(Respectfully snipped for space)
Thank you for this helpful information. And explained in much simpler, more easily understood terms. I look forward to your additional research & feedback (as mentioned), if you have the time.
<snipped with respect also>
Here are the results of my informal research. I'll post photos & details of my equipment and setup later. Basically I used a recording of a woman's scream, played back at calibrated levels with my partner as the operator and I was the blind listener:
Background noise of my room in the evening: 31 dBA SPL
Average SPL of scream at listening positioning:
25 dBA - barely audible, barely intelligible - really need to listen out for it with lab test style attention, could tell was a wailing, high-ish pitch, continuous sound like maybe very faint alarm or machine, but not really identifiable as a scream.
35 dBA - clearly audible, readily identifiable as a woman's scream, very quiet - sounds far off in the distance
42 dBA - same as above but the slightly louder nature made it quite a bit more alarming - more than the numbers suggest - if I heard this in the night, I would immediately wake up my partner and be very concerned
50 dBA - easily audible, definitely alarming, I would waste no time calling the police straight away
The 35 dBA scream played with 45 dBA pink noise cut through very well and was still clearly identifiable - therefore I believe an accurate estimation of the natural background noise on the night is not significant.
My calculation of the expected levels of the scream at 177m is that a reasonable range to predict would be worst-case 23 dBA, best-case 53 dBA, and that a typical figure would be 43 dBA.
(The variation comes almost all from the attenuation characteristics of the toilet door at the source and the window at the receiver. So, although it is probable for a witness to hear the scream and its character, on the other hand it is also entirely and reasonably possible for a witness to not have been able to hear anything - but it all depends, and you need measurements of, the door and window. This could have been easily performed as access to similar doors and windows and buildings was available, but this was never done and the modelling leaned towards the worst-case attenuation and worst-case level of scream. The more likely values were never considered at all, and the best-case levels similarly omitted - this lack of a range suggests possible bias imo. I cannot for the life of me remember Lin's table and ranges for audibility/intelligibility, and have not checked it, so don't know how this matches with my findings)
