The observer is part of the circuit
By now you can drive the bench: guide 1 gave you the DMM, supply, and generator; guide 2 gave you the scope and its x10 probe. But every one of those instruments shares a secret the glossy manuals gloss over. The moment you touch a node to measure it, you have added a new component to the circuit. The instrument is not a silent witness peering in from outside — it is wired in, drawing a little current or charge, and the number it reports is the number of the circuit-plus-instrument, not the circuit you meant to study. This is the loading effect, and learning to see it is the difference between trusting your bench and being fooled by it.
Think of voltage as water pressure at a point in a pipe network. To read that pressure you tap in a gauge — but any real gauge bleeds off a trickle of water to move its needle. If the point you are probing can supply a torrent, that trickle is nothing and the reading is honest. But if the point is fed through a thin, restrictive pipe — a high source resistance — then your gauge's little trickle is enough to make the pressure sag, and you read a value lower than the true one. The instrument did not lie; it changed the thing it measured. Every measurement is a negotiation between how stiff the node is and how greedy your instrument is.
Voltage loading: the meter steals a sip
Put numbers on it. Take a plain voltage divider: two 1 MΩ resistors in series across 10 V. With nothing attached, the midpoint sits at exactly 5 V, and the node's Thevenin source resistance is 1 MΩ in parallel with 1 MΩ, or 500 kΩ — a fairly 'soft' node. Now read it with a good digital multimeter whose input resistance is 10 MΩ. That 10 MΩ sits in parallel with the bottom 1 MΩ, and the careful 5 V you designed for quietly slips away.
10 V
|
[ 1M ] top resistor
|
node A ---------+----------+-----[ meter ]
| |
[ 1M ] R_in (meter input resistance)
| |
GND ------------+----------+----- GND
ideal (no meter): V_A = 10 x 1M/(1M + 1M) = 5.00 V
10M DMM: bottom = 1M || 10M = 0.909M
V_A = 10 x 0.909/(1 + 0.909) = 4.76 V (-4.8 %)
1M scope x1: bottom = 1M || 1M = 0.5M
V_A = 10 x 0.5/(1 + 0.5) = 3.33 V (-33 %)Look at the damage. A pristine 10 MΩ meter still pulls the reading from 5.00 V down to 4.76 V — a 5% error from doing nothing wrong. Worse, point a scope's x1 input (only 1 MΩ) at the same node and it collapses to 3.33 V, a 33% lie. The rule of thumb that falls out: your instrument's input resistance should be at least ten times the node's source resistance for a reading within about 10%, and a hundred times to be within about 1%. On a stiff 50 ohm node a 10 MΩ meter is flawless; on a 1 MΩ node it is already suspect. This is exactly why the x10 probe from guide 2 exists: it trades away nine-tenths of your signal to raise the scope's input resistance from 1 MΩ to 10 MΩ, loading the circuit ten times less.
At speed, capacitance is the thief
Input resistance is only half the story, and at high frequency it is the smaller half. Every probe, cable, and meter input also presents a little capacitance to ground — the scope's own input plus a metre of coax adds up to perhaps 100 pF for a plain x1 probe. A capacitor is a tiny bucket that resists sudden voltage change, and to a fast signal it looks like a low-impedance path to ground. So the probe's capacitance teams up with the node's source resistance to form an unintended low-pass filter, rounding off your edges and shrinking your highs — the high-frequency face of probe loading.
The corner frequency is the familiar f = 1 / (2 times pi times R times C). Probe a 10 kΩ node with a x1 probe's 100 pF and f = 1 / (2 times pi times 10000 times 100 times 10^-12) ≈ 159 kHz. Try to view a 1 MHz square wave through that and you are not seeing your circuit — you are seeing your probe's filter. Now switch to the x10 probe: its tip network drops the capacitance the circuit feels to roughly 12 pF, so f = 1 / (2 times pi times 10000 times 12 times 10^-12) ≈ 1.3 MHz, nearly ten times higher. The x10 probe is not just an attenuator for big signals; its real day job is loading the circuit ten times less, in both resistance and capacitance, so what you see is closer to what is really there.
One more speed trap hides in the probe's own ground lead. That little wire from the probe barrel to your ground clip has inductance — the flywheel that hates sudden current changes — and together with the probe tip capacitance it forms a resonant loop. On a fast edge it will ring, painting decaying wiggles on your screen that live in the probe, not the circuit. The fix is the short spring-tip ground that came in the probe box: a 5 cm clip lead can ring badly where a 5 mm spring is clean. As guide 2 warned, an uncompensated probe distorts amplitude; here a long ground lead distorts the edges. On a breadboard, where strays are already large, both effects are at their nastiest — fast, precise work belongs on a real board.
Measuring current means breaking in
Voltage you measure by tapping across two points; current you cannot, because current flows through things. To read it the old-fashioned way you must break the wire and insert the ammeter into the path so all the current funnels through it. That alone is a bigger disturbance than a voltage probe: you have opened the circuit and dropped a new part into the live current path. And that part is not free — every ammeter develops a small voltage across itself as current passes, called the burden voltage.
Return to the LED you sized in guide 1: a 5 V supply, an LED dropping about 2 V, and a 300 ohm resistor giving (5 - 2) / 300 = 10 mA. Slip an ammeter with 10 ohm of burden resistance into that loop and the current becomes (5 - 2) / (300 + 10) = 9.7 mA — a tidy 3% you can live with. But burden bites hardest on sensitive ranges. Measuring a sleeping microcontroller's supply on the microamp range, the meter's internal shunt might be 1 kΩ; at 1 mA that is a full 1 V of burden, and a 3.3 V chip now sees only 2.3 V. It may brown out, run slower, or skip the very sleep mode you were trying to measure. Your reading is then perfectly accurate — of a different circuit.
Two honest ways out. First, always use the lowest meter range that still reads, and prefer a meter that publishes a low burden voltage — good ones quote it in the datasheet. Second, when you must not break the circuit at all, measure current without contact: a clamp meter reads the magnetic field around a wire (great for amps, blind to milliamps), or you deliberately design in a small known sense resistor and measure the voltage across it with a high-impedance probe, turning a current measurement back into a gentle voltage one. The lesson repeats: you cannot watch for free, but you can choose how small the bill is.
Where you clip the ground decides everything
Here is the disturbance that does not just skew a reading — it can blow up your board. A bench scope's probe ground clip is not a neutral reference floating wherever you like; it is wired, through the scope's chassis and mains plug, hard to earth ground. Clip it to a node that sits at, say, 12 V above earth, and you have just connected that node to earth through the braid of your probe cable — a dead short. At best the supply's current limit trips; at worst a trace vaporises or the probe's thin ground wire melts. The scope did not measure the node; it shorted it.
When you need the voltage across something that is not sitting at ground — a high-side current-sense resistor, a half-bridge output, anything mains-referenced — the answer is a differential measurement: read both ends with respect to the same reference and take the difference, never assuming either end is ground. The clean way is a differential probe, or two matched probes into a scope's A−B math channel. The dangerous shortcut people reach for — lifting the scope's mains earth pin so its ground clip can 'float' — is how benches kill people: it makes the whole scope chassis live at whatever you clipped to. Never do it. This is the one place where the electrical-safety habits from the very first rung are not optional.
- Before you trust any reading, ask how stiff the node is. Estimate its source resistance; if your instrument's input resistance is not at least 10x larger, expect loading error and switch to a higher-impedance probe.
- For fast or precise signals, mind capacitance, not just resistance: reach for the x10 probe and use its shortest ground spring, so the probe's own RC and ringing do not masquerade as your circuit's behaviour.
- To measure current, use the lowest range that reads and check the burden voltage — or sense a small resistor and measure its voltage instead, so you barely disturb the loop.
- Never assume the probe ground clip can go anywhere. If neither end of what you want sits at earth ground, measure it differentially — and never float the scope by defeating its earth.