Sound Loudness, DBFS, SPL

Hello,

I manipulate an audio file in Audacity to get the results I desire. I use an Audacity plugin and get the weighted value (A weighted). I utilize a an algorithm that analyzes a sound file and provides the result in DBFS (A and C weighted). The results are the same. What I am now trying to determine is how to relate this to (what I believe) is SPL and what is safe for hearing. In my manipulation of the audio, everything below 14Khz is filtered out, leaving 14K-20K intact. The user listens to the audio file using headphones, so I need to be VERY sure my measurements and usage are correct.

Currently my goal is to understand what is safe for hearing exposure and be well below those thresholds but my tools for measurement, Audacity and the algorithm, express the results in DBFS. How can I utilize these values or determine how loud they are relative to the customary expression in dB.

Results such as -31.45 DFBS mean something to me when they are relative to another DBFS measurement because I can compare the two. However, I don’t know how to make the comparison of the audio is expressed in dB as is done in OSHA exposure recommendations to the DBFS output in the tools I am using. How do I go about making the connection between the two expressions of sound loudness?

https://www.osha.gov/dts/osta/otm/new_noise/appendixc.pdf

Thank you,


W.

There are several problems that would need to be solved, and I’m not sure that the all can be solved.

  1. A-weighted filters are generally designed for accuracy over the audio spectrum, and (unless scientifically calibrated and certified) are not likely to be very accurate at the ultrasonic (or infrasonic) end of the spectrum.

  2. To convert from dBFS to dB SPL, the system needs to be calibrated, and the frequency response needs to be flat across the frequency band that is being measured.
    The first part can be accomplished quite easily for a single frequency (say 1000 Hz) by measuring the SPL with an SPL meter, though it becomes a lot more complicated for headphones because the resonance between headphone and head needs to be taken into account. Achieving a flat frequency response from the system will be extremely difficult without specialist equipment as even good quality headphones will usually have peaks and troughs in their high frequency range in excess of +/- 10 dB.

  3. Intermodulation distortion is also a major problem when trying to measure very high frequencies. Any harmonic distortion that occurs in the analog stages of the equipment will create ultrasonic frequencies above 20 kHz, which will interact with each other to create sub-harmonics within the range that is being measured, which could thereby introduce substantial errors in the measurements. To counter this, the measuring equipment would need to have an accurate measuring range that extends well beyond 20 kHz so that the magnitude of such errors can be estimated. Without specialist equipment, it is unlikely that you will be able to measure frequencies above 20 kHz accurately.

While this sounds like an interesting project, I doubt that it will be possible to achieve reliable results without using expensive specialist equipment, and even then it will require a thorough understanding of the specifications and standards involved.

I’ve never heard of anybody doing it with headphones. That’s hard.

I have an older SPL meter and I know someone with a more modern digital sampling unit, but neither of those will work with headphones.
Screen Shot 2017-08-03 at 15.55.02.png
Koz

There are several problems that would need to be solved, and I’m not sure that the all can be solved.
I didn’t realize that I had that many problems to solve.

  1. A-weighted filters are generally designed for accuracy over the audio spectrum
    The one good thing is the filter I’ve been using have been accurate to (as I remember) to less than a .5 dB over the entire spectrum
  1. To convert from dBFS to dB SPL, the system needs to be calibrated
    headphones will usually have peaks and troughs in their high frequency range in excess of +/- 10 dB.

I’ve been really worried about the measurements. I had been wondering if I need one of those head measurement dummies. As I think about it more, this kind of thing is probably a lot more than I need

  1. Intermodulation distortion is also a major problem when trying to measure very high frequencies.
    Without specialist equipment, it is unlikely that you will be able to measure frequencies above 20 kHz accurately.

Oh (gasp of air, followed by a very surprised look), so that also means that I need to increase my sampling rate a lot higher than 44.1k to account (measure) for this.

Excellent information.

Thank you,


W.

I used to own one from Radio Shack but it was only roughly accurate but I understand what you mean. It gives me something to think about.

Thank you,


W.