Defect screening method for electronic circuits and circuit components using power spectrum anaylysis

1. A method involving non-destructive testing of a sample device having a specified function in the context of electronic circuitry, comprising:

applying one or more off-normal stimuli to the sample device, wherein each off-normal stimulus is a periodic electrical signal, devoid of logical information, that is applied between a power pin and a ground pin of the device while each other pin is electrically connected to a power pin, is floating, or is biased at a constant voltage;

under each of the one or more off-normal stimuli, measuring at least one power spectrum at a selected pin of the sample device, wherein the measured power spectrum represents a response of the sample device to the respective off-normal stimulus, and wherein each power spectrum is taken over a multiplicity n of spectral bins;

performing a Principal Component Analysis (PCA) on each power spectrum, thereby to obtain a respective set of principal components of each power spectrum;

for each power spectrum on which a PCA has been performed, selecting at least one subset of the set of principal components, wherein the subset consists of fewer than n of the principal components;

comparing the subset or subsets obtained from the respective power spectra on which a PCA has been performed to stored reference data, thereby to obtain a comparison result, wherein the stored reference data include representations in terms of principal components of one or more reference populations of devices having the same specified function as the sample device;

classifying the sample device relative to reference populations, wherein the classification is based at least in part on the comparison result; and

placing the sample device in a batch of devices confirmed for further processing if the sample device is classified as similar to a reference population of acceptable devices, but placing the sample device in a batch for disposal if it is not classified as similar to a reference population of acceptable devices.

2. The method of claim 1 , wherein each selected subset of a respective set of principal components consists of no more than ten of the principal components.

3. The method of claim 1 , wherein each selected subset of a respective set of principal components consists of two or three of the principal components.

4. The method of claim 1 , wherein the classifying step comprises assigning the sample device to one of the represented populations.

5. The method of claim 1 , wherein the classifying step comprises determining either that the sample device is imputable to one of the reference populations, or that the sample device is imputable to none of the reference populations.

6. The method of claim 1 , wherein the classifying step comprises classifying the sample device as a member of either an accepted category or a rejected category, and physically segregating the sample device with other sample devices of the same category.

7. The method of claim 6 , further comprising integrating the sample device into a product on condition that the sample device is accepted.

8. The method of claim 1 , wherein:

each measured power spectrum is provided to a computer;

the computer performs the PCA on the measured power spectrum or spectra; and

the comparing step is performed in the computer using reference data obtained by the computer from a digital storage medium.

9. The method of claim 1 , wherein, in regard to at least one set of principal components:

the selecting step comprises selecting at least two different subsets of the set of principal components, each selected subset consisting of fewer than n of the principal components; and

the comparing step is repeated in respect to each of the selected subsets, thereby to produce a respective comparison result for each selected subset.

10. The method of claim 3 , wherein the classifying step is performed with joint reference to all of the respective comparison results, thereby to produce a classification that is jointly responsive to all of the respective comparison results.

11. The method of claim 9 , wherein the classifying step is repeated in respect to each of the respective classification results so as to classify the sample device with respect to a reference population at least twice, with each resulting classification being based at least in part on a different one of the respective comparison results.

12. The method of claim 1 , wherein:

the method further comprises calculating at least one statistical parameter of at least one designated power spectrum of the sample device; and

in the classifying step, the classification of the sample device is based in part on a comparison between the at least one calculated statistical parameter and stored reference data.

13. The method of claim 12 , wherein the at least one calculated statistical parameter comprises at least one of a mean and a standard deviation of the at least one designated power spectrum.

14. The method of claim 1 , wherein:

the stored reference data include information concerning the statistical distributions of individual principal components in the reference population or reference populations; and

in the classifying step, the classification is based in part on a comparison between at least one principal component of a power spectrum of the sample device and stored reference data pertaining specifically to the same principal component.

15. The method of claim 1 , wherein:

the method further comprises calculating at least one statistical parameter of a power spectrum of the sample device, thereby to provide at least one spectral statistical parameter calculation;

the stored reference data include spectral statistical parameter distribution information concerning the statistical distribution, in the reference population or reference populations, of the at least one statistical parameter of a power spectrum of the sample device; and

in the classifying step, the classification is based in part on a comparison between the spectral statistical parameter calculation and the spectral statistical parameter distribution information.

16. The method of claim 1 , further comprising:

applying at least one normal stimulus to the sample device, wherein the normal stimulus is an electrical signal applied to an input pin of the device so as to place the device in one or more known functional states;

under the normal stimulus, measuring at least one power spectrum at a selected pin of the sample device, wherein the measured power spectrum represents a response of the sample device to the normal stimulus, and wherein the power spectrum is taken over a multiplicity n of spectral bins;

performing a Principal Component Analysis (PCA) on the normal-stimulus power spectrum, thereby to obtain a normal-stimulus set of principal components of the normal-stimulus power spectrum;

for the normal-stimulus power spectrum, selecting at least one subset from the set of normal-stimulus principal components, wherein the resulting normal-stimulus principal component subset consists of fewer than n of the principal components; and

comparing the normal-stimulus principal component subset to at least part of the stored reference data, whereby the comparison result is based, in part, on the normal-stimulus principal component subset.

17. The method of claim 1 , wherein:

the classifying of the sample device relative to reference populations is performed while the sample device is on a wafer among a plurality of similar devices on the wafer;

the method further comprises dicing the wafer;

the placing of the sample device in a batch of devices confirmed for further processing is performed after dicing the wafer; and

the placing of the sample device in a batch for disposal is performed after at least partially dicing the wafer.

18. The method of claim 17 , wherein:

the wafer is subdivided into blocks, each containing a plurality of devices;

the sample device is situated in one of the blocks, whereby all other devices in that block are the block neighbors of the sample device;

if the sample device is classified as similar to a reference population of acceptable devices, the sample device and its block neighbors are placed in the batch of devices confirmed for further processing; and

if the sample device is not classified as similar to a reference population of acceptable devices, the sample device and its block neighbors are placed in the batch for disposal.