Sigma-Delta OFDM Bitstream Filtering for Low-Power Transmission
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing neural networks, particularly those used in IoT and edge devices, require analog-to-digital converters (ADCs) for operation, which complicates the integration of digital neural networks and introduces challenges in managing weighted addition and connectivity between layers, especially in analog implementations like convolutional neural networks (CNNs).
Innovation Solution
The system encodes OFDM data into a sigma-delta modulated bitstream, filters it to remove high-frequency content, and transmits it at a reduced data rate using a cascaded integrator-comb (CIC) filter, allowing for efficient data transfer and accurate recovery of OFDM symbols through FFT processing, utilizing analog circuits with transistor configurations that multiply and rotate complex numbers to maintain signal integrity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If OFDM data is transmitted at full data rate without filtering, then data transfer speed is maintained, but power consumption increases and signal integrity deteriorates due to parasitic effects
Solution Approach 1:
The patent applies a low-pass filter to the sigma-delta modulated bitstream before transmission to remove high-frequency content above the highest frequency OFDM tone. This preliminary filtering action reduces the bandwidth of the transmitted signal, which in turn reduces power consumption and minimizes parasitic effects during transmission, while still preserving all necessary data for accurate OFDM symbol recovery at the receiver
Solution Approach 2:
The patent changes the frequency spectrum parameters of the transmitted signal by filtering out high-frequency components above the OFDM band. This parameter change reduces the signal bandwidth to match the actual data rate requirements, thereby reducing power consumption and improving signal integrity without losing essential information
2Measurement precision
If analog circuits are used for FFT calculations in neural networks, then computational accuracy improves by preserving signal integrity, but circuit complexity increases
Solution Approach 1:
The patent replaces digital signal processing operations with analog circuit implementations for FFT calculations. Analog circuits naturally perform weighted additions and multiplications through current summation and transconductance, eliminating the need for complex digital arithmetic units and ADCs, thereby achieving high computational accuracy while managing circuit complexity through elegant physical analogies
3Ease of manufacture
If external ADCs are used in digital neural networks, then signal conversion is achieved, but integration complexity increases and signal integrity is compromised
Solution Approach 1:
The patent merges the ADC function with the neural network computation itself by using analog circuits to perform weighted additions and multiplications directly on analog input signals. This integration eliminates the need for separate external ADCs, simplifying the overall system architecture and improving signal integrity by avoiding additional conversion stages and their associated errors
Data Source
AI summary
An example system for encoding OFDM data may comprise a processor and a transmitter. The processor may be configured to encode OFDM symbols into a first sigma-delta (ΣΔ) modulated bitstream, such that spectral content of the first sigma-delta (ΣΔ) modulated bitstream contains tones at subcarrier frequencies corresponding to the encoded symbols. The transmitter may be configured to transmit the sigma-delta (ΣΔ) modulated bitstream to a receiver.