Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

# Method for Obtaining the Mean from Multiple Measurements with Possible Errors

IP.com Disclosure Number: IPCOM000115810D
Original Publication Date: 1995-Jun-01
Included in the Prior Art Database: 2005-Mar-30
Document File: 2 page(s) / 82K

IBM

## Related People

Mahlbacher, JC: AUTHOR [+2]

## Abstract

Disclosed herein is an invention that provides a systematic approach to identifying and eliminating defective data in circuit boards using an algorithm that calculates a final mean based on all samples within the threshold of a reference mean based on the last two(2) samples. By successively removing samples, a more accurate norm is established. This algorithm also provides the additional feature of the ability to detect error conditions.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Method for Obtaining the Mean from Multiple Measurements with Possible
Errors

Disclosed herein is an invention that provides a systematic
approach to identifying and eliminating defective data in circuit
boards using an algorithm that calculates a final mean based on all
samples within the threshold of a reference mean based on the last
two(2) samples.  By successively removing samples, a more accurate
norm is established.  This algorithm also provides the additional
feature of the ability to detect error conditions.

In prior art, current algorithms throw out the minimum and
maximum reading in an effort to eliminate the defects from the
samples, this can result in a learned process deviation that allows
defects to go undetected.

Detection of defects in circuit boards can be identified
through the use of electrical measurements.  Theses measurements are
made on a sampling of products to determine the normal and expected
value and deviation in measurements.  A significant deviation in the
expected value will then indicate the presence of a defect.
Typically it is desirable to take readings from several products in
an attempt to learn empirically what this Mean value should  be and
to characterize the process deviation.  A problem exists that there
are defects in the sample data that result in a learned process
deviation that allows defects to go undetected.  This method, a
software algorithm, provides a robust means for identifying and
rejecting defect data.

Capacitive measurements are made on circuit boards, and more
generally interconnect devices, to identify opens and shorts.  In the
presence of shorts the capacitance will at least double and in the
presence of an open will at least be reduced by half.  The
manufacturing process by its very nature is controlled to minimize
process deviation.  A determination of characteristic capacitance of
the interconnect for each network is easily established and is a good
indication of the presence of a defect when it changes.

It is desirable to establish the normal capacitance reading by
taking measurements from several produc...