01Executive summary
Broadcasting, television and radar are three uses of the same physics, and each one forced a different piece of engineering discipline into existence.
Regular commercial broadcasting began around 1920, turning point-to-point wireless telegraphy into one-to-many distribution. John Logie Baird demonstrated televised moving images in 1926 by mechanical scanning, and a regular high-definition electronic service began in 1936. Radar detection of an aircraft by reflected radio energy was demonstrated in 1935, with comparable programmes running independently in several countries at the same time. Broadcasting created spectrum management, television created bandwidth engineering, and radar created the entire field of signal detection in noise.
02Broadcasting and the invention of spectrum management
Wireless telegraphy was conceived as a substitute for a cable: one sender, one receiver, a message between them. Broadcasting inverted the model. One transmitter served an unlimited number of receivers who neither identified themselves nor replied. The receiver could therefore be made simple and cheap, which is what made the medium mass.
The engineering problem that followed was not technical in the usual sense. Radio spectrum is a finite shared resource, and two transmitters on the same frequency in the same area destroy each other’s service. There is no technical fix available to an individual operator, because the interference is caused by someone else. This is the classic structure of a common-pool resource, and it required allocation, licensing and enforcement — an administrative solution to a physical constraint.
Asymmetric cost placement
Putting complexity and cost in the transmitter, where there is one, rather than the receiver, where there are millions, is why broadcast systems scale. The same reasoning governs content delivery networks and read-optimised database design.
Shared resources need governance
Where one user’s behaviour degrades another’s service and no local remedy exists, coordination is not bureaucracy but a necessary part of the engineering. Spectrum, radio channels, network bandwidth and shared airspace all work this way.
Modulation as a design choice
Amplitude modulation varies carrier amplitude with the signal. It is simple to demodulate, which suited early receivers, but noise is predominantly amplitude noise, so the signal degrades audibly. Frequency modulation varies carrier frequency instead, occupies more bandwidth, and in exchange rejects amplitude noise almost entirely. This is a bandwidth-for-quality trade, and it is one of the earliest clear instances of a general principle in communications: spectrum, power and fidelity can be exchanged for each other, and the designer chooses the exchange rate.
03Television: bandwidth as the governing constraint
Television has a genuinely contested history. Baird demonstrated working televised moving images in 1926 using a mechanical scanning disc, and achieved the first public demonstrations and early broadcasts. Fully electronic systems, with contributions from Philo Farnsworth, Vladimir Zworykin and the EMI research group among others, produced the higher-definition service that began in 1936 and superseded the mechanical approach. Both lines were necessary; neither is adequately described as the invention of television.
A rotating disc with a spiral of apertures scans the image one line per aperture. Increasing resolution requires more apertures, so each is smaller and admits less light, and the disc must spin faster. Light throughput falls as resolution rises, and mechanical speed becomes impractical. The limitation is inherent to the method, not to its execution — which is the distinguishing mark of an approach that must be replaced rather than improved. Electronic scanning has no moving part and no such coupling.
The deeper constraint is bandwidth. A moving image is a stream of picture elements, and the required channel bandwidth scales with lines times lines times frames per second. Doubling resolution in both axes quadruples the data rate. Every design decision in analogue television was a manoeuvre within this budget.
| Technique | What it exploits | Modern equivalent |
|---|---|---|
| Interlaced scanning | Perceived flicker depends on refresh rate more than on full-frame rate | Temporal subsampling and frame interpolation in video codecs |
| Reduced colour resolution | Human vision resolves luminance detail far better than colour detail | Chroma subsampling in essentially all digital video |
| Vestigial sideband | Both sidebands carry the same information | Single-sideband and spectrally efficient digital modulation |
| Compatible colour encoding | Colour added within the existing monochrome channel | Backwards-compatible layered encoding and scalable video |
The colour case is the most interesting as engineering. Colour had to be added without obsoleting existing monochrome receivers, so the colour information was encoded onto a subcarrier positioned so that monochrome sets would ignore it. That is a backwards-compatibility constraint solved in the frequency domain, and it produced a system measurably worse than a clean-sheet design would have been. Compatibility with an installed base almost always costs technical quality; the judgement is whether the installed base is worth more than the quality.
04Radar: detection in noise
Radar transmits energy, and measures the small fraction reflected by a target. Range comes from time of flight, bearing from antenna directivity, and radial velocity from Doppler shift. The engineering difficulty is entirely in the smallness of the returned signal.
Energy spreads over a sphere on the way out, so power density at the target falls with the square of range. The echo spreads over a sphere on the way back, falling with the square of range again. Received echo power therefore varies as the inverse fourth power of range. Doubling detection range requires sixteen times the power, all else being equal — which is why radar progress came from receiver sensitivity, antenna gain and signal processing rather than from transmitter power alone.
Pulse length: range against resolution
A short pulse resolves closely spaced targets but carries little energy. A long pulse carries energy but smears range. Pulse compression resolves the conflict by transmitting a long coded pulse and correlating on receive, obtaining both.
Wavelength: resolution against weather
Shorter wavelengths give narrower beams from a given aperture and better angular resolution, but attenuate more in rain and require higher-precision hardware. Band selection is an application decision, not a performance ranking.
Threshold: detections against false alarms
Setting the detection threshold low finds weak targets and generates false alarms; setting it high does the reverse. There is no setting that avoids both, only a chosen operating point on a curve.
Integration: sensitivity against update rate
Averaging successive returns lifts the signal out of noise, but takes time and assumes the target has not moved. Every sensitivity gain by integration costs responsiveness.
Radar established that detection is a statistical decision, not a measurement. The question is never simply whether a signal is present, but whether the evidence is strong enough to declare it, given the cost of being wrong in each direction. That framing — a threshold, two error types and an explicit operating point — is now standard in condition monitoring, medical screening, quality inspection, fraud detection and machine learning. It began as a problem of finding aeroplanes.
The system, not the sensor
The frequently repeated claim that radar won the Battle of Britain overstates the device and understates the engineering. Detection was useful only because it fed a system: a chain of stations, filter rooms that fused and de-duplicated tracks, a plotting and reporting structure, and communications to direct aircraft. The sensor was necessary but the integration was decisive. This is one of the earliest examples of a deliberate system-of-systems design, and it belongs with Apollo as a milestone in engineering organisation rather than in engineering physics.
05Takeaways for current practice
- Place cost where the count is low. Complexity in the single transmitter, not in a million receivers.
- Identify limits that are inherent rather than executional. Mechanical scanning coupled resolution to light throughput; no amount of refinement would have fixed it.
- Backwards compatibility is a priced decision. It reliably costs technical quality, and that cost should be stated rather than absorbed silently.
- Treat detection as a decision with two error types. Ask what each costs before setting a threshold.
- A sensor is not a capability. The processing, fusion and response chain usually contributes more than the device.
Australian references relevant to this part include the Radiocommunications Act and ACMA spectrum planning instruments, AS/NZS CISPR standards for electromagnetic compatibility, and AS/NZS 61000 for EMC limits and techniques. Cited by number for orientation only — verify currency.