Tower Controller described a loss of separation on two successive departures when anticipating normal operating characteristics would be used by the second departure aircraft.

Narrative

[I was] working Local during the departure push. I cleared Air Carrier X; a heavy jet; when Air Carrier Y was 1.5 NM off the departure end of the runway. The runway is 2 NM long. I was anticipating the required 4 NM separation between two heavy jets. I have performed this operation hundreds of times and have found that the resulting separation is usually somewhere between 4 and 6 NM in trail; depending on how long it takes the second aircraft to start his roll. Today however; Air Carrier Y was very slow and Air Carrier X must have had his power up and the parking break engaged. As I finished the takeoff clearance; Air Carrier X immediately started rolling while the pilot was reading back. I have seen this in the past; but it is a pretty rare event. The combination of Air Carrier X 'getting all over it' and Air Carrier Y operating at the extreme low end of his climb envelope; resulted in 3.6 NM separation. When I observed Air Carrier X's very aggressive takeoff; I knew that separation was going to be at or near minima. As soon as he got airborne; I instructed him to turn right heading 242 (15 degrees right) to try and get him out of the wake of the proceeding departure. Of course; the deal was already sealed; separation was lost as soon as he lifted off. I used the splat T function of the STARS to measure the distance; but the display was obscured by his data tag. It looked like it read out 3.9; but it was hard to tell. When I moved the tag; it was reading 4. In my experience; 90% of this Air Carrier's pilots take a significant amount of time to start departure roll. In the past; when our facility got TARP; I waited to achieve the full 4 NM separation before clearing the second heavy for takeoff. I found this resulted in excessive departure delays and spacing of 6 to 10 NM between departures. In trying to mitigate delays to the user and balance my separation responsibilities; I thought that using a 3.5 NM interval was extremely safe; without being excessive. Obviously; I was wrong. I have not listened to the tapes to see if the two minute interval could be applied in lieu of the RADAR minima. I was not trying to use two minutes; I was trying to use the 4 NM radar minima. Given that I have used this technique hundreds and hundreds of times and others use similar; if not the same technique; I am not sure what to change. Even had I waited the full 4 NM minima; I am unsure that would have maintained the required spacing between two aircraft performing at around 100 KTS difference in airspeed. In this case; I was able to diverge the aircraft after takeoff. However; that might not always be possible. Without the turn; I don't know what separation might have shrunk to; given the 1.5 NM/minute closure rate. Given the nature of this particular problem; I intend to use the full 4 NM separation requirement before issuing clearance to the second aircraft. Getting pilots to standardize their departure procedures would greatly help. Having a pilot report the need for a high speed climb before he takes the runway would also help. Changing my attitude about 'service to the user' and minimizing delays is going to be required. I 'cringe' inside when I use 100% more spacing than required. I guess it is time to give myself an attitude adjustment and start accepting huge departure delays.