Dear Analog Lovers,
being True Believer that many LPs sound better than their CD counterparts I have been spending quite some time digitizing my vintage LP collection and, int the intent of getting the best possible transcription, I've been measuring what comes out of the system and trying to figure out what can be improved. Also being a curious animal I tried to get an understanding of why is it that in many cases I do like vinyl better than CDs. I'd like to share some of my conclusions with you. What follows is a little long and somewhat disorganized, so please be patient with me.
One of the claims of analog records is that they have a more extended frequency content than CD records, which are cut off at 22kHz. A question for me has been what ADC sampling frequency should I use to get all the content from the LP?
From the sample of records in my possession (mostly rock and jazz with some classical from 70-80s) it turns out that the actual bandwidth is actually not really very extended, rarely in LP records you see any significant amount of signal above 15kHz, even less at 20kHz. Some (very few) records do have significant high-frequency content, one example is Brian Eno's "Before and After Science", where cymbals are well recorded and sometimes prominent (see "Energy Fools the Magician"). In this case there is lots of harmonic content well above 20kHz.
Later LP records from the 80s which were mostly "Digitally Mastered" have the same frequency response of today's CDs, since they were mastered using old digital systems that would not go beyond 16 bits/44kHz. An example is "Ghost In The Machine" of the Police, good recording, but definitively digital (i.e. a 16 bit/44kHz master tape transcribed to vinyl). Looking at the music signal spectrum you can see that high frequencies extend all the way to 20kHz and then brusquely stop there.
So far my understanding of the actual difference between the sound of LP records and today's CDs is mainly due to the different degree of compression and by the different global equalization used in their mastering. Old LP records had very little or none at all compression (aside from the "natural" amount of compression that the analog tapes used for recording would provide by saturation). Mainstream (pop/rock/jazz) LPs also had a fairly limited low frequency content (i.e. nothing below 40Hz and a good deal of attenuation below 150Hz), due to concerns that the pick up of cheap systems (the vast majority) would mistrack and literally jump out of the groove. High frequencies were also much less forward than what we have on CD, producing a much warmer sound. I think there are two issues here: first the music publishing industry wanted CDs to sound "differently" from LPs, so that consumers could easily tell them apart and find out that CDs sounded "better". The second issue is that analog cartridges (even the good ones) have high distortion values in the high frequencies and just cannot reproduce them correctly at high signal levels (more about this later on).
It is very hard to find good measurements for phono cartridges nowadays, the only place where I could actually find any on the web is on the Miller Audio Research web site, which hosts the measurements for the british "HI-FI News & Records Review Magazine".
I also made a bunch of measurements on my own on a (small) variety of good cartridges I happen to own: two Shure V15 Type V, a Pickering XSV 3000 and a Dynavector 10X3. I acquired a number of test LPs and measured traceability, linearity and distortion in various conditions.
Looking at the numbers it is strange that we actually ever considered this stuff as HI-FI, given that the amounts of distortion they produce is really very high (anywhere between 0.5 and 10-20%) and linearity is approximated between +/- 3dB. Traceability depends on the cartridge technology. The Shure and the Pickering are quite good, but the Dynavector is pretty horrible instead (my arm might be too light for it to perform well though).
For good cartridges, IMHO, the major source of distortion (and of non-linearity) is the high-frequency resonance of the cantilever-pickup system. The reason why these are made very small, light and rigid is that the manufacturer tries to have the resonance as high as possible so that it falls out of the audio band. Some cartridges, like the Shure V5 Type V, have a hollow beryllium cantilever, the Audio Technicas have it made out of boron, high end manufacturers would use ruby or even diamond for the cantilever (rigid indeed but not very light). Information about the actual resonance frequencies attained by the manufacturers are sparse, Shure claims that the V15 series had something like 24kHz for the Type IV and 33kHz for the Type V.
What I think often happens in practice is that most cartridges have the resonance squarely within the audio band (somewhere between 15kHz and 20kHz) or just past it, and the frequency response is "flattened" by adding an electric counter resonance that would compensate for the mechanical one, which explains why often the high frequency linearity of cartridges is often so challenged.
The existence of this (quite pronounced) high frequency resonance also is in my opinion the source of the high distortion levels of most phono cartridges: the music read by the pick up will inevitably excite some of the resonance modes of the cantilever, producing harmonics and spuriae that don't belong to the original signal. From the numbers published on the Miller Audio site (and for what I could measure myself) you can see that, for the good cartridges, distortion rises kind of linearly with frequency up to a few kHz (staying around 2-4%), after that the steepness of the distortion curve becomes significantly higher and quickly reaches 10-20% above 15kHz. It is also interesting to notice how the distortion curve is often quite different for the two channels. My opinion is that this is due to the fact that the signal of the two channels is extracted from addition or subtraction of the vertical and horizontal displacements of the stylus, which might indeed have quite different resonance modes.
The significance of this high frequency resonance and its associated distortion modes depend on how we look at them. First of all we can notice that all distortion products for frequencies above 10kHz fall out of the audio band (i.e.: above 20kHz), so you probably can't hear them anyway. Also from the energy spectrum point of view, in LP records there isn't much signal at all above 10-15kHz (aside from cymbal crashes), so the amount of distortion that reaches the rest of the system is quote negligible.
Sometimes though, as I was pointing out before, there is quite a bit of high frequency content, which leads to the generation of quite significant harmonics, for a tone at 15kHz the first harmonic is at 30kHz and the second at 45kHz. For a tone at 20kHz you get 40kHz and 60kHz harmonics and so on...
The consequence of this distortion when digitizing LPs is that If your ADC has a sampling frequency not sufficiently high you will end up with possibly significant aliasing of the distorted signal in the audio band. That is kind of nasty.
Another consequence I would derive from this discussion is that we probably don't hear frequencies above 20kHz and, if we actually do, what we pick up out there from our analog reproduction chains is the distorted garbage generated by the high frequency resonance modes of our beloved phono cartridges...