A topic in the Open Knowledge Graph — a free, open map of 15,290 topics and the order to learn them in.
Complex baseband representation decomposes a real bandpass signal (modulated at carrier frequency fc) into I (in-phase, real) and Q (quadrature, imaginary) components at baseband via mixing with cosine and sine. This representation halves the required sampling rate compared to passband sampling, enables efficient digital processing, and simplifies modulation/demodulation. The analytic signal (Hilbert transform output) naturally produces I-Q components through multiplication by complex exponential.
Take a modulated signal and generate I-Q components by mixing with cos and sin at carrier frequency, then lowpass filtering. Verify that I-Q signal is complex baseband representation of original. Recover original by upconverting I-Q.
You've already worked with amplitude, frequency, and phase shift keying — modulation schemes that encode information by varying some property of a high-frequency carrier wave. In all of these schemes, the information lives in a relatively narrow band of frequencies centered on the carrier frequency fc, which might be hundreds of MHz or GHz for wireless systems. The complex baseband representation is a mathematical transformation that strips the carrier away and leaves just the information-bearing part, shifted down to zero frequency. Everything that matters about the signal — its modulation, its noise, its distortion — is captured in the baseband representation at a fraction of the original bandwidth.
To extract the I (in-phase) and Q (quadrature) components from a real bandpass signal x(t), multiply x(t) by cos(2πfct) and by sin(2πfct) separately, then lowpass filter both products to remove double-frequency terms. The result is two real-valued signals: I(t) and Q(t). Together they form the complex baseband signal x̃(t) = I(t) + jQ(t). The I component is in-phase with the carrier; the Q component is 90° out of phase (quadrature). These are orthogonal projections, not polar coordinates — a common source of confusion. Magnitude and phase of x̃(t) — that is, √(I² + Q²) and arctan(Q/I) — describe the instantaneous amplitude and phase of the modulated signal, but I and Q themselves are the Cartesian components.
The motivation for this representation is sampling efficiency. The bandpass signal x(t) centered at fc with bandwidth B must be sampled at fs > 2(fc + B/2) to satisfy Nyquist — extremely high when fc >> B. For example, a 5 MHz wide signal riding a 2.4 GHz carrier requires passband sampling above 4.805 GHz. The complex baseband signal x̃(t) has bandwidth B (centered at zero), requiring only fs > B — about 5 MHz for the same example. This 1000× reduction in required sample rate is why every modern digital radio (WiFi, 4G/5G, GPS, Bluetooth) processes signals in complex baseband. The hardware downconverter mixes with cosine and sine at the carrier frequency and lowpass filters, producing I and Q streams at baseband before the analog-to-digital converter — exactly the mathematical operation above, implemented in RF circuits.
The representation makes modulation and demodulation algebraically transparent. BPSK maps bits to x̃(t) = ±1 (Q always zero). QPSK maps two bits to x̃(t) ∈ {1+j, −1+j, −1−j, 1−j}. QAM-64 places 64 points in the I-Q plane, each encoding 6 bits. All of this arithmetic happens in complex baseband; the carrier is reintroduced only at the final upconversion stage before transmission. When you see a constellation diagram — the scattered dots used to diagnose signal quality in digital communications — you are literally looking at a plot of I(t) vs. Q(t) sampled at symbol times. The spread of the dots relative to the ideal constellation points directly reveals noise power, phase error, amplitude imbalance, and other impairments, making complex baseband the natural language for communications system analysis and debugging.