A. Akiyama. T. Katoh, K. Kudo, Y. Mori, T.T. Nakamura, J. Odagiri, M. Tejima, N. Yamamoto
Various kinds of VME modules are used in the KEKB accelerator control system and it is very difficult to fix the hardware problems happen in the complicated systems where a common bus and various kinds of modules are used. In the KEKB control computer system, it happened when we put a VME-MXI and a CPU modules in a subrack of an IOC. The CPU stopped about once or less a day depending on the configuration of the VME subrack. We have 20 such computers and the frequency of the CPU halts was too high for stable KEKB operation. We observed the noises among the signals on the VME backplane when the VME-MXI driver module was accessed. We first tried to make the shape of the waveform better and decrease the number of system halts. For that purpose, we put additional load on to the bus by putting a bus extender module into a slot. It gave us a preferable decrease of system halts but was not the final solution. Then we started to analyze bus signal carefully and found abnormal bus cycles happened when the CPU module requested a bus cycle to the VME-MXI module and the CPU module did not complete the bus cycle. We reported the fact to the manufacturer and in reply, they sent us patch information about the module. Since we put the patch on all the VME-MXI modules we have not observed any halt. The process of this experience will be described.
Last updated: September 13, 2002
Website Problem? Contact Technical Publications