A modification of the steady-state (direct) calibration method is described where-by the phase response of a long-period seismograph system can be determined with a precision at least one order of magnitude higher than has been reported in the literature so far. In a period range of 5 to 200 sec this means determining the time delay of a system to within ±0.05 sec or better. An improvement is also suggested in the method to determine the system parameters from which the phase can be calculated theoretically. The improvement consists in the ability to reduce any bias in the estimate of the parameters. A comparison between the improved versions of the steady-state (direct) and parameter (indirect) methods indicates that for certain seismograph systems the agreement in the phase delay can be within 0.1 sec but for others the difference can be more than 0.5 sec. The difference apparently results from the fact that certain seismometers are not ideal instruments and cannot be represented by theoretical parameters exactly, rather than from any bias in the two methods of calibration. In addition, the steady-state method is used to study the temperature effect on the phase response of two types of long-period systems: (a) a conventional seismometer-galvanometer system and (b) a new seismometer-electronic amplifier-filter system. It was found that, in a period range of 5 to 200 sec, a change in temperature of 8°C causes a change in phase delay of about 0.4 sec for the first system and only 0.05 sec for the second system.