Automotive RF

DDM-MIMO

Q: How does DDM-MIMO separate the transmit channels without losing time on the chirp sequence?

Each transmitter is given a unique progressive phase shift that advances by a fixed increment from chirp to chirp. For Tx number k out of Nt transmitters, the phase increment is typically 2 pi k / Nt radians per chirp. After the slow-time Doppler FFT, this phase ramp shifts that transmitter's energy to a distinct Doppler sub-band, so the Nt channels appear at evenly spaced velocity offsets within the unambiguous Doppler window. Because all transmitters fire on every chirp, no chirps are spent cycling through transmitters one at a time, which is the time penalty that time division multiplexing MIMO pays.

Q: What is the maximum unambiguous velocity penalty in DDM-MIMO and how is it recovered?

With Nt transmitters sharing one Doppler window, the usable unambiguous velocity per channel drops by a factor of Nt because the window is partitioned into Nt sub-bands. For 4 transmitters that means roughly one quarter of the single-transmitter unambiguous velocity. Designers recover range by leaving an empty guard sub-band so a target's true Doppler can be disambiguated by checking which sub-band stays vacant, or by adding a slightly non-uniform phase increment so the aliasing pattern is unique. TDM-MIMO does not have this partitioning loss but it reduces velocity unambiguity through a longer effective chirp spacing instead.

Q: Why does DDM-MIMO raise the noise floor compared with a single-transmitter measurement?

When Nt transmitters share the total power budget and the same pulse repetition interval, each transmitter's energy lands in only one of the Nt Doppler sub-bands while the others contribute to interference in the remaining sub-bands. The per-channel signal-to-noise ratio therefore falls by about 10 log10(Nt) dB relative to a single transmitter using the full power, roughly 6 dB for 4 transmitters and 9 dB for 8. This is offset by the larger virtual array gain and by the fact that DDM keeps the full coherent processing interval, so the frame can be longer for a given frame rate.

/dee-dee-em my-moh/ · Doppler Division Multiplexing MIMO

A transmit-coding scheme for automotive MIMO radar in which every transmitter fires on every chirp at the same time, but each is assigned a unique linear phase ramp that advances by 2πk/N_t radians per chirp. After the slow-time Doppler FFT, those ramps shift each transmitter's return into a distinct, evenly spaced Doppler sub-band, so the N_t channels can be sorted out and combined into the full virtual array. Unlike time division multiplexing, DDM spends no chirps cycling between transmitters, which preserves the coherent processing interval and supports the high frame rates that 77 GHz FMCW corner and imaging radars need. The cost is an N_t-fold split of the unambiguous-velocity window and an SNR penalty of about 10log₁₀(N_t) dB per channel.

Category: Automotive RF

Band: 76 to 81 GHz

Phase step: 2πk/N_t per chirp

How Doppler-Coded Transmit Separation Works

In a MIMO radar the number of independent angle measurements scales with the product of the transmit and receive element counts, because each transmit-receive pair forms one element of a synthetic virtual array. The challenge is making the N_t transmitted signals separable at every receiver so the processor can reconstruct which transmitter produced each echo. DDM solves this in the Doppler domain. All transmitters emit identical chirps simultaneously, but the phase of transmitter k is advanced by a fixed increment Δφ_k = 2πk/N_t from one chirp to the next. A constant phase progression across slow time is mathematically equivalent to a frequency offset, so after the Doppler FFT each transmitter's energy is parked at a different Doppler bin.

Because the Doppler axis is periodic, the unambiguous velocity span is partitioned into N_t equal sub-bands, one per transmitter. A real moving target also carries its own Doppler shift, which slides every channel copy by the same amount. The processor recovers the true velocity either by reserving one empty guard sub-band (the vacant slot reveals the alias offset) or by using a slightly irregular set of phase increments so the wrapping pattern is unique. This is the core trade against time division multiplexing MIMO, where transmitters are pulsed one at a time and the velocity ambiguity instead comes from the stretched effective chirp interval.

DDM is attractive for imaging radars with 12 or more transmitters because the alternative TDM approach would multiply the frame time by N_t, smearing fast-moving targets. Binary phase coding (BPSK-MIMO) and slow-time outer codes such as Hadamard sequences are related simultaneous-transmit methods; DDM is the special case where the codes are pure complex exponentials, which keeps the separation a single FFT bin shift rather than a correlation step.

Governing Relations for DDM Channel Spacing

Per-chirp phase increment (transmitter k):
Δφ_k = 2πk / N_t (k = 0, 1, …, N_t−1)

Unambiguous velocity (single Tx):
v_max = λ / (4 × T_c)

Per-channel velocity window in DDM:
v_max,DDM = v_max / N_t

Per-channel SNR penalty:
ΔSNR ≈ −10 log₁₀(N_t) dB

Where λ = wavelength (≈ 3.9 mm at 77 GHz), T_c = chirp repetition interval, N_t = transmitter count. Example: 77 GHz, T_c = 50 μs → v_max ≈ 19.5 m/s; with N_t = 4, v_max,DDM ≈ 4.9 m/s and the per-channel SNR drops ≈ 6 dB.

DDM Versus Other MIMO Multiplexing Schemes

Scheme	Transmit timing	Frame-time cost	Velocity ambiguity	Per-channel SNR	Best fit
DDM (Doppler division)	All Tx simultaneous	None (full CPI kept)	Split by N_t	−10log₁₀(N_t) dB	Imaging / corner radar, many Tx
TDM (time division)	One Tx per chirp	×N_t longer	Reduced by N_t	Full (one Tx at a time)	Low Tx count, slow scenes
BPSK / code division	All Tx simultaneous	None	Full	Code cross-talk limited	High dynamic range needs
FDM (frequency division)	All Tx simultaneous	None	Full	Full per band	Wide instantaneous bandwidth

Common Questions

Frequently Asked Questions

How does DDM-MIMO separate the transmit channels without losing time on the chirp sequence?

Each transmitter gets a progressive phase shift that advances by Δφ_k = 2πk/N_t per chirp. After the slow-time Doppler FFT this ramp parks that transmitter at a distinct Doppler sub-band, so all N_t channels appear at evenly spaced velocity offsets. Because every transmitter fires on every chirp, no chirps are wasted cycling through transmitters one at a time, which is the penalty TDM-MIMO pays.

What is the maximum unambiguous velocity penalty in DDM-MIMO and how is it recovered?

With N_t transmitters sharing one Doppler window, usable unambiguous velocity per channel drops by N_t; for 4 transmitters that is about one quarter of the single-Tx value. It is recovered by leaving an empty guard sub-band so the vacant slot reveals the alias offset, or by using slightly non-uniform phase increments so the wrap pattern is unique. At 77 GHz with T_c = 50 μs, that takes v_max from ≈ 19.5 m/s to ≈ 4.9 m/s per channel.

Why does DDM-MIMO raise the noise floor compared with a single-transmitter measurement?

When N_t transmitters share the power budget and pulse repetition interval, each one's energy lands in only one Doppler sub-band while the others spill into the rest. Per-channel SNR therefore falls by about 10log₁₀(N_t) dB, roughly 6 dB for 4 transmitters and 9 dB for 8. The larger virtual-array gain and the preserved coherent processing interval offset much of this loss.

Related Terms

← DCI Format DDS + PLL →