Reading a Signal's Health — Weak, Distorted, or Lost

Section Overview

Tracing and injection find where a signal fails; this section reads how it fails, because the manner in which a signal is wrong points at what the stage did to it (signal-tracing-following-a-signal-stage-by-stage). A signal can be present yet wrong in several ways. A weak signal points at lost gain or attenuation, and a distorted one — clipped, rounded, or misshapen — points at overdrive or nonlinearity, each a different fault read from the waveform (the-signal-chain-as-a-diagnostic-path). Noise is judged against the signal. A signal-to-noise ratio compares the wanted signal with the noise riding on it, so a signal well above the noise is healthy and one buried in it, or a stage that has added noise, is a fault a bare amplitude reading would miss. A signal can sit at the wrong level. A DC offset shifts the whole waveform up or down from its correct DC level, so a signal of the right shape and size on the wrong voltage points at a bias or coupling fault, not the signal path. A stage can add a signal from nowhere. A parasitic oscillation is an unwanted oscillation a stage produces on its own from instability, so a signal fouled by a fast ripple or a tone that appears with no input points at an unstable stage. Read each signature — weak, distorted, noisy, shifted, oscillating, or absent — against what the node should carry, and how a signal is wrong names the fault (signal-tracing-following-a-signal-stage-by-stage).

Why This Matters

Finding where a signal fails is half the job; reading how it fails is the other half, because the manner of failure often names the fault before a single part is touched (signal-tracing-following-a-signal-stage-by-stage). This matters because the signature points at the cause: a weak, clipped, noisy, shifted, or oscillating signal each points at a different fault, so reading the manner narrows the search inside a stage (the-signal-chain-as-a-diagnostic-path). This matters because a present signal can still be the fault: a signal that is there but wrong is missed by a present-or-absent check, so reading its health catches a stage that passes but corrupts the signal. It matters because noise needs a reference: a noisy signal is judged by its signal-to-noise ratio, not its bare size, so knowing to read the ratio catches a stage drowning the signal. It matters because level and shape both carry information: a DC offset or an oscillation is invisible if only presence is checked, so reading the whole signal — level, shape, noise — finds faults a quick look misses. And it matters because it speeds the repair: knowing a clipped signal means overdrive, or an oscillation means instability, points straight at the kind of part to check. Read how a signal is wrong, and its failure describes the fault that made it.

Required Prerequisites

• The Signal Chain as a Diagnostic Path — Section 7.1 established judging a signal at each node against what should be there; this section reads the specific ways a signal can be wrong.
• Signal Tracing — Following a Signal Stage by Stage — Section 7.2 read a signal at a node with a scope; this section reads the health of that signal — weak, distorted, noisy, shifted, or oscillating.

Recommended Consumables

• An oscilloscope — to see a signal's shape, level, and noise at each node (signal-tracing-following-a-signal-stage-by-stage)
• A known-good signal or board — to compare a node's signal against a healthy one (the-signal-chain-as-a-diagnostic-path)
• A schematic with expected signals — to know what a healthy signal should be at each point
• A notebook for the signatures — to log how each node's signal is wrong — weak, clipped, noisy, shifted
• The scope's measurement and spectrum tools — to read amplitude, noise, and distortion numerically

Recommended Practice Hardware

• An amplifier driven into clipping — to see a clipped waveform and read overdrive (signal-tracing-following-a-signal-stage-by-stage)
• A stage with lost gain — to see a weak signal and read attenuation
• A board with a noisy stage — to judge a signal by its signal-to-noise ratio
• A stage with a bias fault — to see a DC offset shift the waveform
• A stage prone to instability — to see a parasitic oscillation appear (the-signal-chain-as-a-diagnostic-path)
• A known-good identical board — to compare each node's healthy signal against

Real-World Applications

Reading a signal's health is how a technician names a fault from the waveform alone. A repairer with a quiet amplifier sees the signal weak after one stage and reads lost gain, pointing inside that stage (signal-tracing-following-a-signal-stage-by-stage). A technician with a harsh, buzzy output sees the waveform clipped flat at the top and reads overdrive or a collapsed supply rail. Someone with a hissy signal reads its signal-to-noise ratio and finds a stage that has drowned the signal in noise. A repairer of a stage that will not bias reads a DC offset shifting the whole waveform and looks at the bias network, not the signal path. And a technician chasing a squeal finds a parasitic oscillation riding the signal and traces it to poor decoupling rather than a bad signal stage (the-signal-chain-as-a-diagnostic-path). The failures this prevents: calling a present-but-wrong signal good, judging noise without a reference, and missing a DC offset or an oscillation that a presence-only check would never see.

Common Challenges

• A healthy signal changes shape legitimately along the chain. Each stage transforms the signal, so what is distortion at one node is the intended shape at another, and telling a fault from a transform takes knowing the design — the expected signal at each point has to be learned, not assumed (the-signal-chain-as-a-diagnostic-path).
• Some faults show only under real conditions. A signal may be healthy at idle and distort only under load, at full level, or at a certain frequency, so a fault can hide from a gentle test — exercising the stage as it really works is often needed to see it.
• A measurement artefact can look exactly like a signal fault. A long ground lead or a poor probe can add noise or provoke ringing that is not in the signal at all, so what shows on the trace may belong to the setup rather than the stage — and the two can look identical until the measurement itself is cross-checked, which stays a genuine trap (signal-tracing-following-a-signal-stage-by-stage).

Safety Notes

Risk Level: Medium. Reading a signal's health is done on a powered board with a scope and probe, so it is live work and this section is Medium risk.

Professional Tips Before Starting

• Read level and shape, not just presence. A present signal can still be the fault — read its amplitude, shape, noise, and DC level, not just whether it is there (signal-tracing-following-a-signal-stage-by-stage).
• Compare against the expected signal. A transformed signal looks wrong out of context — know what a healthy signal is at each node, from a known-good board (the-signal-chain-as-a-diagnostic-path).
• Rule out the probe for noise and ringing. A bad ground can add noise the signal does not have — check the measurement before blaming the stage.

Reading How a Signal Is Wrong

Recap and Frame

Sections 7.1 and 7.2 found where a signal is lost; this section reads how it is wrong, and the frame is that each way a signal fails — weak, distorted, noisy, shifted, oscillating, absent — points at a different fault (signal-tracing-following-a-signal-stage-by-stage). A signal has several qualities to read. Amplitude, shape, noise, DC level, and any extra oscillation are all read, so a signal is judged on more than presence (the-signal-chain-as-a-diagnostic-path). Weak and distorted are the common signatures. A weak signal points at lost gain, and a distorted one at overdrive or nonlinearity, the two most frequent ways a signal is wrong at a node. Noise is read against the signal. A signal is judged healthy or not by how far it stands above the noise, so the signal-to-noise ratio, not the bare size, tells whether a stage passes it cleanly. Level and instability add two more. A DC offset shifts the whole waveform off its level, and a parasitic oscillation adds a signal that should not be there, each a distinct fault signature. The manner names the fault. Because each signature points at a kind of fault, reading how a signal is wrong narrows the search inside the failing stage (signal-tracing-following-a-signal-stage-by-stage). Hold the frame — read amplitude, shape, noise, level, and oscillation, and let the manner name the fault — and a signal's health becomes a diagnosis.

A Weak Signal — Attenuation and Lost Gain

The commonest signature is a weak signal — smaller than it should be — which points at attenuation where there should be gain (signal-tracing-following-a-signal-stage-by-stage). Read the amplitude against expected. A signal is weak only relative to what the node should carry, so its amplitude is compared to the expected level or a known-good board, since a small signal is normal early and a fault late. Suspect lost gain in a gain stage. A signal that enters a gain stage healthy and leaves smaller than it should shows the stage is not amplifying — a lost-gain fault in that stage (the-signal-chain-as-a-diagnostic-path). Suspect a lossy path. A passive path — a coupling, a filter, a connector — that attenuates more than it should also weakens a signal, so a weak signal across a passive section points at a loss there. Tell design attenuation from a fault. Some attenuation is by design — a volume control, a divider — so a weak signal is judged against the attenuation the stage should have, not assumed a fault. Check the stage that should amplify. A weak signal at a gain stage's output sends the search into that stage — its bias, its device, its feedback — for why the gain is low, tying the signature to a part. Watch for a weak-and-loaded signal. A signal dragged down by a load, rather than a weak source, is told apart by lifting the load, so a loaded signal is not mistaken for lost gain. Amplitude read against expected, lost gain suspected, a lossy path considered, design attenuation excluded, the gain stage checked, and loading ruled out — and a weak signal is read. Read a weak signal as lost gain or loss, and the search goes to the stage that should have amplified it.

A Distorted Signal — Clipping, Crossover, and Nonlinearity

A distorted signal — present but the wrong shape — points at a stage handling the signal nonlinearly, and the kind of distortion names the fault (the-signal-chain-as-a-diagnostic-path). Read clipping as overdrive or a supply limit. A waveform flattened at the top, bottom, or both is clipped, meaning the signal hit a limit — an overdriven stage or a supply rail it cannot exceed — so clipping points at too much level or too little supply. Read crossover distortion as a bias fault. A distortion at the zero-crossing of a push-pull stage — a kink where the waveform crosses zero — is crossover distortion, pointing at the output stage's bias being wrong, a specific and recognisable signature. Read rounding as lost bandwidth. A fast signal that arrives rounded or with slow edges has lost high-frequency content, pointing at a bandwidth limit, a slow stage, or excessive capacitance. Read general nonlinearity as a stressed device. A waveform distorted in shape without clipping can mean a device operating nonlinearly — a wrong bias, a failing transistor — so the distortion points inside the stage. Compare against the healthy shape. Distortion is judged against what the signal should look like at that node, from a known-good board, so a legitimately transformed signal is not called distorted (signal-tracing-following-a-signal-stage-by-stage). Vary the conditions to reveal it. Some distortion shows only at full level or under load, so driving the stage as it really works brings out a distortion a gentle test misses. Clipping, crossover, rounding, nonlinearity read, compared to healthy, and conditions varied — and a distorted signal is read. Read the kind of distortion, and it names whether the fault is overdrive, bias, bandwidth, or a stressed device.

A Noisy Signal — Noise and the Signal-to-Noise Ratio

A noisy signal — the wanted signal buried in unwanted hash — is judged not by its size alone but by how far it stands above the noise, its signal-to-noise ratio (signal-tracing-following-a-signal-stage-by-stage). Understand the ratio. The signal-to-noise ratio compares the wanted signal with the noise riding on it, so a signal well above the noise is healthy and one nearly as small as the noise is unusable even if present. Read where the ratio falls. A signal clean at one node and noisy at the next means that stage added the noise or lost the signal, so a falling ratio along the chain points at the noisy stage (the-signal-chain-as-a-diagnostic-path). Recognise the kind of noise. Hum at the mains frequency points at a ground loop or hum pickup, hum at twice the mains frequency at power-supply ripple from the rectifier and filter, hiss at a failing or high-gain stage, and bursts at an intermittent fault, so the character of the noise hints at its source. Suspect grounding and supply. Much added noise comes from a poor ground, a noisy supply rail, or coupling, so a noisy signal sends the search to grounding and the stage's power as much as the signal path. Rule out the measurement. A long ground lead or a poor probe adds noise that is not in the signal, so the probe is checked before the stage is blamed for noise (signal-tracing-following-a-signal-stage-by-stage). Read the ratio, not the level. Because a strong noisy signal and a weak clean one are different faults, the ratio is read rather than the bare amplitude, so the health is judged correctly. The ratio understood, where it falls read, the noise character recognised, grounding suspected, the probe ruled out, and the ratio read over level — and a noisy signal is read. Judge a signal by its signal-to-noise ratio, and a stage drowning the signal in noise gives itself away.

A Shifted or Oscillating Signal — DC Offset and Parasitic Oscillation

Two signatures are missed by a presence-only check — a signal shifted off its level and one fouled by an oscillation — and both point at specific faults (the-signal-chain-as-a-diagnostic-path). Read a DC offset. A signal of the right shape and size but centred on the wrong voltage has a DC offset, so reading the signal's DC level as well as its shape catches a waveform riding too high or low. Trace an offset to bias or coupling. Because a stage sets the DC level its signal sits on, an offset points at a bias fault, a failed coupling capacitor, or a leaky part, so the offset sends the search to the stage's DC conditions, not its signal path. Read a parasitic oscillation. An unwanted oscillation — a fast ripple, a burst, or a tone that appears with no input — riding on or replacing the signal is a parasitic oscillation, a signal the stage makes on its own. Trace an oscillation to instability. A parasitic oscillation comes from instability — poor decoupling, stray feedback, a marginal design — so its cure is restoring stability rather than replacing the signal path, a distinct fault class (signal-tracing-following-a-signal-stage-by-stage). Tell an oscillation from noise. An oscillation is a coherent waveform at a definite frequency, while noise is random hash, so reading whether the unwanted content is structured tells the two apart and points at different causes. See both need level and shape read. Neither an offset nor an oscillation shows if only presence is checked, so reading the whole signal — its level and any extra waveform — is what catches them. A DC offset read and traced to bias, a parasitic oscillation read and traced to instability, told from noise, and both needing the full signal read — and a shifted or oscillating signal is read. Read the signal's level and any oscillation, and a bias or instability fault reveals itself.

An Absent Signal and Reading the Manner to the Cause

The last signature is a signal simply gone, and pulling all the signatures together, the manner of failure is read back to the kind of fault (signal-tracing-following-a-signal-stage-by-stage). Read an absent signal as a broken path or dead stage. A signal present at one node and gone at the next means the stage between lost it entirely — an open path, a dead device, a missing supply — the strongest and simplest signature. Distinguish absent from tiny. A signal that is not truly gone but very small is a weak-signal fault, not an absent one, so the scope's sensitivity is raised to tell a lost signal from a badly attenuated one (the-signal-chain-as-a-diagnostic-path). Map each signature to a fault class. Weak points at lost gain, clipped at overdrive, crossover at bias, noisy at grounding or a failing stage, shifted at bias or coupling, oscillating at instability, and absent at a broken path — so the signature chooses the fault class. Confirm with the stage's power. Any bad signal is checked against the stage's supply and bias, since a starved stage produces a weak, distorted, or absent signal that mimics its own failure, tying signal back to power (signal-tracing-following-a-signal-stage-by-stage). Read the manner before the part. Because the manner names the fault class, it is read first, so the search inside a stage is aimed rather than random. Hand the class to isolation. With the signature read to a fault class in one stage, the block is diagnosed at part level, so reading the health has narrowed both where and what. The absent signature read, told from tiny, each mapped to a class, power confirmed, the manner read first, and handed to isolation — and the manner of failure is read to the cause. Read how a signal is wrong before reaching for a part, and its failure points at the kind of fault to find.

Common Mistakes

• Calling a present signal good. A signal can be there but wrong — read its amplitude, shape, noise, and level, not just presence (signal-tracing-following-a-signal-stage-by-stage).
• Judging noise by the bare signal size. A strong noisy signal is not healthy — read the signal-to-noise ratio, not the amplitude.
• Missing a DC offset or oscillation. A presence-only check never sees them — read the signal's level and any extra waveform (the-signal-chain-as-a-diagnostic-path).
• Blaming the stage for probe noise. A bad ground adds noise the signal lacks — rule out the probe before the stage.
• Ignoring what the distortion looks like. Clipping, crossover, and rounding mean different faults — read the kind of distortion, not just that it is there.

Troubleshooting Guidance

Reading a signal's health comes down to read amplitude, shape, noise, level, and oscillation, then map the manner to a fault. If a signal is weak: read it against the expected level and suspect lost gain or a lossy path (signal-tracing-following-a-signal-stage-by-stage). If it is clipped: suspect overdrive or a collapsed supply rail. If it is distorted at the zero-crossing: suspect a push-pull output stage's bias — crossover distortion. If it is buried in noise: read the signal-to-noise ratio and suspect grounding, supply, or a failing stage — and rule out the probe. If it rides too high or low: read the DC offset and suspect a bias or coupling fault (the-signal-chain-as-a-diagnostic-path). If a coherent oscillation rides the signal: suspect a parasitic oscillation from instability, not a lost signal. If a signal is gone: distinguish absent from very small, then suspect a broken path or dead stage, and check the stage's power. The throughline: read how the signal is wrong, map the signature to a fault class, and check the stage's power before condemning it.

Verification & Testing Methods

Confirm you read the signal's health:

• [ ] I read each node's amplitude, shape, noise, and DC level against a known-good signal, not just whether the signal was present (signal-tracing-following-a-signal-stage-by-stage).
• [ ] I judged a noisy signal by its signal-to-noise ratio — how far it stood above the noise — rather than by its bare amplitude.
• [ ] I read the signal's DC offset where the waveform rode high or low, and traced it to the stage's bias or coupling.
• [ ] I recognised a parasitic oscillation — a coherent unwanted oscillation — and told it from random noise, tracing it to instability.
• [ ] I mapped the manner of failure — weak, clipped, crossover, noisy, shifted, oscillating, or absent — to a fault class, and checked the stage's power (the-signal-chain-as-a-diagnostic-path).

Then try the practice exercises below — signal-health-reading practice on powered boards; scenarios differ from the quiz.

Practice Exercises

1. Read a weak signal (5 minutes, hands-on). On a stage with low gain, read the signal against the expected level and confirm it is attenuated where it should be amplified (signal-tracing-following-a-signal-stage-by-stage).
2. Read distortion (5 minutes, hands-on). Drive a stage into clipping and read the flattened waveform, then find a crossover or rounding distortion elsewhere and name what each points at.
3. Judge the signal-to-noise ratio (4 minutes, hands-on). On a noisy node, read how far the signal stands above the noise and reason whether the stage or the probe added it.
4. Spot an offset or oscillation (3 minutes, hands-on). Read a signal's DC level for an offset, and look for any coherent oscillation riding it, tracing each to bias or instability.

These core steps — weak, distorted, noisy, shifted, oscillating, and absent — are tested in the Chapter Quiz at the end of this chapter, where a score of 80% is required to continue.

Key Takeaways

• A signal can be present yet wrong, so its health is read from its amplitude, shape, noise, DC level, and any oscillation — a weak signal points at lost gain, a clipped one at overdrive (signal-tracing-following-a-signal-stage-by-stage).
• A noisy signal is judged by its signal-to-noise ratio — how far it stands above the noise — not its bare size, so a stage drowning the signal in noise is caught even when the signal is present.
• A DC offset shifts the whole waveform off its correct level, pointing at a bias fault, a failed coupling, or a leaky part rather than the signal path.
• A parasitic oscillation — a coherent unwanted oscillation a stage makes on its own — points at instability such as poor decoupling or stray feedback, a distinct fault class from a lost or distorted signal.
• Each signature maps to a fault class — weak, clipped, crossover, noisy, shifted, oscillating, or absent — so reading how a signal is wrong names the fault, checked against the stage's power (the-signal-chain-as-a-diagnostic-path).

Skills Learned

• You can now recognise a weak signal from attenuation and lost gain.
• You can now read distortion, from clipping to crossover, and what it points at.
• You can now judge a noisy signal by its signal-to-noise ratio.
• You can now spot a DC offset and a parasitic oscillation and their causes.
• You can now read the manner of a signal's failure back to the kind of fault.

Glossary Additions

• signal-to-noise ratio — the size of a wanted signal compared with the unwanted noise riding on it, which measures how healthy a signal is against the noise floor. A signal can be present and the right shape yet useless if noise is nearly as large as it, so the signal-to-noise ratio captures a health a bare amplitude reading misses: a strong signal well above the noise is healthy, while a signal buried in noise, or a stage that has added noise so the ratio has fallen, is a fault. Reading the signal-to-noise ratio at each node — how far the signal stands above the hash on the trace — tells whether a stage is passing the signal cleanly or drowning it, so a falling ratio along the chain points at a noisy or failing stage rather than a lost one.
• DC offset — an unwanted shift of a signal away from its correct DC level, so the whole waveform rides higher or lower than it should. A stage sets both the signal and the DC level it sits on, so a signal that is the right shape and size but centred on the wrong voltage has a DC offset, pointing at a bias fault, a failed coupling, or a leaky part rather than a problem with the signal itself. Reading a signal's DC level as well as its shape catches an offset that a signal-only view would miss, and because the offset points at the stage's bias or coupling, it is a distinct signature from a weak or distorted signal and sends the search to the stage's DC conditions.
• parasitic oscillation — an unwanted oscillation a stage produces on its own, adding a signal that should not be there, usually from instability such as poor decoupling, stray feedback, or a marginal design. A stage that should simply pass or amplify a signal but instead breaks into oscillation shows an extra, often high-frequency, waveform riding on or replacing the wanted one, so a signal fouled by an oscillation points at an unstable stage rather than a lost or distorted one. Reading a signal for a coherent oscillation that is not part of it — a fast ripple, a burst, a tone that appears with no input — identifies a parasitic oscillation, whose cure is usually restoring stability with decoupling, layout, or a compensation part rather than replacing the signal path.

Suggested Next Sections

Must read next:

• Tracing Digital and Clock Signals — Section 7.5 turns from analog signals to digital ones: reading logic levels, edges, and clocks, and the ways a digital signal fails — a stuck line, a missing clock, a marginal level, a bus fight — that stall a digital board.

Recommended:

• The Signal Chain as a Diagnostic Path — judging a signal at each node against what should be there, which this section reads in detail.
• Signal Tracing — Following a Signal Stage by Stage — reading a signal at a node with a scope, whose health this section learns to interpret.