HOW-TO GUIDE
How to Measure Propagation Delay Using an Oscilloscope
By the OhmSavvy Research Team ·
Safety first: both ground clips go to circuit ground
- On a mains-powered bench scope, every probe's ground clip is connected to mains earth through the power cord. For a two-probe delay measurement, clip both grounds to the same circuit ground. Clipping either one to a point that isn't at earth potential shorts that point to earth, which can damage the circuit, probe or scope, or injure you.
- Never float the scope by removing or defeating its earth connection. For circuits that aren't earth-referenced, such as anything on the mains side of a non-isolated supply, use an isolation transformer on the device under test, differential probes, or an isolated-channel scope rated for the job, and stay within the voltage and CAT ratings on your probes.
What propagation delay is
Every real circuit takes time to respond. When a gate's input switches, its output follows a few nanoseconds later; an optocoupler or a relay driver can take microseconds; a microcontroller reacting to an interrupt adds its own firmware latency. Propagation delay puts a number on that lag. Because edges are not perfectly vertical, the convention is to time it between the point where the input crosses 50% of its swing and the point where the output crosses 50% of its swing.
| Term | Meaning |
|---|---|
| tPLH | Delay to an output that goes low to high |
| tPHL | Delay to an output that goes high to low |
| tpd | Often quoted as the average: (tPLH + tPHL) ÷ 2, or the worse of the two; check the datasheet |
| Rise / fall time | How fast one edge moves (usually 10% to 90%), not a delay between signals |
| Skew | A timing offset between two channels or probes that isn't part of the circuit |
The two output directions often have different delays, so measure both edges. If you are checking a part against its datasheet, note the test conditions it lists, such as supply voltage, load capacitance and reference levels, because delay changes with all of them.
1Connect the input to CH1 and the output to CH2
Probe the input signal with CH1 and the output with CH2. Using CH1 for the input is a habit worth keeping: it becomes the trigger source and the time reference, so every reading is "how long after the input".
Matched probes
Use two probes of the same model and cable length, both on 10x, so their delays cancel.
Compensate both
A miscompensated probe rounds the edge and shifts its 50% point. Adjust each one first.
Short grounds
Long ground leads add ringing to fast edges. Use ground springs where you can.
Same ground
Both ground clips go to the circuit's ground, close to the part you are timing.
Probe ratio set
Each channel's probe setting must match its probe so the 50% levels are read correctly.
DC coupling
Keep both channels DC coupled so the logic levels sit where they really are.
If compensation is new to you, our guide to compensating an oscilloscope probe takes a couple of minutes per probe. A probe also adds capacitance to the node it touches (the figure is in the probe's specifications), which can slow a weakly driven output slightly.
2Trigger on the input edge
Set the trigger source to CH1, type Edge, and choose the slope that matches the input transition you want to time: rising for one direction, falling for the other. Put the trigger level at the middle of the input swing, so for a 0 to 5 V signal, about 2.5 V. The output then appears at a fixed position after the trigger point on every sweep. If the input only switches occasionally, use Normal or Single mode so the scope waits for the real edge. Our guide on how to trigger an oscilloscope covers level, slope and modes in detail. To compare the two signals' gain and shape rather than their timing, see how to compare input and output signals on an oscilloscope.
3Set the timebase so the delay spans several divisions
Move the horizontal position so the trigger point sits near the left of the screen, then turn the time/div down until the gap between the input and output edges covers at least one or two divisions. A delay that is a fraction of a division is hard to read accurately by eye and gives the scope fewer samples to work with. Set each channel's V/div so its signal fills a good part of its half of the screen, and use the vertical position controls to stack CH1 above CH2.
4Measure between the 50% points
Find where the input edge crosses its 50% level, then where the output edge crosses its own 50% level, and count the divisions between them. Multiply by time/div. The 50% level is measured on each signal separately: if the input swings 0 to 5 V and the output 0 to 3.3 V, the reference points are 2.5 V and 1.65 V.
Real edges have a slope, so where you measure on them matters. Timing from the very start of the input edge to the very top of the output edge would add part of both rise times to the result. The 50% convention removes most of that and makes your number comparable with datasheets that use the same definition.
5Use cursors or an automatic delay measurement
Counting divisions is the method to understand, but cursors are more precise. Turn on the time (vertical, often labeled X) cursors, put the first on the input's 50% crossing and the second on the output's 50% crossing, and read Δt. Zoom in horizontally first so each edge is spread across the screen.
Many digital scopes also offer an automatic delay measurement between two channels, under names such as Delay, Rise-Rise, Fall-Fall or Phase. Check your manual for which edges it uses, whether the reference level is 50% or adjustable, and which channel counts as the source. For an inverting part, measure from a rising input edge to a falling output edge.
Averaging helps with jittery edges: many scopes can show the mean and spread of a measurement over many captures. Repeat the measurement on the other edge, and if you are characterizing a part, at the supply voltage and load the datasheet specifies.
Probe skew and deskew
Each probe and cable adds its own delay, and signals travel along typical coax at roughly 5 ns per meter. Two identical probes mostly cancel out, but different models or cable lengths can show a fixed offset between channels that looks like circuit delay. When you are measuring nanoseconds, check for it:
01
Probe one edge
Both probe tips on the same fast signal
02
Read the offset
Δt between the two 50% points
03
Deskew or note it
Enter it in the channel menu, if offered
04
Re-check
Same edge now lines up on both
Many scopes have a per-channel deskew or delay setting that shifts one trace to remove the offset; check your manual for whether yours has one and its range. If it doesn't, subtract the skew by hand: a measured 15 ns with 2 ns of CH2 skew is a 13 ns delay. For delays of microseconds, such as optocouplers, skew of this size rarely matters.
Worked examples
The arithmetic is always divisions × time/div, then any skew correction. These examples show how the timebase scales with the kind of circuit you are timing:
| Circuit | Time/div | Divisions | Delay |
|---|---|---|---|
| Buffer, rising edge | 10 ns/div | 1.5 | 1.5 × 10 ns = 15 ns |
| Same buffer, 2 ns CH2 skew removed | 10 ns/div | 1.5 | 15 ns − 2 ns = 13 ns |
| Inverter, tPHL | 5 ns/div | 2 | 2 × 5 ns = 10 ns |
| Optocoupler output | 2 µs/div | 2.4 | 2.4 × 2 µs = 4.8 µs |
| Microcontroller pin responding to an input | 1 µs/div | 3.2 | 3.2 × 1 µs = 3.2 µs |
| Average of tPLH 12 ns and tPHL 8 ns | - | - | (12 + 8) ÷ 2 = 10 ns |
The microcontroller case isn't a gate delay but is measured the same way, and it is a useful check on interrupt latency. For a slow signal like that, triggering in Single mode and reading cursors on one capture is often clearer than a live display.
Bandwidth and resolution limits
A scope has its own rise time, roughly 0.35 ÷ bandwidth. Edges faster than that are displayed slower than they really are. Because both channels are slowed in a similar way, a 50%-to-50% delay is affected less than a rise-time reading, but once the delay you are measuring is close to the scope's own rise time, treat the result as approximate.
Sample rate sets how finely the edge is resolved in time, and many scopes divide it between channels when several are on, so check your manual. Our explainer on what oscilloscope bandwidth means covers the 0.35 rule and the 3 to 5 times margin, and what GSa/s means covers sample rate.
Delay vs phase
Which scope you need for delay measurements
The one hard requirement is two channels. A single-channel pocket scope like the FNIRSI DSO152 (1 channel, 200kHz) can't show input and output together. For nanosecond logic delays, the more bandwidth the better; for optocouplers and firmware latency, any of the bench scopes below is plenty. Compare them in our best beginner oscilloscopes roundup, or start with how to use an oscilloscope if the controls are new.
Two-channel-and-up scopes in our reviews
50MHz, 4 channels, 1 GSa/s. The default beginner oscilloscope recommendation for years.
100MHz, 4 channels, 1 GSa/s, with more bandwidth and a better UI than the Rigol.
100MHz, 2 channels, 1 GSa/s, with a built-in waveform generator.
As an Amazon Associate I earn from qualifying purchases. Prices and availability are shown on Amazon.
Frequently asked questions
Can you measure propagation delay with a single-channel oscilloscope?
Not directly. You need to see the input and output edges on the same time axis, which takes two channels. A workaround is to trigger the single channel from an external trigger input driven by the input signal, if the scope has one, then probe the output, but trigger-path delay adds an error. A two-channel scope is the practical choice.
What is the difference between propagation delay and rise time?
Rise time describes how fast one edge moves, usually from 10% to 90% of its swing. Propagation delay is the time between two different signals, measured from the input edge to the output edge, usually at their 50% points. A slow output edge increases the measured delay, but the two are separate measurements with separate definitions.
Why is my measured delay longer than the datasheet value?
Datasheet figures are measured under stated conditions: supply voltage, temperature, load capacitance and reference levels. Your probe adds capacitance to the output, your supply may differ, and your edges may be slower. Uncorrected probe skew and a scope that is too slow for the edges can also add time. Match the test conditions before comparing.
Should I measure propagation delay at 50% or at the logic threshold?
Use whatever reference level the datasheet you are comparing against uses. Many specify 50% of the supply swing, while some logic families specify a fixed voltage. For your own comparisons, 50% of each signal's swing is the usual default, and the important thing is to use the same rule every time.
Do longer probe cables change the delay measurement?
Yes, if the two probes differ. A signal takes time to travel along a cable, roughly 5 ns per meter in typical coax, so two probes with different cable lengths or designs show a fixed timing offset. Use matched probes and measure the skew by probing one edge with both, then correct for it.
