I have been giving the noise survey a lot of thought. It is critical because if outside noise is not adequately contained, the weakest point in the sound isolation will make the rest of the work meaningless. If the noise is not a problem then going with a floating floor will be a massive waste of time and money. Paul wrote somewhere recently that the floating floor in his studio is 40% of the construction cost. The noise question, as I understand it, is train tracks about 500' distant. Other sounds, and neighbours are not an issue.
Working from what we have spoken about, my first thing was to check that the wall construction is adequate. I modeled a wall with:
1/2" gypsum board, 2 layers
2x4 studs
5" airspace
insulation in airspace
2x4 studs
1/2" gypsum board, 2 layers
in Insul. The results are shown in the attached JPG. For those reading this who can't see the graphic, the STC is 67 with the TL 30 dB@50 Hz. The wall system will not be a problem with isolating airborne sound from the outside. This confirms our concern being if a floating floor is needed.
What we need for the sound survey is at one level fairly simple.
1) microphone with good low frequency response
2) a way of calibrating the microphone
3) electronics to accurately record the microphone output
3) system to analysis the the recording from 3).
Where it starts to get difficult is 1) and 3) need low enough noise figures that the self noise does not mask the desired signal. It can be quite expensive to achieve this. That is why the Radio Shack SLM only goes down to 50 dB SPL.
Good low frequency response is primarily with omni microphones. Do you have any good omni mics or friends with that you can borrow?
Assuming that you do, hook the test mic to your recording system with a calibrated level control,
With the test microphone and RS meter side by side outside, record a pink noise source with the RS meter indicating 55 dBC. Note the recording channel settings.
Take the test mic to the quietest place you can think of. Put the mic in a heavy walled sound absorbent lined box ( do like that fancy phrase for a road case?) with the just enough open to run the cable.
Record again and increase the gain until the level equals what you got from the calibration to the RS meter. This will hopefully give an indication of the self noise level of the mic/electronics.
Put that data through software with at least 1/3 octave or preferably FFT analysis capability.
Calculate the self noise spectrum.
Post/email (hopefully both) the results for review. We will be looking at the overall level and the spectrum to determine the masking.
Assuming the above is acceptable, find out when a train is going by and record noting the recording levels sos that we have a calibrated reference to the RS meter.
We can then Analise the results.
The RS calibration is not great, within 2 dB, but for our work we can take that into account.
Another option, and one that would remove any problems with technique band-aid systems would be to rent a suitable meter.
Bruel and Kjaer would have setups for that. I do not know the prices of the rental unfortunately. Something like their 2260 Investigator with software.