Application Note

Electronic Package Fault Isolation Electronic Package Fault Isolation Using TDR Using TDR

This application note was published in EDFAS MicroelectronicFailure Analysis Desk Reference, 5th edition (2004).

IntroductionTime Domain Reflectometry (TDR) measurementmethodology is increasing in importance as a non-destructive method for fault location in electronic pack-ages [1-4]. The visual nature of TDR makes it a verynatural technology that can assist with fault location inBGA packages, which typically have complex inter-weaving layouts that make standard failure analysistechniques, such as acoustic imaging and X-ray, lesseffective and more difficult to utilize [5].

In this paper, we will discuss the use of TDR for non-destructive package failure analysis and fault isolationwork. We will analyze in detail the TDR impedancedeconvolution algorithm as applicable to electronicpackaging fault location work, focusing on the opportu-nities that impedance deconvolution and the resultingtrue impedance profile opens up for such work. We willdiscuss the place of TDR in the overall failure analysisprocess, and present examples of proper fault isolationtechniques.

TDR FundamentalsTDR was initially developed for fault location of longelectrical systems such as cables. Currently, high-performance TDR instruments, coupled with add-onanalysis tools, are commonly used as the tool ofchoice for failure analysis and signal integrity charac-terization of board, package, socket, connector andcable interconnects. Such high-end TDR equipment iscurrently available from two manufacturers - Agilentand Tektronix.

Based on the TDR impedance measurements, thedesigner can perform signal integrity analysis of thesystem interconnect, and the digital systemperformance can be predicted accurately. A failureanalyst can use TDR impedance measurements tolocate an interconnect fault more accurately and quick-ly, allowing the analyst to focus on understanding thephysics of the failure at this failure location. A typicalsystem required for such work will consist of a TDRoscilloscope, a probing or fixturing setup (for example,from Cascade Microtech), and analysis software, suchas IConnect® from TDA Systems.

The TDR instrument is a very wide bandwidth equiva-lent sampling oscilloscope (18-20 Ghz or even more)with an internal step generator. It is connected to theDevice Under Test (DUT) via cables, probes and fix-tures. It delivers a fast rise time to the DUT, andbased on the reflection from the DUT the failure ana-lyst can perform the fault isolation analysis of the DUT.

Copyright © 2004 TDA Systems, Inc. All Rights Reserved.

1

Figure 1. TDR instruments from Agilent and Tektronix

Figure 2. A typical failure analysis setup includes a TDRinstrument, a probing or fixturing setup, and analysis soft-ware

Introduction 1TDR Fundamentals 1

TDR Multiple Reflection Effects 3TDR Resolution and Rise Time 4TDR Measurements of "Splits" and "Stubs" ………………. 5TDR Probing and Fixturing 5Using Good Measurement Practices 7

Failure Analysis Goals and Methods 8Signature Analysis 9Comparative Analysis 9Additional Considerations for Package

10Signal-to-Ground Short Failures 12Plane-to-Plane Short 13Plane-to-Plane Short Text Example 13

Summary, Conclusions and Future Work ……………… 14References 15……………………………………………………….

…………………………………………

……………………………………………………………………

…………………………

…………………………

Open Failure Analysis

………………………………………………………………………………

…………………………………………

…………………………………………………………………………………………………

……………………………………………………

TDR is very similar to X-ray and acoustic imagingtechniques in that it sends the signal to the DUT andlooks at the reflection to obtain the information aboutthe DUT. The difference between X-ray or acousticimaging and TDR is in the type of signal and the typeof propagation media for the signal. X-ray and acousticimaging use X-ray and acoustic stimuli corresponding-ly, propagating through the free space to the DUT,whereas TDR uses fast-electrical-step stimulus, deliv-ered to each trace in the DUT via electrical cables,probes, and fixtures.

A direct electrical contact between the TDR instrumentand the DUT is required to perform the measurement.In addition, not only the signal, but also the groundcontact must be provided in order for the TDR signalto provide meaningful information about the DUT.Without a good ground contact, the TDR signal will nothave a good current ground return path, and the TDRpicture will be extremely hard to interpret.

Because of the wide bandwidth of the oscilloscope,and to ensure that this bandwidth and fast rise timecan be delivered to the DUT, one must use high-quali-ty cables, probes, and fixtures, since these cables,probes and fixtures can significantly degrade the risetime of the instrument, reduce the resolution, anddecrease the impedance measurement accuracy.

A typical TDR oscilloscope block diagram is shown inFigure 4 below. The fast-step-stimulus waveform isdelivered to the DUT via electrical cable, probe, andfixture interconnects. One way to think of the incidentTDR step is as a wave front propagating through theinterconnect and reflecting back from thediscontinuities. The superposition of all the wavefronts, reflected from all discontinuities, is what isdisplayed on a TDR oscilloscope.

The waveform reflected from the DUT is delayed bytwo electrical lengths of the interconnect between theDUT to the TDR oscilloscope, and superimposed withthe incident waveform at the TDR sampling head(Figure 5). The incident waveform amplitude at theDUT is typically half the original stimulus amplitude (V)at the TDR source. The smaller DUT incidentwaveform amplitude is due to the resistive dividereffect between the 50 Ohm resistance of the sourceand 50 Ohm impedance of the coaxial cables connect-ing the TDR sampling head and the DUT.

Equivalent resistance of the TDR source Rsourcedefines the characteristic impedance of themeasurement system. Since Rsource is 50 Ohm forhigh-performance TDR instruments available today,using non-50 Ohm cables and probes can produceconfusing results. Unlike with a regular oscilloscope,no active probes or resistor divider probes are allowedfor use with TDR.

TDR does not provide an optical image of the pack-age, but rather an electrical signature of the trace inthe package. Because of the nature of the informationthat TDR provides, it is important to be aware of typi-cal TDR signatures that correspond to simple packagefailures, such as a short or an open connection (Figure6).

2

TDR SAM X-rayStimulus type Electric Acoustic X-ray

Stimulus delivery medium

Electrical wires Water Air

Direct contact required?

Yes, signal and ground

No No

Output presented for analysis

Package trace reflection profile

Optical image

Optical image

Ability to locate failures between package or board layers

Good Poor Poor

Table 1. Comparison between TDR, SAM and X-Ray failureanalysis techniques

TDRInstrument

TDR probe withsignal and ground

connections

Cables

Figure 3. TDR is connected to the DUT via cables, probesand fixtures. A direct electrical contact to the DUT isrequired for both signal and ground pins of the TDR probe

Ztermination

Cable: Z0, tdRsource

TDR

Osc

illosc

ope

Fron

t P

anel

Rsource = 50 ΩZ0 = 50 Ωthen Vincident = ½V

VincidentVreflected

V

DUT: ZDUT

Figure 4. TDR oscilloscope equivalent circuit

V

½ V

0

TDR Open

TDR Short

2•tcable

Vincident = ½ V

0

ZDUT > Z0

ZDUT < Z0

V •ZDUT / (ZDUT + Z0)

Vmeasured=Vincident+Vreflected

Figure 5. The incident waveform is delayed by twice thelength of the interconnect between the DUT and the TDRoscilloscope, and is divided in half by the resistor dividereffect between resistance of the TDR source and resistanceof the interconnect to the DUT

Note that everything in TDR is round trip delay. Thisapplies not only to the cable, probe and fixture inter-connecting the TDR oscilloscope to the DUT, but alsoto all delay measurements on the DUT itself. In orderto obtain an accurate delay readout, the designer hasto divide the measured delay by 2.

After the round trip delay of the cable, the voltagereflected from the DUT arrives back to the oscilloscopeand is added to the incident voltage on the oscillo-scope to produce the measured voltage values. Theoscilloscope then converts these voltage values intothe values for reflection coefficient and impedance. Itis the impedance and delay that the failure analyst ismost interested in, and the accuracy and resolution ofthe impedance and delay measurements is whatdetermines the accuracy of fault isolation.

The faster the rise time that the TDR interconnect candeliver to the package under test, the smaller the sizeof the discontinuities that can be resolved with a TDRoscilloscope. Available TDR instrumentation providesvery fast rise times; reflected signal rise times of theorder of 25-35 ps can be observed at the TDR oscillo-scope. However, poor quality cabling and fixturing canquickly degrade the TDR instrumentation rise time anddecrease the instrument resolution.

It is important to wear good personal ESD protectionwhen working with high-performance TDRoscilloscopes. TDR instruments are ESD-sensitive,high-precision and high frequency instruments.Personal ESD protection (e.g., anti-static strap con-nected to the instrument) will protect the instrumentwhile maintaining its performance. Add-on ESD mod-ules in the signal path will degrade the rise time of theinstrument and degrade its resolution. It is alsoimportant to discharge possible charge accumulatedon your probe or cable before making the connectionto the DUT.

In addition to measuring impedance, the TDR oscillo-scope is capable of providing L, C and R signaturesfor the DUT. For example, an experienced TDR usercan, without difficulty, recognize a "dip" in a TDRwaveform as a shunt capacitance, and a "spike" as aseries inductance. Any L and C combination can alsobe represented as shown in Figure 7.

A series C or a shunt L, however, will represent a high-pass filter for the TDR signal, and the resulting reflec-tion from the elements beyond such series C or shuntL can not be interpreted without prior knowledge of theDUT topology.

Additional information about TDR measurement tech-nology and TDR oscilloscopes can be found in refer-ences [6] through [8].

TDR Multiple Reflection Effects

One of the limitations of TDR is the effect of multiplereflections, which is present in multi-segmentinterconnect structures, such as an electrical package.The accuracy of the DUT signature observed at theTDR oscilloscope is dependent on the assumption thatat each point in the DUT, the incident signal amplitudeequals the original signal amplitude at the probe-to-DUT interface. In reality, however, at each impedancediscontinuity, a portion of the TDR incident signal prop-agating through the DUT is reflected back, and only aportion of this signal is transmitted to the nextdiscontinuity in the DUT. In addition, the signal reflect-ed back to the scope may re-reflect and again arrive atthe next discontinuity at the DUT. These so called"ghost" reflections are illustrated on the lattice diagramin Figure 8.

3

V

½ V

0

TDR Open

TDR Short

2•tcable

50 Ω

0 Ω

2•tcable

>1000 Ω

ImpedanceShort

ImpedanceOpen

Figure 6. Open and short connection TDR and impedanceprofile signatures. V is the full voltage amplitude of the TDRstep source; tcable is the electrical length of the cable andprobe interconnecting the TDR oscilloscope and the DUT

L-C discontinuity

C-L discontinuity

L-C-L discontinuity

C-L-C discontinuity

Z0Z0

Z0Z0

Z0Z0

Z0Z0

Shunt C discontinuity

Series L discontinuity

Z0 Z0

Z0Z0

Inductive termination

V Capacitive termination

Z0

Z0

½ V

0

½ V

0

½ V

0

½ V

0

½ V

0

½ V

0

½ V

0

½ V

0

Figure 7. Visual lumped (LCR) interconnect analysis usingTDR

As a result of these re-reflections, the signature of theDUT becomes less clear, and additional processing isrequired using the impedance deconvolution algorithm([9] and [10]), which is currently not available in TDRoscilloscopes. The impedance deconvolution algorithmdeconvolves the multiple reflections from the TDRwaveform and provides the true-impedance-profile forthe DUT, significantly improving the clarity of the DUTsignature and simplifying further analysis of the TDRdata.

For example, if a failure analysis technician were look-ing for an open failure in an electrical package, TDRdata by itself would not have been sufficient to locatethe position of the failure (Figure 9) as there appearsto be multiple potential fault locations. The true-impedance-profile provides an exact location of theopen in the DUT whereas the TDR waveform by itselfprovides confusing information about the location ofthis open. In addition, the impedance profile, being anexact signature of the DUT, is relatively easy to corre-late to different layers in a BGA package. Such corre-lation is practically impossible with a TDR waveformalone.

An additional advantage that the true-impedance-profile provides is that it is very easy to evaluatecapacitance or inductance of an impedance profilesegment using the following equations:

The type of discontinuity (inductive or capacitive) thatwe observe in the impedance profile, can also beeasily identified - "dips" in the impedance profile cor-respond to the capacitive discontinuity, and "peaks"correspond to inductive discontinuity. Being able toestimate the value of capacitance or inductance forany given segment can be a significant help in under-standing which package segment is being analyzedand in locating the failure more accurately.

Before discussing package failure analysis techniquesusing TDR in further detail, it is imperative to note theimportance of obtaining a good quality TDRmeasurement and a clean impedance profile. Withouta good TDR measurement for the DUT and thereference, the true-impedance-profile is likely to becomputed incorrectly, and both TDR data for the DUTand the true-impedance-profile will provide a confusingpicture.

TDR Resolution and Rise time

The issues of TDR resolution are often misunderstoodor misrepresented, because the TDR resolution isbelieved to be completely governed by the followingrule of thumb. Two small discontinuities, such as twovias in a PCB, can still be resolved as two separateones, as long as they are separated by at least ½ theTDR rise time:

If these two vias are not separated by half the TDRrise time as it reaches the vias, they will be shown byTDR as a single discontinuity. Assuming we use goodcables, probes, fixtures, and we can deliver the full 30-40ps rise time of the instrument to the discontinuitiesin question, the minimal physical separation betweenthese vias will be 15-20ps. For FR4 board materialwith dielectric constant εr=4, this results in 2.5-3mm(0.1") resolution. Often this number (or other similarcalculation) is quoted as the TDR resolution limit.

Z0 Z1 Z2 Z3 Z4

Vincident1(1) Vtransmitted1(1)

Vreflected1(1) Vreflected2(1)

t0 Vreflected1(2)

Vreflected1(3)

Time Direction of forward propagation

4

Figure 8. Lattice diagram of TDR waveform propagatingthrough a DUT with multiple impedance discontinuities

Figure 9. True-impedance-profile vs. the raw TDR waveformfor a BGA package. The true-impedance-profile providesmuch more accurate information about the failure location

∫⋅=2

1)(

121 t

t

dttZ

C ∫⋅=2

1

)(21 t

t

dttZL (1)

tseparate To resolve a1 and a2 asseparate discontinuities:

tseparate > tTDR_risetime /2 a1 a2

tsingle a1 is not observed iftsingle << tTDR_risetime

(tsingle < tTDR_risetime / 10) a1

Figures 10a and 10b. TDR resolution rules of thumb

(10a)

(10b)

However, in real-life situation, the designer typically islooking to observe or characterize a singlediscontinuity, such as a single via, or a single bondwirein a package, rather than several of such vias or bond-wires! In this case, the above rule is totally irrelevant,and TDR can allow the designer to observediscontinuities of 1/10 to 1/5 of the TDR rise time,bringing the numbers above to 5ps or less than 1mm(25milliinches) range (Figure 10b).

Furthermore, there are well-developed relative TDRprocedures for observing and characterizing evensmaller discontinuities, such as the golden devicecomparisons for failure analysis [1-4].

In addition, faster TDR modules are available fromPicosecond Pulse Labs, which makes it easier toresolve some of the finer discontinuities and faults.

TDR Measurements of "Splits" and "Stubs"

If the package trace under test splits into two or moredirections, the TDR instrument shows the sum of allreflection from all the N legs in the split, but cannotseparate which reflection came from which leg in thesplit. This means the failure location in case of thetraces with splits or stubs can be extremelychallenging.If the splits are of the same impedance and delay (assometimes is the case in a "star" interconnect topolo-gy), they can be simply represented by transmissionlines running in parallel, and the impedance measuredby the TDR oscilloscope equals Z1 / N, with the delayof each trace being equal to the delay measured byTDR (Figure 12).

In case of a stub (which often takes place in a daisychain configuration), if the length of each stub on themain bus is much shorter than the rise time of the sig-nal propagating through the bus, the stub can be treat-ed as lumped capacitances loading the main bus, thussimplifying the measurement problem.

TDR Probing and Fixturing TDR is delivered to the DUT via electrical cable,probe, and fixture interconnects. The quality of theseinterconnects is the key to obtaining a goodmeasurement. As noted before, poor quality cablingand probes can degrade the TDR rise time anddecrease the resolution of the instrument. In addition,when computing the impedance profile, it is necessaryto have a clean reference short or open waveform;without a good reference, we are not likely to get aclear signature of the DUT. Because of these factors,good quality microwave probes and cables arerequired to obtain a good quality TDR measurement.

Fixtures, probes, and probing stations for package fail-ure analysis work are available from various manufac-turers. A full-featured failure analysis probing stationcan provide easy viewing and access to the packagewith a probe, and enable a failure analyst to performat-temperature analysis of the package failures.

5

Figure 11. Picosecond Pulse Labs fast TDR add-on module

Star topology

Z0

Z1

Z1

Z1

Z1

Z1

Daisy-chain topology

Cstub5

Z1

Cstub2

Z1

Cstub3

Z1

Cstub4

Z1

Cstub1

Z0

td stub << trise

Figure 12. Taking TDR measurements of splits in a startopology and stubs in a daisy-chain topology

TDR cables and probes will degrade the rise time ofthe signal measured on the TDR oscilloscope approxi-mately as follows:

where tTDR is the rise time measured on the TDRscope with no cable connected, and f3dB is the 3dBbandwidth of the cable and probe. The factor of 2 inthis equation is due to the fact that the signal has totake a roundtrip through the cable before it is observedand measured on the oscilloscope. Specifying a cablewith a 3dB bandwidth (f3dB) of about 10 Ghz for thescope with its own rise time of 30ps, will result in therise time at the cable end of about 58ps. Specifying3dB bandwidth of 17.5 Ghz will give the rise time endof the cable of about 40ps.

In an application where the TDR cable length can belimited to less than 2 ft, requesting a "lowest-loss" flex-ible cable from your favorite high quality low cost coax-ial cable manufacturer would be sufficient. If you areworking with a 3-4 ft cable, however, or require fullresolution and rise time that the oscilloscope can offer,you will have to work with a high-end microwave cablemanufacturer. Semi rigid cables can provide betterperformance than flexible ones, but are more difficultto use. SMA connector is commonly used in TDRcables, since it provides acceptable performance, andcan be mated directly to the 3.5mm connector foundon 20Ghz TDR sampling modules. For even faster risetime, a higher bandwidth microwave connector, suchas 2.92mm or even 2.4mm may be required1.

When using a probe for taking a TDR measurementon a package, the designer has to define a groundlocation near the signal location. If such ground loca-tion is not available, or if the spacing from signal toground varies widely across the PCB, the designermay have to use a probe which has a long groundwire, or a variable length wire. For a probe with a longground wire, the parasitic inductance will be very large,and will not allow the designer to obtain a good qualityTDR measurement. Variable length ground wires, andvariable pitch (signal-to-ground spacing) probes do notprovide sufficient measurement repeatability, and willnot provide accurate impedance measurement resultsor signal integrity interconnect models.

In many cases, a simple, inexpensive and convenientTDR probe can be obtained by using a 3 inch length ofsemi-rigid coaxial cable with an SMA connector,

exposing the center conductor of such cable, andeither using the sleeve of the semi-rigid coax as theground contact, or attaching a ground wire. Using dif-ferent diameter coax will result in different probe pitch,and making the center and ground conductors shorteror longer can provide the right trade-off between con-venience of use and performance2. Such probes arealso available commercially, Figure 13.

A package fixture may be used to conveniently providea connection to the balls or contact pads of the pack-age. Such fixture must also ensure contact for the sig-nal connection, and it also must provide a groundplane to allow accurate impedance measurement.(Figure 14). An automated version of such fixture maybe even more beneficial for high-throughput fault isola-tion work.

A homemade version of the same fixture can be alsomade. However, the designer of such a fixture needsto ensure both good low-inductance connection to thesignal pins and a good ground connection, such asprovided in a commercial fixture shown on Figure 14.

2

3

2 35.02

⋅⋅++==

dBTDRmeasured ftt

6

(2)

13.5mm connector is specified to operate to a 26.5 Ghz,whereas a typical SMA is rated to 12.5 or 18 Ghz. 2.92 mmconnector is specified to 40 Ghz, and 2.4mm to 50 Ghz.

Figure 13. TDA Systems QuickTDR™ probe supported byCascade Microtech EZProbe™ positioner. The spacingbetween signal and ground in the tip of the probe isextremely small to ensure best performance

2However, a ground lead that is 10mm long will probablyproduce a 10nH parasitic inductance and pretty muchdestroy the measurement accuracy.

Using Good Measurement Practices

To obtain good quality impedance, signal integritymodeling, and failure analysis data, it is important tofollow general good measurement practices whenusing a TDR oscilloscope. The instrument should beturned on and its internal temperature should beallowed to stabilize for 20-30 minutes before perform-ing any measurements. Calibration, compensation andnormalization for the instrument must be performedregularly, as specified by the instrument manufacturer.The internal instrument temperature must be within thespecified range from the calibration points for the giveninstrument.

To maximize the resolution of the scope, particularly inthe time axis, it is important to zoom in on the DUT -but at the same time to allow a window that is suffi-ciently long to include all the reflections related to theDUT. A window that is too short may prevent thedesigner from obtaining complete and accurate infor-mation about the DUT. When the designer intends toperform true impedance profile analysis, asimplemented in IConnect TDR software, it is alsoimportant to window out the transition related to thesampling head to the cable interface, and focus on theDUT portion of the waveform, so as to ensure that theimpedance deconvolution algorithm, discussed above,could perform correctly.

It is critical to use a torque wrench when mating anytwo connectors in your probing, cabling and fixturingsetup. Such torque wrenches ensure a repeatableconnection between the connectors, thus providingbetter measurement repeatability. These torquewrenches are available from most TDR manufacturersand many microwave component suppliers. It is alsoimportant to clean the RF connectors used in the prob-ing / fixturing setup with isopropyl alcohol and lint freeswab.

7

Figure 14. Altair Microwave BGA package probing fixtureprovides a high-frequency connection for four pins of thepackage simultaneously. The body of the fixture serves asthe ground plane

Figure 15. Proper windowing of the TDR waveform for fur-ther analysis in IConnect TDR software. The properly win-dowed waveform will exclude the first transition, correspon-ding to the interface between the TDR sampling head to thecable, but will keep the window sufficiently long to allow allthe reflections corresponding to the DUT, to be included inthe window

Failure Analysis Goals and MethodsThe goal and the task of the failure analyst is todetermine whether there is a possible connection fail-ure in the given package trace, and what the exactposition was when the failure occurred. Once the posi-tion of the failure is determined, further analysis canbe performed to determine the physical cause and thenature of the failure, possibly with destructive analysismethods. Thus, in this scenario TDR is a fault isolationor a fault locator technique, allowing the failure analystto quickly find the failure, and analyze it using other,possibly destructive, analysis techniques.

For example, the following picture illustrates how TDRwas used to locate a short in a package.

And the picture that follows shows an X-ray image ofthe same short failure.

The following four pictures are additional examples ofTDR open signatures in a package, a short signaturein a die, and corresponding optical images.

8

Figure 18. TDR result showing an open in the package sub-strate

Figure 19. Optical image showing a broken trace in thepackage substrate

Figure 20. TDR result showing I/O short in the die

Figure 17. X-ray image of two solder bump shorts

Figure 16. TDR result showing I/O short (AF9 and AE8) in apackage

Typical approaches that can be used to determinewhether there is a failure present are signature analy-sis, where the package trace true-impedance-profiledata is analyzed for known failure signatures, andcomparative analysis, where the package trace data iscompared to the data of a trace in a known good pack-age. Both approaches will be applied to the true-impedance-profile data obtained from the TDR usingthe impedance deconvolution algorithm as it isimplemented in TDA Systems' IConnect® software.

As Figure 9 on page 4 indicates, the true-impedance-profile provides a much clearer picture of the failuretype, and also enables the user to easily determine theexact position of the failure in an electrical sense, i.e.,in terms of electrical length of the interconnect inpicoseconds. Additional analysis must be performed todetermine the physical location (in millimeters or milli-inches) of the failure with the goal of locating the pack-age element that is failing. The true-impedance-profileprovides the user with a way to correlate the TDR datato the specific layers in the package, as well as pro-vide an estimate of a constant that would allow theuser to convert the electrical length in picoseconds intophysical lengths in mils.

Signature AnalysisIn the true-impedance-profile, open and short failurescan be easily identified as 0 Ohm or very lowimpedance readout for the short and very high (1000Ohms or more) impedance readout for the open(Figure 6, Page 3). The exact electrical position of ashort or an open can be easily identified in the true-impedance-profile, even in the presence of multiplereflections, as previously described.

In the following example (Figure 22), the known goodBGA package (ZlineGood.wfm) was analyzed along-side a suspect package (ZlineBad.wfm). The fixture-impedance-profile (ZlineFixture.wfm) is shown forreference. The known good package impedanceprofile ends with a large capacitive dip, corresponding

to the input package capacitance. An open failure isclearly observed in the BGA package at about 80 psinside the package (160 ps roundtrip delay).

So called "soft" failures, i.e., partly shorted or partlyopen leads, can also be identified using the signatureanalysis, but their impedance profile and TDR signa-tures must be identified beforehand. The only alterna-tive to knowing the soft failure signature beforehand isto observe the changes in capacitance of the knowngood device compared to the failing device.

TDR has specific signatures for the open and shortconnections, as shown in Figure 6, and can also beused for identifying the failures. However, in multi-seg-ment structures, such as BGA packages, the exactlocation of the failure can be difficult to determinebecause of the multiple reflection effects. This is whywe want to emphasize the importance of using the trueimpedance profile when performing fault isolation onelectronic packages.

Comparative AnalysisComparative package failure analysis, as the nameimplies, relies on comparison of the known goodwaveform to the suspect waveform. Even thoughsome discrepancy between different measurementsmay still be observed due to measurement repeatabili-ty, comparative analysis utilizing the true-impedance-profile waveforms, computed using IConnect, yieldsvery quick and intuitive results.

Consider the following example. In Figure 22, thepackage failure is identified as an open failure. InFigure 23, the analysis is continued by comparing thefailed waveform to the package substrate waveformonly, without connection to the die. The challenge is todetermine what package component is failing basedon this comparative analysis. Because the failedimpedance profile waveform overlays directly over thesubstrate waveform, it is easy to deduce that the likely

Figure 21. Optical photo of an EOS short in the die on pinG3

Figure 22. Signature analysis of a BGA package failureusing the true-impedance-profile in IConnect

9

failure source is the broken connection between thepackage and the die. Again, the large capacitive dip isdue to the input capacitance of the die.

Based on this analysis, a failure analyst can focus onthe connection to the die area, and use additional fail-ure analysis techniques to determine the physics ofthe failure.

Thus, in comparative analysis at the very least weneed to have a known good device. For a typical pack-age, one of the key impedance profile features differ-entiating an open fault in the package or bondwirefrom an open fault in the die itself is a dip in theimpedance profile corresponding to an input diecapacitance and presenting itself shortly before theopen impedance signature. The presence of this char-acteristic dip in the impedance profile indicates a goodconnection to the die, whereas its absence indicates aproblem in the package structure. Such a capacitivesignature can be observed much more readily on theimpedance profile waveform than on a raw TDRwaveform. An additional known good bare packagesubstrate can also be useful in identifying the exactlocation of the failure.

An important issue when performing comparativeanalysis is measurement repeatability. Following goodgeneral measurement practices, such as:

· maintaining TDR instrument calibration

· keeping the instrument well-warmed in a lab withconstant ambient temperature

· maintaining the probe or cable position and spacingbetween the probe signal and ground during themeasurement

will enable the analyst to minimize any non-repeatabili-ty errors. However, a failure analyst must be awarethat small differences between different impedanceprofiles may actually result from measurement non-

repeatability, rather than failures in the package undertest.

For example, because of the differences between thegood package impedance profile (ZlineGood.wfm) andbad package impedance profile (ZlineBad.wfm) in theoutlined region of Figure 24, a failure analyst may viewthe differences between the good and bad waveformsin the selected region as the cause for the failureobserved in the later portion of the impedance profile.However, because we are working with the impedanceprofile and not the TDR waveform, any effect of thereflections in the selected region on the rest of theimpedance profile waveform is minimal. With that inmind, the differences between the two impedance pro-files are too small to be viewed as the cause of thefailure. And, one can comfortably conclude that thefailure occurred in the later portion of the package (inthis case, again, it is a failure of the package-to-dieconnection.)

Additional Considerations for PackageOpen Failure AnalysisThe true-impedance-profile is very powerful because itopens up other interesting venues for FA on electronicpackages. For example, because the true-impedance-profile represents an exact signature of the DUT, onecan now analyze the package impedance profile andquite easily correlate it to the physical layers in theBGA package, which can be observed in the packagelayout or drawing.

Consider the following two package samples with thefollowing simplified trace layouts in Figure 25. The twopackages are quite similar, except that the trace lead-ing to the via connecting the package trace to the dieis significantly longer for package 1. Both of thesepackages were analyzed with a TDR instrument andan impedance profile in IConnect. In both cases agood package sample and a sample with a failure ofthe connection between the package trace and the diehas been analyzed.

Figure 23. Comparative analysis for a BGA package. Thebad impedance profile waveform clearly indicates an openfailure signature. Comparing it to the package substratewaveform only without connection to the die, allows pin-pointing the likely failure source - a broken connection tothe die

Figure 24. Measurement repeatability considerations

10

The impedance profile enables a simple correlation tothe package geometry (Figure 26). In package 1, theknown good waveform (Zline1good.wfm) shows a seg-ment with inductive behavior (estimated to be about2nH in inductance), correlating to the long packagetrace, then a short segment correlating to the via, andthen a segment correlating to the input capacitance ofthe die. When the connection to the die is broken, thecorresponding waveform (Zline1bad.wfm) still showsthe long trace in the package, but does not go into thecapacitance of the die (estimated to be 800fF). Finally,the shorter second package trace correlates to theshorter section in the impedance profile waveform(Zline2good.wfm), whereas for the failed trace in pack-age 2, the impedance profile goes up to highimpedance at a much earlier point. The estimates forthe inductance of the trace and input capacitance ofthe die match the expected numbers well, which pro-vides further confirmation for the accuracy of theanalysis of the failure type and location.

Once the correlation from the physical package struc-ture to the impedance profile waveform has beendetermined, the location of the fault in the packagecan be found easily.

In addition, since the overall physical length of thepackage trace can be quickly found from the packagelayout, and the impedance profile provides exact infor-mation about the electrical length of the package trace,this correspondence can provide a reasonably goodestimate of the physical location of the failure. Forexample, if the package layout software gives a read-ing for the overall package trace length of ltotal meters,and the true-impedance-profile shows that the pack-age length is td total seconds, then the average relativevelocity of propagation through the package can beestimated as:

where VC is the speed of light. For example, the differ-ence between the length of the traces in package 1and package 2 is 45 ps (90 ps roundtrip). Based onthe layout file data, the corresponding physical lengthis 10 mm, which provides an estimated relative veloci-ty of propagation of 4.5 ps/mm, or 0.74 the speed oflight.

In addition, if a correlation between an electrical posi-tion in seconds to the physical position in metersneeds to be estimated, it can be done using the follow-ing equation:

Using equation (4), one can estimate the relative posi-tion of the failure within a layer, if it is suspected thatthe failure actually occurred within a layer.

One can extend the computation above to determinethe average dielectric constant εr through the packageusing the following equation:

where Vprop average is the average signal propagationvelocity through the package, Vc is the speed of lightin vacuum. We can rewrite this equation as:

Now, the computed εr value can be entered into thesoftware, and the results can be displayed asimpedance vs. physical length (millimeters, inches, orfeet).

Simplified package 1 trace geometryBall

Trace onlayer 1

Via to die

Die

Simplified package 2 trace geometry

Ball

Trace onlayer 2

Via to die

Die

Figure 25. Sample package trace geometries used for corre-lation to the impedance profiles

Figure 26. Layer correlation and distance analysis inIConnect based on the impedance profiles of two packageswith similar layouts

Ctotald

totalaverageprop Vt

lV 1⋅= (3)

totald

totald t

ltl ⋅= (4)

averager

caverageprop

VVε

= (5)

2

=

averageprop

caverager V

Vε (6)

11

Clearly, equations (3) through (6) are only estimates.The propagation velocity will vary through the differentlayers in the package. To get a more accurate valuefor the propagation velocity one needs to do extensivecharacterization of the package substrate material, aswell as other characteristics. Such characterization isvery time consuming and requires that special teststructures be laid out on the material under test ([11,12]). Because of such complexity, the exact data aboutthe velocity of propagation through the separate pack-age layers is rarely available to a failure analyst. Amuch easier approach is to correlate the layers in thepackage to the segments in the true-impedance-profileand use equation (3) to estimate the propagationvelocity in each layer. However, sufficient resolution ofthe TDR instrument is required to resolve the layers,which can be on the order of 10 ps or less in length.

An attractive approach for a failure analyst could be tomodel the package under test, and then attempt topredict the TDR waveform of the package trace viaSPICE or full-wave circuit simulations. The problemwith this approach is, again, that the properties of thepackage material must be known with a reasonablyhigh level of accuracy in order to ensure that thesimulation predicts the TDR waveform correctly, unlessthe package model has been directly extracted fromTDR measurement.

Signal-to-Ground Short FailuresExperiments have demonstrated that a signal toground plane short can be located relatively easilyusing the same techniques as those used for locatingan open fault. Since the trace has a certain amount ofcharacteristic impedance, a short failure will clearlyexhibit itself on the true impedance profile, as a rapiddecrease in impedance until this impedance reacheszero Ohm.

Since in Figure 27 we can observe that the shortwaveform goes towards zero Ohm right where thebare package substrate waveforms goes into an open(high impedance), and right before the good devicewaveform goes to the die capacitance, we concludethat the short is located at or near the connection tothe die.

In this example, the electrical length is 63ps, and thephysical length is 8.7mm, which gives an averagepropagation velocity of 7.24ps/mm or 0.138mm/ps.Using this number, and knowing that the speed of lightin vacuum is 0.305 mm/ps, we obtain the average εr of4.9. Then, one can display the data in the software asimpedance versus physical time.

The actual dielectric constant (εr) for the substrate hasbeen measured to be 4.2 at 1 MHz, which comes fromthe material property data provided by the substratevendor. Using εr=4.2 to recalculate the physicallength, the measured total length of the trace is 9.4mm, which is very close the actual CAD layout lengthof 8.7 mm (~8% error). It is a very good correlation,considering that the εr would be higher at higher fre-quency for the TDR measurement, and that we arelooking for the average εr through the whole package.The localized εr variations, non-homogeneity, effects ofdifferent conductor shape, all contribute to the differ-ence between the vendor supplied εr and the meas-ured value. However, the overall correlation weobserve is considered to be quite good.

Figure 27. Locating a signal to plane short failure using theimpedance profile signature. By comparing the shortwaveform (ZlineShortB6.wfm) to the substrate waveform(ZlineSubstrate.wfm), one easily concludes that the short islocated near the connection to the die

Figure 28. Impedance vs. distance in IConnect TDR soft-ware. The fault occurred about 63ps or 14mm inside thepackage

12

Plane-to-Plane ShortLocating a plane-to-plane short, such as ground planeto power plane short, is a more difficult task. In thecase of a signal trace, the characteristic impedance ofsuch a trace is normally in the range of 30 to 80Ohms, and observing the change from this range tozero Ohm in the impedance profile waveform is rela-tively easy. The power plane, however, presents muchlower impedance to the TDR signal, typically on theorder of less than 0.5-2 Ohm, and the change fromthat impedance to zero does not always allow the fail-ure analyst to use the impedance profile effectively tofind the exact location of the short between the planesusing either time or distance.

In this paper, however, we propose two comparativetechniques for plane-to-plane short location, bothbased on secondary information in the TDR data. Onetechnique looks for the difference in the secondaryreflections in the TDR waveform, and can be per-formed with the raw TDR data or using the trueimpedance profile. The second technique looks at theinductance of the current return path, which can becomputed using IConnect software, based on theJEDEC standard described in [13, 14]. Smallerinductance indicates a shorter distance to the short,and by comparing the failing device measurement tothat of the good device and a shorted package sub-strate, one can determine the relative position of theshort failure. For both techniques, repeating themeasurements multiple times to ensure good repeata-bility is key to finding a fault.

Plane-to-Plane Short Test ExampleConsider the following simple test board example(Figure 29):

This test board consists of two planes, which have viaprobe points at multiple locations on the board. Usingthese vias, the failure analyst can perform TDRmeasurements of the two planes on one side of theboard, while at the same time shorting the planes atthe other side and attempting to locate the position ofthe short. The board is about twice the size of a typicalBGA package, which makes it an easier test case. Thefollowing table summarizes the expected closeness ofthe shorted probe point to the top left probe point,where the TDR signal is applied, based on visualanalysis of the board.

We applied signal every time at the same location (topleft probe point), while creating a short between theplanes by connecting the via from the bottom plane tothe top plane contact. We shorted the two planestogether at each of the remaining probe point locationsand acquired the corresponding TDR waveform(Figure 30).

As one can see from Figure 30, there is little delaybetween the different waveforms. The inset, however,demonstrates, that the waveforms do exhibit a differ-ence in position and waveshape. If a failure analysttried to analyze this difference and determine whichshort is closer to the point where the TDR signal isapplied, the following order could be established asshown in Table 3 (from closest short to the probe pointto the farthest).

3"

1 3/8" (35mm

)

Top planeVias to bottom plane

Center probe point

Bottom center probe point

Bottom left probe point Bottom right probe point

Typical probe placement

Center left probe point

Top leftprobe point

Figure 29. Test board for plane-to-plane short failure loca-tion. The board consists of two planes that can be probedwith TDR at multiple locations at one side of the board andshorted at another side

Left Center RightTop X 2 6Center 1 4 7Bottom 3 5 8

Table 2. Physical closeness of short point to the top leftprobe point (closest, 1 to farthest, 8), based on visual analy-sis

Figure 30. Locating plane-to-plane short on the test board.The fault location is based on the secondary TDRmeasurements, such as secondary reflection delay andinductance measurement

13

By comparing these results with the expected resultsfrom Table 2, a failure analyst can observe a good cor-relation. The only questionable result is that the centertop probe point comes after the center point. The dif-ference between the two waveforms corresponding tothose probe points, however, is small, and can beattributed to measurement repeatability issues.

To confirm our conclusions, we compute the totalinductance of the plane for each measurement inIConnect software. This measurement is a good figureof merit for the current return path through the plane,which, in turn, is a good indicator of where the shortmay have occurred.

The following table summarizes the inductance data,listing the shorted probe point inductancemeasurements from the smallest to the largest. All theinductance measurements have been in 2-4 nH range.

Again, the center point and the bottom left point areoutliers, but otherwise this table correlates well to theTable 2 of expected physical closeness.

These "outliers," or incorrect data points, may comefrom some minor details and changes in the currentreturn path between the planes. Vias and other planeopenings, such as those on the test board in questioncan affect this return path. All these issues indicatethat the failure analyst must apply these techniques toplane-to-plane short analysis with the good under-standing of the TDR measurement technique, andmust study the layout and structure of the packagecarefully. Repeated measurements may be necessaryin order to prove the actual location of the short.

As a final note, additional analysis using triangulation(making TDR measurements from three differentpoints on the plane) would enable even easier locationof the position of the short between the planes.

Summary, Conclusions and FutureWorkIn this paper, we discussed TDR measurement tech-nology as it applies to the failure analysis of electronicpackaging. We analyzed the impedance deconvolutionalgorithm, and demonstrated the advantages that thetrue-impedance-profile (resulting from applying thisalgorithm to the TDR data), provides for a packagefailure analyst over a simple TDR data set, for bothsignature and comparative package failure analysis.Additional analyses were presented, which can be per-formed on the true-impedance-profile, and that can fur-ther simplify the location of the failures in electronicpackaging.

We have analyzed the application of the TDRmeasurement techniques and the true impedanceprofile to finding the location of signal-to-ground andplane-to-plane shorts in electronic packages. Locatinga signal-to-ground short has been shown to present lit-tle difficulty over a comparable open fault locating task.Plane-to-plane shorts, however, present additionalchallenges, which require more attention to therepeatability and accuracy of the measurements.However, with the true impedance profile and planeinductance analyses, the claim of impossibility of locat-ing a plane-to-plane short is effectively challenged inthis paper.

For future work, more automated fixturing is required,which would enable the failure analyst to increase thethroughput and locate more failures with less timeinvested. Faster rise time TDR modules can producebetter resolution results and allow the analyst to locatethe failure even more easily.

Left Center RightTop X 4 5Center 1 3 7Bottom 2 6 8

Table 3. Electrical closeness of short point to the top leftprobe point (closest, 1 to farthest, 8), based on TDRmeasurement

Left Center RightTop X 3 4Center 1 5 7Bottom 6 2 8

Table 4. Electrical closeness of short probe point to the topleft probe point (from smallest inductance, 1 to largest, 8),based on inductance measurements in IConnect TDR soft-ware

14

References

1. C. Odegard and C. Lambert, ComparativeTDR Analysis as a Packaging FA Tool, -Proceedings from the 25th International Symposiumfor Testing and Failure Analysis (ISTFA), 1999

2. D. Bethke, W. Seifert, TDR Analysis ofAdvanced Microprocessors, - Proceedings from the26th International Symposium for Testing andFailure Analysis, 2000

3. D.A. Smolyansky, Electronic PackageFailure Analysis Using TDR, - Proceedings fromthe 26th International Symposium for Testing andFailure Analysis, 2000 (TDA Systems applicationnote PKFA-0700)

4. D.A. Smolyansky, D. Staab, P. Tan, SignalTrace and Power Plane Shorts Fault Isolation UsingTDR, - Proceedings from the 27th InternationalSymposium for Testing and Failure Analysis, 2001(TDA Systems application note PKSH-0102)

5. P. Viswanadham, P. Singh, Failure Modesand Mechanisms in Electronic packages, -Chapman and Hall (1998)

6. TDR Impedance Measurements: AFoundation for Signal Integrity, - TektronixApplication Note 55W-14601-0

7. Agilent High Precision Time DomainReflectometry, - Agilent Application Note 1304-7

8. D.A. Smolyansky, TDR Testing Primer, -Printed Circuit Design Magazine, March 2002 (TDASystems application note TDRP-0402)

9. L.A. Hayden, V.K. Tripathi, Characterizationand modeling of multiple line interconnections fromTDR measurements, - IEEE Transactions onMicrowave Theory and Techniques, Vol. 42,September 1994, pp.1737-1743

10. C.-W. Hsue, T.-W. Pan, Reconstruction ofNonuniform Transmission Lines from Time-DomainReflectometry, - IEEE Transactions on MicrowaveTheory and Techniques, Vol 45, No. 1, January1997, pp. 32-38

11. D.A. Rudy, J.P. Mendelsohn, P.J. Muniz,Measurement of RF Dielectric Properties withSeries Resonant Microstrip Elements, - MicrowaveJournal, March 1998, pp. 22-39

12. D. I. Amey, S.J. Horowitz, TestsCharacterize High-Frequency Material Properties" -Microwaves and RF, August 1997

13. Guideline for Measurement of ElectronicPackage Inductance and Capacitance ModelParameters, - JEDEC Publication #123, JC-15Committee, October 1995

14. D.A. Smolyansky, TDR Techniques forCharacterization and Modeling of ElectronicPackaging, - High Density Interconnect Magazine,March 2001, Part 1 of 2 (TDA Systems applicationnote PKGM-0101)

15

© 2004 TDA Systems, Inc. All Rights Reserved4000 Kruse Way Pl. #2-300, Lake Oswego, OR 97035 USATelephone: (503) 246-2272 Fax: (503) 246-2282E-mail: info@tdasystems.com Web site: www.tdasystems.com

EPFI-0106

Data subject tochange without notice

The Interconnect Analysis Company™