While reading up on MP3 compression, I read through this concept: Auditory masking.
What this basically means is that if two sounds (tones) are played simultaneously, the tone with higher intensity will "mask" the tone at lower intensity, as long as they are within a few Hz of each other.
The listener will be unable to distinguish between these two, and hear the combination of the two tones.
For illustration, take a look:
This means that even if I remove these 'quiet' sounds from the signal, the listener will not notice. Hence the signal can be 'compressed', to take up less space. This is the concept used in MP3, and the reason why its hard to distinguish between high bitrate MP3 and lossless.
But here's where it gets interesting. Masking follows a pattern, based on the Masker vs Signal Frequency:
Here, the masker is centered at ~220Hz.
The curves represent the threshold of this masking, both in intensity, and as it varies with the frequency.
Also represented is the auditory threshold, the point below which any such signal is inaudible.
Take a look at this pattern as well, this represents the variation in the masker intensity:
Now, consider this from the point of view of Frequency Response of an Amp/Headphone/Speaker, taking the two diagrams above as example, and a typical frequency response of Shure SE425 vs Etymotic ER-4S:
If you see at ~200Hz, the two curves differ by about 5dB.
I can think of the following implications:
- If there is some signal just at the threshold, and being masked, elevating it by 5dB will result in it becoming discernible, meaning you'll hear something you might not have heard with other headphones. Thus, the Shures will be revealing in the low/mids, while the Etymotics in the highs.
- Conversely, the attenuation of the signal at high frequencies also means the masker will be attenuated, and based on the curves, it'll reduce the masking effect as well. Hence, some of the signal being masked at other (higher/lower) frequencies, may be revealed.
- If compression algorithms like MP3 can remove these masked signals, does that mean the possibility of hearing these masked signals is also reduced? In lossless that should not happen.
So, I put forth the following questions:
- Why are some headphones more revealing? Does it have to do with this phenomenon? Revealing here means you hear sounds you've never heard before, missed on other headphones.
- How can we enjoy our music fully? Headphones and speakers with a large variation in the frequency response from the original have the effect of revealing some sounds, and masking others, so there's no perfect sound. We are always hearing something at the expense of missing something else.
- Should we be compressing our music? Even if the compression model is based on safe assumptions, signals that could've been heard are lost by compression.
- Do we need to buy a few headphones with varying response? If every headphone hides some, reveals some, then it seems you need both kinds of headphones.
- How much role does music production have to play? Recording music is different from playback, as the levels of different instruments and sounds are adjusted after recording.
Also, please feel free to correct me if I'm wrong.
Edited by proton007 - 5/15/12 at 6:37pm