In modern wireless communications, data is typically transmitted by modulating both the amplitude and the phase of a carrier wave. To decode this data, a receiver must know exactly how the wireless channel has altered the signal's phase along the way. Traditionally, systems solve this by wasting a significant portion of their bandwidth on "pilot symbols"—pre-agreed signals that allow the receiver to map out the channel.
However, as we scale to Massive MIMO, grant-free multiple access, and massive device networks, the overhead of sending these pilot signals becomes unsustainably high. This has driven researchers toward blind and semi-blind estimation, where the receiver attempts to reconstruct both the channel state and the transmitted data simultaneously from the received signal alone.
But blind estimation introduces a fundamental mathematical blind spot: phase ambiguity.
The Mathematical Blind Spot
To understand why phase ambiguity happens, imagine a basic received signal model where the observation $Y$ is the product of an unknown channel $H$ and an unknown data symbol $X$:
$$Y = HX + \text{noise}$$
If a receiver knows neither $H$ nor $X$, it faces an identifiability problem. Suppose the true channel is $H$ and the true symbol is $X$. If we rotate the channel by an arbitrary phase angle $\theta$ (multiplying it by $e^{j\theta}$) and rotate the data symbol by the exact opposite angle (multiplying it by $e^{-j\theta}$), the observed product remains completely unchanged:
$$Y = (H e^{j\theta})(X e^{-j\theta}) + \text{noise} = HX + \text{noise}$$
Because the receiver only observes $Y$, it cannot distinguish between a phase rotation introduced by the physical communication channel and one belonging to the transmitted data constellation. As noted by Abla Kammoun et al. in Fundamental limits via CRB of semi-blind channel estimation in Massive MIMO systems, purely data-driven channel estimation techniques inherently suffer from these rotational or phase ambiguities, making it impossible to uniquely determine absolute channel states without an external reference mechanism.
Without resolving this ambiguity, a receiver might successfully recover the shape of a signal constellation (like a QAM grid) but will have no idea how it is oriented. A 90-degree channel rotation could cause the receiver to mistake every "00" bit pair for a "01" bit pair, rendering the communication link useless.
Strategy 1: Minimal Pilot Anchoring
The most common way to break this mathematical symmetry is to introduce a tiny, low-overhead phase reference—a technique often called semi-blind estimation. Instead of using long pilot sequences that consume massive amounts of bandwidth, the transmitter sends just enough known data to act as a "phase anchor."
Traditional Pilot-Aided Frame:
[ Pilot ] [ Pilot ] [ Pilot ] [ Pilot ] [ Data ] [ Data ] [ Data ] [ Data ]
Semi-Blind Anchored Frame:
[Anchor] [ Data ] [ Data ] [ Data ] [ Data ] [ Data ] [ Data ] [ Data ]
The intuition here is simple: use a blind algorithm to do the heavy lifting of separating the signal from the channel up to an unknown rotation, and then use a single known point to snap the orientation into place.
- Single-Bit/Symbol Anchors: Tareq Y. Al-Naffouri et al. demonstrate in Low-Complexity Blind Equalization for OFDM Systems With General Constellations that a single pilot bit is sufficient to establish an absolute phase reference for complex QAM constellations. Similarly, Soung Chang Liew et al. in Sparsity Learning-Based Multiuser Detection in Grant-Free Massive-Device Multiple Access integrate a single reference symbol at the start of each packet. This single symbol allows the receiver to correctly map recovered symbols to the constellation by distinguishing channel-induced phase shifts from data modulation.
- Scaling to Complex Networks: This anchoring philosophy scales effectively to advanced architectures. Youngchul Sung et al. apply it to space-time block coding in Blind Channel Estimation for Space-Time Coded WCDMA, proving that a minimal number of pilots can achieve coherent detection with far less overhead than traditional training. For next-generation ultra-reliable low-latency communications (URLLC), Malong Ke et al. in Next-Generation URLLC with Massive Devices: A Unified Semi-Blind Detection Framework for Sourced and Unsourced Random Access combine a tiny amount of reference information with message-passing algorithms to resolve both phase and permutation ambiguities simultaneously.
Strategy 2: Sidestepping Phase via Differential Modulation
If a system requires absolutely zero pilot overhead, it can bypass the phase ambiguity problem entirely by changing how information is encoded. Instead of mapping data bits to an absolute phase position on a constellation diagram, the transmitter maps data to the difference in phase between the current symbol and the previous one.
If the channel rotates the entire signal by an unknown angle $\theta$, it rotates both consecutive symbols by the same angle $\theta$. When the receiver calculates the phase difference between them, the $\theta$ terms subtract out perfectly:
$$\Delta \phi = (\phi_{\text{current}} + \theta) - (\phi_{\text{previous}} + \theta) = \phi_{\text{current}} - \phi_{\text{previous}}$$
In Blind Grant-Free Random Access With Message-Passing-Based Matrix Factorization in mmWave MIMO mMTC, Fei Liu et al. leverage this property to solve the phase ambiguities inherent to blind matrix factorization. By utilizing differential modulation, the receiver decodes signals without ever needing to calculate the absolute phase of the channel, trading a minor non-coherent performance penalty for maximum spectral efficiency.
Strategy 3: Parameter Decoupling and Structural Constraints
A third approach relies on exploiting the physical and mathematical structure of the communication system to isolate or constrain the ambiguity.
Disentangling Channel Phase from Hardware Noise
Phase ambiguity becomes particularly destructive when multiple independent sources of rotation mix together. For example, an OFDM receiver faces both the initial phase shift of the wireless channel and the rapidly changing phase noise of its own local oscillators.
If an algorithm tries to estimate these jointly, they can easily confuse each other. As explored by Omar Hazim Salim et al. in Channel, Phase Noise, and Frequency Offset in OFDM Systems: Joint Estimation, Data Detection, and Hybrid Cramér-Rao Lower Bound, failing to resolve this parameter confusion causes errors to propagate rapidly through data detection. The authors resolve this by mathematically separating the initial channel phase from the starting phase noise sample, decoupling the parameters so the estimator can uniquely identify and compensate for each source of rotation.
Leveraging Sparsity and Manifold Optimization
In Massive MIMO systems, the physical environment often forces signals to travel along a limited number of spatial paths, creating structural sparsity. Algorithms can exploit this sparsity to narrow down the possible solutions for $H$ and $X$, reducing the ambiguity from an infinite continuum of rotations down to a discrete set of permutations.
- Manifold Constraints: Ye XUE et al. in Blind Data Detection in Massive MIMO via ℓ₃-Norm Maximization Over the Stiefel Manifold formulate data detection as an optimization problem over a Stiefel manifold. Their $\ell_3$-norm maximization approach inherently restricts the system's search space, isolating the uncertainty to a predictable phase-permutation ambiguity that can be cleanly swept away with a single common reference symbol and user IDs.
- Sparsity Learning: Similarly, Jianwen Zhang et al. (Blind Signal Detection in Massive MIMO: Exploiting the Channel Sparsity) and Xiaojun Yuan et al. (Super-Resolution Blind Channel-and-Signal Estimation for Massive MIMO With One-Dimensional Antenna Array; Double-Sparsity Learning-Based Channel-and-Signal Estimation in Massive MIMO With Generalized Spatial Modulation) use compressed sensing and hidden Markov models to enforce sparsity constraints. These constraints ensure robust signal recovery despite the lack of explicit pilot training.
- Pre-Assignment Workarounds: In specialized hardware environments like spaceborne phased arrays for LEO satellites, Wei Zhang 0191 et al. (Over-the-Air Phase Calibration of Spaceborne Phased Array for LEO Satellite Communications) manage the phase ambiguity between estimated channel state information and hardware phase deviations by pre-assigning specific values to anchor the calibration process.
The Fundamental Capacity Cost of Ambiguity
Phase ambiguity is not just an algorithmic nuisance that complicates receiver design; it represents a fundamental loss of information capacity.
This capacity penalty is most visible in reflective communication systems, such as those utilizing Reconfigurable Intelligent Surfaces (RIS). An RIS acts as an intelligent mirror, dynamically shifting the phase of incoming waves to steer them toward a receiver.
[Transmitter] ----(Phase θ_tx)----> [ RIS Mirror ] ----(Phase θ_ris)----> [Receiver]
(No Direct Path)
If there is no direct line-of-sight path between the transmitter and the receiver, the receiver only sees the combined phase shift:
$$\theta_{\text{total}} = \theta_{\text{tx}} + \theta_{\text{ris}}$$
Because the receiver cannot separate the phase contribution of the transmitter from the phase contribution of the reflecting surface, a global phase ambiguity emerges. As proven by Hei Victor Cheng et al. in Degree-of-Freedom of Modulating Information in the Phases of Reconfigurable Intelligent Surface, this inability to distinguish the two sources of phase modulation results in a strict 0.5 degree-of-freedom penalty on the system's capacity.
This theoretical limit underscores why resolving phase ambiguity remains a central focus of wireless research: whether through clever hardware calibration, structural optimization, or minimal pilot anchoring, pinning down the absolute orientation of a signal is the key to unlocking high-efficiency, pilot-free communication.
Key Foundational Work
Low-Complexity Blind Equalization for OFDM Systems With General Constellations (2012) — contributed 1 concept elements to this direction.
Blind Channel Estimation for Space-Time Coded WCDMA (2004) — contributed 1 concept elements to this direction.