AoNArt of Nanoelectronics
My space
05 · Foundational

From time plots to defensible signal descriptions

Synthesize SG-01 into a repeatable engineering workflow: identify the physical quantity and reference, read units and observation window, classify behavior without overclaiming periodicity, extract vertical and temporal descriptors, separate average and zero-mean variation, fit a sinusoidal model only when justified, validate phase and RMS independently, and document what came from the mathematical model versus a finite measurement record.

140–180 minFoundational
In this lesson
  • Apply a fixed checklist that starts with physical quantity, reference direction, units, time origin and observation window before extracting numerical parameters.
  • Classify constant, time-varying, periodic and sinusoidal behavior without treating these categories as synonyms or inferring infinite repetition from a finite record.
  • Extract extrema, peak-to-peak value, midpoint, period average, period and frequency with explicit definitions and observation intervals.
  • Decompose a periodic signal as x(t)=x̄+x̃(t) with x̃(t) having zero period average, while keeping this time-domain decomposition distinct from later phasor/Fourier representations.
  • Construct a sinusoidal model X0+A cos(ωt+φ) only when waveform evidence supports it and state the cosine phase convention/time origin.
  • Cross-check an extracted sinusoid using extrema, period/frequency, initial value, slope direction, average and RMS rather than relying on a single graphical cue.
  • Recognize failure cases such as clipping, drift, multi-tone/beating, non-sinusoidal periodic waveforms and finite-window ambiguity.
  • Produce a reproducible signal-description record that links waveform parameters, estimation method, instrument context and uncertainty to the measurement evidence.
Course synthesis

A time plot can look simple while supporting several incompatible interpretations. The engineering task is not merely to “read amplitude and frequency,” but to separate what is observed, what is calculated, and what is assumed by a model.

This final chapter turns SG-01 into a reusable workflow for hand calculations, oscilloscope work, simulation output and later AC analysis.

05.1Keep three layers separate: physical quantity, mathematical model, measurement record

LayerQuestion
physical quantityWhat voltage/current/other quantity exists, between which references and under which circuit conditions?
mathematical modelWhat function x(t) are we using to represent that quantity?
measurement recordWhat finite data did the instrument actually return?

A sinusoidal equation can be exact as a mathematical model while the measured trace is noisy, bandwidth-limited or clipped. Conversely, a trace that looks sinusoidal does not prove that the physical source is an ideal sinusoid outside the observation window.

Figure 5.1 — Signal description has an order

The one question is: at which step are we allowed to choose a sinusoidal model?

1 · Identify2 · Measure / read3 · Choose model4 · Validate

First identify the quantity, reference and observation window; then extract descriptors; only then choose a model; finally validate that model against the data.

05.2Step 1 — identify the quantity before reading numbers

Before extracting any parameter, write down:

  1. the physical quantity: voltage, current, displacement, power, etc.;
  2. its reference: terminals, polarity/current direction, node/reference point;
  3. vertical and horizontal units;
  4. the chosen time origin;
  5. the observation interval or acquisition window;
  6. whether the source is a mathematical expression, simulation output or instrument record.

Without these items, even a numerically correct “5” or “2 ms” can be physically meaningless.

05.3Step 2 — classify conservatively

Ask the classification questions in increasing order of commitment:

  1. Is the quantity constant over the stated interval?
  2. If not, how does it vary: monotonic ramp, pulse, oscillatory, noisy, drifting?
  3. Does the available evidence support repetition?
  4. If repetition is supported, is the waveform actually sinusoidal or merely periodic?

Do not label a square wave “sinusoidal” because it has a frequency. Do not label one pulse “periodic” because another similar pulse appears nearby. Do not infer infinite repetition from a finite screen capture.

05.4Step 3 — extract vertical descriptors before fitting a waveform family

For a bounded waveform over a stated interval or one period, determine when useful:

xmax,   xmin,   xpp=xmax−xmin

Then compute the time average if it is relevant and the record/model supports it. For a periodic signal:

x̄=(1/T)∫t0t0+Tx(t)dt

A useful time-domain decomposition is

x(t)=x̄+x̃(t)
(1/T)∫t0t0+Tx̃(t)dt=0

Here x̃(t) means the zero-mean variation around the average. This is still an ordinary real time function, not a phasor or Fourier coefficient. Later courses may call the average the DC component and the zero-mean variation the AC part in a stated context.

05.5Step 4 — extract temporal descriptors only after repetition is justified

If the model is periodic, estimate or use the fundamental period T0, then

f0=1/T0
ω0=2πf0

For measured data, prefer multiple-cycle timing when appropriate:

T̂=(tn+m−tn)/m

State what landmark was timed: rising center crossing, peak, trigger edge, etc. Different landmarks can react differently to noise or distortion.

Phase is introduced only after a sinusoidal model or another explicit periodic reference has been selected. A phase angle without a reference waveform/time origin and convention is incomplete.

05.6Step 5 — construct a sinusoidal model only when the evidence supports it

Suppose a voltage record shows approximately: maximum 5 V, minimum −1 V, successive maxima at 0.25 ms and 1.25 ms, and at t=0 the trace crosses 2 V while rising.

From the extrema:

X0=(5+(−1))/2=2 V
A=(5−(−1))/2=3 V

From successive maxima:

T=1.00 ms
f=1.00 kHz
ω=2000π rad/s

At t=0, the trace is at its center and rising. Under the AoN cosine convention this corresponds to φ=−π/2. A candidate model is therefore

v(t)=2 V+3 V·cos(2000πt−π/2)

The word candidate matters for measured data. We still need to check whether the trace actually follows this model within the expected measurement uncertainty and distortion.

05.7Step 6 — validate with independent checks, not the same cue twice

For the candidate model above, check:

  1. extrema: 2±3 V gives 5 V and −1 V;
  2. period: 2000π rad/s gives 1 ms;
  3. initial value: v(0)=2 V;
  4. initial slope: positive, matching the observed rising crossing;
  5. average: 2 V over a complete period;
  6. total RMS: √(2²+3²/2)≈2.915 V.

If raw data are available, inspect the residual

r(t)=xmeas(t)−xmodel(t)

A structured residual — clipping near peaks, slow drift, another oscillation — is evidence that the chosen model omits relevant behavior. A small residual alone does not prove physical truth, but it is a stronger check than visual similarity alone.

05.8Failure cases: when the four-parameter sinusoid is the wrong model

Observed featureWhy a single sinusoid can fail
flat tops / bottomsclipping or saturation changes waveform shape and creates non-sinusoidal distortion
slowly changing centeroffset drift makes X0 time-dependent over the record
slow envelope / beatingmore than one nearby frequency may be present
triangular or square shapeperiodic does not imply sinusoidal
only part of one cycle visibleperiod/frequency/phase can be weakly constrained by the window
rounded or delayed traceinstrument/probe/source bandwidth may alter amplitude and phase

The correct response is not to force a prettier fit. Expand the model or improve the measurement only when the engineering question requires it.

05.9A reproducible signal-description record

For AoN's future Real Measurement Library, a time-domain record should preserve at least:

  1. quantity, nodes/reference direction and units;
  2. time origin and acquisition window;
  3. instrument, channel, probe/attenuation, coupling and relevant bandwidth settings;
  4. sample/timebase information and raw data file when available;
  5. observed extrema/mean/period landmarks and the method used to estimate them;
  6. the chosen model, its assumptions and parameter convention;
  7. fit/check residuals or independent consistency checks;
  8. uncertainty or at least the dominant resolution/specification limitations;
  9. script/version/provenance when automated processing is used.

This is the bridge between SG-01 and MS-02: waveform mathematics becomes measurement evidence only when the acquisition context is documented.

05.10Mastery synthesis and handoff to AC-01

  1. A trace ranges from −4 V to +8 V. Find midpoint and peak-to-peak value. Can you infer its time average from extrema alone?
  2. A record shows 12 repeated cycles in 3 ms. Estimate period and frequency.
  3. A periodic signal has average 2 V. Define its zero-mean part x̃(t) and state its period average.
  4. A sinusoidal-looking trace has maximum 7 V, minimum 1 V and T=200 μs. Find X0, A, f and ω.
  5. The trace in exercise 4 is at X0 and falling at t=0. Give the cosine phase and write the model.
  6. For that model, compute the total RMS value.
  7. Name three observations that would make you reject a single-sinusoid model even if the waveform is periodic.
  8. Explain why an oscilloscope screenshot alone is insufficient for a reproducible phase measurement.
Solutions

1. Midpoint=2 V; peak-to-peak=12 V. No: the average also depends on how the waveform occupies values over time.

2. T≈3 ms/12=0.25 ms; f≈4 kHz.

3. x̃(t)=x(t)−2 V; its average over a complete period is zero.

4. X0=4 V, A=3 V, f=5 kHz, ω=10,000π rad/s.

5. Falling through the center corresponds to φ=+π/2 under the cosine convention: x(t)=4+3 cos(10,000πt+π/2) V.

6. √(4²+3²/2)=√20.5≈4.528 V.

7. Examples: clipping, triangular/square shape, drift, beating/multiple frequencies, non-repeating envelope.

8. Phase requires a time/reference convention and the acquisition/probe/channel timing context; a screenshot may omit those and does not preserve raw timing/uncertainty evidence.

Volume ITable of contents