CA2047426A1

CA2047426A1 - Product defect detection using thermal ratio analysis

Info

Publication number: CA2047426A1
Application number: CA002047426A
Authority: CA
Inventors: Eldon E. Cox, Jr.; Michael P. Rolla
Original assignee: Digital Equipment Corp
Current assignee: Digital Equipment Corp
Priority date: 1990-09-14
Filing date: 1991-07-19
Publication date: 1992-03-15
Also published as: JPH04282445A; AU626146B2; EP0475570A3; US5032727A; EP0475570A2; AU7818991A

Abstract

ABSTRACT OF THE DISCLOSURE
A method and apparatus for indicating defects in manufactured products employs, instead of the conventional thermal image subtraction, "thermal ratio analysis", which involves ratios of thermal data and their analysis including statistical analysis. Various techniques for "image" enhancement and for suppression of known artifacts are employed to facilitate the decision as to when a defect is detected. The thermal ratio analysis technique is particularly useful for detecting hidden defects in electronic circuitry, such as integrated circuits.

Description

1~ 2 ~ 2 ~
¦FIELD ~ TEB I ~ IO~
The pre~ent invention relate~ to product defect detection using ~hermaL analysis.
BACgGROUND OF TH~ I~V~TION
The early discovery of hidden defects in parts and products is of increa~ing concern to manufacturer~ a~ they strive to obtain ~uperior product quality Particularly, therQ i5 a need ~¦for the early discovery of defects which could remain latent, or l undi~covered, for an inde~erminate time.
i Thermography, or thermal analysi~, has attrac~ed ¦¦considerable recent attention as one way of discovering ~uch ¦ defect3. All ob~ect~ "glowl~ ~rom thermal radia~ion with an ¦ intensity and ~color~ which i dependent upon ~he temperature.
¦ ~t room temperatur~ thi~ "color" i~ within a range known a~
infrared and cannot be seen with th~ unaided eya. At extreme temperature~ an ob~ect will glow vic~ibly as in the ca~e of iron heated in a fire. This property can be l~ed to mea~ure the temperatur~ of a ~urface without need f or any kind of contact.
Any or ~everal type~ of equipment can conver~ this temp2rature infor~ation in~o a black and whi~e or color image that r~presents thQ temperature~ within th~ 3cene. Such equipm~nt can be called a "thermal im~ger" and can be u~ed to study non-vlsible prop3rtiQs of electronic as~emblia~ in th~ hop~ of locating defective devices.
r ~ rl~c~ , ~AN, HENDER50N
R.~O~, ~RREl~
~ DUNNER
00 t ~Tn~ W, C ~000!~
00~000 ~ `
? -.
, ` ' ' ' ' ~ ~ .
- -I1 2 ~ l~, 7 ~ 2 ~

~ t ha~ long been known that pattern~ of heatiny effects (e.g., pattern of the infrared glow) in a product may be affected by a latent defect; but the heating effect may not be readily detectable for some type~ of defect . Particularly in bipolar semiconductor circuit3, prior thermal analy~i~
thermography~) techniques have been only marginally effective in locating defect~, except in sertain limited ~ituations.
Frequently, the analy~i~ technique3 employed with such equipment involve elevating the temperature of the ob~ect for at least one of several image~. Then image3 obtainad undar diferent conditions are compared in an attempt ~o r~move averything from the image which i~ normal and leavQ only the imago featur~3 that relate to the defect. In the case of s~miconductor circuit~ or component~, the di~ferent conditions can be the normal powcred state and no~mal non-powered sta~e.
One of these prior analysis ~echniques i~ known a~ image subtraction. Gonerally, in this techniqu~ an image, compri~ing a regulax array of value3 rapres~nting infraxed radlation, is obtainad from a refer~nce ~ample, whlch i~ a high-quality sample of the p~oduct, and i~ ub~equently ~ubtracted ~rom a ~imilar image obtained f rom a tes~ sample, which is a sampla, of unknown quaIlty, of the product. The p~rpo~e i3 to remova feature3 from i the~ difference im~ge which are kno~ to b6~ normal, ~o as to increase tho likalihood that any re~iduo in ~he differance image, is indicatiYe of a def~ct in the product. Available thermal , -, ~A~l o~l~le~
~EC~N HENDE~N analy~is techni~ue~ use image ub~rac~ion in ene fo~m or another.l 8 DU~INE~
~OO~Ta~ For example, ee the description in the article by C.G. ~a~i, ooo I
_ 2 :
~ .

. ,, , , ~ , " . . ~ , :', '~ . ' ' ; ' . ' 2~4 ~42~
j "Finding Board Faults With Thermal Imaging", TEST AND MEASUREMENT
. WORLD~ ~arch, 1989 pp. lOO, 111, 112. The image ~ubtraction l technique in this article, a~ described in connection with a i' circui~ board, start~ with the board in a known thexmal state l~ (e.g., the entire board at 22C). The test operators then apply '¦ a given power source to the board and monitor changeR in the thermogram as the hoard hea~s up to operating temperature.
The basic phenomenon employed in image subtraction ¦ thermography for such productq i~ black-body radiation from ohmic heating of current-carrying traces and components, as explained in the ar~icle by C.G. Ma~ What Can Thermal Imaging Do For You?~ TEST AND MEASUREMENT WOR~D, May, 1988. It iB al~o known, however, that image subtraction thermography can be applied to non-curxent-carrying produc~s, to the extan~ such products can be ~ub~ected to con~rolled thermsl change~.
The monitoring of thermal changes by image ~ubtraction i~
I ba~ed on the premi~e that the therm~l change~ for a non-defective, product ~hould b~ different from ~he thenmal change~ for a defective product. Thu~, if ~he thermogr~m for a known i non-defectiv~ pxoduct (sample) i~ ~ub~racted fro~ the thenmogram for a psoduct ( or ~ample) being te~ted, ~he differenc~ any, are hoped to b~ indica~ive of a de~ect. Conversely, if a hidden defoct doe~ ex~8~ in a ~ample, it i~ hoped ~h~t i~ will produce a thermogra~ which is differ3nt frem the ~hermogr~m of the non-defective 3ample. The greate~t uccesse~ in the prior art ~E~N HENDE~N techniqu~g have be~n fo~ products which produca rslati~ly little ~R.'IBO\~. G~RRETT
~D~NER hQat. ~owever, products such a~ transi~tor-tran~istGr-logic 300 1 3 R~n~. N. ~V.
31 1NO ON~ OC ~0003 ~ OoO 1~ _ 3 -~` ~'' . ;~' , , , - : . .. : .

circuit3 which produce much heat have yielded marginal success in diagno3tic te3ting using pxior art techniques. The problem appears to be that the variability of heating among non-defec~ive (normal) sample3 can be much larger than the effect upon heating p~oduced by a sample having a subtle defect. Thi~ tendency make~
such defect difficult, if not impossible, to detect by previously known image subtraction technique~.
SUMNARY OF 'rE~ N~IO~
It is an object of the present invention to overcome ~he limitation~ of image subtraction thermography. It i3 desirabl~ ¦
to provide a method and apparatus ~o empha~iza tho~
thermographic difference~ between a refer0nce ~ample and a te~t ample which are likaly to be indicative of a defect, relative to differences which are not likely to be related to a defect.
Additional ob~ect~ and advantage3 of the invention will be set forth in part in the de~cription which follow~ and in part will be obviou~ from the descrip~ion, or may be learned by practice of the invention. The ob~ect~ and advantaye~ of the invention may be r~alized and obtainad by m~an~ of the instrumen~al$tiQ~ .~nd combinations particularly pointed out in the appended clalm~.
The invention i3 based upon the re~ogni~ion ~hat moro defect-sen8itiva an~ly8i8 can b~ achieved by obtaining a ra~io of ~ariable~ related ~o thermographic ro8pon~es to ~wo differing non-equilibrium thermal stimuli, appliad first to a pxeviously ~w orr~ct t~ l test~d product ("refare~ce ~-~mpl~"), known to b~ good, and then ~AgO~. ~ARRETT
~ ~DuNNER applied ~ubsequen~ly to a ~e~t Rample of the sama product. The 30Ct ~ ~TN~CT, N. W.
31<1NOTON. DC 2000 1~202 400-4000 ~ I _ 4 -~- .
.
:` ;; :

.;~ .

2 ~ ~ 7 L 2 ~ ~
I term "thermal ~timulu~" or l~thermal s~imuli~l as usad herein l refer~ to any ~timulus, not nececsarily thermal in origin, which when applied to a sample of a product, ultima~ely ha~ an effect ~ of changing the temperature thereof. A ~table ba e level, or I equilibrium, thermogram fox each sample also is involved in l determining the variables.
!I Four difference record~ relating ~o the reference sample and li the test 3ample are generated, and therefrom at least one ratio ¦¦ record is derived. A composite record is formed from the ratio record and the unu~ad difference record~; and a defect indication is generated when the compoxite record yi21d3 a stati tically 3ignificant deviation from an expected value.
Ths 3~atistical ba~is for such defect detec~ion will be discussed in detail later.
The technique of thi3 invention ha~ been ~ermed Thermal Ratio Analy~i~ (TRA). The invention reqide bo~h in a method and in test equipment for obtaining at lea~t one pattern of ~hPrmal ratios from a plurality of thermal difference record~ obtained from a reference ~ampla and rom a tes ~ample, and for forming a composit~ record including at lea~t one pattorn of thermal ratio~. Since the~ ~ample~ are like samples of the ~ame part or product, they will be termed a "reference 8ample" and a "te~t sample~.
Accordin~ ~o a prinoipal fea~ure of the i~vention, a method of detec~ing a defact in a te~t 8ampl~ of a manufactured device ~wO~,c~. i8 provided which compris~ the ~tep~ of:
~R~ RRETT
~ DtrNNER
300 t OTI~CI', N. W.
gNlNOTON. DC 20003 1~02'400-4000 ,_ 5 establishing a reference record for at least one reference sample of the device, comprising the sub-steps of:
(1) making a base thermal record of the reference sample at a base value of a thermal stimulus;
(2) making a plurality of changed value thermal records of the reference sample at a plurality of respective changed values, compared to the base value, of the thermal stimulus, including applying a respective changed-value thermal stimulus to the sample; and (3) making a first difference record from the base-value thermal records and one of the plurality of changed value thermal records and a second difference record involving at least another of the plurality of changed-value thermal records, the first and second difference records each comprising a plurality of data points in an image-related array;
generating a test record for the test sample by repeating sub-steps (1)-(3) with the test sample replacing the reference sample;
deriving at least one ratio record from the four difference records consisting of the first and second difference records for the test sample and the first and second difference records for the reference sample;
forming a composite record including at least the derived one ratio record; and generating a defect indication when the composite record yields a statistically significant deviation from an expected value.

?~7 ~2~ ~
l In the preferred embodiment, the thermal ~timuli which are ! changed in the ~teps of producing first and ~econd thermographs I are voltageY of first and second magnitudes that aro applied to the samples in such a way as to produce heating therein; and the l¦base-level thermograph is an ambient temperature thermograph.
! One embodiment is i~mediately applicable to electronic components on a populated circuit board and can be extended by l use of masking and filtering ~echniques, Yometimes called image ! enhancement techniques, and by use of robotic scanning control, to integrated semiconductive circuits~ in particular, p~ckaged one~, mounted on ceramic ~ub~trate3 or printed circuit board3-In anothar embodiment of the invention, th~ thermal ~imuliare voltage3 which are changed in duration rather than magnitude.
This embodiment is advantageou~, among other r~a~on~, if electronic component~ or circuitry ~uch a~ ~emiconductor components or circuit~y are to be tested under conditions more realistic a~ r~lated to an intended mod~ of operation in a computer or a timing program control. In fact, it i3 certainly rea~onablQ to consid~r doing som~ c~refully-timed oparatlonal test$ng simultaneou~ly with ~hs thermal ra~io analysi~.
Th~ principles of the invention ex~end to all product~ that can be theDmographlcally tested. For non-curren~-carrying ' :
product3, microwave radiakion can be con~idered a~ one likely stimulu~ to be ~mployed. In 3uch manner, these product3 or, more ¦ specifically, ~est ~amples of thes~ product~, can be te~ted by .HENDERsoN methods and equipmen~ within the ~cope o~ the appended cl~im~. i ~RA90~, C~R~ETT

~OD I ~rRCCr, N~ W.
3-~1NOTON. DC 2000 ~ 40~ ~000 . _ 7 _ ' .

~7~
¦~ The accompanying drawings, which are incorporated in and ¦ constitute a part of this specification, illustrate one preferred embodiment of th~ invention and, together with the description, s~rve to explain the principle~ of the invention.
I BRIEF DESCRIPTIC)~I OF q~HE DE~ ING~;
,¦ Fig~ 5 are flow diagramq illu~trating s~eps of the method of the invention;
, Fig. 6 is a diagram of a tes~ se~-up for practicing the .
method of the invention; and Fig. 7 i a bar graph ~howing t~mperature relationships for comparable point~ for a referencQ ~ampla and a te~t sampl~
DESCRIP~IO~ 0~ TE~ PR~RR~D~E~BODI~E~T5 An example of the prRferred method i~ ~hown in ths flow diagrams of Fig~ 5.
The method of the invention over~omes th~ limitations of image qubtraction techn~qua~ by compensating fox the expected thormal variation~ among like ~mple~ of a given product. The inYention work~ on the premi~e that th~ ratio of heating of like sample~ under diffarent thermal 3timuli ~hould be a con~tant. If .
a functional defect or other abnormality i~ pre~ent in a sample, i~ i8 exp4cted to c~u~e the thermal re~pon~e for at lea3~ one of tha therm~l stimuli to deviat~ from the expected value; and this expoctation h~ b~en borne out in practice. ~hi~ alter~ the ratio of heating ob~erved for that sample at a particular data point in an image-related array of data poin~ and thu~ indicates ~E~N HENDER50NI thak data point a~ th~ locu~ of either an actual d~fect, a ~R~ OW, C~RRETT ¦
DUNN E R
100 1 !ITQtrT, N. ~V~
I-IINOTON. DC 20001S
`202 ~01!1 ~000 ~7~
,Idefect-relatsd iymptom, or a laten~ defect. For purpo~es of thi~
~application, a dsfect iR any deviation from acceptable quality or , propertie~ of a product.
Normal thermal variations cancel when ratios of two array~
. of image difference~ are derived for measurements from the same sample or from the reference sample and the te~t sample ~ To complete the technique, the meaRurement.C are formed into a composite record including two other array~ of imaga differences, ¦ for corre~ponding data point~ in each array, for the reference Ij sample and the te~t ~ample. Illu~tratively, to form a composits ¦1 ~et of mea~urement , a different ratlo record can be derived, and then the rstio of tha ratio r0cord~ can be taken a~ a meanB of ~I diqcerning correct from incorrect ~emperature profile3. In ! general, a defect i8 indicated when an anomAlou~ region or datum appear~ in the composite record.
At present two type3 of stimuli have been iden~i~ied as being u~eful wi~h TRA in accordanc~ with the pre~en~ invention.
The fir~t and ~imple~t i8 to va~y the volta~e applied to power tha 9ample - typ~ cally u~ing the maximumi and minimum voltages specified for the ~ample. The 9ample can be a ~tand-alone sample or p2rt of a lar~ar functional group. It can be te~ted in a totally ~tatic test, or in a full ~imulation of it~ exp~cted operation, if timing constrain~ of the thermal ratio analysis are compatible therawith. Sta~ic te~ting do~, however, aliminate the pos~ible interaGtion of a simul~tion and hea~ing ~wO~c~. effect, that i~, opera~ion altersd by sofkware ~aking non-normal ~D~;NNER deci8ion-lQop branche~ during th~ 3imulation dua to ~he defect.
~00 1 !~Ti7~1Cr, N~ W.
I~NOTON. OC 2000 1'20i~ ~01~-4000 ~,_9_ ' ~ ' . ' , : ~
- ' . . ' ' ' 1. ' ' ~' , ' . ' , 2~7~
~The ~imulation t~pe of test purposPly dep~nd3 upon the heating ¦effect~ introduced by two software xoutines or clock~d verse~
non-clocked operation -- in fact, this type of te~t simply can be designed to simulate operation of tha sample and even to simulate the effec~ of the more unu~ual durational ~tre~3e~ of ~ignal~
upon th~ sample, by primarily affecting the duration of each stimulus instead of its magnitude.
. The thermal variation observed be~ween differen~
non-defective sample~ of a product is largely due to manufacturing and design tolerance within the ~amples themselve~
For electronic circuitry, this v~riation ha~ both a linear and non-linear ohmic component with re3pect to the applied volta~e and i3 also subj~ct to any dynamic ~timulation applied. The amount of temperature rise above amhien~ is relatsd linearly to the power di~3ipated multiplied by ~ome constant, in tha case of an integrated circui~, due to the packaging ~hereo~. Thi3 packaging can vary from circuit to circuit but i3 no~ a functi3n , of any extern~l ~timuli. Thu~, the temperature riso exp~cted to bo ob~erved i~, int~rnallyt a func~ion of the packaging and the linoar nd non-l~n~ar re~istance~ and, externally, a function of the ~pplied voltage and dynamic stimulation (i.e. t software).
Defecta c~n manif~t them~elve~ as modification~ of the intorn~l re~istive structurs3 of an electronic ~ample and are difficult to diccern from normal variation~ u3ing ~tandard image i ~ubtraction. It is not understood from prior art thermography .HENDER~NI technique~ that ~aking the rat$o be~waen two tempera~ure rises uNNER ¦ will produce a characteristic thermal ignature of a produc~, or ~Q0 1 ~TRt~r, N. ~'/.
~tllt~OlOtl, DC 200Q~
`~1`202`-0a`.~000 , - 10 ~` I

: ` ,. . .
,, , `' ' ' ~

`' ' ` : ' .,, :' 2~7~2~
a point in a product. Defects also have a strong tendency to be l functions of tha external variables which differ from those i func~ions exhibi~ed by their normal counterpar~s~ For example, ~¦ the non-linear re istance of normal sample~ of a product usually l have the same non-linear relationship and differ only by a linear ! con~tant. If a sample i stimulated by applying power at two different voltage~, and ~he sub~equent steps o~ our invention are employed, the two measured temperatures at any corre~ponding data points will ~e related by two term~: (1) the non-linear resistance expressed a~ a function of the ratio of the applied voltage~ (the ratio raised to tha power ~, whers ~ doe~ not equal 0), and (2) the square of th~ ratios of tha two voltage~.
All the linear relationship~ which are equally hared will drop out.
Thus the ratio of temperature rise sbova ambient for ~wo different voltage level~ ~hould be a con tant from sample to ~ample of the same product. Evan the thsrmal con~tant of the package drop~ out of tho calculation except in ca~e~ o~ extreme l changes in packagin~. It i8 therefore po~sib~e ~o m~asure a 'I
l reference ~ample and a test sample of th~ ~ame product and subs~nti~lly to cancel or eliminate the expec~d variation, by following analy~is qteps appropriately iacluding at lea~t one ratio, and, illustra~ively, tap~ including taking a ratio of ratios. When this is done, as it wlll ba mathemat~ally ~hown below, then all of certain t~rms in the expres~ion for the values ~wor~lct~ of the ima~e-related data point~, even the term related to the ~0~, Gt~RRETT
gDuNNER s~uare o~ the voltaga~, drop ou~. Wha~ i3 left i~ an image T~UI~r, N. ~. : :
5~ 0TOr~. OC 20003 .:
~`~0~--0~--<00 ~ ~

: ' . ~, : - .` .

: ~ `
;~

2~L/~J~
highlighting only thoRe pixels (data point~ in the resulting array) for which ~he raf~rence sample and ~he te3t 4ample exhibit .
sub~tantial differences in their "non-linear~ characteri~tic~.
A mathematical derivation of the principle~ ~ust described concerning the use of a dual-lev~l voltage stimulu~ for thermal ratio analy~is, in accordance wi~h the inventiont followc~:
The temperature ris~ ~T above the ambient of a sample is equivalent to the power dissipation (Pd) multiplied by a constant (Cp) related to the packaging and heat-sinking of the ~ample.

( ) Tam~ient = Cp x P~(E), where T(E) i~ the elevated temperature produc~d by applied voltage E.
Sinc~: p - E x I(E) (l)¦
~T 5 Cp X ~ X ~(E) (2)¦
Sinco: I(E) - E / R(E) (3)1 I ~T ~ Cp x E / R(E) (4)1 i But re~i~tanc~ can b~ replaced by i~ reciprocal G, conductance:
~T - Cp x E2 x &(E) (5) Conductance can be modeled a3 a function of E with linear and non-linea~ ter~s . . .
G~E) - ~ x E~ (6) ~T ~ Cp x E2 x ~ x ~, (7) the relationship being non-linear to a degree dependant upon' tho degr~e to which ~ deviates rrom ~ero.

If image~ are taken at two voltage level~ and ~hs temperature ~w~r~e~ difference ratio, ~ takan between th~m, ~hen . . .
:E~N, HENDER~N 2 ~BO~. ~RRETT

.oo t ~ .
nl~o~ c 2000 ~20~400-4000 ~ 12 -, L~7 ~ 2~

T2 1 ~ (AT2/~Tl) ~ (Cp x E22 x ,g x E2a)/~Cp x E12 x Ig x l~ ) (8) ¦ ~2:1 = E x _ (9) At this point, many of the unwanted terms have already cancelled out.
. As3ume a referenca sample is imaged and has a temperature ratio of Tr~ and a test sample ha~ a ratio o~ Tt~. Taking the . ratio of the ratios yields. . .
! Tts 2:1 E2 ( ts r~
__ (10) TrY 2:1 El Thu~ tha resultant image i~ a function only of the applied , voltages and the difference in the non-linsar component of the I internal re~istance~. Th~t i~, detectable difference~ exi~t at i th~ array of data point when ~t~ does not equal ~r~ which in general wiLl ba true when the ~'~ do not equal 0. I~ the non-linear resi~tance~ are equal, then the re~ult goe~ te unity.
Thus, thi~ can become a ~en~itive measure of non-linear defect~, becau~ the mall, but r~dily variable, non-linaar term i~ not swamp~d by larg0r linear ~ariation~, whi~h have cancQlled out.
It i~ al~o pos~ible to hold the vol~age constant and to vary the dyn~ml~ stimuli, usually by changlng ~he sof~ware or by turning a clock on or OfL'. Ths mechanism u~ed hor0 i3 that the heat generated by a defect will depend upon how the duration of a thermal ~timuluq i3 changsd a i w21} a~ upon it magnitude. ~ -~AN~ENDERSON The general ~omperatura rela~ionship~ for any TRA, ~RA~W. ~RREl~' : I.
~DuNNER regardless of ~timulu~, can be seen in the diagram in Fig. 7.
71~htl', N. W.
I~IN070N. OC 2000 ~ 20~-401~-4000 13_ l , , , " ... . ............... . .

~ 2 ~
The temperatures mea3uxed in respon3e to the two stimuli are Tl and T2. The original ambient temperature is Ta.
In the cas~ of two non-defective sample~ of the same product, the ratio of temperatures due to the two ~timuli should be equal be~ween the reference sample and the te~tt sample, within the limits of noise and measuremen~ error. From the mathematical developmen~ above and by reference to Fig. 7, we have:
A/B
= 1 (11) C/D

_ = _ or _ = _ (12) B D C D
The e ara obtained directly by mea~uxing Ta, Tl, and T2 for both samples and performing the appropriate ~ub~raction~. (For example, D = T2-Tl for th~ reerence sample).
In the ca~e of a defective sampl~, the original tempera~ure component~ (A, B, C and D) can be con~idered to be identical to the ~cenario ralating ~o non-defective sample~. However~ they arQ not diroctly measurable due to the temperature componant addad by ~he d~fect (~) - 3ee Fig. 7. ~q ~hown, the de~ect tempersturo component F could ~how up a3 a function of either or both ~her~al ~timuli.
Without carrying th~ analysi3 f urthor, on~ can then arrange thermal ratio analy~$s to involv~ the u~e of ~hree ratio~, fir~t, following either of ~he alternativ~ ~oxm8 of e~uation (1~) to obtain two o~ the ratio t and then ~aking a ratio o~ ratios for ~WO~C~ each data point a~ 8ugg~3ted by ~uation (11).
W, C~RRE~T

000 t OTIIICT. N. W.
0111N~lTON, OC 20t)00 1 ~0~ 400-~000 , - ;

:.
~' ' In prac~lcal terms, once one ha~ th~ four dlfferencP records g~neratsd a~ described above, TRA can be carried out fundamenta11y by the fo11Owing: deriving first and ocond ratio I record from the first and second difference records for the reference sample and the first and second difference records for the te~t ~ample, either by taking the ratio of the fir~t difference record~ and the ratio for the second difference records in the same order, af~er which one takes the ra~io of the ratio~, or by taking the ratio for the difference records for the reference sample and the ratio for the differ2nce records for the te~t samp1e in the ~me order, after which one take~ the ratio of the ratio~. j An ~stimat~ of F can bc made by a~suming that F only modifie~ tho va1ue of C. (Othsr approximating as~umptions can also bs made ~o 1ikc sffect). ~y rearranging the equation above, an estimate of what C ~hould be, based on the mea~ured values of A, B and D, can bs made. Thi~ i8 then ~ubtrac~ed from the value (C~ ) measured for T2-T1 for the test ~a~ple.

C = D x A (13) B
C' = C + F (14) F = C' - C = C' - D x A (15) ~ quation (15) is ~u8t one of many approximate algebraic transfonmaticn8 of e~uation (11).
In fact, a~ equation (15) sugge~x, in thi~ algebraic ;E~N.HENDE~N tran~formation of e~uakion (11~, TR~ can be con~idered to be a 3~RRE~ particular typ0 of ims~e ~ubtraction which ha~ been normalized to gT~ , N. ~V.
NOSON. DC 2000 1`202`-01~'4000 - 15 - ~
l 1, ., :

~" ~ . ! . ", - . ':
, ` " ,.

~ J ~ . 7Si ~J ~
compensatP for predictable varia~ion~ between liks components.
Thi3 iQ a particularly useful embodiment of the invention.
Looked at another way, this approximate form of T~A
i requires, in it8 3imple~t form, that only one ratio be taken, the ratio being taken for comparable diff~rance record~ from the reference sample on the one hand~ and the te~t ~ample on the other hand, providing one is willing to accomplish the mul~iplication and ultimat~ or ~normalized~ difference taking, involving the other difference records.
The preceding analysis demonstra~es the ~implicity of TRA, showing why it provid~s an image in which many otherwi~
expectable axtifact~ are removed becausQ of the use of ratio~.
From equation~ (lQ) and (11) it i~ e~tabli~hed that a ratio o~ ratio~ calculated from ~he mea~ured thermal ~alue~ should equal one (1) in cases whora no defect exists. If desired, it is also pos~ible to pre~ent the thermal ratio relation~hip in an alternate ~orm by m~ans of any applicable algebraic tran~formation. Equation (15) is an example of a tran~formation that affec~s the thermal information ~uch tha~ value~
corrosponding to no defect tend ~oward zero (O) in~tead of one.
This form ~l~o h~s the advantag~ that calculated values will be aymma~rical on ~i~h~r ~ide of tho no defect l~vel.
In either ca~et there i3 ~om~ natural varia~ion, or 3catter, in all mea~ur~mon~3 thak will follow stati3tical la~. ~or thi~
rea~on the re~ults of a TRA calculation will never ~xactly go to ,HENDE~N ono for equation (11) ~or zero for equa~ion (15)]. In~tead there ~R.~90~. G~RRErr ~DUNNER will be random di~tribution of data value~, or sc~tter, which ~00 1 ~tQ2tT, N~ W.
~NINOtON, DC 2000 ~`~~'-'~~ ~

'` `: ' ~' ' '` `' ': , `

: ` :` ` ` ~: :;: :

2 ~ ~ ~; 2 ~
will b~ centered around the central ~ralue. Thi~ di~tribution will be ~trongly center weighted and will have ~ome characteri~tic spread (i~a~ standard deviation) that i_ dependent on environmental factors and the quality of the component in the TRA system. Presented as an image, this would li appear as some degree of noise or non-uniformity in an otherwise uniform background.
The TRA value~ as~ociated with defects will tend to not approach the centxal value (one or zero, depending on the equation u~ed). The method for the detection of defect-related thermal information iq a 3imple ~tati~tical te3t. A value for the spread, or scatter, o~ the mea~urement data i calculated tstandard de~iation i~ a good mea3uxe3. Thi~ 4pread may encompa~s the entire te~t fiald, or only a local area of it.
thre~hold i9 ~et ba~ed upon a multiple of thi~ ~pread value.
Common ~tatis~cal practice u~u~lly pick~ a multiple between two and three times the standard deviation (" igma"). Thi~ threshold ¦
i8 thon set above and below the central valua~ If a multiple of two i8 u~ed~ th~n 9S% of the norm~l noi3e ~hould fall between the !
two threshold~ the multiple i8 ~hr~e, then 99~ of th~ normal noi~o ~hould fall be~ween th~ limit~. Any TRA valua that all~
beyond on~ c~ the~e rangss r~present~ a po~ible indication of a faul~. !
The certainty of the defect indication can be bia~ed by two considorations~ F~rst, valu~ that far exceed the ~he~holds usually can b~ con5id0r~d to b~ 8tronger def~ck indications than ~R~ . G~RRE~T
aD~NER ones which ~u~t exceed the limit 5econdly, when the T~A ~alue~ 1 ~00 t ~Q~Cr, N~
~NOl'ON. OC 2000 '`''''' 11 . -17- 1, ,.

.
, form an image, mo~t real component3 axe largs enough that at lea~t several distinctive adjacent thPrmal ratio analy~i~ values will correspond to a given component. The uncertainty of the . defect indication is inversely proportional to the number of related TRA image values that exceed the threshold~. Thu3, a general principle i~ that the more image poi~ts that correspond to a component, where these image point~ have substantial contrast with re~pect to the gener~l background, the lower the threshold can be to maintain a given cartain~y for the defect.
For any given ~ize of component (in ~erms of number of image values) it i~ possible to et a dynamic te~. thre~hold that maintain~ a fixed probability of falsely indica~ing a good device ¦
or not detecting a defective device. The~e ar~ ~tandard stati~tical method~ for detecting de~ired feature~ in the presence of noise.
~ he smpha i3 on daf~ct d2taction up te thi~ point has been in regard to the preferred embodiment of the inven~ion in which the TRA data value~ compri~e an imaye. In ~hat ca~e, the formal statistical approach to defsct detection presen~ed above can often b~ abandoned in favor of a more intui~ive approach based on vi8u~1 in~pection o ~he image. Simply ~tated, defect~ will app0~r a~ point~ or r~gions which ltand out by virtu~ of a ~ignificant (or ~reater than average) contrast in di~played inten~ity (or color) from the re~t Q~ ~he image.
The invention how~ver is not re~ricted to imaging ~C~N, HENDER~ application alons. An embodiment o~ th~ invantion i~ po~ le ~ ao~ C~RRETT
~DuNNER in which di~crete temperature mea~uremen~ are made only at ~0 1 ~TRttT. N. W.
OTON. DC 20005 `202 40~-4000 . - 18 ~ ;
., single points, usually corresponding to individual components on a unit. Obviously such measurements are not spatially related to each other and do not form an image array of data values. In this case the formal statistical analysis is required to discern defect-related thermal values from normal ones. However, much less data is available to establish a reference threshold because the non-component areas of the unit are not available for examination. The statistical method described above is still valid, however, due to the fact that in a real manufacturing environment most components will indeed be good (non-defective) and the data values of the good components correspondingly will reflect this. Thus the data values of the good components will provide sufficient statistical information to enable the detection of defective component(s). The ability to detect thermally subtle defects will not be as accurate as the imaging information available for processing. As in all cases of statistical testing, accuracy is directly related to the amount of data available.
Error Factors Some explanation needs to be given to the factors that contribute to the introduction of error into the TRA image. The amount of quantization error prior to any image pre-processing can be estimated from the form of the approximate TRA formula:

(16) 2 ~
3efor~ thi~ can be u~ed however, the variable~ that ¦introduce ~rror should first be con~idered. The uncertainty T) in any temperature mea~urement T i~ equivalent to the 'minimum temperature resolution of the imager Tr ~u~ually ~/-0.05 ¦degree C), plus any uncertainty due to taking the image prior to l1therm 1 stabilization. This will be called the temperature i ~tabilization error, T~e, ~o th~t:
i ~T = Tr ~ T~ (17) To estimate Tse, the ~empexature s~abilization function must be modeled. This can be con~idered to roughly approxima~e an exponQntial which has an a~sociated time-con3tant r and that ultimately reach~3 ~ome upper temperature limit, Tmax. Expressed a~ a function of the time tt) and the time-con~tant (r ), the t~mperature ~tabilization func~ion i~s T3(r,t) ~ Tmax x (1 e t/r) (18) To expre~ the uncertainty of Ts in terme oP the value3 and uncertainty of r and t, we can u~es Tse - T~l(r+dr,t+~t) - T8(r,t) (19) Tse Y ~max x (e +t/r _ ~ ~(t~t)/(r+dt) ) (20) In connection with the T~A formula, each of the term~ A, B, C', and D i~ ac~ually ths di~fere~ce betw2en either Tl or T2 and tha original ambient temperakure Ta. Therefore, the net cumulative uncertainty for each of the~e terms can be expres~ed as a function of tho uncertaintie~ of the ~econd mea~ured temperature Tm (Either Tl or T2) and the original ambient o~et- t~mperature ~at ~EO`~;T, G~RRETT
DUNNER
ITI~lCT, N. \'1.
<INOTON, DC 2000 1`~0~ 000 - 2~ -- ~ . . . .

.: :
:~ ' .' ~ . :', ' ~ r~
~A = ~B = ~C' = ~D = ~Tm ~ ~Ta (21) Howev~r, ~Tm = ~Ta = Tr + Tse (22) Therefore, ~A = ~B = ~C~ = ~D = 2xTr + 2xT3e ( 23 ) This is due to the fact that error term~ always add, even when the normal ~non-error) terms are subtracted. The ne~ effect of adding error term~ i~ to approximate a normal distribution function, with an average of zero and +/-3~ points equal to +/-the sum of all the error term In terms of imaging, thiCt produce~ an image where, on average, th~ fea~ure3 are preserved but the speckle content i~ exaggerated.
From thi~, the uncortain~y of tha calculated image approximately can be expres3ed a~D
~F = ~C' + ~D x ~ ~L~ L (24) B +/- dB
wherein the signs are cho~n such that they maximize the ~D
term. For all practical purpose~, the uncertainty o~ ~ can probably b~ ignored. In fact, the dA, ~B and ~ term~ can be ignored as long a3 A~A and B~>~B. Two generalizations on the error in the derived defect-related image tha~ de~cribes tho~e ara~ o~ the image where ~igni~icant heating occur~ versu~
thosa area~ where lit~la heating occur~ can thexefore be made.
Whexe tha heating is ~ignificant, the ~rror term approaches:
~ .

~AW ~r-c~
`~E~N, HENDERsoN
\~30W, G,~RRET1' DuNNER
~oo t ~T~ . ~.
s~ oror~. oc 2000 1~2Q~ 0~ 4000 . ~ 2~ - :

':

-- . . . . ~ .
.. :,, ;
- . ~ - . : . . .

- , . .
,, ~ .

~q7~l~2~

~F = ~C ' + ~D ~ _ 2xl~ (25 = 2 + _ x(Tr + Tse) Where the heating is minimal, the error t~rm approaches:
~A + ~
~F = ~C ' ~ ~D x _ _ ( 26 ) ~3+ ~

But with minimal heating ~>~A = aB, so a good approxima~ion is ~F = ~C ' ~ ~D = 4x ( Tr + Tss ~ ( 27 ) But, then again, wi~h minimal heating ~r~>T~ o thi reduc~s to:
~F = 4 x Tr (28) It ~hould be ob~iou~ that noi~a in the TRA image can be reduced by minimizing Tr and Ts~. ~rr can be reduced significantly by interpolatin5~ (a~eraging) the value~ of ~he pixQls o~ the TR~ lm~go with their neare~t neighbor~. Thi~
lessen~ the effect of quantizing the image. Likewi~e, Tse can be minimizad by prop~rly controlling th~ accuracy of the timing inte~val and m~king the~ timing interval as long as fea~ibly posEi~l~. Prom the~ fonnula or ~se, it ~an b~a seen that tho offact o~ ~t and ~ can ba minimized by m~king t~3r which allows the temperatuxa to achiev~ greater than 95% of it~ steady i stata value. It was found that recording a ~he~nogram four ,HENDERsoN minut~ aftar a ~tep funct~on the~rmal ~ti~ulu~ wa~ fir~t applied ~IOW. (;~I~Ri~TT
aDu~NER produced adl3quata re~ults.
~00 t ~ r, N~ ~V.
3111NOtON. DC 3000--1`~0~ . 00-4000 i, ' .
:
. . ~ , . . . :
,. .

~7 ~2~
The flow diagram of the method o the invention ~hown in Fig~ 5 ha~ been developed using commercially available thermal imager~ such a~ the MIKRON$ 6T62 Thermo Tracer. Such imagers rely upon an electro-mechanical sc~nning system to route received infrared radiation to a fixed di~cre~e infrared sen~or and it produces a raster image of thermal da~a which can be viewed by a standard video monitox or by mean~ of a computer interface and monitor.
For testing integrated circuit~ mounted on board~, th~
technique could be modified to u~e a robo~ically contsolled sensor to obtain data only at component (e.g., resistor, active dQvice, or even an entire "chip") location~. This latter fea~ure i~ desirable because it simplifie3 def~c~ recognition data ¦
processing when used a~ an automat~d ~y~te~.
In the flow diagram of Fig. 1, three two-dimensional arrays or image-like matric~s of da~a point , that i~, thres thsrmal image~, ar~ acquired for a reference sample, in s~ep 11. ~he first image i~ for the re~erenc2 ampl8 at stabl~ ambient temperaturs.
It has baan found tha~ blowing compressed air at or just below ~mbient t~mperature over the sample will allow it to st~bi}$~ at ambi~nt temperaturQ relatively ~uickly. The second image i8 ~or the reference 3ample at a firs~ elevat~d temperatura, or in re~ponse to a fir~t ~timulu~, which i~, for example, a firYt voltag~ producing a fir~t amount of heating in ~E~N~HENDER~N the samplo. ~ second, typically mor~ int~nse, stimulu~ is then ~30\~, ~RRETT
~DuNNER applied to the refer~nce sample to obtain a ~hird image twhich is I
~00 1 ~ Cr, N. W. ::
'0! . OC 2000~
0~`40~-4000 ~ :

, , :. ..

.
.
.- ' '' ~7~
the second image at elevated temperature). It has been found that a Ytimulu~ producing 5-10% mor~ heating of the sample than the first stimulus is suitable. For each of the two different step-function timuli, it wa~ ound that equal application times l¦(e.g,, for four minutes) produced 3ubstantially complete heating i and was advantageous. ~egardle~s of whether th~ ~wo ~timuli application time~ are equal, the application timas for like-value stimuli mu~t be the ~ame a~ between the test and reference sample. Step ll can be repeated for at lea~t a 3econd reference sample, unles~ one has some highly ~andardiz~d previou31y tested samples of known good quality.
In ~tep 12, ~he initial analy~is ~teps characteri tic of T~A
are performed upon the ~ets of referencQ data, for aach reference 3ample, the acqui~ition of which data was just de cribed.
Step 12 make3 a first difference record B from the ambient temperature thsrmograph and one of the elevated temperature thermographs, and a ~econd differencs record D from the elevated temperature thermographs. In ~ep 13, procedures as in teps 11 and 12 are repeated for the te~t sample replacing ~he ref erence ~ampl~. Before any ra~io~ are taken, preliminary image proces8ing ~teps ars performed a~ indicated in ~Qp 14, if de~lre~, as dlscu3~ed b~low. A~ ~hown in Fig. 2, ratio analysis is then performQd as shown in ~tep 15, according to any one of the alternato formulas. In ~he fir~t ormula, "~" and "C" can be intarchangad becau~e division i~ commutative. ~ach ratio of L~W or~lc~- step lS for all th~ data point~ in Qach of two o~ ths four ~RUOW G~RRTT
~DUNNE~ difference records corre~pond~ to a poin~ in the sample. Since ~00 1 ~ . w.
OTO~l, DC 20000 .' '' . :. :

:" : ' ' -' '~ ' .

~0~7~26 the TRA has proved to be a superior "normalizing~' technique for thermogxaphy, i~ iY feasible to do a quick ~sanity" check in step 15, for example, by comparing ra~io images for ~upposedly very similar llreference sample~ll. Any markedly dissimilar , re~ults of the comparison ~ill peint to a failure to follow the proper testing technique, or failure to hav~ obtained suitable reference test units.
Proper testing technique may require tho following:
--No air drafts except perhaps controlled air flow to maintain ambiant temperature.
--No heat ~ource~ near by (including the sample~ own power supply)., I
--Blow off with air ~et to stabilize thermally.
--Handle with in3ulating gloves to avoid warming by handling.
--Pro-te3t staging loca~ion ~hould be ad~acent to ~he TRA
fixture to provide natural thermal equaliza~ion to actual test ambi~nt conditions. This reduce~ th~ stabilLzatlon time achi~ved by mean o~ tha air ~et.
--Tape critical reflectivc point3 (being careful to prevent el2ctro~tatic def~ct generation upon later removal of ~he tape).
It ~hould be rem~mbered tha techni~ue d~p2nds upon infrared radiation from the sample, and reflections are not de~ired. It is not required to ~o trea~ component leads.
--Edge connector~ typically u~ed with ~emiconductor circuits c~N.HENDER~N should be ~hermally main~ained ak ambie~ ~o coun~ract warming ~R.~OW~ GARRETT
~DuNNER from sampl~.
~00 1 ~TRC~, N~ ~.
3~1NOn~N~ DC ~Q00!~
1'~0~`40~-4000 ,- 25 -.

~7~

--Avoid area that exparience larye thermal shift~ from air conditioning or sun-lit windows.
--Block reflections from ~urrounding surfaces and bodie~
using baffles of low thermal mass and high emissivity.
l --Seek viewing angl~ that prevents the imagar or detector I from se~ing itself in reflection3 rom modul~ (Thi~ goal implies long~r viewing dis~ance and a ~mall tilt o~ tha sample relative to the sen30r).
In some instance~, would-be reference sample~ thought to be normal or non-defective are found not to be, and mu~t bs replaced.
Notice that the first formula, the fundamental formula, involves two ratio~ derlved from the difference records in an ordered way, as may be 3e~n by reference to Fig, 12, which ratio records are then further re~p~ctively sub~ec~ed to the taking of ratio3, point by point. The compo~ite record then at each point compri3e~ a ratio of ra~io~.
The first direct indi~ation of devia~ion between ~he reference sample and the test sample can be obtained frem th~
compo~ita record of ~tep 15, according to step 17. In many case~
no further ~s~ting i~ nee23~ary; and the procedure~ of ~tep 16 may b~ skipped. An immed~ate deci~ion can be made a~ to whether the tes~ sample i~ sufficiently ~imilar to the reference sample.
In fact, step 17 can be performed automatically, by tes~ing whsther the compo3i~e record i~ feakurele-~s in re~pect of having ~w or~c~
NECAN, HENDE~50N
~R~BO\~ RRETT
UNNER
~O0 1 5$R5~. N. W.
~rn~OSOI-. ~C 20ooa ~`102`.-01~ .~000 . : , , ,. , -.......................... .

~7~
no va~iation~ from the background which exceed a statistical limit ~uch as a multiple of the standard deviation. The . indication iR then tha~ there is no defect.
If an undesirable level of ambiguity persists after step 15 . is performed, i.~., if a suspectsd defect produces differences ',~from an image for the reference sample which do not exceed a i first prede~ermined level (llthree-sigmal~ level), but are still above a second, lower, predetermined level (~sigma level)~
various image enhancement technique~ can be empl~yed, particularly with respect to the thermal ratio image for ths test sample. In such techniques, the thermal ratio data i3 organized in the 3ama two-dimensional way as in the original thermograph and, for example, can be viewed on the computer moni~or. The3e are the optional post-ratio-analy~is imag~ proce~sing steps shown in step 16.
For exampl~l in step 16, or a~ ~hown in Fig. 4, optional noise filtering is done. If ths results ax~ inc~ufficient to reduce the l~vel of ambiguity to make a deci~ion pos~ible, step 20 then r~moves by ma~king techniques any known non-defeet-type ar~ifac~ found in either of both of th~ referenca ~ample and the te~ sample. Thi~ i~ a sp~cific example of techniques which arQ known in as~ronomy and reconnaissance photography as image enhancem~n~ techniques.
For further examplas of possibly relevant enhancem2nt t~chnique3, ~ee ~ , by R. Gonzalez et al., soN Addi~on-We~ley (1987), pp. 162-163 (median filtaring, pertaining ~R~aOW. C~RRETT
` ~ D~ER to noise filtaring); pp. 158-160 ~local enhancement, pertaining ~00 1 ~TRCTT, N. W.
NOTON~ OC 2000~
~`202`40~-4000 - 27 `

to ma~k generation from original image data). As duly enhanced, the thermal ratio images are re-compared in ~tep 17; and a deci~ion is made as to the test sample, e.g., whether the first predetermined level or amount (~'three-sigma~ level) is exceeded ! at any point. The test apparatus ~nd technique are ready to be applied to the next te~t samples, e.gO, tho~e coming down an assembly line for semiconductive integrated circuits, as indicated generally at step 18.
i Fig. 3 shows the detailed steps in a portion of the ll procedure of Figs. 1 and 2, a~ applied ~o a current-consuming i product, and illustrate~ specific implamenta~ion of ~ome of the point~ of good te~ting practice listed above. Of particular note in the datailed ctep3 of Fig. 3/ which are partially repetitive of tho~e of Figs. l and ~ is that the fir~t and second l elevated-tempera~.ure thermograph~ can be produced, in the case of ¦ the typical electronic component, by the application of voltage~, clo~Q to normal operating voltage~, which can be u~ed for operating parformance test~. For exampla, the succe~sive thermal ~timuli might be provided by 35% of normal operating voltage and 105% of normal oper~ting voltag~, as illu~trated by step~ 34, 36, 38 and 440 Thi ver~tility of the method is greatly facilitated by it~ wide dynamic rangs, which al80 ~igni~iet that thes~ are no~ critical value3 of ~timuli.
Fig. 4 COh~i3t~ of diagram~ ~hat deal with certain aspects of imAg~ enhancement, particularly noi~e filterlng. Steps 50-52 .HENDE~oNI rolate to a relat1ve ~moothing between ad~acent pixels of each ~R.~BO~r. c~RRErr ~00 t ~ cCr, N~ W.
0~ . DC 2000 `~02 ~0~ 000 _ 28 -, , ~ L~7 ~2 ~
l~thermal ratio image and for the resulting comparative image that ,Iwill tend to filter speckle noise.
Fig. 5 show~ yet another aspect of image enhancement that can be performed, per steps 61-63, with onP or more of the original elevated temperature thermographs and i9 useful for generating one or more mask images fox spatial filtering.
Fig. 6 shows the interrelationships among the items of equipment used in TRA.
The thermogram data-gatherer 91 is configured to hold a . sample 96 from which infrared data i~ to be acquixed. The data i i~ acquired by the infrared sensor 97, that has typically a . single, fixed discrete sensing element, which effectively 3amples i¦or scans the expo~ed portion of sample 96 via an interposed jlscanning apparatu~ in sen or 97. In the preferred embodiment, the infrared ~ensor 97 i~ a thermo-tracer, for example, a ~IKRON
6~62 thermo-tracer.
Alternati~ely, sen~or 97 could be a semiconductor charge-coupled-device camera, thereby avoiding the electro-machanical scanner of the ~I~RON 6T62 thermo~tracer. The sequencing of image-type data point~ i~ then obtained purely electronically, for example, a~ i8 ~ypically providad for ~uch a camera.
~ he thormal ratio logic and storaga unit 92 i~ ba3ically a central proce~sing unit which in addLtion ~o performing the calculation~ lndlcated aboYe, coordinate~ the functioning of the image ~nhancement mean~ 93, the de~i3io~ circui~ 94, as well as .HENDER50~ tho timing through duration control 99, and ~ize, through p~ak ~R.~90~, C~RRETT
~ ER control 98, of thormal ~timuli applied from ~timulu~ source 95 to ` ~00 ~ 9TR~CT, N. W.
3r~1NO~ON, DC 2000 1 202 ~0~ 000 ~ - 29 -.~
.. ...
.
. .
. . .: . .
.. . .. . .
.

7t~
¦ ~ample 96, and the scanning of the sen~or 97. Thu~, the logic ¦ and torage mean~ 92 obtains the variou~ difference records, as well as the reference sample ra~io image and the test sample ratio image. Ima~e enhancement meanq 93 can use any of the above-described filtering, smoothing and other image enhancement techniques, as well a~ any of those known in the prior art.
;l Indeed, again ac described abo~e, a~ well as performing in the position shown, image enhancement can ac~ direc~ly on thermograph~ before entering logic and storage unit 92, as well ~l as on difference records obtained therein. ~hus, the arrangement ,1 of Fig. 6 is merely illustrative.
¦ Decision circuit 94 accepts or re~ects te~t ~amples a~
Il having no defect or a dafec~ of sufficient magnitude ba~ed upon I the sta~istical analy~i~ of the composite record, which for ! example, may show one more pixel3 deviating from the ~urrounding . background by one or more standard deviation3.
It should be under~tood that TRA is more broadly applicable l to product3 than ~ust semiconductor circui~s or even electronic ! assemblie~. All manufac~ured products (non-living) can be caused to experience controlled tempera~ure excur~ion~. While use on an assembly line may be a preferred use, it can also be used at any time in the lie of a product, provided data, e~en archived data, from a reference sampla of the product i3 available.
. Additional advantage~ and modifications will readlly occur to tho~e ~killed in tha ar~. There~ore, the invention in its u~HENDE~oNi broad2r a5pects i~ not limited to the sp0ci~ic det~il~, ~30w, G~RETT ¦
~ D~ER representative device~, and illu~rative example~ ~hown and ~o I OTRtlT, N. W.
~NOTON, OC ~OOO~ :
0~-400 4000 _ 30 -~.
. . - . .

. ~ , ~ ~7~2~
dascribed. Accordingly, departure~ may be made ~rom such details withou~ departing from the spirit or scope of ~he general i inventive concept as defined by the appended claim~ and their quivalents.

~w o~ct~
;E~N, HENDERSON :
R~30~, C~RRETT

101~ 1 ~TR~TT, N~ W.
NOTON~ I~C 2000 ~`~O~`~OtJ ~OOO
: - 3~ ~

.
, :
:

~,: :

Claims

1. A method of detecting defects in a test sample of a product, comprising the steps of:
generating a reference record for at least one reference sample of the product that does not have defects, comprising the sub-steps of:

(1) making an ambient-temperature thermal record of the reference sample at ambient temperature;
(2) making a plurality of elevated temperature thermal records of the reference sample at a plurality of respective elevated temperatures, including applying a respective thermal stimulus to the reference sample, and (3) making a first difference record for the reference sample from the ambient-temperature thermal records and one of the plurality of elevated temperature thermal records and at least a second difference record for the reference sample involving at least another of the plurality of elevated temperature thermal records, generating first and at least second difference records for the test sample by repeating sub-steps (1)-(3) with the test sample replacing the reference sample;
deriving at least one ratio record from the difference records for the reference sample and the test sample;
forming a composite record including at least the derived one ratio record; and generating a defect indication when the composite record yields a statistical significant deviation from an expected value.

2. The method of detecting a defect according to claim 1, in which the sub-steps of making a plurality of elevated temperature thermal records include employing computerized control of the temperature.

3. The method of detecting a defect according to claim 1, in which the sub-steps of making a plurality of elevated temperature thermal records include allowing the passage of a fixed interval of time while each stimulus is applied.

4. The method of detecting a defect according to claim 1, in which the sub-step of making an ambient-temperature thermal record comprises stabilizing the thermal state of the sample by rapidly flowing gas from a compressed gas source around the sample at an effective temperature substantially equal to ambient temperature.

5. A method of detecting a defect in a test sample of a product, comprising the steps of:
establishing a reference record for at least one reference sample of the product that doss not have detects, comprising the sub-steps of:
(1) making a base thermal record of the reference sample at a base value of a thermal stimulus;
(2) making a plurality of changed valve thermal records of the reference sample at a plurality of respective changed values of the thermal stimulus, including applying a respective changed-value thermal stimulus to the reference sample; and (3) making a first difference record for the reference sample from the base-value thermal record and one of the plurality of elevated value thermal records and a second difference record for the reference sample involving at least another of the plurality of elevated value thermal records;
generating first and second difference records for the test sample by repeating sub-steps (1)-(3) with the test sample replacing the reference sample;
deriving at least one ratio record from the four difference records consisting of the first and second difference records for the reference sample and the first and second difference records for the test sample;
generating a defect indication when the composite ratio record yields a statistically significant deviation from an expected value.

6. A method of detecting a defect in a test sample of a product, comprising the steps of:
establishing a reference record for at least one reference sample of the product that does not have defects, comprising the sub-steps of:
(1) making an ambient-temperature thermal record of the reference sample at ambient temperature, (2) making a plurality of elevated temperature thermal records of the reference sample at a plurality of respective voltages applied to the sample to produce heating therein, including applying a respective voltage to the reference sample, (3) making a first difference record from the ambient-temperature thermal records and one of the plurality of elevated temperature thermal records and a second difference record involving at least another of the plurality of elevated temperature;
generating a test record for the test sample by repeating sub-steps (1)-(3) with the test sample replacing the reference sample;
(4) deriving at least one ratio record from the four difference record consisting of the first and second difference records for the test sample and the first and second difference records for the reference sample;
forming a composite record including the at least one ratio record; and generating a defect indication when the composite record yields statistically significant deviation from an expected value.

7. The method of detecting a defect according to claim 1, in which the sub-step of applying a respective thermal stimulus to the sample comprises employing the stimulus for a respective period of time different from a period of time for another respective thermal stimulus.

8. The method of detecting a defect according to claim 5, in which the sub-step of making a plurality of changed value thermal records include employing computerized control of the value of the stimulus.

9. The method of detecting a defect according to claim 5, in which the sub-step of making a plurality of changed value thermal records include allowing some stabilization period after the stimulus is first applied.

10. The method of detecting a defect according to claim 5, in which the sub-step of making a base thermal record comprises stabilizing the sample by rapidly flowing gas from a compressed gas source around the sample at an effective temperature substantially equal to ambient temperature.

11. The method of detecting a defect according to claim 6, in which the sub-steps of making a plurality of elevated temperature thermal records include employing computerized control of the voltage.

12. The method of detecting a defect according to claim 6, in which the sub-steps of making a plurality of elevated temperature thermal records include allowing a period temperature stabilization after a respective voltage is first applied.

13. The method of detecting a defect according to claim 6, in which the sub-step of making an ambient temperature thermal record comprises stabilizing the sample at ambient temperature by rapidly flowing gas from a compressed gas source around it at an effective temperature substantially equal to ambient temperature.

14. The method of detecting a defect according to claim 7, in which the sub-steps of making a plurality of elevated temperature thermal records include employing computerized control of each respective period of time.

15. The method of detecting a defect according to claim 7, in which the sub-steps of making a plurality of elevated tempera-ture thermal records include allowing some temperature stabilization after an elevated temperature is reached.

16. The method of detecting a defect according to claim 7, in which the sub-step of making an ambient-temperature thermal record comprises stabilizing a sample at ambient-temperature by rapidly flowing gas from a compressed gas source around it at an effective temperature substantially equal to ambient-temperature.

17. A system for thermally detecting a defect in a test sample of a product, comprising means for recording infrared data of a reference sample that does not have defects at ambient and elevated temperatures and for recording infrared data of the test sample at ambient and elevated temperatures;
means for forming a reference sample difference record from the infrared data of the reference sample at the ambient and elevated temperatures, and for forming a test sample difference record from the infrared data of the test sample at the ambient and elevated temperatures;
means for deriving at least one ratio from two of four difference records comprising said reference sample and test sample difference records;

means for converting the ratio into a composite record involving all the difference records; and means for generating a defect indication when the composite record yields a statistically significant deviation from an expected value.

18. The system of claim 17, further comprising means for enhancing the ratio record to remove potentially false defect indications.

19. The system of claim 18 further comprising means for removing speckle noise from the ratio record.

20. The system of claim 19 further comprising means for preprocessing the thermal records, or the difference records, prior to deriving the ratio record.

21. A method of detecting defects in a test sample of a product, comprising the steps of:
generating a reference record for at least one reference sample of the product that does not have defects, comprising the sub-steps of:
(1) making an ambient-temperature thermograph of the reference sample at ambient temperature;
(2) making a plurality of elevated temperature thermographs of the reference sample at a plurality of respective elevated temperatures, including applying a respective thermal stimulus to the reference sample, and (3) making a first difference record for the reference sample from the ambient-temperature thermograph and one of the plurality of elevated temperature thermograph and a second difference record for the reference sample involving at least another of the plurality of elevated temperature thermographs, the first and second difference records each comprising a plurality of image-type data points in an image-related array, generating first and second difference records for the test sample by repeating sub-steps (1)-(3) with the test sample replacing the reference sample;
deriving at least one second ratio record from the four difference record consisting of the first and second difference records for the reference sample and the first and second difference records for the test sample, the ratio record including data points corresponding to respective data points in at least one of the difference records;
generating a defect indication when the composite record has at least one discrete region of substantially greater than average contrast with respect to surrounding regions.

22. The method of detecting a defect according to claim 21, in which the sub-steps of making a plurality of elevated temperature thermographs include employing computerized control of the thermal stimulus.

23. The method of detecting a defect according to claim 21, in which the sub-steps of making a plurality of elevated temperature thermographs include allowing the passage of a fixed interval of time while each stimulus is applied.

24. The method of detecting a defect according to claim 21, in which the sub-step of making an ambient-temperature thermograph comprises stabilizing the thermal state of the sample by rapidly flowing gas from a compressed gas source around the sample at an effective temperature substantially equal to ambient temperature.

25. A method of detecting a defect in a test sample of a product, comprising the steps of:
establishing a reference record for at least one reference sample of the product that does not have defects, comprising the sub-steps of:
(1) making a base thermograph of the reference sample at a base value of a thermal stimulus;
(2) making a plurality of changed value thermographs of the reference sample at a plurality of respective changed values of the thermal stimulus, including applying a respective changed-value thermal stimulus to the reference sample; and (3) making a first difference record for the reference sample from the base-value thermograph and one of the plurality of elevated value thermographs and a second difference record for the reference sample involving at least another of the plurality of elevated value thermographs, the first and second difference records each comprising a plurality of image-type data points in an image-related array;
generating first and second difference records for the test sample by repeating sub-steps (1)-(3) with the test sample replacing the reference sample;
deriving at least one ratio record from the four difference records consisting of the first and second difference records for the reference sample and the first and second difference records for the test sample, the ratio record including data points corresponding to the respective data points in at least one of the difference records;
forming a composite record including at least the one derived ratio record; and generating a defect indication when the composite record has at least one discrete region of substantially greater than average contrast with respect to surrounding regions.

26. A method of detecting a defect in a test sample of a product, comprising the steps of:
establishing a reference record for at least one reference sample of the product that does not have defects, comprising the sub-steps of:
(1) making an ambient-temperature thermograph of the reference sample at ambient temperature, (2) making a plurality of elevated temperature thermographs of the reference sample at a plurality of respective voltages applied to the sample to produce heating therein, including applying a respective voltage to the reference sample;
(3) making a first difference record from the ambient temperature thermographs and one of the plurality of elevated temperature thermographs and a second difference record involving at least another of the plurality of elevated temperature thermographs, the first and second difference records each comprising a plurality of image-type data points in an image-related array;

generating a test record for the test sample by repeating sub-steps (1)-(3) with the test sample replacing the reference sample;
(4) deriving at least one ratio record from the four difference records consisting of the first and second difference records for the test sample and the first and second difference records for the reference sample, the ratio record including data points corresponding to respective data points in at least one of the difference records;
forming a composite record including the at least one ratio record; and generating a defect indication when the composite record has at least one discrete region of substantially greater than average contrast with respect to surrounding regions.

27. The method of detecting a defect according to claim 21, in which the sub-step of applying a respective thermal stimulus to the sample comprises employing the stimulus for a respective period of time different from a period of time for another respective thermal stimulus.

28. The method of detecting a defect according to claim 25, in which the sub-step of making a plurality of changed value thermographs include employing computerized control of the value of the stimulus.

29. The method of detecting a defect according to claim 25, in which the sub-step of making a plurality of changed value thermographs include allowing some stabilization period after the stimulus is first applied.

30. The method of detecting a defect according to claim 25, in which the sub-step of making a base thermograph comprises stabilizing the sample by rapidly flowing gas from a compressed gas source around the sample at an effective temperature not exceeding ambient temperature.

31. The method of detecting a defect according to claim 25, in which the sub-steps of making a plurality of elevated temperature thermographs include employing computerized control of the voltage.

32. The method of detecting a defect according to claim 26, in which the sub-steps of making a plurality of elevated temperature thermographs include allowing a period temperature stabilization after a respective voltage is first applied.

33. The method of detecting a defect according to claim 26, in which the sub-step of making an ambient-temperature thermograph comprises stabilizing the sample at ambient temperature by rapidly flowing gas from a compressed gas source around it at an effective temperature not exceeding ambient temperature.

34. The method of detecting a defect according to claim 27, in which the sub-steps of making a plurality of elevated temperature thermographs include employing computerized control of each respective period of time.

35. The method of detecting a defect according to claim 27, in which the sub-steps of making a plurality of elevated temperature thermographs include allowing some temperature stabilization after an elevated temperature is reached.

36. The method of detecting a defect according to claim 27, in which the sub-step of making an ambient-temperature thermograph comprises stabilizing a sample at ambient-temperature by rapidly flowing gas from a compressed gas source around it at an effective temperature substantially equal to ambient temperature.

37. A system for thermally detecting a defect in a test sample of a product, comprising means for forming infrared images of a reference sample that does not have defects at ambient and elevated temperatures and for forming infrared images of the test sample at ambient and elevated temperatures, each of the images comprising a plural-ity of image data points;
means for forming reference sample difference records from the images of the reference sample at the ambient and elevated temperatures, and for forming test sample difference records from the images of the test sample at the ambient and elevated temperatures;
means for forming at least one ratio record from two of four difference records comprising said reference sample differ-ence records and test sample difference records;
means for converting the ratio record into a composite record involving all the difference records; and means for generating a defect indication when the composite record has at least one discrete region of substantially greater than average contrast with respect to surrounding regions.

38. The system of claim 37, further comprising means for enhancing the ratio record to remove potentially false defect indications.

39. The system of claim 38 further comprising means for removing speckle noise from the ratio record.

40. The system of claim 39 further comprising means for preprocessing the thermographs, or the difference records, prior to deriving the ratio record.