×
About Doppler & MTI
How a pulse radar sees speed, and how it uses speed to pick a moving target out of the echoes from land, sea and rain. Set the radar's frequency (pick a NATO band or type one) and its pulse rate along the top: PRF in pulses a second, or its PRI , the time between pulses in microseconds (each box follows the other, like the Python script PRF PRI Converter.py that sits beside this app). Then pick one of four modes on the key panel. Speeds are in knots, ranges in nautical miles. Keys 1 to 4 pick the modes.
SHIFT : a target's speed and its aspect (0° heading straight for the radar, 90° crossing, 180° going straight away). Only the part of its speed along the line of sight, the radial speed, makes a Doppler shift. The right-hand side shows what the radar actually gets: one sample of the echo's phase per pulse, turning by the same angle each time.
BLIND SPEEDS : the response of a two-pulse and three-pulse canceller against radial speed, its blind speeds, and staggered PRFs that fill the gaps. CANCELLER : a simulated A-scope, one sweep before and after the canceller, with clutter whose spread comes from the wind and the aerial turning. AMBIGUITY : what the PRF does to unambiguous range and unambiguous speed, and how several PRFs together pin down the true range.
On the screen, yellow is always the plain, single-PRF picture and cyan what the radar makes of it: in SHIFT the speed along the line of sight (yellow, the only part that counts) and the echo's phase sample at each pulse (cyan), with the slower wave the samples seem to follow, when the shift is more than half the PRF, in magenta; in BLIND SPEEDS one steady PRF (yellow) against the staggered train (cyan), its first true blind speed in magenta and the target as a green dot; in CANCELLER the sweep before (yellow) and after (cyan) the canceller, the target marked by a green arrowhead.
THE STORY
Christian Doppler, an Austrian physicist, proposed in 1842 that the colour of a star should change with its motion towards or away from us. He was wrong about the colours of double stars, but right about the effect. In 1845 C. H. D. Buys Ballot tested it with sound: musicians on an open railway carriage near Utrecht played a steady note while trained listeners by the line heard it rise as the train came and fall as it left.
Radar echoes are Doppler shifted too, by a tiny amount: a few hundred to a few thousand hertz on a carrier of thousands of millions. A pulse radar can't measure that in one pulse; it compares the echo's phase from pulse to pulse. During the Second World War, radar engineers in Britain and America built the first moving target indicators (MTI): each sweep was stored for exactly one pulse interval, in a mercury delay line, and subtracted from the next. Hills, buildings and the sea return the same echo every time and cancel; an aircraft has moved a little and doesn't. Digital memories replaced the delay lines from the 1970s, and the same subtraction became one sum among many in a radar's signal processor.
The catch, found straight away, was blind speeds: a target that moves exactly half a wavelength (or a whole number of half-wavelengths) between pulses looks stationary and is cancelled too. Staggering the pulse intervals moves the blind speeds far out; a textbook example uses four intervals in the ratio 25 : 30 : 27 : 31, which pushes the first true blind speed out to 28¼ times that of a single PRF. The Chinese remainder theorem, used here to untangle range with several PRFs, goes back to the Chinese mathematical text Sunzi Suanjing , written between the 3rd and 5th centuries AD: a number leaves 2 when divided by 3, 3 when divided by 5 and 2 when divided by 7; what is it? (23.)
HOW IT WORKS
The shift. The echo travels out and back, so a target closing at radial speed vr shortens the path by 2vr every second: the shift is fd = 2vr /λ. Between pulses, T = 1/PRF apart, the echo's phase turns by 360° × 2vr T/λ. The radar sees only one sample per pulse, so it can't tell a turn of 30° from 390°: shifts fold into the band from −PRF/2 to +PRF/2, and speeds fold into ±λ·PRF/4. (The exact two-way shift is f·2β/(1 − β) with β = vr /c; at any real speed the difference is about one part in a million.)
Cancellers. A two-pulse canceller takes each pulse minus the one before: its power response is 4 sin²(π fd T), zero at fd = 0 and at every multiple of the PRF, so the blind speeds are v = nλ·PRF/2. A three-pulse canceller (weights 1, −2, 1) is the two-pulse one twice over: 16 sin⁴(π fd T), with a wider, deeper notch round zero. The curves here are drawn against the peak (0 dB). With staggered intervals the page averages the power response over the stagger cycle, the usual way of drawing it; the first speed where every interval holds a whole number of Doppler cycles is the first true blind speed.
Clutter. Real clutter isn't still. Its echoes spread over a band of speeds, taken here as a Gaussian with a standard deviation from two causes added in quadrature: the clutter's own motion (trees swaying, waves, rain blowing), and the aerial sweeping past, which chops each echo into a burst of n pulses and widens its spectrum by σs = 0.265 × PRF ÷ n (from a Gaussian beam). The motion figures are typical values from radar textbooks: wooded hills about 0.02 m/s when calm up to about 0.3 m/s in a gale; rain 2.5 m/s (1.8 to 4 m/s is usually quoted); for the sea the page uses a rough rule of a tenth of the wind speed, in line with the 0.5 to 1 m/s measured in moderate to fresh winds. The page centres all clutter on zero speed; real sea and rain clutter drift with the wind as well, which a fixed canceller handles worse.
Improvement factor is the signal-to-clutter ratio after the canceller divided by the ratio before, averaged over all target speeds. For Gaussian clutter with correlation ρ = exp(−2π²σ²T²) between pulses it is 1/(1 − ρ) for a two-pulse canceller and 6/(6 − 8ρ + 2ρ⁴) for three, close to the textbook 2(PRF/2πσ)² and 2(PRF/2πσ)⁴. The A-scope is a simulation: random clutter with exactly that correlation, receiver noise at 0 dB and the target, and the page measures the improvement it actually got. The clutter levels along the trace are illustrative, not a model of any place. The target's echo is set 25 dB above the receiver noise.
The radar's own limit. However still the clutter, a canceller can't remove more than the radar's own pulse-to-pulse steadiness allows: changes in the transmitter's phase and frequency, oscillator noise, timing jitter and the A/D converter leave a residue that doesn't cancel. The residues add, so the improvement you get is 1/(1/Iclutter + 1/Ilimit ). The page offers typical figures of the kind radar textbooks quote, a rough guide rather than any particular set: about 30 dB for a magnetron radar (each pulse starts at a random phase, which the receiver has to lock to), about 45 dB for a coherent amplifier chain (klystron or travelling-wave tube driven from a stable oscillator), and about 60 dB for a solid-state transmitter with a digital receiver. In the simulation the limit is applied as random pulse-to-pulse phase jitter on the echo, sized to give exactly that limit.
Ambiguity. An echo must be home before the next pulse leaves to be read at the right range: Ru = c/(2·PRF). The speed limit grows with the PRF: ±λ·PRF/4. Multiplied together they make c·λ/8, whatever the PRF, so a radar that needs both long range and high speeds has to give one up: low PRF keeps range clear and folds speed, high PRF keeps speed clear and folds range, medium PRF folds both and untangles them with several PRFs. In the demo the range is cut into 1 µs cells (150 m); each PRF's interval is a whole number of cells, the numbers sharing no factor, so the folded cells fix the true one uniquely out to their product. With many targets at once pairs of PRFs can line up by chance (ghosts), which is why real medium-PRF radars use more PRFs than strictly needed and accept a target only when enough of them agree. The page also leaves out eclipsing, where an echo arrives while the radar is transmitting and can't hear it.
FURTHER WATCHING AND READING