Pitch Scaling Research Articles

A compact pitch and time representation for melodic contours in Indian art music.

Predominant melody offers a complete representation of melodic contours of heterophonic Indian Art Music (IAM). A compact representation of melodic contours while preserving raga characteristics is proposed. Such representations have applications in music transcription, analysis, and synthesis. Contours are quantized on a pitch-time grid after normalizing critical points with tonic and rhythmic pulse period estimates. Non-uniform quantization intervals are selected from pitch and time scales prevalent in IAM, while accommodating pitch and inter-note-interval variations on pitch-time grid. An evaluation of quantized-reconstructed contours through listening tests by trained musicians shows raga preserving capabilities of the proposed approach in spite of alterations in contour shapes.

Association Between Flat-Panel Computed Tomographic Imaging–Guided Place-Pitch Mapping and Speech and Pitch Perception in Cochlear Implant Users

Cochlear implant users generally display poor pitch perception. Flat-panel computed tomography (FPCT) has recently emerged as a modality capable of localizing individual electrode contacts within the cochlea in vivo. Significant place-pitch mismatch between the clinical implant processing settings given to patients and the theoretical maps based on FPCT imaging has previously been noted. To assess whether place-pitch mismatch is associated with poor cochlear implant-mediated pitch perception through evaluation of an individualized, image-guided approach toward cochlear implant programming on speech and music perception among cochlear implant users. A prospective cohort study of 17 cochlear implant users with MED-EL electrode arrays was performed at a tertiary referral center. The study was conducted from June 2016 to July 2017. Theoretical place-pitch maps using FPCT secondary reconstructions and 3-dimensional curved planar re-formation software were developed. The clinical map settings (eg, strategy, rate, volume, frequency band range) were modified to keep factors constant between the 2 maps and minimize confounding. The acclimation period to the maps was 30 minutes. Participants performed speech perception tasks (eg, consonant-nucleus-consonant, Bamford-Kowal-Bench Speech-in-Noise, vowel identification) and a pitch-scaling task while using the image-guided place-pitch map (intervention) and the modified clinical map (control). Performance scores between the 2 interventions were measured. Of the 17 participants, 10 (58.8%) were women; mean (SD) was 59 (11.3) years. A significant median increase in pitch scaling accuracy was noted when using the experimental map compared with the control map (4 more correct answers; 95% CI, 0-8). Specifically, the number of pitch-scaling reversals for notes spaced at 1.65 semitones or greater decreased when an image-based approach to cochlear implant programming was used vs the modified clinical map (4 mistakes; 95% CI, 0.5-7). Although there was no observable median improvement in speech perception during use of an image-based map, the acute changes in frequency allocation and electrode channel deactivations used with the image-guided maps did not worsen consonant-nucleus-consonant (-1% correct phonemes, 95% CI, -2.5% to 6%) and Bamford-Kowal-Bench Speech-in-Noise (0.5-dB difference; 95% CI, -0.75 to 2.25 dB) median performance results relative to the clinical maps used by the patients. An image-based approach toward ochlear implant mapping may improve pitch perception outcomes by reducing place-pitch mismatch. Studies using a longer acclimation period with chronic stimulation over months may help assess the full range of the benefits associated with personalized image-guided cochlear implant mapping.

A Review of Nanoporous Metals in Interconnects

Interconnections in semiconductor devices need to provide a seamless electrical, thermal, and mechanical link to different parts of the circuit. Since the move away from lead-based solders due to toxicity concerns, the search has been on for materials that can provide desirable attributes before, during, and after bonding. While lead-free solders have been extensively used as a replacement, they are now reaching their limits in terms of pitch scaling, as well as, electrical and thermal properties, forcing the semiconductor industry to look for alternative material systems for the next interconnection node. This review provides an overview of key properties needed in metal interconnections and possible use of nanoporous metals due to ease of fabrication, nanostructured ligaments, and other key physical attributes.

(Invited) Very Low Temperature Epitaxy of Group-IV Semiconductors for Use in Finfet, Stacked Nanowires and Monolithic 3D Integration

Microelectronic devices enable the continuous improvement in mobile data exchange and mobile communication. This exciting technological progress has been achieved by transistor scaling, resulting in an increased performance for lower energy consumption and cost. As CMOS device scaling proceeds, the number of epitaxial growth steps in device fabrication schemes increases. Examples are SiGe or Ge as high mobility channel materials [1-3], SiGe/Si or SiGe/Ge multi-layers to define nanowires in lateral Gate All Around (GAA) devices [3-5], and Source/Drain (S/D) stressor layers which also allow to reduce the S/D contact resistance [3,6-8]. In this contribution we will describe the characteristics of the epitaxial growth schemes as used for Fin and Gate-All-Around (GAA) devices. For Si- and SiGe-channel GAA devices, strained SiGe/Si multi-layers, which are used to define lateral nanowires, can be grown with conventional precursors (SiH4 or SiCl2H2 and GeH4). The compositional gradient is sufficiently steep to allow efficient wire release by selective SiGe or Si removal [2,5,9]. In case of Ge GAA, epitaxial growth of Ge-rich SiGe/Ge on SiGe Strain Relaxed Buffers is more challenging. Extremely low process temperatures are required to avoid strain relaxation. These layers need to be grown with higher order precursors, since conventional precursors become ineffective [4]. The epitaxial growth of ultra-thin Si layers on (strained) Ge fins or lateral nanowires used to passivate Ge surfaces in the high-k gate module has been discussed at the ECS Fall meeting of 2017 [10] and will not be covered in this presentation. A major challenge is to increase the active doping concentration in epitaxial S/D layers. Current scaling of gate and fin pitch reduces contact area and the contact resistance becomes a key contributor to device parasitics [7,11]. In addition, monolithic 3D integration, requires to reduce the thermal budget of the complete epitaxial growth scheme (including pre-epi surface cleaning). For Si:P, we demonstrated that P doping activation can be increased by post-epi anneals (>1e21 cm-3) without major strain loss [8]. Nearly similar active doping concentrations (up to 8e20 cm-3) are obtained by using Si3H8 instead of SiCl2H2 and low growth temperatures without the need for a post epi thermal treatment. For SiGe:B, we obtained similar active boron concentrations using conventional precursors [12,13]. Because of the low boron solubility in Ge, it is more challenging to fabricate S/D Ge pMOS contacts with the right lattice constant (to act as a stressor) and with a sufficiently high active doping concentration. Effort is ongoing to increase the active boron concentration by moving the growth process further away from equilibrium. This is done at low growth temperatures by either using Ge2H6 as high-order precursor or by assessing new materials such as GeSn. On blanket wafers, higher active doping concentrations have been reported but it is extremely challenging to translate this into a selective growth process without Sn precipitation. This triggered the research community to assess alternative dopants. For Ge, the Ga doping solubility is significantly higher than that of B. Epitaxial growth schemes are preferred above ion implantation as the latter one requires an unwanted high thermal budget for doping activation and on patterned wafers it does not allow conformal doping profiles. First Ge:Ga layers with a resistivity below 0.4 mΩ.cm have been grown without any detectable C incorporation. In addition, Ga + B co-doping enables to reduce the resistivity of SiGe:B:Ga, which confirms the potential of Ga as an add-on. Acknowledgements The imec core CMOS program members, European Commission, the TAKEMI5 ECSEL project, local authorities and the imec pilot line are acknowledged for their support. Air Liquide Advanced Materials is acknowledged for providing advanced precursor gases.

Acceleration and deceleration at constant speed: systematic modulation of motion perception by kinematic sonification.

Many domains of human behavior are based on multisensory representations. Knowledge about the principles of multisensory integration is useful to configure real-time movement information for the online support of perceptuomotor processes (motor perception, control, and learning). A powerful method for generating real-time information is movement sonification. Remarkable evidence exists on movement-acoustic real-time information being effective in behavioral domains (music training, motor rehabilitation, sports). Here, we investigate whether and how biological motion perception can be enhanced, substituted, or modulated by kinematic sonification, with a focus on pitch coding. We work with gross motor cyclic movements and investigate the effectiveness of pitch scaling and consistent transposition on audiovisual motor perception accuracy (Experiment A). Beyond that, a new kind of audiovisual stimulus with inconsistent pitch transposition is used to produce a directed modulation of the integrated audiovisual percept (Experiment B). Results from Experiment A indicate pitch being powerful for mediating kinematic information to enhance motor perception and substituting information between perceptual modalities, even exceeding visual performance. Beyond these findings, results from Experiment B indicate that visual estimations of movement velocity can be enhanced or reduced auditorily. Movement sonification used for reshaping intermodal adjustments should be a powerful new tool for subconsciously shaping human movement patterns in the future.

Advanced FDSOI Device Design: The U-Channel Device for 7 nm Node and Beyond

In this paper, we propose the extendibility of ultra-thin body and box (UTBB) devices to 7 and 5 nm technology nodes focusing on electrostatics. A difficulty in scaling traditional UTBB is the need for SOI scaling to about one fourth of the gate length. We propose a U-channel fully depleted silicon on insulator architecture that starts off with a thicker SOI (8–11 nm) and has a U-shaped channel enabled by a recessed metal gate. This device improves the electrostatics by increasing the overall gate length at fixed metal gate opening, mitigating drain field coupling to the source due to the recessed metal gate region and having thin SOI below the center of the device (4–5 nm). Modeling shows that good electrostatics can be maintained at small metal gate opening to enable pitch scaling. This device provides lower cost options for mobile and IOT technologies.

The prospects of transition metal dichalcogenides for ultimately scaled CMOS

MOSFET gate length scaling has been a main source of progress in digital electronics for decades. Today, researchers still spend considerable efforts on reducing the gate length and on developing ultimately scaled MOSFETs, thereby exploring both new device architectures and alternative channel materials beyond Silicon such as two-dimensional TMDs (transition metal dichalcogenides). On the other hand, the envisaged scaling scenario for the next 15 years has undergone a significant change recently. While the 2013 ITRS edition required a continuation of aggressive gate length scaling for at least another 15 years, the 2015 edition of the ITRS suggests a deceleration and eventually a levelling off of gate length scaling and puts more emphasis on alternative options such as pitch scaling to keep Moore’s Law alive. In the present paper, future CMOS scaling is discussed in the light of emerging two-dimensional MOSFET channel, in particular two-dimensional TMDs. To this end, the scaling scenarios of the 2013 and 2015 ITRS editions are considered and the scaling potential of TMD MOSFETs is investigated by means of quantum–mechanical device simulations. It is shown that for ultimately scaled MOSFETs as required in the 2013 ITRS, the heavy carrier effective masses of the Mo- and W-based TMDs are beneficial for the suppression of direct source-drain tunneling, while to meet the significantly relaxed scaling targets of the 2016 ITRS heavy-effective-mass channels are not needed.

Effect of Prosody Modification on Children's ASR

Transcribing children's speech using acoustic models trained on adults’ speech is very challenging. In such conditions, a highly degraded recognition performance is reported due to large mismatch in the acoustic/linguistic attributes of the training and test data. The differences in pitch (or fundamental frequency) between the two groups of speakers is one among several mismatch factors. Another important mismatch factor is the difference in speaking rates. To overcome these two sources of mismatch, prosody modification is explored in this letter. Prosody modification is done by using glottal closure instants (GCIs) as anchoring points. The GCIs, in turn, are determined using zero-frequency filtering (ZFF). The ZFF-GCI-based prosody modification is fast and results in highly accurate scaling of pitch and speaking rate. The experimental evaluations studying the effect of prosody modification resulted in a relative improvement of $50\%$ over the baseline.

Current trends in changing the channel in MOSFETs by III–V semiconducting nanostructures

AbstractThe quest for high device density in advanced technology nodes makes strain engineering increasingly difficult in the last few decades. The mechanical strain and performance gain has also started to diminish due to aggressive transistor pitch scaling. In order to continue Moore’s law of scaling, it is necessary to find an effective way to enhance carrier transport in scaled dimensions. In this regard, the use of alternative nanomaterials that have superior transport properties for metal-oxide-semiconductor field-effect transistor (MOSFET) channel would be advantageous. Because of the extraordinary electron transport properties of certain III–V compound semiconductors, III–Vs are considered a promising candidate as a channel material for future channel metal-oxide-semiconductor transistors and complementary metal-oxide-semiconductor devices. In this review, the importance of the III–V semiconductor nanostructured channel in MOSFET is highlighted with a proposed III–V GaN nanostructured channel (thickness of 10 nm); Al2O3dielectric gate oxide based MOSFET is reported with a very low threshold voltage of 0.1 V and faster switching of the device.

STORY OF TJONG A FIE: Programmatic Music Composition Combining Chinese, Malay and Western Music Elements

Many composers have combined Western music with Asian music such as Turkish, Arabic, Berber, Persian, Indian, Japanese, Chinese, Indonesia and others. The most significant characteristic of Asian music is the use of pentatonic and gong. Pentatonic and gong are also used in western music along with other instruments. This describes the relationship between Western and Asian music. This music composition is an idea from the story of the life of TjongA Fie. Tjong A Fie (1860-1920) or Tjong Yiauw Hian was a Hakka Chinese businessman, banker and kapitan (Chinese Major) who built a large plantation business in Sumatera, Indonesia. This composition examines the TjongA Fie story which combines Chinese, Indonesian (Malay rhythm: inang, joget and zapin) and Western music techniques. The researcher created a music composition as entitled above, using the six-pitch scales or modes, Huowu. The six-pitch scale is then combined with the twelve tones of the Western music system. Then the tones are transposed within twelve tones in the Western music system that will produce a new pitch scale. With the new set of scales, the music has more aesthetic expressions in terms of visual elements which include melodic motif, color, sound, atonal harmony, rhythm patterns, and texture.

STRAIGHT-Based Emotion Conversion Using Quadratic Multivariate Polynomial

Speech is the natural mode of communication and the easiest way of expressing human emotions. Emotional speech is expressed in terms of features like f0 contour, intensity, speaking rate, and voice quality. The group of these features is called prosody. Generally, prosody is modified by pitch and time scaling. Emotional speech conversion is more sensitive to prosody unlike voice conversion, where spectral conversion is the main concern. Several techniques, linear as well as nonlinear, have been used for transforming the speech. Our hypothesis is that quality of emotional speech conversion can be improved by estimating nonlinear relationship between the neutral and emotional speech feature vectors. In this research work, quadratic multivariate polynomial (QMP) has been explored for transforming neutral speech to emotional target speech. Both subjective and objective analyses were carried out to evaluate the transformed emotional speech using comparison mean opinion scores (CMOS), mean opinion scores (MOS), identification rate, root-mean-square error, and Mahalanobis distance. For Toronto emotional database, except for neutral/sad conversion, the CMOS analysis indicates that the transformed speech can partly be perceived as target emotion. Moreover, the MOS and spectrogram indicate good quality of transformed speech. For German database except for neutral/boredom conversion, the CMOS value of proposed technique has better score than gross and initial–middle–final methods but less than syllable method. However, QMP technique is simple, is easy to implement, has better quality of transformed speech, and estimates transformation function using limited number of utterances of training set.

(Invited) Probing the Intrinsic Limitations of the Contact Resistance of Metal/Semiconductor Interfaces through Atomistic Simulations

Over the last technology nodes, the continuous reduction of the CMOS dimensions has been pushing the physical limits of the transistor into numerous ends. Consequently, a straightforward linear scaling of device dimensions does not apply anymore to all the aspects of the CMOS transistor. For instance, the aggressive scaling of the transistor gate pitch requires that both the gate length and the S/D contact dimensions are reduced in size; constraining the gate, the sidewall spacers and the S/D contacts to compete continuously for space. On the other hand, the transistor gate length has also to follow a different scaling law and is kept as large as possible to ensure a proper electrostatic control. A direct consequence of this approach is the need to keep on decreasing the S/D contact resistance, which currently constitutes a new bottleneck in the scaling of modern CMOS. Aside from being able to reduce the physical dimensions of the contact, it is paramount to minimize the metal/semiconductor intrinsic resistivity which has now become a major roadblock to be conquered. Recently, the cumulated efforts of the scientific and engineering communities have substantially extended the physical limits of the S/D contact resistance thanks to the increase in the concentration of activated dopants in the semiconductors and to the engineering of their Schottky barrier. The latter is achieved by integrating metals with a weak Fermi level pinning at the interface with the semiconductor. As a result, contact resistivities as low as 2x10-9 Ω.cm2 [1] and 8.4x10-10 Ω.cm2 [2] have been reported for n-Si-TiSi(amorphous) and p-Si30Ge70-TiSi(amorphous) with dopant concentrations as high as 6x1020cm-3 and 1x1021cm-3[1,2]. Although these accomplishments are very impressive, little is known on the intrinsic limits that will ultimately determine the contact resistivity and it remains unclear whether even higher doping concentrations would be desirable or not. In this contribution, we will report a fundamental study of the factors that set the contact resistivity between metals and highly doped semiconductors. We will investigate the case of n-type doped Si contacted with amorphous TiSi combining first-principles calculations with Non-Equilibrium Green functions transport simulations [4][5]. We will show that, whereas the metal/semiconductor intrinsic contact resistivity initially scales with the doping concentration, it is found to saturate at ~2x10-10 Wcm2, as of a doping concentration of ~5x1021 cm-3. From this concentration on, electron injection is governed by Ohm’s law. It turns out that, in such high doping regimes, the electron injection does no longer depend on the barrier height set by the interface potential to be crossed. In turn, it becomes less sensitive to the chemical composition of the interface/degree of disorder and is limited by the intrinsic transmission probability of electrons being injected from the metal into the semiconductor. The latter depends on both the effective masses of the metal and of the semiconductor [3] and imposes an intrinsic physical limit to the contact resistance. We will show that, in this regime, contacting metals with a heavy electron effective mass are favorable to increase the transmission probability and, hence, to reduce of the semiconductor-metal intrinsic contact resistivity. Acknowledgements: The authors thank the imec industrial affiliation core CMOS program and quantumwise for financial and technical supports, respectively.

Effect of Varying Pitch and Overlap Scale on Enhanced Heat Transfer of Half-coil Jackets

Effect of Varying Pitch and Overlap Scale on Enhanced Heat Transfer of Half-coil Jackets

Spatially Selective, High-Density Placement of Polyfluorene-Sorted Semiconducting Carbon Nanotubes in Organic Solvents.

High-performance logic based on carbon nanotubes (CNTs) requires high-density arrays of selectively placed semiconducting CNTs. Although polymer-wrapping methods can allow CNTs to be sorted to a >99.9% semiconducting purity, patterning these polymer-wrapped CNTs is an outstanding problem. We report the directed self-assembly of polymer-coated semiconducting CNTs using self-assembled monolayers that bind CNTs into arrays of patterned trenches. We demonstrate that CNTs can be placed into 100 nm wide HfO2 trenches with an electrical connection yield as high as 90% and into 50 nm wide trenches with a yield as high as 70%. Our directed self-assembly method is an important step forward in pitch scaling.

Device Exploration of NanoSheet Transistors for Sub-7-nm Technology Node

In this paper, lateral gate-all-around nano-sheet transistors (NSH-FETs) are explored from intrinsic performance to dc and ring oscillator (RO) benchmark compared with FinFETs and nanowire transistors (NW-FETs) for sub-7-nm node. The band structure calculated technology computer aided design results show comparable intrinsic performance to FinFETs at same channel cross section. On top of that, dc and RO are evaluated by taking into account electrostatics, parasitic components, and layout configurations. The NSH-FETs show an advantage in drive current with the NSH width but their RO performance is limited by the device capacitance. The multiple narrow NSH-FET shows ~5% higher drive current compared to the NW-FET at similar subthreshold swing, allowing heavier capacitive loaded circuit. In addition, NSH-FETs can provide the device design freedom from aggressive fin pitch scaling.

Statistical Timing Analysis Considering Device and Interconnect Variability for BEOL Requirements in the 5-nm Node and Beyond

In an increasing interconnect resistance era and aggressive metal pitch scaling, the elevating RC delay could significantly shadow the improvements from advanced device architectures and become a severe design issue. This paper will holistically analyze the interplay between transistors and interconnect delay and the variability induced by back-end-ofline (BEOL) process for the 5-nm node. A global sensitivity analysis using Monte Carlo simulation is employed as a powerful tool for understanding the significance of different variation sources and propagating these process uncertainties to circuit performance and parametric yield. For the BEOL integration process, our results show that dielectric κ-value is the most sensitive parameter. Regarding the patterning options, the BEOL process using self-aligned quadruple pattering with positive tone process requires more than a 4× process margin and suffers from 50% parametric yield loss. The required guardband for lithoetch litho-etch becomes as critical as for the self-aligned double patterning process when the overlay control is 6× higher than the critical dimension control. For trench patterning using spacerdefined techniques, a negative tone process is required to achieve a large process window. From a design perspective, the wire length in SoC can be optimized using a disruptive architecture as a vertical FET, which could potentially reduce the average wire length by 11%.

Fabrication of high-quality GaAs-based photodetector arrays on Si

We report on fabrication and characterization of high-quality 32 × 32 GaAs photodetector (PD) arrays on Si substrates fabricated by wafer bonding and epitaxial lift-off (ELO) techniques. Fabricated GaAs PD arrays showed good crystal quality on Si substrates with Raman spectra and X-ray diffraction measurement. Also, pitch scaling gave us faster ELO process time as well as high-density PD arrays. Furthermore, we investigated electrical and optical characteristics of fabricated GaAs pin PD arrays on Si substrates. Especially, the components of dark current characteristics were also evaluated, because it is very important to explore further pitch scaling.

Simulation of Small-Pitch HgCdTe Photodetectors

Recent studies indicate as an important technological step the development of infrared HgCdTe-based focal plane arrays (FPAs) with sub-wavelength pixel pitch, with the advantage of smaller volume, lower weight, and potentially lower cost. In order to assess the limits of pixel pitch scaling, we present combined three-dimensional optical and electrical simulations of long-wavelength infrared HgCdTe FPAs, with 3 μm, 5 μm, and 10 μm pitch. Numerical simulations predict significant cavity effects, brought by the array periodicity. The optical and electrical contributions to spectral inter-pixel crosstalk are investigated as functions of pixel pitch, by illuminating the FPAs with Gaussian beams focused on the central pixel. Despite the FPAs being planar with 100% pixel duty cycle, our calculations suggest that the total crosstalk with nearest-neighbor pixels could be kept acceptably small also with pixels only $$3\,\mu \hbox {m}$$ wide and a diffraction-limited optical system.

Adult human perception of distress in the cries of bonobo, chimpanzee, and human infants

Understanding the extent to which humans perceive the emotional state of animals has both theoretical and practical implications. While recent studies indicate that natural selection has led to some convergence of emotion coding among vertebrate species (including humans), highlighting the interspecific value of emotional signals, it has also been argued that interspecific communication of emotions can fail due to species-specific signalling traits impairing information decoding and/or absence of familiarity with heterospecific communication systems. In this article, we show that human listeners pay attention to the mean pitch of vocalizations when asked to rate the distress level expressed by human baby cries, and that they use a similar pitch scale to rate the emotional level of baby non-human ape (bonobo and chimpanzee) distress calls. As a consequence, the very high-pitched bonobo infant calls were systematically rated as expressing overall high distress levels despite being recorded in contexts eliciting various stress intensity. Conversely, chimpanzee infant calls – which are characterized by a relatively lower pitch – were systematically rated as expressing relatively lower distress levels. These results indicate that, in the absence of exposure/familiarity, our spontaneous ability to range the emotional content of vocalizations in closely related ape species remains biased by basic frequency differences, suggesting that the absolute interspecific value of emotional signals should not be overestimated.

Experimental and Theoretical Assessment of Thin Glass Substrate for Low Warpage

This paper compares the warpage after lead-free solder assembly of ultrathin glass and organic substrates for microelectronic packages. Smart mobile devices demand packages to be thinner, which exacerbates warpage, which in turn affects interconnect pitch scaling and reliability. Although low-coefficient-of-thermal-expansion (CTE) organic laminates have been developed to reduce the warpage during assembly, glass substrates offer added benefits of much higher modulus and thermo-mechanical stability, with the potential to reduce warpage beyond low-CTE organic materials. The focus of this paper is on modeling and measurements to quantify the warpage of glass and organic laminates at 100- $\mu \text{m}$ core thickness after thermo-compression bonding at 260 °C peak temperature. In the experiments, four-metal-layer glass substrates were fabricated at the panel level, diced into $18.4 \times 18.4$ -mm2 coupons, 10-mm $\times $ 10- $\text {mm}\,\,\times 630-\mu \text{m}$ silicon dies were thermo-compression bonded, and the warpage of the assembled packages was measured using shadow moire interferometry. In parallel to the experiments, numerical models that account for the viscoplastic behavior of the solder as well as the sequential build-up materials and processes have been developed. The predicted warpage from the models was compared to the experimental measurements. It was seen that the low-CTE glass had lower warpage than the organic core of equivalent CTE, while increasing the CTE of the glass increased the warpage. Also, the underfill fillet size was found to influence the warpage for thin substrate packages, and an optimal fillet size has been identified for minimizing warpage. B-staged and capillary underfill materials were compared and found to have similar warpage values.