Iranian strong motion data are studied to find the spectral demand ordinates of the selected records having better signal to noise ratios. The capacity spectrum method is used in this study and the spectral demand curves are estimated for 89 records. The records having low signal to noise ratio in the frequency range of less than 0.3Hz are excluded, and the rest of records with appropriate quality in long periods and a PGA (at least for one of the three-components) greater than 50cm/sec2 are processed and analyzed. The site classification for these records is performed based on the receiver function method (estimating H/V ratio for any recorded motion). According to this procedure, the number of selected records were 22, 16, 25 and 26 for the site classes 1, 2, 3 and 4, respectively. The frequency contents of most records show dominant amplitudes between frequencies 0.2 and 10Hz. This selected catalog of 89 accelerograms is obtained from 45 earthquakes. The demand curves for these records are classified for horizontal and vertical components and for two major seismotectonic regions of Iran, Alborz-Central Iran and Zagros. The demand curves for 2 near-fault recorded motions in Bam and Tabas are classified separately while the average demand curves for such conditions are presented as well. These curves show significant difference between near and far fault motions as well as between rock and soft soil sites. However, no significant difference is distinguished between hard alluvium and deep incoherent soil classes which could be due to fewer available data for some site classes and unequal available data for different conditions.