Do I Need Isolation or Differential Measurement?#
Most bench instruments — scopes, DMMs in some modes, signal generators — have their ground terminal connected to earth ground through the power cord. If the measurement point isn’t at earth potential, connecting a ground-referenced instrument either shorts something to earth, gives a meaningless reading, or both.
The Ground Problem#
Most oscilloscopes connect the ground clip (the outer shell of BNC, the probe ground clip) directly to earth ground through the three-prong power cord. The ground clip is not a floating reference that can go anywhere — it’s hard-wired to the building’s earth. Connecting it to any node that isn’t already at earth potential creates a short circuit through the scope’s ground path. This can destroy the circuit, destroy the scope, blow fuses, or in mains-connected circuits, create a lethal shock hazard.
Scenarios Where This Fails#
| Scenario | What happens with scope ground clipped to it |
|---|---|
| Half-bridge midpoint (motor drive, class-D amp) | Ground clip shorts the low-side switch through earth — smoke and/or dead FETs |
| Primary side of mains SMPS | Ground clip connects mains live or neutral to earth — massive fault current, tripped breaker, possible fire |
| Bridge rectifier output (no isolation transformer) | Ground clip shorts a diode — blows the diode or the scope’s ground path |
| Floating battery in series stack | Ground clip forces one terminal to earth — disrupts the circuit, potentially dangerous in high-voltage stacks |
| Across a current-sense resistor not at ground | Ground clip shorts out the sense resistor — removes current sensing, changes circuit behavior |
Solutions#
Three approaches exist for measuring voltage between nodes that aren’t at earth potential.
Differential Probes#
A differential probe has two input pins (+ and −) and no ground clip connection to the circuit. It amplifies the difference between the two inputs and rejects the common-mode voltage (the voltage both inputs share relative to earth). The output connects to a single-ended scope input.
Differential probes allow true voltage measurement between floating nodes without disturbing the circuit, and safe measurement of nodes at high common-mode voltage (e.g., high-side gate drive signals). Common-mode ratings are typically 600V–1000V peak.
Isolation Transformers#
An isolation transformer breaks the galvanic connection between the DUT’s mains power and earth ground. With the DUT floating, its internal nodes can be at any potential relative to earth — connecting a ground-referenced scope probe no longer creates a short.
The scope ground clip connects to the DUT’s local ground (circuit common), and signals are probed as usual. This approach is particularly useful for primary-side SMPS measurements where real scope bandwidth and multiple channels are needed.
CH1 − CH2 Math Subtraction#
With no differential probe available, two scope channels and math subtraction provide a rough approximation. CH1 connects to the “+” node, CH2 to the “−” node, and both ground clips connect to a common earth-ground point on the circuit (not to either measurement node). The math trace (CH1 − CH2) shows the differential voltage.
This technique works for low-frequency, low-common-mode situations — such as measuring across a sense resistor near ground — but has significant limitations compared to true differential measurement.
Tips#
- Set the scope’s probe attenuation to match the differential probe’s actual ratio
- Verify common-mode voltage is within the differential probe’s rating before connecting
- Use matched probes (same attenuation, delay, bandwidth) for CH1 − CH2 math measurements
- Size the isolation transformer for the DUT’s power draw and voltage — undersized transformers saturate and stop isolating
- When using math subtraction, connect both ground clips to the same earth-ground point, never to the measurement nodes
Caveats#
- Scope ground clips are hard-wired to earth through the power cord — not a floating reference
- Differential probes have finite CMRR (typically 60–80 dB); at 60 dB (1000:1), 100V of common mode leaks through as 100 mV of error, which can dominate small-signal measurements
- Differential probe bandwidth varies widely — cheap probes may be 25–50 MHz while high-end active probes reach 1 GHz+
- Isolation transformers have parasitic capacitance between windings that leaks at high frequencies, limiting common-mode rejection
- Everything downstream of an isolation transformer floats — including the DUT’s chassis; touching the DUT and earth simultaneously creates a shock hazard
- Never float the scope by lifting its earth prong — the scope chassis becomes live at whatever potential the circuit puts on it
- CH1 − CH2 math has poor CMRR (20–30 dB at DC, degrading rapidly with frequency) — 100x worse than a real differential probe
- Both channels receive full common-mode voltage with math subtraction; if common-mode exceeds the scope’s input range, clipping occurs before subtraction
- Probe cable length differences create phase errors at high frequency that appear as false differential signal
In Practice#
- Smoke or dead FETs when probing a bridge circuit indicates the ground clip shorted a switch node to earth
- Tripped breaker when probing mains-connected equipment indicates the ground clip connected live or neutral to earth
- Unexpected DC offset on a small differential signal suggests CMRR leakage from high common-mode voltage — check differential probe specifications
- Math trace that looks phase-shifted or distorted at high frequency indicates probe mismatch or cable length difference — this technique is only trustworthy at low frequencies
- A differential probe reading that changes when the probe orientation is reversed indicates a connection or calibration issue
- Isolation transformer that runs hot or hums loudly is saturating — the DUT draws more power than the transformer can supply while maintaining isolation