Informative Topic: Spectral Content Of Music
Above we see measurements of background noise in recordings, chosen to range from reissues from 60-year-old unprocessed analogue tape to modern digital recordings. Obviously these analyses embody the microphone and room noise of the original venue, but in some, analogue tape-recorder noise. Even the best recorder's noise floor is above that of an ideal 16-bit channel. It is worth noticing that a 20-bit PCM channel is more than adequate to contain these recordings and that consequently 32-bit precision offers no clear benefit.
For any individual recording, we find that the peak and noise curves form a triangle. The type of triangle (acute, obtuse, etc) depends on the specific material. However this insight allows us to consider that for every practical recording there are these elements:
A background noise which persists before and after the music is playing
A peak level that infers a finite rate of innovation (almost a slew-rate limit)
A frequency above which the variable part of the signal appears to be below the noise
An area, which when plotted on a Shannon diagram, informs us of the actual information content in the music.
Purple shows surface noise from a 78rpm shellac transcription;
Olive shows the peak levels of the Hypersonic music box in a special transfer supplied by Dr. Nishina (footnote 5).
Red is the peak spectrum of a close-microphone recording of a harpsichord, put together as a torture test with a Sanken 100kHz microphone under the lid of the instrument close to the plectrum area.[45]
Not shown here are the background noise levels for the added signals, but they are high (fewer than 10 bits in the Music box and 12 bits for the harpsichord). The test signals are discussed later.
Footnote 5: The music box was a vintage Polyphone made in Germany around 1920; recording was made originally using a microphone with 100kHz bandwidth placed inside the instrument adjacent to the reeds. A less severe version of the recording is available at [38]. Audio above 50kHz is rapidly attenuated by air and this technique allowed powerful harmonics to be captured up to 100kHz using an experimental high-speed DSD converter.
Figure 14. Peak spectral level gathered over a corpus of 96kHz and 192kHz recordings.
There is significant content above 20kHz in many types of music, as an analysis of high-rate recordings summarized in Figure 14 has revealed. One notable and common characteristic of musical instrument spectra is that the power declines, often significantly, with rising frequency.
Even though some musical instruments produce sounds above 20kHz [27] it does not necessarily follow that a transparent system needs to reproduce them; what matters is whether or not the means used to reduce the bandwidth can be detected by the human listener.
Figure 15: Examples of background noise in 192kHz, 24-bit commercial releases. Also shown is TPDF dither noise for 192kHz, 16- and 20-bit quantization. Curves plotted as noise-spectral-density in 1Hz bandwidth.
Figure 16: Showing the peak and noise plots for two recordings.
Figure 17: Showing peak levels in a large survey of commercially available 192kHz recordings.
In the figure above we see a black curve which links the peak spectral peak levels of a large survey of commercial 192kHz music files. We see that in general, recordings issued at 192kHz show lower levels than those mastered at 96kHz (evident from the step at 48kHz in Figure 14) and in this we have detailed in Blue the peak spectrum of a cymbal crash in an excellent wideband recording by Mick Sawaguchi [40].
We have also added the peak spectra of three extreme non-commercial recordings we have used for testing:
Footnote 5: The music box was a vintage Polyphone made in Germany around 1920; recording was made originally using a microphone with 100kHz bandwidth placed inside the instrument adjacent to the reeds. A less severe version of the recording is available at [38]. Audio above 50kHz is rapidly attenuated by air and this technique allowed powerful harmonics to be captured up to 100kHz using an experimental high-speed DSD converter.
