BER Test Method Using Dynamic Statistical Boundaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current statistical tests for devices under test (DUTs) can only be optimized by two parameters: Test time and Confidence level, without consideration for Selectivity, leading to inefficiencies and variable decision quality, especially in determining Bit Error Ratio (BER) and distinguishing between good and bad DUTs.
Innovation Solution
The method involves constructing early pass and early fail limits using empirically or analytically derived distributions to separate good and bad DUTs, ensuring a high probability of passing DUTs with a specified BER while failing slightly worse DUTs, by simulating a large population of DUTs and adjusting decision probabilities to optimize test time and selectivity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional statistical tests are used to determine BER, then test simplicity is maintained, but test quality and selectivity are insufficient
Solution Approach 1:
The patent transforms the traditional single-parameter BER threshold test into a multi-parameter statistical test involving confidence levels, selectivity factors, and dynamic decision boundaries. By changing the test parameters from simple pass/fail thresholds to statistically derived boundaries based on Poisson distributions, the measurement precision and test quality are significantly improved while maintaining manageable complexity through systematic parameter management.
Solution Approach 2:
The patent introduces statistical distributions (Poisson distribution) and derived parameters (decision boundaries, confidence intervals) as intermediaries between the raw error counts and the final pass/fail decision. These intermediaries enable more nuanced and accurate BER assessment by incorporating statistical reasoning, thereby improving test quality without requiring fundamentally complex test architecture.
2Measurement precision
If more samples are tested to improve BER determination accuracy, then measurement precision improves, but test time increases
Solution Approach 1:
The patent implements early decision mechanisms that allow the test to terminate before completing the full sample set when sufficient statistical evidence has been gathered. By using dynamically adjusted decision boundaries and confidence levels, the test can achieve adequate BER determination accuracy with fewer samples than traditional methods, thereby reducing test time while maintaining measurement precision.
Solution Approach 2:
The patent performs preliminary statistical analysis to establish decision boundaries and confidence intervals before the actual BER measurement begins. This preliminary action enables the test to make informed decisions during execution, allowing early termination when results are statistically significant, thus avoiding unnecessary testing time while preserving accuracy.
3Measurement precision
If traditional pass/fail criteria are used, then test simplicity is maintained, but selectivity between good and bad DUTs is insufficient
Solution Approach 1:
The patent applies different decision criteria and statistical parameters to different regions of the BER distribution. By establishing separate early pass boundaries, early fail boundaries, and continuation conditions based on local statistical characteristics, the test achieves high selectivity in distinguishing between good and bad DUTs. Each region of the decision space has optimized quality characteristics tailored to its specific statistical properties.
4Reliability
If a fixed number of samples is used, then test time is controlled, but the ability to make confident decisions is reduced
Solution Approach 1:
The patent transforms the static fixed-sample test into a dynamic adaptive test where the sample size and decision boundaries adjust based on observed error rates and statistical confidence. The test continuously evaluates whether sufficient confidence has been achieved to make a reliable decision, allowing early termination when confidence is high and extending testing when more evidence is needed, thus optimizing both reliability and test time.
Data Source
AI summary
A method for testing the error ratio BER of a device under test against a specified allowable error ratio comprises the steps: measuring ns samples of the output of the device, thereby detecting ne erroneous samples of these ns samples, defining BER(ne)=ne/ns as the preliminary error ratio and deciding to pass the device, if the preliminary error ratio BER(ne) is smaller than an early pass limit EPL(ne). The early pass limit is constructed by using an empirically or analytically derived distribution for a specific number of devices each having the specified allowable error ratio by separating a specific portion DD of the best devices from the distribution for a specific number of erroneous samples ne and proceeding further with the remaining part of the distribution for an incremented number of erroneous samples.