(idea) by hardly Fri Jun 18 2004 at 12:00:14

The debate of CD vs. vinyl is a popular religious issue amongst music fans, audiophiles, DJs and techies. The arguments are usually peppered with scientific factoids ranging from the pedantically correct to the hilariously mistaken. Here I intend to explore the technical issues in a balanced manner. Reasoned scientific feedback, debate and corrections are most welcome.

I shall try to avoid the quagmire of subjective aesthetic judgement; having never owned any vinyl records I couldn't say which format sounds "better".

The Frequency Domain

Frequency is pitch - the difference between a low, deep sound and a high piercing sound. As we hear sounds as a sum of frequencies, most of the important information is in this domain, and accurate reproduction is correspondingly important.

The sample rate of a CD is 44,100 samples per second. According to the Nyquist theorem this is sufficient to capture all the information in a signal that contains no frequencies above 22,050 Hz.1 This bandwidth is more than enough. The human ear is most sensitive at around 3,000 - 4,000 Hz. (The fundamental tones of a piano range roughly 25 Hz to 4,200 Hz)  The upper frequency limit for human hearing is roughly 20,000 Hz, while for a middle aged male this figure may be nearer 15,000 Hz. At both extremes of the audible frequency range the perceived loudness relative to the signal power is less than that for moderate frequencies.

The reason for this extra sampling overhead is actually to make the CD player's job a little easier. When recreating an analogue signal from digital data it is necessary to filter artefacts generated above the Nyquist frequency (in this case 22,050 Hz). Analogue filters with sharp cut-off slopes are expensive, so a 2,050 Hz gap to ramp down is quite helpful. Still, a poorly designed filter might unduly attenuate the audible high end or have a "ringing" frequency response across the full range.

Sound behaves linearly, which means that frequencies above 20,000 Hz cannot somehow affect, alter or "shape" the lower, audible sounds. Nevertheless, audiophiles often believe that these high frequencies actually do make a difference. The best scientific support I have seen for this view is a paper2 which suggests that brain activity may be somehow affected by the presence of hypersonic sounds.

Frequency domain issues with vinyl have an entirely different nature. First off, vinyl records are not recorded with a constant frequency response. To properly reproduce low frequencies would require larger grooves, so records are pressed with the low frequencies reduced. In addition, as a strategy to drown out noise, higher frequencies are boosted on the recording. A RIAA equalisation amplifier mostly corrects the frequency response on playback. Problems introduced by cheap turntables include "wow and flutter", which refers to frequency shifts caused by variation of playback speed and "rumble", low frequency noise from the motor.

Dynamic Range

The dynamic range is the resolution of the recording - the ratio of the loudest possible sound to the softest, expressed in the logarithmic decibel (dB) scale.

For CDs, each sample has a 16-bit range (65536 values). This might not seem to be a great deal but it actually translates to roughly 96 dB dynamic range. I would argue that this should be sufficient for any music, when we consider the range of human hearing.

By definition, decibel levels are always comparative - the difference between the loudness of two sounds. When you see an absolute figure quoted, ("A lawn mower measures 90 dB") this is in relation to a defined zero point - 0 dB is defined as the softest sound audible to a person with perfect hearing. This means that if you set the volume of your CD player such that you can hear all the sound it contains, the loudest point on the track will measure at least 96 dB on the absolute scale. Listening at a higher volume than this for too long will reduce the need for a large dynamic range.3

The biggest general misconception about analogue technologies is that they have infinite resolution, which is never true. (In this example, a PVC molecule is small, but not infinitesimal). In practice, the dynamic range of vinyl records is affected by the age and quality of the recording, and the density of the tracks. The best-case seems to be around 60 dB. The dynamic range will be adversely affected at low frequencies because of the equalisation issue detailed above.

Errors and Noise

CDs are recorded with interleaved redundant information built in. This allows the decoding machinery to correct or interpolate bad data caused by damage to the CD surface. However, should the damage be too great, the CD will simply skip.

Any wear to the surface of a vinyl record will cause a gradual deterioration in the quality of the recorded signal. This manifests as hiss and reduces the effective dynamic range. Scratches and other damage can result in pops and clicks that momentarily drown out the recorded sound.

Conclusions

CDs technology is mostly an improvement on vinyl technology. Stop worrying and just enjoy the music.4

  1. Frequencies above the Nyquist point will cause interference if not removed before sampling. This is one motivation for sampling at higher rates in the studio.
  2. Oohashi, Tsutomu, Emi Nishina, Manabu Honda, Yoshiharu Yonekura, Yoshitaka Fuwamoto, Norie Kawai, Tadao Maekawa, Satoshi Nakamura, Hidenao Fukuyama, and Hiroshi Shibasaki. Inaudible high-frequency sounds affect brain activity: hypersonic effect. J Neurophysiol 83: 3548–3558, 2000.Also available online at: http://jn.physiology.org/cgi/reprint/83/6/3548.pdf
  3. In UK workplaces hearing protection is mandatory for sound levels above 90dB.
  4. I've never listened to so much music as I did when I was 17, using a cheap cassette player. There are factors more important than bidirectional copper cable.


I like it! 1 C!