论文部分内容阅读
过去几年,IBM在纽约的Yorktown Heights实验室发展了一种电子束刻蚀系统。现有几种型号在Yorkown的实验室运行,有一个型号在IBM San Jose实验室工作。每一个这样的系统均采取的许多新的措施,保证系统运行的高度自动化。本文叙述一个图形发生器,它从一个小计算机传送存储的掩模信息时,采用了高度的数据压缩方法,与此同时,计算机还控制系统的其它功能。图形发生器可以用最少量的数据传送扫描单元图形,在赋以初始值之后,仅需规定第一个单元图形的原点和尺寸,即可开始扫描。对于下一个单元图形而言,所需传送的参数仅为相对于上一个单元图形的变化的部分,送完即可恢复扫描操作。在图形发生器中予先为一组单元图形(矩形,平行四边形,三角形,……)编好了微程序,在扫描掩模时可直接调用。系统的一个更为灵活得多的特点是可以动态地装入一个微码来定义图形单位(Cell),图形单位可以包含若干单元图形,而且其边缘可以是不规则的或曲线。此外,只要少数几条机器指令就可将全部图形按一个有序阵列绘出,这就使复制数百万个相同的单元图形只要传送几个字。文中叙述了图形发生器的结构,以及整个16K×16K点阵高速寻址系统的其余部分。
In the past few years, IBM developed an electron beam lithography system at the Yorktown Heights Laboratory in New York. Several models are available in the labs in Yorkown and one in the IBM San Jose lab. Many of the new measures taken by each of these systems ensure a high degree of system automation. This article describes a pattern generator that uses a high degree of data compression when transferring stored mask information from a small computer, and at the same time, the computer controls other system functions. The graph generator can send the scanning unit graph with a minimum amount of data. After giving the initial value, only the origin and size of the first unit graph are specified to start the scanning. For the next cell graph, the parameters to be transmitted are only those changes relative to the last cell graph, and the scan operation resumes when the cell is finished. In the pattern generator, a set of cell patterns (rectangles, parallelograms, triangles, ...) are pre-programmed in advance and can be called directly while scanning the mask. A much more flexible feature of the system is the ability to dynamically load a microcode to define a graph cell that can contain several cell shapes and whose edges can be irregular or curvilinear. In addition, all the graphics can be drawn in an ordered array with just a few machine instructions, which makes it possible to copy millions of identical unit figures for just a few words. The article describes the structure of the graphics generator, as well as the rest of the entire 16K × 16K high-speed dot matrix addressing system.