I am
ABSOLUTELY going to avoid "yet another discussion about what cannot possibly be audible".....
HOWEVER, I feel obligated to point out something about the graph included here....
We have a single graph representing the energy spectrum summed over a 60 second period of the sample.
By definition, this means that what we see on that graph is an average, taken over that 60 second interval.
So, for example, the amount of energy in that sample at 20 kHz,
WHEN SUMMED OVER A SIXTY SECOND PERIOD, is about 60 dB lower than the average for the entire audio spectrum.
This shows that the
average amount of energy at 20 kHz is quite low (and probably insignificant.).
However......
- by definition, averaging the result eliminates our ability to see short term variations
- we could get that result if the energy level at 20 kHz was a smooth consistent -60 dB over the entire 60 second period
- we could
ALSO get the
SAME result if the energy level at 20 kHz was -80 dB for 58 seconds, but our sample also included a 20 kHz tone at -20 dB for 2 seconds
If you want to suggest that "in the entire recording there is nothing significant at 20 kHz" then you need to show that the level of 20 kHz content never reaches a significant level
AT ANY SINGLE TIME DURING THE RECORDING.
For that you would want to use a
PEAK indicator of some sort.
Most programs that offer signal analysis will offer the ability of tracking peaks....
- so you would want to show that "the highest peak level of signal content at 20 Khz
NEVER EXCEEDED" some low number (maybe -60 dB).
- simply showing that the average is very low may suggest that "there isn't much there", but doesn't suggest that "there isn't occasionally something significant there for short periods of time".
In layman's terms, if I were to sit here and tap on my coffee mug with my pencil every few seconds, and record and analyze the resulting sound....
The
AVERAGE sound level would be very very low.... but the individual taps, each of which was quite loud, but only for a very brief interval of time, would be clearly audible.
So, if you were attempting to prove "whether there were audible sounds or not", a graph of the average would be the wrong tool for the job.
If the recording itself is limited to frequencies below 20 kHz, then we would expect the
PEAK level at 20 kHz to
NEVER significantly exceed the noise floor.
(Or, if it does exceed the noise floor on occasion, we would want to show that those excurssions were caused by noise spikes, perhaps caused by surface ticks, rather than by potentially useful musical content.)
Ten years ago or so I recorded a vinyl record with 88200 Hz sample rate out of curiosity with Olympus LS-5. Here is the magnitude spectrum of the A-side track, 60 seconds section from 2 minute mark to 3 minute mark:
Now, The spectrum starts to drop fast at 10 kHz and at 20 kHz it's about 60 dB below the highest levels between 60-100 Hz. It's ridiculous to think filtering frequencies above 20 kHz would make any kind of audible difference and in my case the age of 47 makes it impossible for me to hear much anything above 16 kHz. The vinyl in question is
The Beatmasters with Merlin -
Who's in the House - LEFT 31T - Rhythm King Records 1989.