"... In PAGE 71: ... These 166 dies were carefully selected. Table6 -1 lists the summary of the packaged dies for the final package test. Table 6-1 Summary of the Packaged Dies for the Final Package Test defect class total number of dies based on wafer probe results number of packaged dies CUT sampling failures 125 122* VLV-only failures 23 20** IDDQ failures NA 1 good dies NA 166 total NA 309 * Three out of the 125 dies were on the four wafers for studying anomalous IDDQ results from the wafer probe ** Three out of the 23 dies were on the four wafers for studying anomalous IDDQ results from the wafer probe 6.... In PAGE 71: ... Table 6-1 lists the summary of the packaged dies for the final package test. Table6 -1 Summary of the Packaged Dies for the Final Package Test defect class total number of dies based on wafer probe results number of packaged dies CUT sampling failures 125 122* VLV-only failures 23 20** IDDQ failures NA 1 good dies NA 166 total NA 309 * Three out of the 125 dies were on the four wafers for studying anomalous IDDQ results from the wafer probe ** Three out of the 23 dies were on the four wafers for studying anomalous IDDQ results from the wafer probe 6.2 Experimental Setup This section describes only the experimental setup that is related to IDDQ and VLV testing.... In PAGE 72: ... The very slow test timing is 3 times slower than the rated test timing. Table6 -2 lists the clock speeds used at nominal supply voltage. Table 6-2 Clock Speeds at the Nominal Supply Voltage for the Final Package Test test timing clock speed r-rated timing rated (worst-case) speed of each CUT s-slow timing slower than rated timing (2/3 rated) ss-very slow timing much slower than rated timing (less than 1/3 rated) f-fast timing faster than rated timing (15% for MUL and SQR, 5% for others) We used two clock speeds for all Boolean tests at 2.... In PAGE 72: ... Table 6-2 lists the clock speeds used at nominal supply voltage. Table6 -2 Clock Speeds at the Nominal Supply Voltage for the Final Package Test test timing clock speed r-rated timing rated (worst-case) speed of each CUT s-slow timing slower than rated timing (2/3 rated) ss-very slow timing much slower than rated timing (less than 1/3 rated) f-fast timing faster than rated timing (15% for MUL and SQR, 5% for others) We used two clock speeds for all Boolean tests at 2.... In PAGE 72: ...iming dependent. Tables 6-3 and 6-4 list the clock speeds used at 2.5V and 1.7V. Table6 -3 Clock Speeds at 2.5V for the Final Package Test test timing clock speed r-rated timing 1/3 rated speed at 5V ss-very slow timing... In PAGE 73: ... Table6 -4 Clock Speeds at 1.7V for the Final Package Test test timing clock speed r-rated timing 1/5.... In PAGE 76: ...Figure 6-3 Maximum IDDQ Distribution of All 1515 CUTs for IDDQ1 Test Set Figure 6-4 Maximum IDDQ Distribution of the 930 CUTs that Passed All the Boolean Tests at Nominal Voltage for IDDQ1 Test Set Table6 -5 shows the IDDQ measurements for CUTs with different defect classes. The table shows only the CUTs that either were identified in some defect classes or had IDDQ values larger than 3 m A.... In PAGE 76: ... The table shows only the CUTs that either were identified in some defect classes or had IDDQ values larger than 3 m A. The last column in Table6 -5 lists the number of CUTs that had IDDQ values within the current range but passed all other tests. This number could be larger than the ones listed in the table because we only packaged 309 dies and no IDDQ information was available for selecting dies for the final package test.... In PAGE 78: ... Table6 -6 Test Results of the 9 VLV-Only Failures CUT 5V 2.5V 1.... In PAGE 78: ...7V, which we chose based on the Shmoo plot of a good device. The results in Table6 -6 also show that testing at 1.7V was more effective than testing at 2.... ..."

"... In PAGE 23: ... Excitation conditions in confocal microscopy are several orders of magnitude more severe, however, and restrictions imposed by characteristics of the fluorophores and efficiency of the microscope optical system become the dominating factor in determining excitation rate and emission collection strategies (1, 7, 93). Because of the narrow and wavelength-restricted laser spectral lines employed to excite fluorophores in confocal microscopy (see Table1 ), fluorescence emission intensity can be seriously restricted due to poor overlap of the excitation wavelengths with the fluorophore absorption band. In addition, the confocal pinhole aperture, which is critical in obtaining thin optical sections at high signal-to-noise ratios, is responsible for a 25 to 50 percent loss of emission intensity, regardless of how much effort has been expended on fine-tuning and alignment of the microscope optical system (7).... In PAGE 23: ... Traditional Fluorescent Dyes The choice of fluorescent probes for confocal microscopy must address the specific capabilities of the instrument to excite and detect fluorescence emission in the wavelength regions made available by the laser systems and detectors. Although the current lasers used in confocal microscopy (see Table1 ) produce discrete lines in the ultraviolet, visible, and near-infrared portions of the spectrum, the location of these spectral lines does not always coincide with absorption maxima of popular fluorophores. In fact, it is not necessary for the laser spectral line to correspond exactly with the fluorophore wavelength of maximum absorption, but the intensity of fluorescence emission is regulated by the fluorophore extinction coefficient at the excitation wavelength (as discussed above).... ..."

"... In PAGE 1: ...See Jurafsky et al. (1997a), Shriberg et al. (1998), and Stolcke et al. (1998) for other publications describing aspects of this work). Table1 shows a sample of the kind of discourse structure we are modeling and detecting. Besides the usefulness of discourse structure detection for speech understanding, discourse structure can be directly relevant for speech recog- nition tasks.... In PAGE 14: ... Surely it is important to model not only the DA sequence, but also which speaker said what. For example, in the sample conversation of Table1 , the grammar should compute the probability of the second utterance in Channel A being a Ack-Answer, given that the previous utterance was a Statement on Channel B and before that was a Wh-Question on Channel A. We can think of the events in the sequence as consisting of pairs (Ui; Ti), where Ui is a DA label and Ti is a speaker label.... In PAGE 15: ... Results are shown in Table 10. Table1 0: Perplexity when guessing both the DA and turn information. Discourse Grammar Perplexity None 84 Unigram 18.... In PAGE 16: ...end-of-conversation tags: Table1 1: Perplexity of speaker/turn stream only. Discourse Grammar Perplexity None 3 Unigram 2.... In PAGE 16: ... The results are shown in Table 12. Table1 2: DA Perplexity conditioned on turn information. Discourse Grammar Perplexity None 42 Unigram 9.... In PAGE 17: ...Table1 3: Perplexities of maximum entropy discourse grammars. Features Unigram Speaker- Anychannel Samechannel Otherchannel Trigger Perplexity change Bigram Trigram Skip1-Bigram Bigram Bigram 5.... In PAGE 18: ...ombined with the N-gram discourse grammar described in Section 4.3. Results for discourse grammars of various orders are summarized in Table 14. Table1 4: Results for DA detection from words. Discourse Grammar Accuracy (%) Full conv.... In PAGE 18: ...9 Since the devtest conversations had been truncated to five minutes for recognition purposes, and a few utterances had been lost as a result of the segmentation process, we ran this experiment both on the full conversation transcript, and on the 5-minute subset of utterances used in later recognition experiments. As shown in Table1 4, the accuracies on the full conversations were slightly better, probably because the discourse grammar was somewhat mismatched to the truncated conversations. The fact that the differences were minor is important, however, since we will later use the same discourse grammars in combination with recognizer outputs.... In PAGE 19: ... We observe about a 7% reduction absolute reduction in accuracy compared to using the true words. Table1 5: Results for DA detection from N-best lists. Discourse Grammar Accuracy (%) None 42.... In PAGE 20: ...Table1 6: Duration features. Feature Name Description Duration ling dur duration of utterance (linguistically-segmented) Duration-pause ling dur minus min10pause ling dur minus sum of duration of all pauses of at least 100 msec cont speech frames number of frames in continuous speech regions ( gt; 1 sec, ignoring pauses lt; 100msec) Duration-correlated F0-based counts f0 num utt number of frames with F0 values in utterance (prob voicing=1) f0 num good utt number of F0 values above f0 min (f0 min = .... In PAGE 21: ...Table1 7: Pause features. Feature Name Description min10pause count n ldur number of pauses of at least 10 frames in utterance, normalized by dura- tion of utterance total min10pause dur n ldur sum of duration of all pauses of at least 100msec in utterance, normalized by duration of utterance mean min10pause dur utt mean pause duration for pauses of at least 10 frames in utterance mean min10pause dur ncv mean pause duration for pauses of at least 10 frames in utterance, nor- malized by same in convside cont speech frames n number of frames in continuous speech regions ( gt; 1 sec, ignoring pauses lt; 10 frames) normalized by duration of utterance capture the amount of information in the utterance, by counting accents and phrase boundaries.... In PAGE 21: ...7.3 F0 features F0 features, shown in Table1 8, included both raw and regression values based on frame-level F0 values from ESPS/Waves+. To capture overall pitch range, mean F0 values were calculated over all voiced frames in an utter- ance.... In PAGE 22: ...Table1 8: F0 features. Feature Name Description f0 mean good utt mean of F0 values included in f0 num good utt f0 mean n difference between mean F0 of utterance and mean F0 of convside for F0 values gt; f0 min f0 mean ratio ratio of F0 mean in utterance to F0 mean in convside f0 mean zcv mean of good F0 values in utterance normalized by mean and st dev of good F0 values in convside f0 sd good utt st dev of F0 values included in f0 num good utt f0 sd n log ratio of st dev of F0 values in utterance and in convside f0 max n log ratio of max F0 values in utterance and in convside f0 max utt maximum F0 value in utterance (no smoothing) max f0 smooth maximum F0 in utterance after median smoothing of F0 contour f0 min utt minimum F0 value in utterance (no smoothing); can be below f0 min f0 percent good utt ratio of number of good F0 values to number of F0 values in utterance utt grad least-squares all-points regression over utterance pen grad least-squares all-points regression over penultimate region end grad least-squares all-points regression over end region end f0 mean mean F0 in end region pen f0 mean mean F0 in penultimate region abs f0 diff difference between mean F0 of end and penultimate regions rel f0 diff ratio of F0 of end and penultimate regions norm end f0 mean mean F0 in end region normalized by mean and st dev of F0 from conv- side norm pen f0 mean mean F0 in penultimate region normalized by mean and st dev from con- vside norm f0 diff difference between mean F0 of end and penultimate regions, normalized by mean and st dev of F0 from convside regr start f0 first F0 value of contour, determined by regression line analysis finalb amp amplitude of final boundary (if present), from event recognizer finalb label label of final boundary (if present), from event recognizer finalb tilt tilt of final boundary (if present), from event recognizer numacc n ldur number of accents in utterance from event recognizer, normalized by duration of utterance numacc n rdur number of accents in utterance from event recognizer, normalized by duration of F0 regression line numbound n ldur number of boundaries in utterance from event recognizer, normalized by duration of utterance numbound n rdur number of boundaries in utterance from event recognizer, normalized by duration of F0 regression line... In PAGE 24: ...Table1 9: Energy features. Feature Name Description utt nrg mean mean RMS energy in utterance abs nrg diff difference between mean RMS energy of end and penultimate regions end nrg mean mean RMS energy in end region norm nrg diff normalized difference between mean RMS energy of end and penulti- mate regions rel nrg diff ratio of mean RMS energy of end and penultimate regions snr mean utt mean signal-to-noise ratio (CDF value) in utterance snr sd utt st dev of signal-to-noise ratio values (CDF values) in utterance snr diff utt difference between maximum SNR and minimum SNR in utterance snr min utt st dev of signal-to-noise ratio values (CDF values) in utterance snr max utt maximum signal-to-noise ratio values (CDF values) in utterance victoria holt is that right a a rb 4.... ..."

"... In PAGE 1: ...See Jurafsky et al. (1997a), Shriberg et al. (1998), and Stolcke et al. (1998) for other publications describing aspects of this work). Table1 shows a sample of the kind of discourse structure we are modeling and detecting. Besides the usefulness of discourse structure detection for speech understanding, discourse structure can be directly relevant for speech recog- nition tasks.... In PAGE 14: ... Surely it is important to model not only the DA sequence, but also which speaker said what. For example, in the sample conversation of Table1 , the grammar should compute the probability of the second utterance in Channel A being a Ack-Answer, given that the previous utterance was a Statement on Channel B and before that was a Wh-Question on Channel A. We can think of the events in the sequence as consisting of pairs a3a24a1 a38 a45 a16a15 a38 a7 , where a1 a38 is a DA label and a15 a38 is a speaker label.... In PAGE 15: ... Results are shown in Table 10. Table1 0: Perplexity when guessing both the DA and turn information. Discourse Grammar Perplexity None 84 Unigram 18.... In PAGE 16: ...end-of-conversation tags: Table1 1: Perplexity of speaker/turn stream only. Discourse Grammar Perplexity None 3 Unigram 2.... In PAGE 16: ... The results are shown in Table 12. Table1 2: DA Perplexity conditioned on turn information. Discourse Grammar Perplexity None 42 Unigram 9.... In PAGE 17: ...Table1 3: Perplexities of maximum entropy discourse grammars. Features Unigram Speaker- Anychannel Samechannel Otherchannel Trigger Perplexity change Bigram Trigram Skip1-Bigram Bigram Bigram a11 a11 5.... In PAGE 18: ...ombined with the N-gram discourse grammar described in Section 4.3. Results for discourse grammars of various orders are summarized in Table 14. Table1 4: Results for DA detection from words. Discourse Grammar Accuracy (%) Full conv.... In PAGE 18: ...9 Since the devtest conversations had been truncated to five minutes for recognition purposes, and a few utterances had been lost as a result of the segmentation process, we ran this experiment both on the full conversation transcript, and on the 5-minute subset of utterances used in later recognition experiments. As shown in Table1 4, the accuracies on the full conversations were slightly better, probably because the discourse grammar was somewhat mismatched to the truncated conversations. The fact that the differences were minor is important, however, since we will later use the same discourse grammars in combination with recognizer outputs.... In PAGE 19: ... We observe about a 7% reduction absolute reduction in accuracy compared to using the true words. Table1 5: Results for DA detection from N-best lists. Discourse Grammar Accuracy (%) None 42.... In PAGE 20: ...Table1 6: Duration features. Feature Name Description Duration ling dur duration of utterance (linguistically-segmented) Duration-pause ling dur minus min10pause ling dur minus sum of duration of all pauses of at least 100 msec cont speech frames number of frames in continuous speech regions (a1 a9 sec, ignoring pauses a0 a9 a0a1a0 msec) Duration-correlated F0-based counts f0 num utt number of frames with F0 values in utterance (prob voicing=1) f0 num good utt number of F0 values above f0 min (f0 min = .... In PAGE 21: ...Table1 7: Pause features. Feature Name Description min10pause count n ldur number of pauses of at least 10 frames in utterance, normalized by dura- tion of utterance total min10pause dur n ldur sum of duration of all pauses of at least 100msec in utterance, normalized by duration of utterance mean min10pause dur utt mean pause duration for pauses of at least 10 frames in utterance mean min10pause dur ncv mean pause duration for pauses of at least 10 frames in utterance, nor- malized by same in convside cont speech frames n number of frames in continuous speech regions (a1 a9 sec, ignoring pauses a0 a9 a0 frames) normalized by duration of utterance capture the amount of information in the utterance, by counting accents and phrase boundaries.... In PAGE 21: ...7.3 F0 features F0 features, shown in Table1 8, included both raw and regression values based on frame-level F0 values from ESPS/Waves+. To capture overall pitch range, mean F0 values were calculated over all voiced frames in an utter- ance.... In PAGE 22: ...Table1 8: F0 features. Feature Name Description f0 mean good utt mean of F0 values included in f0 num good utt f0 mean n difference between mean F0 of utterance and mean F0 of convside for F0 values a1 f0 min f0 mean ratio ratio of F0 mean in utterance to F0 mean in convside f0 mean zcv mean of good F0 values in utterance normalized by mean and st dev of good F0 values in convside f0 sd good utt st dev of F0 values included in f0 num good utt f0 sd n log ratio of st dev of F0 values in utterance and in convside f0 max n log ratio of max F0 values in utterance and in convside f0 max utt maximum F0 value in utterance (no smoothing) max f0 smooth maximum F0 in utterance after median smoothing of F0 contour f0 min utt minimum F0 value in utterance (no smoothing); can be below f0 min f0 percent good utt ratio of number of good F0 values to number of F0 values in utterance utt grad least-squares all-points regression over utterance pen grad least-squares all-points regression over penultimate region end grad least-squares all-points regression over end region end f0 mean mean F0 in end region pen f0 mean mean F0 in penultimate region abs f0 diff difference between mean F0 of end and penultimate regions rel f0 diff ratio of F0 of end and penultimate regions norm end f0 mean mean F0 in end region normalized by mean and st dev of F0 from conv- side norm pen f0 mean mean F0 in penultimate region normalized by mean and st dev from con- vside norm f0 diff difference between mean F0 of end and penultimate regions, normalized by mean and st dev of F0 from convside regr start f0 first F0 value of contour, determined by regression line analysis finalb amp amplitude of final boundary (if present), from event recognizer finalb label label of final boundary (if present), from event recognizer finalb tilt tilt of final boundary (if present), from event recognizer numacc n ldur number of accents in utterance from event recognizer, normalized by duration of utterance numacc n rdur number of accents in utterance from event recognizer, normalized by duration of F0 regression line numbound n ldur number of boundaries in utterance from event recognizer, normalized by duration of utterance numbound n rdur number of boundaries in utterance from event recognizer, normalized by duration of F0 regression line... In PAGE 24: ...Table1 9: Energy features. Feature Name Description utt nrg mean mean RMS energy in utterance abs nrg diff difference between mean RMS energy of end and penultimate regions end nrg mean mean RMS energy in end region norm nrg diff normalized difference between mean RMS energy of end and penulti- mate regions rel nrg diff ratio of mean RMS energy of end and penultimate regions snr mean utt mean signal-to-noise ratio (CDF value) in utterance snr sd utt st dev of signal-to-noise ratio values (CDF values) in utterance snr diff utt difference between maximum SNR and minimum SNR in utterance snr min utt st dev of signal-to-noise ratio values (CDF values) in utterance snr max utt maximum signal-to-noise ratio values (CDF values) in utterance victoria holt is that right a a rb 4.... ..."

"... In PAGE 18: ... The MDLs were determined from analysis of samples from a solution matrix containing the analytes of interest. Detection limits for all compounds studied, as given in Table6 , were determining using the EPA procedure for Method Detection Limits (USEPA, 1982). Standards for TCE and PCE were run at the beginning of each day and an additional set of standards was run later in the day or dispersed throughout the analysis run.... In PAGE 18: ... Because of the low sample volumes available for analysis, the detection limit for alkalinity measurements varied from 2 mg/L CaCO3 for the batch experiments to between 5 and 8 mg/L CaCO3 for the column experiments. Detection limits for the inorganic parameters are included in Table6 . The method detection limit for Cr(VI) measured in the lab is between 0.... ..."

"... In PAGE 7: ... We expect recognition of backchannels to be useful because of their discourse structuring role (knowing that the hearer expects the speaker togo on talking tells us somethingabout the course of the narrative) and because they seem to occur at certain kinds of syntactic boundaries; detecting a backchannel may thus help in predicting utterance boundaries and surrounding lexical material. For an intuition about what backchannels look like, Table3 shows the most com mon realizations of the approximately 300 types (35,827 tokens) of backchannel in our Switchboard subset. The following table shows examples of backchannels in the context of a Switchboard conversation:... ..."

"... In PAGE 11: ... Static loads analyses were performed via vector analysis on an inelastic structure and also using the FEM. Resulting loads are summarized in Table5 . Vector Analysis An initial estimate of static loads was hand-computed via vector analysis of a tensioned structure without applied gravity.... In PAGE 12: ... This not only verifies results but also indicates that little distortion of the elastic structure occurs due to such static loading. In 1g, the system static loads increase to accommodate the uniform gravity loading as summarized in Table5 . Panel Geometry Trade Study An analytical tensioned panel trade study via FEM analysis was performed to identify a panel geometry with optimal tension distribution.... ..."

"... In PAGE 2: ... 1998 during the low (hard) spectral state of the source. In total our sample contained 26 obser- vations randomly chosen from proposals 10235, 10236, 10238, 20175 and 30157 ( Table1 ). The energy and power density spectra were averaged for each individual obser- vation.... In PAGE 4: ...1 and 2). The best{ t parameters are listed in Table1 . The accu- racy of the absolute values of the best{ t parameters is discussed in the next section.... In PAGE 5: ... 05{00 and 10238{01{05{000 as a template. The best{ t values of the frequency scaling factor obtained in such a way are listed in Table1 . Fig.... ..."

"... In PAGE 8: ...Table2... In PAGE 8: ...Table2... In PAGE 10: ... The most significant difference in the daily and weekly estimates occurs in 1996 when several large 2-3 day price swings are not detected with weekly observations. The patterns shown in Figure 1 are generally consistent with the standard deviations reported in Table2 , but two additional features of volatility are now observable. First, the time- series estimates in Figure 1 allow us to detect finer variations in volatility.... ..."