rafaelmardojai
Appreciations about sound levels
I'll be updating this file with my thoughts and recommendations about soundclips for Blanket, explaining as simple as possible the characteristics these files should have to be comfortable and suitable for environmental sound.
Disclaimer
This recommendations about sounds will be thought for humans! I suppose that this are the expected listeners of this application, but it is important to claim that this recommendations are not for other animals.
The standard and recommendations used here will be based on EBU R128, the standard of European Broadcast Union about audio normalisation and maximum sound levels.
The main unit used will be the LUFS, Loudness Unit Full Scale, which is a relative unit to humans hearing. More info here.
Characteristics
Here I'll explain each characteristic that should be taken into account to decide whether a soundclip is ideal for the app.
LUFS Integrated
The LUFS Integrated is an average measure which a sound has through all its length. A soundclip does not have the same level at each moment, it changes. With this unit we can measure an average of the clip.
The closer to 0 LUFS, the louder. It is inverted.
Dynamic Range
A sound has a changing level through its duration. With the dynamic range we measure the range between the loudest and the lowest level registered. The higher the range (the difference between the loudest and lowest level), the more comfortable the sound will be. Sounds with poor range exhaust the listener and, therefore, annoys.
This characteristic is almost the most important in ambient sounds, as what we want is to be comfortable, not annoying.
True Peak (dBTP)
This is the maximum level that the sound wave takes through the duration of itself. This is imperceptible because this maximum level lasts only milliseconds. But having too loud peaks produces annoying clipping (snaps, cracks, pops) and distorted sounds on speakers, so it is better to avoid having them.
Max LUFS (momentary)
This, unlike True Peak, is a more perceptible way to meter maximum loudness, as it meters it in a timeframe of 400ms.
How to measure
To measure the sound levels, all faders (sound-level bars) must be to the 100%. The most times this will be the loudest, unless it is indicated with percentage or with a mark of 0 dB.
0 dB = 100%
After setting app the faders, the sound source must be connected to an analyzer. I use x42 Meter Collection, open source.
Appreciations
I measured all the soundclips that are in the current version of the app, 0.3.0.
Here are the levels:
Current
| Sound | LUFS (Int) | Range | True Peak (dBTP) | Max LUFS (mom) |
|---|---|---|---|---|
| Rain | -17,2 | 0,3 | 0 | -16,2 |
| Storm | -17,8 | 9,6 | 0 | -11,5 |
| Wind | -16,8 | 1,2 | -3,1 | -14,1 |
| Summer Night | -14,5 | 2.5 | -3,4 | -11,3 |
| Coffee Shop | -22,7 | 1,9 | -9,3 | -18,7 |
| Fireplace | -38,8 | 1 | -9,2 | -35,8 |
| Pink Noise | -11,5 | 0,2 | -0,5 | -11,2 |
| White Noise | -1,7 | 0,1 | 3,7 | -1,5 |
- As said before, the most important characteristic for ambient sound is the range. So those with poor range,
Rain,Wind,FireplaceandPink and White Noises, will be the more exhausting ones, even more with higher average volumes. - The sounds are extremely loud, making it difficult to adjust with the faders in the app.
- Consequently, the peaks are also too loud.
Ideal approximated louder comfortable levels
| Sound | LUFS (Int) | True Peak (dBTP) | Max LUFS (mom) |
|---|---|---|---|
| Rain | -29 (-11,8) | -11,4 (-11,4) | -28 (-11,8) |
| Storm | -24 (-6,2) | -7 (-7) | -18 (-6,5) |
| Wind | -25 (-8,2) | -11 (-7,9) | -22 (-7,9) |
| Summer Night | -23 (-8,5) | -11 (-7,6) | -19 (-7,7) |
| Coffee Shop | -29 (-6,3) | -15,4 (-6,1) | -25 (-6,3) |
| Fireplace | -38,8 (-) | -9,2 (-) | -35,8 (-) |
| Pink Noise | -34 (-22,5) | -22,5 (-22) | -33 (-21,8) |
| White Noise | -29 (-27,3) | -24 (-27,7) | -29,5 (-28) |
The average of the reduced volume is nearly -7, which is more than the double.
The dBs are a logarithmic scale, 6 is the double
I encourage anyone that want to meter their ideal approximated louder comfortable levels and informing about them here
Recommendations
- The longest the recording, possibly the more comfortable. Sounds in nature vary a lot, they tend to have large dynamic ranges.
- Lower sound levels, specially for those monotonous soundclips.
- Avoid short and sausage clips.
Parameters that every clip should pass
As the maximum comfortable perceived level is directly correlated to the range, there could be variations, but here is a table:
| Criteria | LUFS (Int.) | Range | True Peak (dBTP) | Max LUFS |
|---|---|---|---|---|
| Minimum | -35 | 5-7 | -30 | -30 |
| Ideal | -27 | 7-🔝 | -7,5 | -25 |
| Maximum | -23 | 🔝 | -6 | -18 |
Thanks for the very detailed explanation!
Would be awesome to have some improvement on this for 0.4.0.
What do you think would be the next step to advance in this?
Well, I'm already talking to other sound technicians to have more accurate recommendations, but I have them more or less.
I would add some parameters that a sound should achieve, before accepting them into the project. That test should be public, in a CONTRIBUTING.md file or so, if anyone proposing a sound would like to past the test by themselves, or otherwise me or any other contributor could have the responsibility to pass the test.
I'll create the documentation for that test when I'll have the recommendations cleared!
What do you think?
Also there is a proposal of other technician about entering very low frequency modulation on sound to induce a specific brain state. Seems the neuronal movement frequency varies depending on the alarm-state we are.
So entering modulation in frequencies in which brain frequency is when calm could induce that state. It is very used on psych-trance and new-age music styles.
@rafaelmardojai, I updated the post. I added some more criteria, that could be a definitive one, but maybe it will be difficult to get sounds with such characteristics
Sorry, I had been a little busy with other duties.
Yeah, I took a look at it. Looks pretty good.
I suppose now we should try to fit the current sounds to the recommendations or replace them?
Also I think you can start making the CONTRIBUTING.md doc with this recommendations, later I will update it to add other contributing guidelines.
Hi! I'll now be working on porting sounds from #46 with a adequate criteria. After that I will create the CONTRIBUTING, but you can start with that, I won't make it today.
I'm done. Now I'll PR. About current sounds, my plan would be to try to fit them and otherwise replace them by new sounds.
I'm doing that too.