The modified source codes of NEC-81 antenna analysis program are located in this directory. The Sun-4 Fortran 77 compiler can read this code and the code works in Sparc-based SUN-4 computers. Some f77 compilers (e.g. Stardent) do not accept this code, though. The compiling process is performed as follows (under UNIX OS): f77 -0 nec3000 *.f >& errors which makes an executable file nec3000 and directs the compiler messages to the file named errors. The array dimensions in the source code are modified to allow the maximum number of segments to be 3000. The altered variables are listed below: Variable Original Altered Purpose name value value ------------------------------------------------------------------------ LD 300 3000 Max. number of segments and surface patches IRESRV 8100 1000000 Nr. of words in core storage NETMX 30 100 Max. number of non- NDIMN 30 100 radiating neworks. NDIMNP 31 101 NETMX, NDIMN and NDIMNP must all be specified NSMAX 30 50 Max. nr. of voltage sources LOADMX 30 100 Max. number of loads (LD Cards) The program executes reasonably fast, if the simulation fits in the CPU memory. If the machine has to use disk as a virtual memory, the execution time grows rapidly. The core size is about 20 Mb, when the listed variable values are used. The core can be made smaller by reducing the value of IRESRV. The program then makes temporary files on the disk, if the simulation does not fit in the core. nec_bat is a shellscript-file, which reads the names of the input- and outputfiles from the file necfiles and writes the start- and endtimes of the execution to the file dates. The program messages are directed to the file nec_run_log. The purpose of the batch file is to allow running NEC on background. nec_bat should be used as follows: nec_bat & which starts the process and leaves it running on background. Of course, the NEC program can be used interactively as well, by typing the compiled program name (e.g. nec3000) at the command line, in which case the program prompts for the input- and outputfilenames and continues running on foreground. The results of the program have been compared to the results of the PC version and they seemed to match. Only small differences in 3rd and 4th decimal of the computed gains were noted. The execution time was about 1/10 of the PC version when computing a 250-segment simulation. This is the first modified version, any comments and suggestions are welcome. Petri Kotilainen OH3MCK