I think an aspect of sound that's being ignored, and that Umbran hit on though didn't specifically label it, is the
frequency of sound. The brain differentiates sound based on volume
and frequency - with both being equally important (neither dominant).
In general and only taking into account volume (dB), the science that's been presented in this thread is spot on - but it's not the whole story.
For the most part, it takes a certain range of volume increase over ambient before a new sound is detected - but only if they are basically the same frequency. If your ambient noise is at 50db, but it's mostly in the middle frequency range, and there's suddenly a very high pitched sound - it does not have to exceed 50db to be noticed.
I don't know the science for this but it seems to me that the larger the difference in frequency, the lower the db threshold necessary for detection. Though a
really loud sound can still drown out any new sound - so there's a threshold there also (like if you're standing next to a running jet engine - it doesn't really matter what frequency or volume the new sound is).
Although not scientific, some personal examples are:
- With older "tube" tv's I can hear the very high pitch of the tube when the tv is turned on - even in a fairly noisy and crowded room. The sound itself is not very loud (low db), but very noticeable to me because of the high pitch (frequency).
- In a quiet house with consistent ambient noise (such as a central air system), I can oft times hear someone come up behind me due to changes in the ambient noise (pitch and volume), and not due to them actually making a discernible noise themself (the pitch and volume of the ambient noise shifts due to their position, kind of like varying the length of a resonating tube).
Now, how to model that is another story. But modeling a mechanic with only db as the contributory factor creates an unrealistic mechanic (if realism is what you're shooting for).