Oscilloscope Mastery

Study an annotated diagram

A DMM tells you the average; the scope shows you the truth over time. Anything that moves — clocks, data, ripple, glitches, oscillation — is scope territory. Open visuals/03-test-equipment.html ↗ for the annotated front panel, screen anatomy, probe diagram, and the good-vs-bad waveform gallery.

1. What a scope is

A voltage-vs-time grapher. The screen is a grid (graticule): vertical scale in volts/division, horizontal in seconds/division. Two to four input channels, each with its own probe. Everything else on the front panel exists to answer three questions: how big (vertical), how fast (horizontal), and when to take the picture (trigger).

2. Probes — where most scope mistakes live

• Standard probe is 10:1 passive: divides the signal by 10, which lightens loading on the circuit and extends range. The scope must know (probe setting = 10X) or every reading is off by 10×. Most modern probe/scope pairs auto-detect; verify anyway.
• Compensate every probe on the scope's cal terminal (the square-wave tab) before trusting it: adjust the trimmer until the square wave has flat tops — not rounded (under-compensated), not spiked (over). Two-second ritual, do it whenever you pick up an unfamiliar probe.
• The ground lead matters. The long alligator ground lead is an inductor; on fast digital edges it manufactures ringing that isn't really in your circuit. Seeing ugly ringing on an edge? Re-measure with the short ground spring at the probe tip before blaming the board.
• ⚠️ The ground clip is earth ground. Bench scopes tie the probe ground to mains earth. Clip it to a node that isn't at ground potential and you've made a short through the scope. On floating/off-line circuits (mains-side of power supplies especially), use a differential probe — never "float" the scope by defeating its earth pin.

3. The setup ritual (until it's muscle memory)

1. Probe compensated, set to 10X, scope channel set to 10X.
2. Verify the UUT reference is safe to connect to protective earth; then use the approved short ground connection close to the test point.
3. Coupling: DC coupling by default (shows the true level including DC). AC coupling only when you deliberately want to see a small AC component riding on a big DC level (ripple measurement).
4. Vertical: volts/div so the expected signal fills 50–80% of the screen. For a 3.3V logic signal, 0.5V/div puts it a healthy six-plus divisions tall; 1V/div would leave it cramped in the bottom third.
5. Horizontal: time/div to show a few cycles. For a 1MHz clock (1µs period): 0.5–1µs/div.
6. Trigger: source = your channel, type = edge, rising, level ≈ mid-signal (1.6V for 3.3V logic). Mode Auto while exploring (always draws something), Normal when you want only real events, Single to capture one-shot events (a glitch, a power-up sequence).
7. Or press Autoset and then fix what it guessed wrong — fine for orientation, never for final measurement.

4. Reading the screen

• Amplitude: count vertical divisions × V/div. Better: use the scope's built-in measurements (Vpp, Vmax, Vmin, Vavg, RMS) and cursors for manual checks.
• Time/frequency: one period in divisions × s/div; f = 1/T. Built-in freq measurement is fine when triggering is stable.
• A stable, non-scrolling display = good trigger. A jittering smear = trigger level/source wrong, or the signal genuinely varies (that's information too).

5. The measurements a repair tech actually makes

Power rail ripple

DC-couple first: is the rail at nominal? Then AC-couple, crank to 10–50mV/div, time/div near the converter's switching period, and (if available) 20MHz bandwidth limit to reject ambient hash. Compare ripple against the actual specification at its defined bandwidth, load and probe setup; there is no universal healthy ripple amplitude. Failing bulk cap: ripple grows several-fold, often with big sag at load transients. Linear-regulated rails should be nearly flat — line-related ripple can indicate upstream filtering or loading issues but does not alone condemn a capacitor. Use DC coupling to verify absolute rail minima and supervisor thresholds.

Is the clock alive and clean?

Probe the crystal/oscillator output: correct frequency, healthy amplitude, fast clean edges. A flatline calls for setup, probe-loading, supply and enable checks before suspecting the oscillator. Slow saggy edges = loading/partial short. (Touching a crystal pin with a 10pF probe can disturb or stop a marginal oscillator — probe the buffered/oscillator output side, and treat "it stopped when I probed" as a clue, not a kill.)

Reset behavior

Single-shot capture on power-up: reset line should hold low (for active-low) for a defined time after rails rise, then release cleanly. Reset stuck low = supervisor IC, sagging rail, or something else holding it. Repeated reset pulses require correlation with monitored rails and reset sources; brownout, watchdog, manual reset and other drivers are different hypotheses.

Digital signal integrity

Levels reach proper high/low; edges are crisp; no runts (pulses that don't reach full height — suspect bus contention: two drivers fighting); no excessive ringing/overshoot (after ruling out the ground-lead artifact).

Serial buses

• UART: idle high; each byte = low start bit, 8 data bits, high stop. Eyeball the bit width to confirm baud (104µs/bit = 9600).
• I²C: two lines (SCL, SDA), open-drain so idle = high via pull-ups; if either line sits low forever, a device is hung or a pull-up is missing. Edges are RC-rounded on the rising side — normal.
• SPI: clock bursts plus data lines and a chip-select dropping low around each transaction.
• Modern scopes decode these on screen — learn the decode feature on whatever scope your shop has; it turns waveform-squinting into reading hex.

Hunting intermittents/glitches

Single or Normal trigger armed on the misbehavior (e.g., trigger on reset falling when it shouldn't), then apply the stressor — flex, heat, cold spray, vibration. The scope waits; you provoke. Persistence/infinite-persistence display modes also reveal rare runts inside an apparently healthy stream.

6. Bandwidth honesty

A scope only shows what's within its bandwidth. Rule of thumb: scope bandwidth ≥ 5× the highest frequency of interest; for digital, the edges (not the clock rate) set the requirement. A 100MHz scope showing a 66MHz clock draws a pretty sine wave even if the real signal is square — know what your instrument can and can't show, and don't diagnose "slow edges" at the limit of your bandwidth.

7. Self-check

1. Your 3.3V clock shows 0.33V amplitude. First suspicion? Reveal answer
2. Why AC coupling for ripple? Reveal answer
3. When is the long alligator ground lead unacceptable? Reveal answer
4. Reset line pulses every 1.6s forever. Meaning? Reveal answer
5. Why never clip scope ground to the high side of a current-sense resistor on a live supply? Reveal answer

Next: 06 — Troubleshooting Methodology