首页> 外文OA文献 >Convolutiegebaseerde modeleringsmethodologie voor de snelle thermische analyse van driedimensionaal gestapelde elektronische componenten
【2h】

Convolutiegebaseerde modeleringsmethodologie voor de snelle thermische analyse van driedimensionaal gestapelde elektronische componenten

机译:基于卷积的建模方法,可对三维堆叠的电子元件进行快速热分析

摘要

The relevance of accurate predictions of the thermal behavior of microelectronic systems has been increasing since the introduction of 3D integrated circuits (3D-ICs). Due to the vertical stacking of the active dies the reliability issues related to high temperature and temperature gradients are, indeed, exacerbated. Different thermal modeling strategies have been developed with the aim of providing quick estimations of the device temperature under operating conditions. It is, indeed, important to be able to quickly compare the thermal impact of different design and technological parameters already during the design phase.In this thesis, an easy-to-use fast thermal modeling methodology based on the Green's function theory is presented. It provides highly resolved temperature maps on selected levels (or selected points), avoiding the calculation in locations that are not thermally significant and reducing, therefore, the computational time. The model is able to deal with both the steady state and the transient regime and it proved to be two orders of magnitude faster than conventional finite element methods, maintaining the error on peak temperatures below 5%.The core of the algorithm is constituted by the convolution between 1) the thermal response of the system to localized and impulsive power dissipation and 2) the actual dissipated power map. However, this basic convolution approach is valid only for stack (layered) structures in which multiple layers, of homogeneous material and with infinite horizontal size, are placed on top of each other. The “method of images” is used to take into account the finite dimension of the stack while correction strategies are applied to account for the thermal impact of specific µbump layouts (only in case of a two dies stack in the steady state regime) and of the package. Moreover, an a posteriori mathematical transformation, the Kirchhoff transformation, has been proposed to deal with the temperature dependency of the silicon thermal conductivity. By overcoming the limitations of the basic convolution approach, the developed fast thermal model is able to deal with realistic 3D-IC configurations. This has been proved by a successful experimental validation with respect to measurement data.The model has also been extended to deal with other geometries commonly available in microelectronic applications (side by side integration on an interposer and stack of dies with different sizes). Moreover, to demonstrate the applicability and the easiness-of-use of the developed methodology, the model has been applied to perform realistic analyses that might be needed during the design phase of an IC.It is, therefore, concluded that the developed fast thermal model can be a valid alternative to conventional thermal modeling strategies for 3D-ICs and related geometries: the computational time is, indeed, strongly reduced and high accuracy is maintained.
机译:自从引入3D集成电路(3D-IC)以来,对微电子系统的热行为进行准确预测的相关性一直在增加。由于有源管芯的垂直堆叠,与高温和温度梯度有关的可靠性问题的确加剧了。为了提供对工作条件下器件温度的快速估计,已经开发了不同的热建模策略。确实,重要的是能够在设计阶段就快速比较不同设计和技术参数的热影响。本文提出了一种基于格林函数理论的易于使用的快速热建模方法。它在选定的水平(或选定的点)上提供了高度解析的温度图,从而避免了在不具有热重要性的位置进行计算,从而减少了计算时间。该模型能够处理稳态和瞬态两种情况,并且证明比常规有限元方法快两个数量级,并且在峰值温度下的误差保持在5%以下。该算法的核心是由1)系统对局部和脉冲功耗的热响应与2)实际功耗图之间的卷积。但是,这种基本的卷积方法仅对堆叠(分层)结构有效,在该结构中,均质材料和水平无限大的多层相互叠置。 “图像方法”用于考虑堆叠的有限尺寸,而校正策略则用于考虑特定µbump布局的热影响(仅在稳态状态下为两个裸片堆叠的情况)和包装。此外,已经提出了后验数学变换,即基尔霍夫变换,以处理硅热导率的温度依赖性。通过克服基本卷积方法的局限性,开发的快速热模型能够处理现实的3D-IC配置。通过对测量数据的成功实验验证证明了这一点。该模型也已扩展为处理微电子应用中常见的其他几何形状(并排集成在中介层和不同尺寸的裸片叠层上)。此外,为证明所开发方法的适用性和易用性,该模型已被用于执行IC设计阶段可能需要的现实分析,因此得出结论,已开发的快速散热对于3D-IC和相关几何图形,该模型可以有效替代常规热建模策略:确实大大减少了计算时间,并保持了较高的准确性。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号