Much Is Enough

A 25 MHz SPI clock looks clean on a 100 MHz oscilloscope, so the bench seems fine. Then someone measures the data line’s edge and gets 4 ns, compares it with the driver’s datasheet, which says 2 ns, and starts looking for a layout problem that isn’t there. The scope is the problem. Oscilloscope bandwidth and rise time determine how faithfully an instrument can show a fast edge, and the numbers are easy to misjudge. This article explains what the bandwidth and rise time specifications mean, how the scope, probe and signal combine, and how to size an instrument for the measurement you actually need.

Key Takeaways

• Bandwidth is the frequency at which a sine wave is attenuated to 70.7% of its true amplitude (−3 dB). At that frequency the amplitude error is about 30%, not a few percent.
• Rise time (10% to 90%) relates to bandwidth by t_r ≈ K / BW, with K about 0.35 for scopes with a Gaussian-like response (typical below 1 GHz) and about 0.4 to 0.45 for maximally flat responses (typical above 1 GHz).
• A measured edge combines the signal, the probe and the scope. For near-Gaussian systems, rise times add in quadrature: a 2 ns edge on a 100 MHz scope (3.5 ns) measures about 4 ns.
• Size the scope from the fastest edge, not the clock. The signal’s significant content extends to about the knee frequency, 0.5 / t_r for a 10–90% edge.
• The probe and how it’s grounded can limit the measurement as much as the scope. A long ground lead adds inductance and ringing that has nothing to do with the circuit.

The problem: the instrument shapes what you see

An oscilloscope is a low-pass system. Its front end, amplifiers and digitizer pass low frequencies and attenuate high ones, and a probe adds its own response in series. A fast edge contains high-frequency energy, and when the system removes some of it, the edge on the screen is slower, rounder or ringing differently from the edge in the circuit.

Most of the time this is harmless: a clock at a few megahertz looks the same on any modern scope. It becomes a problem when someone measures a rise time, checks timing margins between fast signals, or diagnoses signal integrity, and trusts the screen without accounting for the instrument.

What the specifications mean

Bandwidth: the −3 dB point

Tektronix’s oscilloscope primer defines bandwidth as “the frequency at which a sinusoidal input signal is attenuated to 70.7% of the signal’s true amplitude,” the −3 dB point. Keysight’s application note Evaluating Oscilloscope Bandwidths for Your Application (5989-5733EN, 2014) points out that this corresponds to about 30% amplitude error at the rated frequency. A sine wave at the scope’s bandwidth is shown with roughly 70% of its real amplitude.

Tektronix describes a “5 times rule”: choosing a scope with five times the bandwidth of the highest frequency component of interest gives less than about ±2% amplitude error. For digital signals, the Keysight note mentions the common rule of thumb of at least five times the fastest clock rate, which captures up to the fifth harmonic, and cautions that it doesn’t account for the fastest edges.

Rise time and the 0.35 constant

A scope’s own 10–90% rise time is related to its bandwidth by:

t_r ≈ K / BW

The Tektronix primer gives K between 0.35 and 0.45: about 0.35 for scopes under 1 GHz and typically 0.40 to 0.45 above. The Keysight note explains the difference: lower-bandwidth scopes typically have a Gaussian-like frequency response, which rolls off gradually, while higher-bandwidth scopes typically have a maximally flat response, which is more accurate in band and rolls off more sharply.

A 100 MHz scope with K = 0.35 therefore has a rise time of about 3.5 ns; a 500 MHz scope, about 0.7 ns; a 1 GHz scope, 0.35 to 0.45 ns depending on its response.

Reference levels and transition duration

Rise time needs a definition. IEEE Std 181-2011, Standard for Transitions, Pulses, and Related Waveforms, defines terms and algorithms for transition duration, state levels and waveform aberrations in two-state waveforms. The usual “rise time” is the transition duration between the 10% and 90% reference levels; some fields use 20% and 80%, which gives a shorter number for the same edge. Always state which levels a rise time uses.

How rise times combine

The signal on the screen is the real signal convolved with the response of the probe and the scope. NIST’s Technical Note 1470 on fast pulse transition measurements (Larson and Paulter, 2005) describes the consequence: the measured transition is slower than the input’s, and for Gaussian waveforms the input transition duration can be recovered as the square root of the difference of the squares of the measured and system durations. The corresponding forward estimate is:

t_measured ≈ √(t_signal² + t_probe² + t_scope²)

This quadrature relationship is exact only for Gaussian responses and is an approximation otherwise. Keysight’s note Bandwidth and Rise Time Requirements for Making Accurate Oscilloscope Measurements (5991-0662EN) shows, by simulation with a modern scope response, cases where the error is smaller than the quadrature rule predicts; for a signal with little overshoot, it found that a signal rise time of about 1.56 times the scope’s rise time gave under 3% error for that response. The quadrature rule is a reasonable conservative estimate for Gaussian-like systems; for a specific scope, use the manufacturer’s guidance.

Sizing a scope, worked through

Illustrative case: a driver with a 2 ns 10–90% edge, measured with a 100 MHz scope (3.5 ns) and a passive probe whose own rise time is small enough to ignore.

• Quadrature estimate: √(2² + 3.5²) ≈ 4.0 ns. The measurement is twice the real value.

How fast must the scope be for a measurement within about 3%?

• √(1 + (t_scope / t_signal)²) ≤ 1.03 means t_scope / t_signal ≤ about 0.25.
• For a 2 ns edge, t_scope ≤ about 0.5 ns, which with K = 0.35 is a bandwidth of about 700 MHz.

The Keysight bandwidth note approaches the same question through the knee frequency, f_knee = 0.5 / t_r for a 10–90% edge: 250 MHz for a 2 ns edge. It recommends a bandwidth of 1.0 × f_knee for about 20% timing accuracy, 1.3 × (Gaussian) or 1.2 × (maximally flat) for 10%, and 1.9 × (Gaussian) or 1.4 × (maximally flat) for 3%. For a 3% measurement with a Gaussian-response scope, that’s about 475 MHz. The two methods use different assumptions and give different answers, but both say the same thing in practice: a 100 MHz scope can’t measure a 2 ns edge, and somewhere around 500 MHz to 1 GHz can.

Note that the clock rate never entered the calculation. A 25 MHz SPI bus with 2 ns edges needs the same scope as a 200 MHz bus with 2 ns edges, if the edges are what you need to see.

Sample rate and probes

Bandwidth isn’t the only limit. The Tektronix primer recommends a sample rate of at least 2.5 times the highest frequency component for sin(x)/x interpolation and at least 10 times for linear interpolation. Most modern scopes meet this at full bandwidth on one or two channels, but sample rate can drop when more channels are on or the record is long.

The probe is part of the measurement system. A passive probe has its own bandwidth and loads the circuit with its capacitance, and its ground lead forms an inductive loop that, with the probe’s capacitance, rings at a frequency that can fall within the scope’s bandwidth. A short ground spring instead of a long clip lead often removes “ringing” that was never in the circuit.

Scope bandwidth vs what it can measure

Scope bandwidth	Scope rise time (K = 0.35)	Fastest edge measurable within ~3% (quadrature, Gaussian)	Typical use
50 MHz	7 ns	≈ 28 ns	Low-speed analog, power, audio, slow logic
100 MHz	3.5 ns	≈ 14 ns	General embedded, I²C, UART, slow SPI edges
200 MHz	1.75 ns	≈ 7 ns	Faster MCU I/O, most SPI edges
500 MHz	0.7 ns	≈ 2.8 ns	Fast logic edges, DDR-adjacent debug, switching power edges
1 GHz	0.35–0.45 ns	≈ 1.4–1.8 ns	High-speed digital, signal integrity

What to actually do

Size from the fastest edge you need to measure

Find the rise time of the fastest edge from the driver’s datasheet or a first measurement on a fast instrument. Compute its knee frequency and apply a margin for the accuracy you need.

Correct, or at least estimate, the instrument’s contribution

Use the quadrature estimate as a sanity check on any rise-time measurement. If the measured edge is within a factor of two of the scope’s own rise time, the measurement is mostly the scope.

Treat the probe as part of the bandwidth

Check the probe’s bandwidth and rise time, use the shortest ground connection available, and consider an active or low-capacitance probe for fast or high-impedance nodes.

Check the sample rate in the configuration you use

Confirm the sample rate with all the channels and record length you actually have on, not the maximum on the datasheet.

State reference levels with every rise time

Report whether a rise time is 10–90% or 20–80%, and which instrument and probe were used.

What not to do

Don’t size the scope from the clock frequency alone. And don’t diagnose ringing until you’ve ruled out the probe’s ground lead.

How to put it to work

If you have an hour

Look up your scope’s bandwidth and response type, your probe’s bandwidth, and the fastest edge rate in your design. Compute the quadrature estimate for that edge.

If you have a sprint

Measure a known fast edge (a fast logic gate or a pulse generator with a specified transition) with your scope and probe, compare with the specification, and write down the correction for your bench.

If you want it to last the product’s lifetime

Keep a short bench note with each instrument’s rise time, probe rise times and recommended ground accessories, and include the measurement setup in every test report that quotes a rise time.

The question to ask yourself: “Is this rise time a property of my circuit, or of my scope and probe?”

Frequently asked questions

What oscilloscope bandwidth do I need?

Enough to pass the significant frequency content of the fastest edge you need to measure, plus a margin for accuracy. A common approach is to compute the knee frequency (0.5 divided by the 10–90% rise time) and use one to two times that, depending on the scope’s response and the accuracy required.

What is the relationship between bandwidth and rise time?

Rise time is about K divided by bandwidth, with K about 0.35 for scopes with a Gaussian-like response and about 0.4 to 0.45 for maximally flat responses. A 100 MHz Gaussian-response scope has a rise time of about 3.5 ns.

Why does my measured rise time look slower than the datasheet?

Because the measurement includes the scope and probe. For near-Gaussian systems, rise times add roughly in quadrature, so if the scope’s own rise time is comparable to the signal’s, the measured edge is significantly slower than the real one.

Is the 5x bandwidth rule enough?

It’s a useful starting point for amplitude accuracy on sine-like signals and for capturing up to the fifth harmonic of a clock, but it doesn’t account for edge rates. For edge and timing measurements, size from the fastest rise time.

Does the probe affect bandwidth?

Yes. The probe has its own bandwidth and rise time, loads the circuit with its capacitance, and its ground lead can add ringing. The system bandwidth is that of the probe and scope together.

Related reading: ADC resolution vs accuracy and ENOB, sampling, aliasing and filtering and explaining measurement uncertainty.

Sources

1. IEEE (2011). IEEE Standard for Transitions, Pulses, and Related Waveforms. IEEE Std 181-2011. https://reference.globalspec.com/standard/11002981/181-2011
2. Larson, D. R., & Paulter, N. G., Jr. (2005). NIST Measurement Services: Fast Repetitive Pulse Transition Parameters, 50 Ω. NIST Technical Note 1470. https://nvlpubs.nist.gov/nistpubs/Legacy/TN/nbstechnicalnote1470.pdf
3. Keysight Technologies (2014). Evaluating Oscilloscope Bandwidths for Your Application. Application Note 5989-5733EN. https://helpfiles.keysight.com/sp/PD1000A/Content/PD1500A%20DPT%20Tests/5989-5733EN.pdf
4. Keysight Technologies (2014). Bandwidth and Rise Time Requirements for Making Accurate Oscilloscope Measurements. Application Note 5991-0662EN. https://xdevs.com/doc/HP_Agilent_Keysight/5991-0662EN%20Bandwidth%20and%20Rise%20Time%20Requirements%20for%20Making%20Accurate%20Oscilloscope%20Measurements%20c20141030%20%5B8%5D.pdf
5. Tektronix. XYZs of Oscilloscopes, “Evaluating Oscilloscopes.” https://tek.com/en/documents/primer/evaluating-oscilloscopes