Keysight Technologies (Singapore)

The emergence of cloud-native architectures and Open RAN (O-RAN) principles has revolutionized the deployment and scalability of mobile network infrastructure. However, the energy efficiency of Virtualized Network Functions (VNFs) operating in such environments remains a critical concern, particularly as 5G and 6G networks scale in complexity and resource demands. This study introduces a modular and reproducible testbed for high-fidelity energy profiling of containerized monolithic srsRAN-based gNB implementations running on Commercial Off-The-Shelf (COTS) server-class hardware. The testbed integrates a commercial-grade Power Analyzer (PA) and a full-stack network emulation framework to measure the impact of key parameters, including CPU frequency and core allocation, on the power consumption of a virtualized gNB. A comprehensive configuration dataset is collected across both high-load and low-load scenarios. The results reveal that CPU frequency throttling consistently reduces energy consumption beyond specific performance thresholds, while core limitation is effective only in low-load scenarios; however, it enables VNF co-location, which contributes to reducing overall infrastructure-level energy consumption. These findings validate the applicability of dynamic energy optimization strategies and provide actionable insights for orchestration frameworks aiming to balance energy efficiency with Quality of Service (QoS) requirements. The emergence of cloud-native architectures and Open RAN (O-RAN) principles has revolutionized the deployment and scalability of mobile network infrastructure. However, the energy efficiency of Virtualized Network Functions (VNFs) operating in such environments remains a critical concern, particularly as 5G and 6G networks scale in complexity and resource demands. This study introduces a modular and reproducible testbed for high-fidelity energy profiling of containerized monolithic srsRAN-based gNB implementations running on Commercial Off-The-Shelf (COTS) server-class hardware. The testbed integrates a commercial-grade Power Analyzer (PA) and a full-stack network emulation framework to measure the impact of key parameters, including CPU frequency and core allocation, on the power consumption of a virtualized gNB. A comprehensive configuration dataset is collected across both high-load and low-load scenarios. The results reveal that CPU frequency throttling consistently reduces energy consumption beyond specific performance thresholds, while core limitation is effective only in low-load scenarios; however, it enables VNF co-location, which contributes to reducing overall infrastructure-level energy consumption. These findings validate the applicability of dynamic energy optimization strategies and provide actionable insights for orchestration frameworks aiming to balance energy efficiency with Quality of Service (QoS) requirements.

Accurate modeling of self-heating (SH) induced junction temperature rise is crucial for circuit design at high current densities required for high speed and high gain, as well as for reliability, as most degradations worsen at higher temperatures. For relatively lower power, the junction temperature rise (Trise) is typically modeled as a linear function of the dissipated power (Pdiss), with the slope being a parameter named thermal resistance (Rth), which can be further modeled as a function of ambient temperature (Tamb).In Mextram, this was the case until the recent implementation of the nonlinear SH model of [1]-[4] in Mextram 505.4 to improve modeling for high-power applications [5] [6]. A notable feature of the Mextram nonlinear SH model implementation is that it uses the same model parameters as its linear SH model, i.e., the low power Rth at reference temperature, rth, and the material thermal conductivity temperature coefficient ath used to describe the Tamb dependence of the linear SH Rth. In the nonlinear SH model, the same ath is used for Pdiss dependence of the effective Rth.Theoretically, if ath is already physical, one can simply turn on the nonlinear SH switch for improved modeling at higher collector current (IC) and collector-emitter voltage (VCE). The key to this simplicity lies in the specific form of the nonlinear functional dependence of Trise on Tamb and Pdiss derived by solving the nonlinear heat diffusion equation using the Kirchhoff transformation [7].In practice, however, retuning these two parameters with high power data is beneficial for better overall fitting, particularly over multiple ambient temperatures. Fitting a larger power range at multiple ambient temperatures can still be challenging, due to the presence of multiple materials with different thermal conductivity temperature coefficients in the heat flow path and heat dissipation through multiple paths, e.g., from metallization in addition to substrate.To provide flexibility in silicon data fitting and extend the model's capability to a wider ambient temperature range and power range, we present an additional semi-empirical model. By deriving the slope of Trise versus Pdiss in the single material single path model problem, we show that it is only a function of the junction temperature Tjunc=Trise+Tamb, which is behind the ability of the model to use only two parameters for both Tamb and Pdiss variation.We propose a semi-empirical modification of the Trise versus Pdiss slope behavior, similar to the low-temperature CMOS measurement-based model in [8] and [9], which has the same issue of having a Tjunc-dependent Trise vs Pdiss slope, but an easier functional form for modification to provide flexibility in better fitting high power region over a wider range of ambient temperature. Our model construction ensures that the low Pdiss behavior is guaranteed to be the same as existing Mextram SH models.Figs. 1-3 show the modeling results of IC-VCE at various base-emitter voltages (VBEs) for a SiGe HBT at -25, 25, 75, and 125°C using 1) Mextram's default linear SH model; 2) Mextram 505.4's nonlinear SH model; and 3) the proposed wide power and temperature range nonlinear SH model. The new model shows the best overall fitting.In the full paper, we will also present results obtained using the nonlinear SH model used in HICUM L2 [10], where the effective Rth modeled as a linear or nonlinear function of Tjunc, and results obtained by implementing the model of [9]. We will make comparisons of the various models and provide best practice recommendations.[1] W. B. Joyce, SSE, vol. 18, pp. 321-322, 1975.[2] K. Poulton et al., JSSC, vol. 27, no. 10, pp. 1379-1387, 1992.[3] B. Yeats, Dig. GaAs IC symposium, pp. 59-62, 1999.[4] J. C. J. Paasschens et al., BCTM, pp. 96-99, 2004.[5] G. Niu et al., The Mextram Bipolar Transistor Model Version 505.4.0, 2023.[6] M. Willemsen, AKB-Workshop, 2023.[7] G. R. Kirchhoff, Vorlesungen über die Theorie der Wärme, 1894.[8] K. Triantopoulos et al., TED, vol. 66, no. 8, pp. 3498-3505, 2019.[9] G. Ghibaudo et al., SSE, vol. 192, article 108265, 2022.[10] M. Schroter, HICUM/L2 Technical documentation of model version 3.0.0, 2020. Figure 1

Globally, colorectal cancer (CRC) is the third most frequently occurring cancer. Progression on to an advanced metastatic malignancy (metCRC) is often indicative of poor prognosis, as the 5-year survival rates of patients decline rapidly. Despite the availability of many systemic therapies for the management of metCRC, the long-term efficacies of these regimens are often hindered by the emergence of treatment resistance due to intratumoral and intertumoral heterogeneity. Furthermore, not all systemic therapies have associated biomarkers that can accurately predict patient responses. Hence, a functional personalised oncology (FPO) approach can enable the identification of patient-specific combinatorial vulnerabilities and synergistic combinations as effective treatment strategies. To this end, we established a panel of CRC patient-derived organoids (PDOs) as clinically relevant biological systems, of which three pairs of matched metCRC PDOs were derived from the primary sites (ptCRC) and metastatic lesions (mCRC). Histological and genomic characterisation of these PDOs demonstrated the preservation of histopathological and genetic features found in the parental tumours. Subsequent application of the phenotypic-analytical drug combination interrogation platform, Quadratic Phenotypic Optimisation Platform, in these pairs of PDOs identified patient-specific drug sensitivity profiles to epigenetic-based combination therapies. Most notably, matched PDOs from one patient exhibited differential sensitivity patterns to the rationally designed drug combinations despite being genetically similar. These findings collectively highlight the limitations of current genomic-driven precision medicine in guiding treatment strategies for metCRC patients. Instead, it suggests that epigenomic profiling and application of FPO could complement the identification of novel combinatorial vulnerabilities to target synchronous ptCRC and mCRC.

Connected autonomous vehicle (CAV) technology has the potential to enable significant gains in energy economy (EE). Much research attention has been focused on autonomous eco-driving control enabled by various methods. In this study, the state of the literature on autonomous eco-driving control is reviewed, an overall systems’ description of eco-driving control for a CAV is provided, and representative methods are evaluated comparatively against each other in simulation. Simulations are conducted using real-world traffic signal data and a validated future automotive systems technology simulator (FASTSim) model. Results indicate that an EE improvement in the range of 5%–15% is attainable depending on the method and cost function used. In this article it is shown that dynamic programming (DP) methods are most effective in improving EE but are significantly more computationally expensive than other methods. The genetic algorithm (GA) methods are shown to present the most potential in terms of EE improvement and run-time. Results also indicate that velocity-sensitive cost functions allow all the methods to perform better than pure acceleration minimization.

Link prediction on knowledge graphs is useful in numerous application areas such as semantic search, question answering, entity disambiguation, enterprise decision support, recommender systems and so on. While many of these applications require a reasonably quick response and may operate on data that is constantly changing, existing methods often lack speed and adaptability to cope with these requirements. This is aggravated by the fact that knowledge graphs are often extremely large and may easily contain millions of entities rendering many of these methods impractical. In this paper, we address the weaknesses of current methods by proposing Random Semantic Tensor Ensemble (RSTE), a scalable ensemble-enabled framework based on tensor factorization. Our proposed approach samples a knowledge graph tensor in its graph representation and performs link prediction via ensembles of tensor factorization. Our experiments on both publicly available datasets and real world enterprise/sales knowledge bases have shown that our approach is not only highly scalable, parallelizable and memory efficient, but also able to increase the prediction accuracy significantly across all datasets.

The project to define Long Term Evolution (LTE) of UMTS cellular technology started in year 2004 is well established to the world. The subsequent evolve of LTE-Advanced however targets for enhancement in average spectrum efficiency and cell-edge user throughput. The limiting factor of the system performance at cell-edge is inter-cell interference (ICI). The challenge that is discussed in this paper focuses on ensuring the overall satisfaction mainly cell-edge system performance through the analysis on Fractional Frequency Reuse (FFR) and hybrid Macro system which involves small cell technology — Femtocell technology. This paper discussed the three frequency allocation schemes. The schemes include without the application of Femtocell, with randomly application of Femtocell and with preplanned Femtocell scheme. The FFR mechanism divides the cell into two regions (inner and outer). The inner radius and inner region resource is determined base on overall average throughput and overall user satisfaction.