HOW-TO GUIDE
How to Measure a Differential Signal With a Two Channel Oscilloscope
By the OhmSavvy Research Team ·

Ground clips, mains and floating scopes
- Both probe ground leads on a mains-powered scope connect to mains earth. Clip them only to the circuit's true ground. Clipped to any other node, they short it to earth.
- Never float the scope by defeating its earth connection. Its chassis and every BNC shell would rise to whatever voltage the ground clips touch.
- Do not use this method on mains or high-voltage circuits. Use a differential probe rated for the voltage and measurement category.
Differential vs common-mode voltage
A differential signal travels on two lines that move in opposite directions, and the information is the voltage between them, not on either line alone. Interference picked up along a cable tends to hit both lines equally, so a receiver that looks only at the difference ignores it. Two quantities describe what is on the pair:
The two voltages
1Match the probes and channel settings
The subtraction is only as good as the match between the two paths. Use two probes of the same model, both set to 10x and both compensated on the scope's calibration output (see how to compensate an oscilloscope probe). Then make both channels identical: same probe ratio, same V/div, DC coupling on both, and the same bandwidth limit setting.
2Connect CH1 to +, CH2 to −, both grounds to ground
Touch CH1's tip to the positive line (A, D+, CANH, the hot audio pin) and CH2's tip to the negative line. Clip both ground leads to the circuit's ground, as close to the measurement point as you can. If the circuit has no ground connection to the scope at all, such as a balanced line from separately powered equipment, its common-mode voltage is unknown and may be large; check each line against ground first.
3Deskew the channels and check for zero
Before trusting the result, touch both probe tips to the same signal, such as the calibration output or one line of the circuit, and turn on CH1 − CH2. In a perfect match the difference is a flat line. Spikes at each edge mean the two paths have different delays: adjust the channel deskew or delay setting, found in the channel menu on many scopes, until they shrink. Any remaining offset or scaled copy of the signal is gain mismatch, which sets your error floor.
4Display CH1 − CH2
With both probes on their lines, turn on the math subtraction. The two input traces mirror each other; the math trace shows the full differential swing, with any shared noise or offset removed. Measure its amplitude and timing with cursors or automatic measurements as you would any channel.
Common-mode limits and CMRR: why the method breaks down
Three things limit a two-channel differential measurement:
- Each channel must hold the whole voltage. If the lines sit at 12 V with a 50 mV difference, both channels need a V/div coarse enough to show 12 V without clipping, so the small difference is spread across very few ADC levels.
- Gain mismatch leaks common-mode. If one channel reads 1% high, 1% of the common-mode voltage appears in the result as a false signal.
- Matching gets worse with frequency. Small differences in probe capacitance and cable delay matter more as edges get faster, so common-mode rejection falls at high frequencies.
A dedicated differential probe subtracts in its own amplifier before the signal reaches the scope, with a specified common-mode rejection ratio (CMRR) and voltage range. For high-speed pairs, also keep the scope's bandwidth in mind; see what oscilloscope bandwidth means.
Never float the scope or clip ground to a live node
The two-channel method exists precisely so that neither ground clip goes anywhere but ground. The tempting shortcut, putting one probe's ground on the negative line, shorts that line to earth through the scope. On a mains-derived circuit that can mean a flash, burned probe leads and live metalwork. Removing the scope's earth pin to stop the short is worse: the whole instrument then floats at circuit voltage and anyone touching it completes the path. Each probe also still sees its line's full voltage to earth, so this is no way to measure mains either.
Examples: RS-485, CAN, balanced audio and shunt resistors
| Signal | Lines | Typical levels | Notes |
|---|---|---|---|
| RS-485 | A and B | At least 1.5 V differential from a driver into a standard load; receivers switch at ±200 mV | Common-mode range −7 V to +12 V by the standard, so check each line first |
| CAN | CANH and CANL | Recessive: both near 2.5 V, about 0 V differential. Dominant: about 2 V differential | Low common-mode voltage suits the two-channel method well |
| Balanced audio | Hot and cold (XLR pins 2 and 3) | Equal and opposite signals around ground | Pin 1 is the shield; use it as the ground reference |
| High-side current shunt | Each end of the shunt | Millivolts across the shunt, sitting at the supply voltage | Fine on low rails; on higher rails resolution and mismatch dominate |
A low-side shunt, with one end at ground, needs no subtraction: a single probe across it reads the current-sense voltage directly.
When you need a real differential probe
- Anything connected to mains, or a switch node in an off-line power supply.
- Common-mode voltage large compared with the differential signal, such as a high-side shunt on a 48 V rail.
- Fast differential pairs where probe matching cannot be good enough.
- Circuits with no safe ground connection to the scope.
Which oscilloscope you need for differential signals
You need at least two channels with a math subtraction function. Four channels let you watch two pairs, or one pair plus the signals that drive it. A one-channel pocket scope cannot do this method at all. Compare the options in our best beginner oscilloscopes roundup.
Multi-channel scopes in our reviews

100MHz, 2 channels, 1 GSa/s, with a built-in waveform generator in a compact unit.

50MHz, 4 channels, 1 GSa/s. Four channels at this price is rare.

100MHz, 4 channels, 1 GSa/s, with a well-regarded, intuitive interface.
As an Amazon Associate I earn from qualifying purchases. Prices and availability are shown on Amazon.
Frequently asked questions
Can I measure a differential signal by putting the ground clip on the negative line?
Not on a normal mains-powered bench scope. The ground clip is connected to mains earth through the scope, so clipping it to the negative line shorts that line to earth. At best that distorts the signal; at worst it destroys the probe, the circuit or both, and on mains-referenced equipment it is dangerous. Use two channels and math, or a differential probe.
Is a pseudo-differential measurement accurate?
For signals with small common-mode voltage, such as RS-485 or CAN on a bench, it is usually good enough to judge levels and timing. Accuracy drops as the common-mode voltage rises, because any gain mismatch between the channels leaks common-mode into the result, and at higher frequencies, where small probe differences matter more. Check by probing one point with both probes and confirming the math trace stays near zero.
What is CMRR on an oscilloscope?
Common-mode rejection ratio describes how well a differential measurement ignores the voltage shared by both inputs. A high CMRR means a large common-mode voltage produces only a tiny error in the result. Dedicated differential probes specify it, usually falling at higher frequencies. Two-channel subtraction has no designed-in CMRR; it depends entirely on how well the two channels and probes match.
Do I need two identical probes for differential measurements?
Ideally, yes. Two probes of the same model and length have similar capacitance and delay, so they track each other more closely across frequency. Mismatched probes give different edge shapes and timing, which shows up as spikes on the math trace at every edge. Compensate both on the same scope and deskew the channels afterwards to reduce the difference.
Can a one-channel scope measure a differential signal?
Only with a differential probe or an isolated, battery-powered scope used within its ratings. A single ground-referenced channel can only measure each line against ground, one at a time, which tells you levels but not the true difference at each instant. For two-channel subtraction you need at least two channels with matching settings.

