ページの本文へ

Hitachi

企業情報研究開発

2013年6月3日

発表者からのレポート

The 2013 Asia-Pacific International Symposium and Exhibition on Electromagnetic Compatibility was held in Melbourne (Australia) between May 20th and 23rd. The symposium covered the entire scope of electromagnetic compatibility and gathered scientists and companies from all over the world.

Yokohama Research Laboratory (YRL), Hitachi Ltd., presented one article titled "LSI Noise Model Extraction with the Printed Reverberation Board", where a new modeling approach and a new measurement method for the electromagnetic noise generated by large scale integrated circuits (LSI) above 1 GHz have been presented.

The simultaneous switching noise (SSN) generated by LSI propagates as an electromagnetic wave between the power supply planes and causes radiation from the printed circuit boards (PCB) edges. This is an important source of electromagnetic interference between electronic equipments. An estimation of the power supply noise generated by an LSI is required to estimate system level radiation, to design countermeasures, to compare different LSIs, and when possible also to estimate power integrity and signal integrity effects.


図1 Single noise source approach
拡大図

Since an LSI noise estimation based on simulations is very difficult and often not practicable, measurement of power supply noise is necessary. However, the available measurement and modeling techniques are usually limited to frequencies below 1 GHz, and their application to packages with many pins or to ball grid arrays (BGA) is problematic.

The proposed original approach presents many novelties. A single current noise source is used to model the noise injected by all the power supply pins into the power supply planes. This noise current source is represented by means of a random variable, whose high percentiles are extracted with a statistical measurement approach that makes use of one special measurement equipment developed at YRL: the printed reverberation board (PRB). The PRB was presented for the first time at the EMC Europe conference in 2012.


図2 Random noise source approach
拡大図

Among the many new aspects of the proposed approach, in the present work only one necessary condition has been verified by means of measurement results, namely the coincidence of two noise models of the same noise source extracted from two different measurement ports. The maximum difference between the resulting noise source currents in the frequency range between 4.6 and 6 GHz was 7.5 dB, and it was much smaller at most of the frequencies. These results are very encouraging, considering that they are affected by the limited accuracy of modeling and design of the first ever PRB. In the test, a special PCB designed at YRL was used to emulate the stochastic noise generated by an LSI, but in a wider frequency range than usual.

Although the application of the model has not been presented yet, it should be clear that the noise model must be used in conjunction with a statistical approach at board or system level, which is the most suitable for frequencies above 1 GHz. The advantage of the present approach is that it very much simplifies both modeling and measurements, making for the first time feasible two tasks that otherwise would be impossible to handle due to their complexity.

(Umberto PAOLETTI    記)

  • ページの先頭へ