Random Number Generator Bias Removal via Partial Sum
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating random numbers using signal sources with deviations result in biasing, where certain values occur more frequently than others, reducing the utility of the generated numbers, and existing solutions either introduce unintended DC deviations or require complex circuits and signal processing.
Innovation Solution
The method involves generating raw bit strings by sampling physical signals from signal sources and performing block or sequential partial sums to produce unbiased random numbers, either by processing individual signals or combining multiple signals, thereby alleviating biasing without complex analysis or feedback circuits.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If feedback control or signal processing methods are used to remove biasing, then biasing is reduced, but device complexity increases
Solution Approach 1:
The patent extracts the biasing effect by separating it from the random number generation process. By using a simple counter that tracks deviations from expected 50/50 distribution and independently adjusts thresholds, the biasing mechanism is isolated and corrected without affecting the core random number generation circuitry, thus reducing overall device complexity.
Solution Approach 2:
The invention employs a simple, lightweight counter mechanism that operates independently of the main random number generation circuit. This disposable-like counter simply tracks cumulative deviations and triggers threshold-based corrections, avoiding the need for complex continuous feedback systems while effectively removing biasing.
2Manufacturing precision
If complex signal processing algorithms are used to remove biasing, then biasing is reduced, but processing time increases
Solution Approach 1:
The patent performs preliminary action by pre-establishing threshold values and counter mechanisms before random number generation occurs. The counter is initialized to track deviations from the expected 50/50 distribution, and thresholds are set in advance. This allows real-time correction without requiring complex algorithms during the actual number generation process, minimizing processing time.
Solution Approach 2:
The invention skips complex signal processing steps by using a direct counter-based approach. Instead of performing lengthy analysis or multiple processing stages, the system rapidly tracks cumulative deviations and applies immediate threshold-based corrections, rushing through the bias removal process in minimal time.
3Manufacturing precision
If DC filtering is applied to remove average value deviation, then biasing is reduced, but unintended DC deviation may be introduced
Solution Approach 1:
The patent implements feedback by continuously monitoring the cumulative distribution of generated random numbers through a counter that tracks deviations from the expected 50/50 split. When the counter detects that a threshold has been exceeded, it triggers a correction mechanism that adjusts the sampling threshold. This closed-loop feedback ensures uniform distribution while avoiding unintended DC deviation through precise, data-driven control.
Solution Approach 2:
The invention changes parameters dynamically by adjusting sampling thresholds based on real-time counter values. Instead of using fixed DC filtering, the system modifies the comparison threshold against a reference value based on accumulated statistical deviations. This parameter adaptation allows the system to correct biasing while maintaining the intended DC characteristics of the signal.
Data Source
AI summary
The present disclosure provides a method and an apparatus for removing biasing due to a signal source during random number generation. At least one embodiment includes a method, performed by a random number generation apparatus, for generating a random number having a signal source biasing removed, including: generating a raw bit string by sampling, at a predetermined cycle, a physical signal input from a signal source; and generating the random number by performing block partial sum or sequential partial sum on the generated raw bit string.