Why Does Connecting an Oscilloscope Probe Change the Signal Being Measured?

What the probe looks like to your circuit

Every measurement takes a little energy from the thing being measured. A multimeter on DC voltage draws a tiny current through its input resistance; an oscilloscope probe does the same, but it also has to charge and discharge a capacitance every time the signal changes. Electrically, the probe tip, cable and scope input can be modeled as a resistance R in parallel with a capacitance C, connected between your test node and the point where the ground lead lands. The node you are probing almost never has zero source impedance, so it now drives a divider made of its own source resistance and that R ∥ C load.

The values depend on the probe. A general-purpose scope input is typically 1 MΩ in parallel with roughly 10 to 20 pF. A typical passive 10x probe adds a 9 MΩ resistor in its tip, so the circuit sees about 10 MΩ, and because the tip capacitance sits in series with the cable and input capacitance, the circuit sees only about 10 to 15 pF. A 1x probe has no tip resistor, so the circuit sees the bare 1 MΩ input plus the full capacitance of the cable and input together, often 50 to 100 pF or more. These are typical figures; your probe's spec sheet or the label on its BNC end gives the real ones. How the 10x tip divider is built is explained in how to compensate an oscilloscope probe.

Resistive loading: the reading drops

At DC and low frequencies the capacitor barely matters and the probe acts as a plain resistor. The voltage you see is the true voltage multiplied by R ÷ (Rs + R). On a low-impedance node, such as a regulator output, the loss is invisible. On a high-impedance node, such as a divider made from megohm resistors, a sensor bias point or the output of a high-value RC filter, it is not.

A 1 MΩ source probed on 1x reads half its real value, which can make a healthy circuit look broken. The same node on 10x reads about 9% low. Neither is a scope fault; the scope is accurately showing the voltage that exists once the probe is attached. The same divider effect explains part of why a scope and a multimeter can disagree on the same node, which is covered in why a scope and a multimeter show different voltages.

Capacitive loading: edges slow down

As frequency rises, the capacitance takes over. A capacitor's reactance is Xc = 1 ÷ (2πfC), so it falls as frequency goes up. At 1 MHz, 15 pF is 1 ÷ (2π × 1,000,000 × 15 × 10⁻¹²) ≈ 10.6 kΩ, and 100 pF is about 1.59 kΩ. At 10 MHz those become about 1.06 kΩ and 159 Ω. Compare that with the 10 MΩ or 1 MΩ resistance, and it is clear that above a few hundred kilohertz the probe's "high impedance" is mostly a myth: a 1x probe at 10 MHz looks like a couple of hundred ohms to ground.

In the time domain, the same capacitance has to charge through the source resistance every time the signal switches, which slows the edge. The 10% to 90% rise time of a simple RC is about t ≈ 2.2 × R × C. A logic output or pull-up with 10 kΩ of effective source resistance driving a 15 pF probe gives 2.2 × 10,000 × 15 × 10⁻¹² ≈ 0.33 µs. Swap to a 100 pF 1x probe and it becomes 2.2 × 10,000 × 100 × 10⁻¹² = 2.2 µs, nearly seven times slower. The equivalent bandwidth of that loaded node, 1 ÷ (2πRC), drops from about 1.06 MHz to about 159 kHz. A slowed edge can change timing, miss a logic threshold, or make a fast signal look like a triangle.

The probe also limits bandwidth on its own. A 10x passive probe is usually rated for the scope it ships with, often 100 MHz or more, but a 1x probe or a 10x probe switched to 1x is typically good for only a few megahertz to around 10 MHz, so edges look slower even on a low-impedance node. Check the bandwidth printed for each position on your probe. Rise-time measurement itself is covered in how to measure rise time with cursors, and the scope side of the equation in what oscilloscope bandwidth means.

Ground-lead inductance: extra ringing

The ground clip completes a loop from the tip, through the circuit, back along the ground lead. That loop has inductance, commonly estimated at around 100 to 200 nH for a typical clip lead of 15 cm or so, and it forms a series LC circuit with the tip capacitance. Its resonant frequency is f = 1 ÷ (2π√(LC)). With 150 nH and 15 pF, √(150 × 10⁻⁹ × 15 × 10⁻¹²) = 1.5 × 10⁻⁹, so f ≈ 1 ÷ (2π × 1.5 × 10⁻⁹) ≈ 106 MHz. Any edge fast enough to contain energy near that frequency rings at it, and the ringing is drawn on screen as if the circuit produced it.

A short ground spring has much less inductance, so the resonance moves well above the bandwidth of most hobby scopes and disappears from view. If you need to know whether ringing is real, the method is in how to measure overshoot and ringing; this page is about removing the part the probe adds.

Symptoms, causes and fixes

1Compare 1x and 10x on the same node

The quickest loading test is to measure twice. Note the amplitude and edge shape on 1x, then switch the probe to 10x, set the channel's probe factor to match, and look again. If the waveform changes shape or size by more than the scope's normal accuracy, the probe is loading the node, and the 10x view is the more trustworthy one. If the 10x reading still looks suspicious, the node is sensitive enough that even 10 MΩ and 10 to 15 pF matter, and the later steps apply.

1. 01

   Measure on 1x

   Note amplitude and edges

   →

2. 02

   Switch to 10x

   Match the channel's probe setting

   →

3. 03

   Compare

   A change means loading

   →

4. 04

   Reduce the load

   Shorter ground, better node, better probe

2Shorten the ground

For anything with edges faster than a few tens of nanoseconds, take the long clip lead off. Most 10x probes come with a small spring or bayonet ground that slips over the tip barrel and touches a ground pad a centimeter or less away. If the board has no nearby ground, a short solid wire soldered to ground near the test point does the same job. The goal is the smallest possible loop between tip and ground. The basic attachment routine, including where the clip should go, is in how to connect an oscilloscope probe correctly.

3Move to a lower-impedance test point

The same signal often appears at several places in a circuit, and some are much harder to disturb than others. A driven output, the far side of a buffer, or the low side of a divider has a lower source resistance than a timing node, a crystal pin or the junction of two megohm resistors. Because rise-time slowing scales with R × C, moving from a 10 kΩ node to a 1 kΩ node cuts the slowing tenfold: 2.2 × 1,000 × 15 × 10⁻¹² ≈ 33 ns instead of 0.33 µs. With a crystal oscillator, probe the oscillator buffer or clock output rather than the crystal pins; the procedure is in how to check a crystal oscillator with a scope.

4Compensate the probe

A 10x probe only divides evenly across frequency when its tip capacitance is matched to the cable and scope input. Moved to a different channel or a different scope, it may not be. An undercompensated probe rounds the corners of a square wave much like capacitive loading does, and an overcompensated one adds overshoot that looks like ringing. Before blaming the circuit, put the probe on the calibration terminal and check for flat tops. If the corners are rounded or peaked there too, the problem is compensation, not loading. The trimmer adjustment is covered step by step in how to compensate an oscilloscope probe.

5Use the right probe for the node

A standard 10x passive probe is the right default for nearly all hobby and repair work. When even its 10 to 15 pF is too much, for example on a crystal pin, a high-speed clock or a very high-impedance analog node, an active FET probe typically puts only around 1 pF or less on the circuit, at the cost of a limited input voltage range and the need for power. Low-capacitance passive probes, sometimes called resistive divider or Z0 probes, are another option for fast, low-impedance signals and are meant for a 50 Ω scope input. For floating or mains-referenced signals, the right tool is a differential probe rather than a lower-capacitance one. Read the input range and capacitance on the maker's datasheet before buying any of them.

1x or 10x: which to use

The 1x position exists for one reason: sensitivity. It is useful for very small, slow signals where the 10x divider would push the trace down near the scope's noise floor and the source can easily drive a 1 MΩ, high-capacitance load. For everything else, 10x loads the circuit less and preserves more bandwidth. The figures below are typical values for general-purpose passive probes, not specifications for any particular model.

Whichever you choose, the channel's probe setting has to match, or every voltage reading is out by a factor of ten. Small-signal work on 10x is easier on a scope with fine vertical resolution, which is one place a 12-bit instrument helps.