Originally Posted by BillD
I've been thinking about and studying this topic for some time. There have been comments that the CD format (16-bit 44.1kHz sampling) is insufficient to produce good playback. If you believe information theory (Nyquist criterion), a sample rate of twice the analog spectrum is sufficient to perfectly reproduce the analog spectrum, so 44. 1kHz meets the criterion, and then some. The sampling frequency was chosen somewhat higher than the Nyquist rate since practical filters needed to prevent aliasing have a finite slope. Now, is 16 bits enough to represent the levels. 16 bits represents about 120dB of dynamic range, pretty good.
Most people will hear a CD as "perfect". There are those that claim both that the overtones and harmonics above 20kHz can be sensed and feel the need for more than 120dB dynamic range will claim otherwise (I actually tend to be in this camp, and digitize at 96kHz and 24 bits), but that is seriously more than any vinyl record can produce.
I have come to believe that this is all the recording engineers fault, pressured by artists and radio stations that want their music to "pop". If you put a scope on the output of a CD and a vinyl recording of the same piece, you will find that they do not resemble each other very much. I've found some CD's to be compressed so much, that the average level for the entire disk was -3dB. Back in the '70s, the average level for rock was -18dB, and this is what is typically found on vinyl (also because the medium doesn't handle the "all loud" format very well - hence the RIAA equalization curve). So, we will play our vinyl louder to make it sound the same level as the CD, and the added dynamic range will give the music much more impact.
The good news on all of this is while some SACD and DVD-Audio releases are affected, nearly all DVD-Audio discs also contain a Dolby Digital (AC3) or DTS sound track to allow the disc to be played in DVD-Video players without DVD-Audio playback capability. Dolby Digital has a defined and calibrated reference average playback level (-20dB), and the DTS track will also follow this level. It is therefore beneficial that the high-resolution DVD-Audio soundtrack will be produced at the same reference leveland this indeed is normally the case.
As these new high-resolution formats are marketed largely at audiophiles, attempts to master them for loudness would almost certainly be counterproductive, as the target audience is likely to be highly critical of sound quality and dynamics.
Sure, something could be done. The producers could adopt a common alignment level, which would be enforced by audio standards organizations. For example, the European Broadcast Union standards call for 18 dB of headroom. This would reduce the need to produce music that is mastered for loudness. Hybrid SACD, which allow two layers of audio data, could offer a compromise: two versions of music would be mastered, one that is compressed and one that is left as-is. The version with compressed audio could be played in environments which benefit from compression, such as on car stereos or portable audio players. The version left as-is can then be played in quiet environments with high-fidelity equipment.
Must correct serious errors in BillD's story.
The internet is full of serious and valid data, so it really is unnecessary to make so many errors and uninformed claims.
1. The analog spectrum is much wider in bandwidth than 20Hz...20kHz.
Transients of up to and over 50kHz have been proved to be of significance as to timbre and localisation of sound sources.
This is not a "claim", this is the outcome and the conclusion of scientific research.
2. Dynamic range of 16 bits is not 120dB. Each bit represents 6,02dB so 16 bits is 96,32dB.
3. Dynamic range on an lp is more than 96dB. It is not defined by the signal to noise ratio as many think, it is defined by the molecular structure of vinyl, and the smallest groove that can be imprinted into the material.
4. Human hearing easily detects discrete tones buried 20dB deep into random noise, also the conclusion of scientific research. A pristine LP read with a good stylus can produce a surface noise-to-signal ratio of way over 80dB.
The 17th harmonic riding on a violin note can reach -90dB but since it is riding on a note of say -10dB it is clearly lifted above the surface noise, provided that there is no noise modulation present.
5. The reproduction of small details in ordinary cd-players is greatly impaired by non-musical artifacts, which are real world spurious and music-alien signals which mask parts of the original content.
You can not measure that when you measure only steady state tones.
But nobody knows how to really measure music, let alone scientifically compare music played back through a variety of digital filters and dacs. It can be done, but would offend many a conceptual thinker.
Now thąt would be the proof that 44k1/16bits digital is fatally flawed.
However look what Bob Stewart of Meridian has wrought: the apodising filter. Brilliant piece of work. Yet not more than a band-aid: the digital format that clearly supersedes analog is already there and it is called 128FS DSD. Unfortunately no-one is willing to recognize it as such and declare it a new industry standard. It's a money game.