Frequency Offset Estimation Using 1-Bit Quantization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional frequency offset estimating methods in digital signal processors face increased circuit size and reduced calculation speed due to floating-point division calculations, especially when dealing with large dynamic ranges, making it difficult to achieve high accuracy and prevent overflow or underflow.
Innovation Solution
A frequency offset estimating method that performs discrete Fourier transform on a reception signal, calculates an average power, sets a threshold, and applies 1-bit quantization to the frequency components, then calculates the centroid frequency by multiplying frequencies with quantized powers and dividing by the sum of powers, allowing the divisor to be a power of 2 or an integer, thereby reducing circuit size.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If floating-point division calculation is used for frequency offset estimation, then calculation accuracy is improved, but circuit size increases and calculation speed decreases
Solution Approach 1:
The patent changes the parameter of the divisor from a general floating-point value to a power of 2 or integer value. This parameter change allows the division operation to be performed more efficiently in hardware, reducing circuit complexity while maintaining sufficient estimation accuracy for practical applications.
Solution Approach 2:
The patent uses a simplified division approach that sacrifices some precision for significantly reduced hardware complexity. Instead of implementing a full floating-point divider, the system uses a simpler integer or power-of-2 division that is computationally cheaper and requires less circuitry, acceptable for the application's accuracy requirements.
2Measurement precision
If floating-point division calculation is used for frequency offset estimation, then calculation accuracy is improved, but calculation speed decreases
Solution Approach 1:
The patent changes the divisor parameter to a power of 2 or integer value, which enables faster division operations in hardware. Power-of-2 division can be implemented using simple bit-shifting operations rather than complex division circuits, significantly improving calculation speed while maintaining adequate precision.
Solution Approach 2:
The patent replaces the mechanical floating-point division operation with a simpler integer or bit-shift-based division. This substitution eliminates the need for complex floating-point arithmetic units and their associated timing overhead, resulting in faster execution suitable for real-time signal processing.
3Device complexity
If 1-bit quantization is applied to frequency components' powers, then circuit size is reduced, but measurement precision may be affected
Solution Approach 1:
The patent applies 1-bit quantization to the frequency components' powers, changing the parameter from multi-bit precision to single-bit representation. This drastic parameter reduction simplifies the hardware required for power representation and subsequent centroid calculation, while the frequency offset estimation remains sufficiently accurate for practical purposes.
Data Source
AI summary
When a circuit that calculates a frequency offset using a shape of a frequency spectrum is implemented by hardware, the circuit size can be reduced. A frequency offset estimating method for estimating the difference between a carrier frequency of a reception signal and the frequency of an output signal of a local oscillator includes performing a discrete Fourier transform on a reception signal previously sampled at a predetermined sampling frequency and outputting a frequency spectrum with a plurality of frequency components, calculating an average power of the frequency spectrum, calculating a threshold by adding a predetermined value to the average power or power obtained by multiplying the average power by a constant, performing 1-bit quantization on powers of the frequency components of the frequency spectrum based on the threshold, and calculating a centroid frequency by multiplying frequencies of the frequency components by powers of 1-bit quantized frequency components, calculating the sum of multiplied products, and dividing the sum of the products by the sum of the powers of the 1-bit quantized frequency components of the frequency spectrum.