The analyst must decide whether to use on-line analysis equipment to process the data at the test site or to record the data for subsequent analysis
on a large, off-site digital computer. Convenience, cost, capabilities of avail­able manpower, and the delay that can be tolerated will dictate the choice.

If a large digital computer is to be used, then a multichannel FM tape recorder may be used to record the input and output signals. The analog tape is subsequently played into an analog-to-digital converter and a digital magnetic tape is written. Another choice is to use a digital data acquisition system that digitizes the data at the time of recording and stores the data in digital form. The digital data (obtained by either procedure) are then used along with an analysis program (probably an FFT program) on the large digital computer. This method rates low on convenience. There is a long delay between the completion of the test and the availability of the results. Also, this method requires considerable data handling and the operation of several different devices. The cost depends on whether the necessary equip­ment is already available or whether it would have to be purchased or rented for the tests. The manpower capabilities will vary greatly from one organiza­tion to another.

The other possibility for data analysis is to build or purchase an on-line analyzer. Analog analyzers have been used, but most on-line analyzers now use small digital computers. The analyzer consists of the minicomputer with appropriate software, a multiplexer, an analog-to-digital converter, and some sort of output device. The cost of the equipment ranges from $25,000 to $75,000, depending on the number of special options.

Most available analyzers that employ a minicomputer use the FFT algorithm. It is certain that the FFT algorithm is the fastest procedure avail­able, but the price paid for this speed is that the sampling rate and signal characteristics must give 2" data points per period and the computer must have enough storage to hold at least a block of data at a time. An analyzer based on the procedure of Section 4.4 would have no restriction on the number of data points and the storage requirements would be very small. Also, the programming for the algorithm of Section 4.4 is much simpler than the programming for the FFT. (However, this is often not a problem because FFT programs are available for most computers.)