Module 02 / Signals & Modulation

Why Modulation Exists

Modulation does not manufacture information or make bandwidth vanish. It places and maps information so an antenna, channel, shared spectrum, transmitter, and receiver can attempt the same engineering job under stated constraints.

01 / 10

The impossible direct-to-antenna proposal

If an ideal amplifier supplies any voltage we ask for, can a 20 kbit/s sensor waveform go straight into a compact antenna and still provide selectable channels?

Illustrative engineering case: the fictional condition-monitoring node produces 20.0 kbit/s at the D0 application boundary. A rushed proposal connects that low-frequency electrical representation to a compact radiator, adds ideal gain, and promises that receivers will select different users later. Gain addresses none of the missing frequency placement, electrical-size, sharing, hardware, propagation, or authorization decisions.

Think about itWhich problems does ideal voltage gain solve: antenna electrical scale, user separation, spectral placement, propagation/hardware context, or all of them?
Answer

None of those system questions is solved by voltage gain alone. The amplifier can scale a signal at one reference plane; it does not choose an RF region, create channel separation, change wavelength, establish an allowed emission, or predict a propagation result.

Calculated schematic, not measured data. The lobes show placement and relative support only; they do not claim antenna efficiency, radiated power, occupied bandwidth, or a legal channel.
The first separation

Carrier choice asks where the waveform belongs. Mapping choice asks howinformation controls the waveform. Multiple access asks how users share resources. They interact, but they are not synonyms.

Evidence Illustrative failure constructed from the approved Path 02 case; wavelength is derived with exact c. No radiation result is asserted.

02 / 10

Baseband information and DC boundaries

What is lost when “baseband” is treated as a synonym for audio, slowness, or a square bit waveform?

Begin with the representation, not the medium. In this lesson, baseband is the information-bearing spectral content around zero frequency at the declared D0, D2, or D3 plane before translation to an RF carrier. It can be wide, fast, complex-valued, coded, or pulse-shaped. ITU-R SM.328-12 supplies the formal emission-vocabulary boundary; it does not define a complete digital-radio architecture for us.

D0 information ≠ one universal voltage waveformA bit sequence still needs a line code, symbol mapping, pulse shape, timing convention, level scale, and reference plane before it becomes a waveform.

Real interfaces also have low-frequency boundaries. Transformer coupling, bias networks, AC-coupled amplifiers, and many sensing or radiating structures cannot reproduce arbitrary DC and near-DC content at their output plane. An AC-coupled transformer is a useful analogy: moving information away from zero can cross a blocked low-frequency region. It is not an antenna model and says nothing about radiation efficiency.

Same nominal bit rate · different D0/D2 waveform consequences
Pattern or decisionTime behaviorSpectral consequence
Long run of ones under NRZNearly constant levelStrong DC / low-frequency content
Alternating 1010… under NRZPattern repeats every two bit periodsFundamental pattern rate is Rb/2, not automatically Rb
Transition-balanced line codeMore guaranteed transitionsChanges low-frequency content and timing observability
Smoothed pulse shapeFinite transition timeChanges spectral skirts/support without changing the information sequence
Common misconceptionRaw bits are a square wave whose fundamental is the bit rate.

Bits are decisions, not a unique continuous waveform. Pattern, line coding, mapping, pulse shape, and observation determine the spectrum. Even the alternating NRZ pattern repeats every two bit intervals, while a long constant run approaches DC.

Go deeperA zero-frequency label depends on the coordinate system

After ideal downconversion, a signal that occupied a passband around fc can again be described around zero as a complex envelope. “Around zero” names the current coordinate representation, not a claim that the physical RF carrier disappeared from the product.

Evidence Definition boundary: ITU-R SM.328-12 §1; waveform dependence and transform reasoning: P02-S1, Chapters 4–5.

03 / 10

Frequency translation preserves the recoverable envelope

What can carrier multiplication move without changing the information underneath?

Module 02.1 showed that multiplication in time becomes convolution in frequency. For the limited double-sideband suppressed-carrier example below, multiplication by a cosine creates two half-scaled copies of the baseband transform. This is a translation result, not a bandwidth reduction.

s(t)=m(t)cos(2πfct)S(f)=12[M(ffc)+M(f+fc)]s(t) = m(t)\cos (2\pi f_c t) \Longleftrightarrow S(f) = \frac{1}{2}[M(f - f_{c}) + M(f + f_{c})]m(t) is a real baseband waveform; fc exceeds its support so the translated copies are separable. The output is the ideal R0 waveform before RF filtering or hardware impairment.
Think about itRaise fc while holding m(t), its information rate, and its pulse shape fixed. Which quantities are invariant?
Answer

The D0 information rate, baseband spectral offsets, and baseband width are invariant. Absolute RF placement, vacuum wavelength, component/channel context, and absolute hertz error for the same ppm tolerance change.

Definition/derived diagram under ideal coherent timing, frequency, phase, multiplication, and low-pass filtering. The spectral image before filtering is part of the derivation, not a measured spur.
s(t)cos(2πfct)=12m(t)+12m(t)cos(4πfct)s(t)\cos (2\pi f_c t) = \frac{1}{2}m(t) + \frac{1}{2}m(t)\cos (4\pi f_c t)An ideal low-pass filter removes the term around ±2fc. Multiplying the retained baseband by 2 restores m(t). A phase/frequency mismatch changes this result; section 8 names that price.
Common misconceptionA higher carrier carries more bits or automatically reduces bandwidth.

Translation changes absolute placement. Capacity, information rate, and occupied support depend on other assumptions: mapping, coding, pulse shape, channel, noise/interference, error target, and implementation. They do not appear merely because fc increased.

Evidence Definition and derivation: P02-S1, Chapters 4–5, and MIT OCW Unified Engineering Signals and Systems modulation material.

04 / 10

Carrier frequency is a physical and institutional choice

Which consequences come from choosing the frequency region rather than the modulation name?

Carrier frequency sets the absolute electromagnetic and hardware context. It changes wavelength, electrical size, material/component behavior, oscillator error in hertz, and which propagation mechanisms and implementation technologies deserve study. It also places the waveform inside an institutional spectrum framework that must be checked for the actual jurisdictions, service, equipment category, emission, and operating conditions.

λ0=cfcΔfclock=fc×ppm×106\lambda _{0} = \frac{c}{f_{c}} \qquad |\Delta f_{\mathrm{clock}}| = f_{c} \times \mathrm{ppm} \times 10^{-6}Use c = 299,792,458 m/s. λ₀ is vacuum wavelength; λ₀/4 is an electrical-scale reference, not a finished antenna dimension. The ppm result is one reference’s absolute bound.
Derived fixture · exact c · information rate held at 20.0 kbit/s
FrequencyVacuum wavelengthQuarter wavelength20 ppm absolute error
10 kHz29.979 km7.495 km0.2 Hz
868 MHz345.383 mm86.346 mm17.36 kHz
2.450 GHz122.364 mm30.591 mm49.00 kHz

Required interpretation: these lengths do not predict finished antenna dimensions or efficiency. Materials, conductor and ground geometry, end effects, loading, matching, nearby objects, balance/common mode, loss, and radiation analysis remain unresolved.

Common misconceptionHigher frequency always has more propagation loss.

“More loss” is incomplete until the comparison holds something specific fixed. A free-space link with fixed isotropic gains, one with fixed physical apertures, and a product with frequency-dependent antennas are different problems. Path 06 owns the propagation and antenna-depth comparison; this lesson records the missing conditions.

Compliance boundary. 868 MHz, 915 MHz, and 2.45 GHz are illustrative orientation points only. They grant no authorization and make no claim about a named service, technology, channel, power, duty cycle, or market. Applicable rules are a later, jurisdiction-specific verification task.

Evidence Wavelength/ppm values are internal derivations. Electrical-size context: Balanis, 4th ed.; current antenna terminology cross-check: IEEE 145-2025 (active, published 2026-03-31).

05 / 10

Sharing spectrum and separating users

If several links use the same place and time, which resource lets a receiver tell them apart?

Translating a baseband signal makes frequency separation possible, but frequency is only one sharing resource. Time, code/sequence, space/beam, and polarization can also help a receiver distinguish intended energy. A modulation family describes how information changes a waveform; a multiple-access method describes how users share resources. One design may use both.

Illustrative separation model. Solid traces mark wanted support; dashed traces mark unspecified filter/skirt regions. No current band plan, mask, occupied bandwidth, or legal guard is implied.

Frequency · FDMA orientation

Place users in distinct frequency regions.

Filter skirts, guards, drift, and adjacent energy determine practical separation.

Time · TDMA orientation

Assign users different time intervals.

Timing error, burst ramps, guard time, and latency matter.

Code / sequence · CDMA orientation

Separate users by distinguishable sequences.

Correlation, timing, near–far behavior, and receiver processing matter.

Space / beam · SDMA orientation

Separate users by spatial response.

Array aperture, channel geometry, calibration, and mobility matter.

Polarization

Use field orientation as another discrimination resource.

Antenna orientation, axial ratio, scattering, and cross-polar response matter.
Think about itClassify five proposals: separate carriers, alternating slots, assigned sequences, different beams, and orthogonal field orientations.
Answer

They are frequency, time, code/sequence, space/beam, and polarization resources, respectively. Real separation also depends on waveform skirts, oscillator error, filters, timing, channel coupling, and receiver dynamic range.

Consequence

Ideal channel centers do not guarantee separability. A waveform can spill through filter skirts, drift across a guard, arrive outside a time slot, lose sequence orthogonality, or couple across beam and polarization boundaries.

Evidence Informative system orientation: P02-S2, introductory system material, and P02-S9, Chapters 1–4. No named multiple-access standard is evaluated.

06 / 10

Carrier degrees of freedom

Once the spectral region is chosen, which carrier properties can carry the mapping?

A real passband carrier offers amplitude and angle as controllable coordinates. A changing angle can be described through phase itself or through its time derivative—instantaneous frequency. The family names below identify what is controlled; they are not standards or product modes.

s(t)=A(t)cos[2πfct+ϕ(t)]s(t) = A(t)\cos [2\pi f_c t + \phi (t)]A(t) is real envelope amplitude at R0, fc is nominal carrier [Hz], and φ(t) is phase [rad]. The physical R0 waveform remains real and continuous in time in this model.
fi(t)=fc+(2π)1dϕ(t)dtf_{i}(t) = f_{c} + \frac{(2\pi)^{-1} d\phi (t)}{d t}This instantaneous-frequency expression requires a differentiable, consistently unwrapped phase over the interval of interest. Discontinuous ideal symbol models need a physically shaped interpretation.
Definition map · family names are not technology selections
Controlled quantityEquation termFamily orientationImmediate implementation question
AmplitudeA(t)AM / ASK; QAM also changes magnitudeA varying envelope may require linearity or deliberate distortion control.
Phaseφ(t)PM / PSK; QAM combines phase and magnitudeCoherent decisions require a phase/frequency reference estimate.
Instantaneous frequencyfc + (2π)⁻¹dφ/dtFM / FSKFrequency deviation spends support and receiver frequency knowledge.
Illustrative time-domain trajectories at the ideal mapping/shaping boundary—not I/Q coordinates, a constellation, or a measurement. Exact complex-envelope implementation is deferred to 02.3.
Common misconceptionA digital modulation waveform must be a square wave.

“Digital” describes a discrete information alphabet, mapping, and receiver processing. The transmitted RF voltage is still continuous in time; transitions are shaped by real bandwidth and hardware. Analog versus digital is not a test of whether the oscilloscope trace is smooth.

Evidence Definition and family orientation: P02-S2, Chapters 4–5, and P02-S9. I/Q sign/scaling remains explicitly deferred.

07 / 10

What a modulation family spends

Which resource becomes harder when a family appears to save another one?

A useful comparison names the exact support definition and then exposes the resources not represented by that width: energy per information bit at a stated error target, PA envelope and linearity, timing/frequency/phase knowledge, receiver complexity, acquisition latency, adaptation, coding, and channel behavior. Without those conditions there is no universal “most robust” ranking.

Rs=Rblog2MBNN=(1+α)RsR_{\mathrm{s}} = \frac{R_{\mathrm{b}}}{\log _{2}M} \qquad B_{\mathrm{NN}} = (1 + \alpha)R_{\mathrm{s}}Teaching proxy for full M-ary uncoded mapping with no framing/pilot overhead and an ideal combined raised-cosine response. BNN is total null-to-null support, not occupied, necessary, or legal channel bandwidth.
Informative comparison · no universal robustness rank
FamilyNamed support relationshipEnvelope orientationKnowledge priceEvidence needed next
AM / ASKNeeds a stated pulse or message bandwidth; analog DSB is 2Bm only under its modelVarying magnitudeTiming; carrier reference depends on detectorLinearity, threshold/dynamic range, and amplitude-channel evidence
Binary FSKB ≈ 2Δf + (1 + α)Rs · teaching support proxyIdeal mapper constant magnitude; filtering can vary itFrequency plus symbol timing; phase coherence may be avoidableΔf, tone separation, acquisition, filtering, and detector choice
M-PSKBNN = (1 + α)Rb/log₂M · uncoded/no overheadIdeal symbol magnitude constant; transitions need a shaped-waveform checkCoherent phase, frequency, and timingReference tracking, phase-noise allocation, pulse shape, and SNR target
M-QAMSame ideal RC support expression under the stated mapping assumptionsMagnitude variesCoherent phase, frequency, timing, and gainLinear PA/backoff, gain accuracy, waveform peaks, and SNR target

Heuristic boundary. “Fewer current conflicts” in the interaction means only that fewer entered ceilings, PA preferences, or evidence fields are presently in tension. It predicts neither range, efficiency, legality, battery life, BER, nor product throughput.

Common misconceptionConstant envelope means narrow bandwidth.

Envelope and spectral support are different properties. FSK can keep constant ideal magnitude while spending support through frequency deviation and transition shaping. A poorly shaped constant-envelope signal can still have wide skirts.

Common misconceptionHigher order always gives more product throughput.

Increasing M raises mapped bits per symbol only under a full mapping. Net throughput also depends on symbol rate, coding/framing/pilots, error performance, channel width, linearity, reference quality, and retransmission behavior. Some of those terms are deliberately absent here.

Go deeperCoherent and noncoherent are receiver assumptions

A coherent decision uses an estimated carrier phase/frequency reference aligned to the signal model. A noncoherent decision avoids some phase knowledge by comparing other observables, often changing bandwidth, energy, complexity, or latency. Those labels do not by themselves determine performance; the detector and channel assumptions must be named.

Evidence Informative trade axes: P02-S2, Chapters 4–5, and P02-S9, Chapters 1–4. Support expressions are derived under the conditions printed beside them.

08 / 10

Frequency and phase knowledge have a price

What happens when the receiver’s carrier and timing coordinates are not the transmitter’s?

The canonical reference tolerance makes the scale visible. At 2.450 GHz, ±20 ppm is ±49.0 kHz for one reference—larger than the future 13.5 kHz ideal raised-cosine null-to-null support of the 20.0 kbit/s QPSK teaching case. That comparison does not prove failure: acquisition and tracking are specifically designed to estimate and correct offsets. It does prove that the offset cannot be ignored when choosing a mapping and burst structure.

2.450×109Hz×20×106=49.0×103Hz2.450 \times 10^{9} \mathrm{Hz} \times 20 \times 10^{-6} = 49.0 \times 10^{3} \mathrm{Hz}Derived one-reference absolute bound. Relative transmitter/receiver error needs its own worst-case or statistical budget. The 13.5 kHz comparison is an ideal D3 support proxy, not an acquisition-range specification.

Static carrier error

A nearly constant offset in hertz over the burst

Acquisition range and frequency tracking

Phase noise

Time-varying phase around the nominal carrier

Short-term reference quality and loop bandwidth

Unknown carrier phase

An arbitrary phase at burst arrival

Coherent phase estimation or a detector that avoids it

Symbol timing error

The receiver samples away from the intended decision instant

Timing recovery, pulse shape, and preamble structure
Think about itCan a short low-latency burst use the same synchronization assumptions as a continuous link merely because both use QPSK?
Answer

No. A burst must spend time and energy on detection, frequency/phase acquisition, timing, and often a preamble before useful symbols can be interpreted. Continuous tracking may amortize that cost differently. The exact loops and preamble are not designed here.

Consequence for the recurring case

Manufacturing’s relaxed-clock request affects more than the oscillator line item. It can change acquisition range, preamble duration, estimator complexity, latency, and the set of mapping/detector candidates worth carrying forward.

Evidence Offset value is internally derived; coherent/noncoherent and synchronization orientation: P02-S2 and P02-S9. Loop performance is not modeled.

09 / 10

Make carrier and mapping decisions separately

Can another engineer tell which evidence supports the frequency region and which supports the mapping family?

Use two columns before opening an interactive model. A carrier-region argument should survive a change from FSK to QPSK. A mapping-family argument should survive a move between frequency regions unless a stated hardware, channel, or synchronization dependency couples them.

Decision column A

Carrier region

Hard constraints
Physical volume/interfaces, available components, applicable authorization, and required channel coexistence once verified.
Preferences
Integration convenience, common reference ecosystem, manufacturability, and reuse.
Unknown evidence
Antenna efficiency/pattern, propagation distribution, adjacent use, materials, regional rules, and production tolerance.
Decision column B

Mapping family

Hard constraints
Entered channel-width proxy, PA operating mode, required information rate/latency, and receiver knowledge that must be attainable.
Preferences
Lower complexity, relaxed acquisition, adaptation headroom, and reusable processing.
Unknown evidence
FSK Δf, pulse shape, coding/overhead, SNR/error target, phase noise, PA distortion, channel response, and detector design.
Static comparison before ranking · illustrative cases, not product recommendations
StudyKnown starting constraintWhat the carrier calculation establishesDefensible next action
Compact 2.45 GHz node30 mm product dimension; saturated PA preferredλ/4 = 30.591 mm is an electrical-scale warning, not an antenna resultKeep binary FSK and QPSK; verify antenna integration, filtering, clock acquisition, channel, and authorization
Sub-GHz range / airtime study868/915 MHz orientation; information rate held at 20.0 kbit/sLonger wavelength and lower absolute ppm error are known; range and airtime are notStudy antenna volume, jurisdiction, path distribution, permitted waveform, and duty/latency requirements before choosing
Wired low-IF lab linkConducted cable fixture; no radiating product claimAntenna-size motivation disappears, while translation can still separate channels or fit an AC-coupled pathChoose mapping from fixture bandwidth, reference sharing, linearity, and the measurement objective
Think about itIn the model below, raise only carrier while holding information rate, order, roll-off, and PA choice fixed. Explain every value that should remain fixed.
Answer

Symbol rate and mapping support remain fixed because they depend on Rb, M, and α—not fc. The carrier move changes λ₀, dimension/λ, absolute ppm error, and the unmodeled physical, channel, component, and institutional context.

Think about itThen change only QPSK to 16-QAM at fixed carrier. Which carrier outputs must remain fixed, and which mapping assumptions become more demanding?
Answer

Wavelength, dimension/λ, and absolute ppm error remain fixed. The uncoded symbol rate and ideal support proxy fall because log₂M rises, while magnitude variation, linearity, coherent gain/phase knowledge, SNR/error target, and waveform-quality evidence become more demanding.

Interactive · carrier-mapping-trade/1.0 · explanatory constraint proxy

Carrier & Mapping Trade Space

Change the carrier and mapping decisions independently. The ranked table orders current constraint friction; it does not choose a product waveform.

If carrier doubles while information and mapping stay fixed, what changes?

Choose a prediction before applying the carrier change.

1 MHz6 GHzEnter 1 to 6,000 MHz; press Enter or leave the field to clamp an out-of-range value.
Enter kbit/s · 1 to 20,000; press Enter or leave the field to commit.
Enter kHz · 10 to 40,000; press Enter or leave the field to commit.
PA constraint
Mapping family to inspect
Explicit model bound: positive and below carrier; blank is canonical.
Vacuum wavelength
122.364 mm
λ₀ = c / fc
Electrical-scale ratio
0.2452 λ
entered dimension / λ₀ · no efficiency verdict
Absolute reference error
±49.000 kHz
fc × ppm × 10⁻⁶
QPSK symbol rate
10.000 ksymbol/s
uncoded · no overhead
PSK / QAM support proxy
13.500 kHz
ideal RC null-to-null

Active cause: Canonical case: 2.450 GHz carrier, 20.0 kbit/s information, 1 MHz channel-width proxy, 30 mm product dimension, ±20 ppm reference, and saturated PA preferred.

Constraint ledger · selected family: PSK / QAM
ConstraintEntered boundaryCalculated or missing evidenceRule result
Product electrical scale30.0 mm maximum dimensionλ/4 = 30.591 mm; dimension/λ = 0.2452Quarter-wave reference exceeds the entered dimension; investigate geometry and loading.
Channel-width proxy1.000 MHzQPSK: 13.500 kHzSelected teaching support proxy is below the entered width; skirts and masks remain unmodeled.
Reference tolerance±20.0 ppm±49.000 kHz (9.80% of entered width for the ± span)Numerically below the entered width; acquisition and tracking remain unmodeled.
PA envelope preferenceSaturated preferredConstant magnitude at the ideal PSK mapper; practical pulse filtering can create envelope variation.No automatic conflict; the shaped waveform still needs verification.
Antenna / channel / authorizationNot suppliedMaterials, ground, loading, propagation, adjacent use, jurisdiction, and equipment category are absent.Deferred — no range, efficiency, legal-channel, or product conclusion.

Binding-constraint proxy: The rule set identifies one binding-constraint candidate: quarter-wavelength electrical-scale reference versus entered product dimension. Dimension/λ and λ/4 are electrical-scale references only. They do not predict antenna size, efficiency, match, gain, or pass/fail behavior.

Inspection rank · lower score means fewer conflicts or evidence gaps in the entered case, not better BER or range
RankCandidateSymbol rateSupport resultEnvelope at stated planeSynchronization orientationCurrent friction
1 (score 1)Binary FSK20.000 ksymbol/s27000.000000 Hz + 2ΔfNeeds frequency deviation Δf. The model keeps the result symbolic because no canonical FSK deviation is specified.Constant magnitude in the ideal symbol-mapper model; practical filtering can still vary the envelope.Frequency and symbol timing knowledge are needed; coherent phase may be avoided by some detectors.1 evidence gap
1 (score 1)QPSK10.000 ksymbol/s13.500 kHzIdeal raised-cosine total null-to-null support proxy under uncoded, no-overhead mapping assumptions.Constant magnitude at the ideal PSK mapper; practical pulse filtering can create envelope variation.Coherent carrier phase/frequency and symbol timing estimates are required.1 evidence gap
2 (score 3)Binary ASK20.000 ksymbol/sNeeds a declared pulse or message-bandwidth modelThis lesson does not assign a DSB or pulse-support number to ASK without another explicit bandwidth assumption.Varying magnitude at the ideal mapper; linearity or controlled distortion needs checking.Symbol timing is required; carrier phase requirements depend on the chosen detector.1 PA-preference conflict; 1 evidence gap

λ₀ = c/fc · |Δfclock| = fc(ppm)10⁻⁶ · Rs = Rb/log₂M · BNN = (1 + α)RsBinary FSK: B ≈ 2Δf + (1 + α)Rs. This is a transparent support proxy only. The ranking score is 4 per entered-width conflict, 2 per PA-preference conflict, and 1 per evidence gap; ties share a rank. It is not a performance, legality, BER, or range score.

Checked carrier fixtures and server fallback
Derived with c = 299,792,458 m/s · 20 ppm
CarrierVacuum wavelengthQuarter wavelengthAbsolute error
10 kHz29.979 km7.495 km0.2 Hz
868 MHz345.383 mm86.346 mm17.36 kHz
2.450 GHz122.364 mm30.591 mm49.00 kHz

Canonical mapping fallback: QPSK gives 10.0 ksymbol/s and 13.5 kHz ideal raised-cosine null-to-null support at α = 0.35. Binary FSK gives 20.0 ksymbol/s and the symbolic proxy 27.0 kHz + 2Δf because no canonical deviation is supplied.

Your shortlist and two separate rationales

Model boundary: D0 information enters a D2 symbol-mapping comparison; the raised-cosine support proxy represents a future D3 shaped waveform; carrier and clock results are stated at ideal R0. No gain, range, antenna efficiency, legal channel, occupied or necessary bandwidth, BER, battery life, acquisition loop, filter skirt, fading, coding, or protocol overhead is calculated.

Wavelength 122.364 mm. Absolute reference error 49.000 kHz. PSK or QAM support proxy 13.500 kHz.

Evidence The interaction is a deterministic derived/heuristic model, carrier-mapping-trade/1.0. Its default wavelength is 122.364 mm and its QPSK support proxy is 13.5 kHz.

10 / 10

Update the waveform decision record

Can the record preserve what 02.1 established while adding two rationales that cannot be swapped?

Version 2 retains the v1 observation and bandwidth questions. It adds an illustrative carrier candidate, a family shortlist, binding constraints, preferences, and deliberately missing evidence. Nothing in this record selects a standard, guarantees a link, or authorizes transmission.

V1 · information / observation
D0 finite framed 20.0 kbit/s payload bursts; D2 teaching observation remains defined separately from the physical RF product.
V1 · spectrum / bandwidth decision
Two-sided convention retained. Filter response, local calculated support, noise bandwidth, channel width, occupied bandwidth, and mask remain distinct fields.
V2 · carrier rationale
Retain 2.450 GHz only as an illustrative candidate. λ₀ = 122.364 mm and λ₀/4 = 30.591 mm expose a product-scale integration question; components, channel, and authorization still require evidence.
V2 · mapping-family shortlist
Carry constant-magnitude binary FSK and coherent QPSK—not named standard modes. FSK preserves a saturated-PA option but lacks Δf; QPSK gives 10.0 ksymbol/s and 13.5 kHz ideal RC support but spends coherent reference knowledge.
V2 · comparison boundaries
1 MHz channel-width proxy; α = 0.35; saturated PA preferred; ±20 ppm reference; uncoded/no framing or pilot overhead in the rate proxy.
V2 · hard constraints vs preferences
Hard once verified: physical envelope, applicable channel/rules, information/latency target. Preferences: saturated PA and relaxed clock. Do not promote a preference into a physical law.
V2 · explicit deferrals
Antenna geometry/efficiency/pattern, propagation/range, allocation/limits, exact channel width, FSK deviation, pulse shaping, coding, synchronization loops, BER, and battery life.
V2 · first source-verification claim
Verify whether the intended product, operating mode, and jurisdictions permit the proposed frequency/channel/emission conditions. No current allocation is claimed here.
Five-sentence recommendation · illustrative model answer
  1. The source remains a 20.0 kbit/s D0 information stream, and its v1 observation and bandwidth definitions remain unchanged.
  2. I retain 2.450 GHz only as a carrier-region candidate because its 30.591 mm vacuum quarter-wave scale exposes a product-integration study, while antenna, channel, propagation, and authorization evidence are still missing.
  3. I shortlist binary FSK for its ideal constant-magnitude mapping under a saturated-PA preference and coherent QPSK for its 10.0 ksymbol/s rate and 13.5 kHz ideal raised-cosine support proxy.
  4. The binding inputs are the 1 MHz channel-width proxy, 30 mm product dimension, saturated-PA preference, and ±20 ppm reference; FSK Δf and shaped-waveform PA behavior can still reverse the shortlist order.
  5. The first external claim to verify is lawful use for the actual jurisdictions and equipment configuration, followed by antenna integration, channel/propagation, synchronization, and measured waveform quality.

Swap test: sentence 2 cannot justify FSK or QPSK; sentence 3 cannot justify 2.450 GHz. If the rationales can be swapped without becoming wrong, the record has mixed the decisions.

Handoff to 02.3 · planned

Carrier translation puts the waveform in the intended spectral region. Independent amplitude and phase control now need an explicit coordinate system; 02.3 will introduce I/Q and the complex envelope without changing the decisions recorded here.

Evidence Illustrative decision record derived from the approved recurring case. It deliberately separates calculated proxies from unverified product and regulatory claims.

Ungraded review

Check your understanding

Answer each question in your own words, then reveal the model answer.

  1. 01Why does raising carrier frequency not raise the information rate by itself?
    Model answer

    Carrier multiplication translates the spectrum to another absolute frequency region. If the D0 information sequence, D2 mapping, symbol rate, and pulse shape are unchanged, the information rate and baseband support are unchanged too.

  2. 02What absolute frequency error corresponds to ±20 ppm at 2.450 GHz?
    Model answer

    |Δf| = 2.450 GHz × 20 × 10⁻⁶ = 49.0 kHz, so one reference can lie within a ±49.0 kHz bound. A two-ended link needs a declared relative-error budget rather than silently doubling or cancelling this value.

  3. 03Classify these concerns as carrier-region or mapping-family concerns: antenna scale, coherent phase reference, symbol geometry, and a national band rule.
    Model answer

    Antenna electrical scale and the applicable national band rule belong primarily to the carrier-region decision. Coherent phase reference and symbol geometry belong primarily to the mapping-family decision, although the complete radio couples the two through hardware and channel constraints.

  4. 04Correct the statement: ‘QPSK transmits four voltage levels.’
    Model answer

    QPSK maps pairs of bits to four ideal phase states at D2. The R0 signal is a continuous real RF waveform; filtering and implementation create trajectories between states. Four symbol states are not four instantaneous antenna-terminal voltage levels.

  5. 05A saturated PA is strongly preferred. Make a defensible family choice without pretending there is one universal winner.
    Model answer

    Binary FSK is a defensible shortlist candidate because its ideal mapper has constant magnitude, but its frequency deviation and resulting support proxy must be supplied. A constant-magnitude PSK candidate can remain on the shortlist if the chosen pulse shaping and PA operating point preserve acceptable waveform quality. A varying-magnitude QAM choice needs a linearity argument and is harder to defend under this preference.

  6. 06Name at least two items that the v2 record must defer rather than invent.
    Model answer

    Examples include antenna geometry and efficiency, propagation and range distribution, applicable allocation and legal limits, channel model, coding/framing overhead, acquisition-loop design, exact filter skirts, BER, battery life, and named technology modes.

Sources and further study

The lesson cites and paraphrases these references. Version-sensitive records were rechecked on 5 September 2026: ITU-R SM.328-12 remains in force, and IEEE 145-2025 is active. All diagrams are original calculated or illustrative teaching graphics; none is measured evidence.

Signals and digital communications

  • P02-S1 · Informative: A. V. Oppenheim, A. S. Willsky, and S. H. Nawab, Signals and Systems, 2nd ed., Chapters 4–5.
  • P02-S2 · Informative: J. G. Proakis and M. Salehi, Digital Communications, 5th ed., Chapters 4–5.
  • P02-S9 · Informative: B. Sklar, Digital Communications: Fundamentals and Applications, 2nd ed., Chapters 1–4.
  • P02-S3 · Informative/open: MIT OpenCourseWare: Unified Engineering Signals and Systems, Fourier-transform and modulation materials.

Terminology and antenna boundary

  • P02-S4 · Normative terminology: ITU-R SM.328-12, Spectra and bandwidth of emissions, approved 2025-09-01 and in force when checked.
  • ANT-1 · Informative: C. A. Balanis, Antenna Theory: Analysis and Design, 4th ed., 2016, for wavelength/electrical-size context—not a product antenna prediction.
  • ANT-2 · Normative terminology boundary: IEEE 145-2025, IEEE Standard for Definitions of Terms for Antennas, active and published 2026-03-31.
  • Derived: wavelength, quarter-wave, ppm error, symbol-rate, and support-proxy fixtures follow the equations printed in the lesson and are regression tested.
Previous in Signals & ModulationSeeing Signals in Time & Frequency
Next in Signals & Modulation · plannedI/Q & the Complex Envelope