The compact disk, its demise(?) and classical music

[Cato]

Many thanks to all for the information!

Some years ago here at GMG we discussed the "decay" of a CD because of an article on "bronzing" and other problems.  I have a 31-year old Schoenberg CD (London label, with Ashkenazy conducting Verklaerte Nacht ) which I won in a contest back then.  (I had to wait 2 years to play it, because we had no CD player!)

It still sounds quite fine on my system, which uses a 6-speaker BOSE Surround Sound unit.

If I last another 30 years, along with the CD, I will be very satisfied!     0:)

On the other hand, some "burned" CD's given to me by my students in the 1990's and early 2000's have developed problems, but not all of them.

[Spineur]

Many thanks to all for the information!

On the other hand, some "burned" CD's given to me by my students in the 1990's and early 2000's have developed problems, but not all of them.

By the way a trick to make burned CD last longer: you force the software to do it at the slowest speed possible (x1 usually).  This way, the laser will maker bigger cleaner holes.

[71 dB]

Of course CD's are not at concert hall level

What is "concert hall level"? Dynamic range? Put it in technical terms. I say CD is "human hearing level" meaning it is also "concert hall level".

because some of the subtler overtones are lost.

Define subtler overtones? It is up to the recording engineer what is lost and what is captured.

If you hear music only or mostly on CD and have never been inside an acoustically splendid hall, then you really are not hearing the full bloom of the instruments.

It's not CD, it's the sound reproduction in your room. The sound waves aren't reflecting from your walls the same way they reflect inside a concert hall unless your living room is a concert hall.  Your setup tries to imitate the sound in a concert hall, but it is an imitation. Having higher sampling rates or bit depth don't change anything. If you want concert hall sound, go to concert hall. If you are happy with an imitation you can have it in your home. The better set up and acoustics etc. the better imitation.

And while there have been SACDs with greater capacity and supposedly improved sampling rate, the degree of improvement has been disputed.

Yes, because CD already is "human hearing level" audio. SACD gives multichannel sound. Generally higher sampling rates and bit depths are marketing gimmicks (in music production things are different).

This is unlike what's happened in home versions of film, where a company like Criterion originally issued DVDs that were sometimes quite poor, and slowly the company has been reissuing and remastering its catalogue to take advantage of the new BluRay format that can hold a far greater amount of digital information in the same space. Even so, the company can issue maybe 30 films a year.

That's because DVD wasn't "eye level". Even Blu-ray isn't, but at least it makes films look like films.

[71 dB]

It does depend on how good you system is, 10 years ago I could easily tell the difference between between a normal CD and one made @ mp3 standard 192, 256 even 320 at times, but with age I struggle to find a difference.

Well, lossy compression is lossy. Those mp3 files contain only information we are is "supposed to hear". CD have ALL the information, a lot of it we can't even hear.

[71 dB]

I generally agree.

There is an issue is that with 44.1kHz sampling. The maximum signal that can be represented is 22.05 kHz.

In theory. In practise 20000 Hz is the limit.

Some will claim to hear beyond 22 kHz, but very few can.

Most dogs and bats can. Humans can't. Even 20000 Hz is possible only for kids. If you are 50, you don't need frequencys above 17 kHz.

Another thing is hearing threshold. Over 100 dB is needed to hear a 20000 Hz tone to beging with. People should stop worrying frequences above 20000 Hz. WE DON'T HEAR THEM PERIOD. You are lucky to hear anything above 16 kHz.

The main issue is that any frequency component above 22.05 kHz gets "aliased" to a frequency below 22.05 kHz.

Yes, if not filtered, but ADCs have an anti-aliasing filter for that reason.

If there is signal at 40 kHz in the audio feed it gets aliased to 44.1 kHz - 40 kHz = 4.1 kHz by the digital recorder. That is a horrible form of distortion and the early recorders needed brick-wall filters to cut off any signal above 22 kHz. The filters themselves could distort the sound.

Yes if someone were that stupid to bypass the anti-aliasing filter. Nowadays they use higher sampling rates recorders and downsample afterworth for CD to minimize problems (avoid them).

Modern recorders sample very fast (192 kHz or faster) and the spurious frequencies can be suppressed by mathematical processing of the signal as it is converted to 44.1 kHz sampling, eliminating the need for the agressive analog filters.

Yes. That's why 44100Hz/16bit works so well now.

80's digital sound often gets criticized, and I think part of the problem is that the sound engineers of the day had developed their bag of tricks to compensate for the weakness of analog recording equipment, and this was not appropriate for digital recorders, which had different limitations.

It was the beginning of digital audio. You can't expect things to be perfect on day one. Now digital audio have many decades perfection time behind it.

[Bogey]

Going back to the "used" pricing for lps. 

For jazz/exotica/space age lps, I simply use EBAY (only the sold section) to get a bearing on what to pay for certain, and I mean certain lps in a shop.   Beyond that, I rarely pay more than $5 for each.  Once and a while I will pop for $10 and on a rare occasions I will pay full "market" price.  If the lp is marked at let's say $30, and the last three on EBAY have sold for $15, I may give the shop $20 for it if I REALLY want it because I will pay that with shipping anyways.  If they still think they have "gold", then I put it back in the bin.  If the lp is no where to be found on EBAY and I cannot find a it on Amazon (do not skip these sellers for used lps) I will sometimes buy it because it probably is a bit difficult to find.  I am also starting to inventory my lps on Discogs and they also have used prices that I am beginning to factor in before a purchase. 

Also, I have found by dealing with the same shop and one main lp dealer that does show here over and over again, they usually cut you a break even if you do not ask for it....especially if you bring a pile to a counter that has dust bunnies on it.

For classical, anything over $3 IMO is pushing it, unless you are a hardcore audiophile for this genre and have specific wants.  My shop has thousands(?) of classical lps and they seem to be more permanent than the Rockies to the west of the shop.

[71 dB]

I can quite clearly hear (and measure) the difference between 16bits/44kHz and the 24/96kHz which is at present the new standard. 

I'd say it is the way the DAC handles the formats rather than the amount of bits making any difference.

There is a new DXD format out which uses 32 bits.  The claim is you can really hear a note dying into silence with it.

32 bits means almost 200 dB of dynamic range (the difference between a quiet room and handgun is less than 150 dB). If the attack of a note is as loud as a gun near your ears, even 24 bits is enough to hear the note die out. Of course after such a loud note your hearing is damaged.

In music production every bit of extra dynamic gives extra headroom and makes things easier, but for consumer audio at home 16 bits is enough because it's dynamic range is the difference of the background noise in a quiet room (about 30 dB) and the pain threshold (about 120 dB). To put things in perspective, dynamic range of vinyl is about 60 dB ("10 bits").

Also, dither noise makes it possible to hear sound die out under the noise floor! The sound is heard dying out with the noise floor until the noise masks the sound completely. I have made tests* and I could hear sound at level about -110 dB "extending" the usable dynamic range about 10 to 20 dB. People just don't know these things because they don't know/understand enough about digital audio.

* I created a logarithmically decaying sinusoidal sound starting at level -90 dB using 24 bits.
Then I converted that sound to 16 bits using dither noise to avoid granulation (all CDs use dither).
Then I made the sound 50 dB louder in order to make it easy to hear the end result.
Yes, loud dither noise and sinusoidal sound decaying with it until the noise masked it.

Anyway, I can still live with the CD standard, but in the same way the industry went from DVD to blu ray, there will be an upgrade to a new format (PCM or DSD).  Its going to take a lot more time because the industry is in poor shape, and also because several generations have gotten used to compressed music, the degree zero of music quality.

CD pretty much fulfills the limits of human perception while DVD does not (unless one watches the picture far away - something not very "cinematic"). DVD is perhaps like having 16 kHz/12 bit audio. No wonder Blu-ray is a big improvement!

Consumer audio doesn't need higher sample rates or bit depths. 44100 Hz/16 bit done well is all we need.

[Jo498]

I have a 31-year old Schoenberg CD (London label, with Ashkenazy conducting Verklaerte Nacht ) which I won in a contest back then.  (I had to wait 2 years to play it, because we had no CD player!)

It still sounds quite fine on my system, which uses a 6-speaker BOSE Surround Sound unit.

If a CD has really become defective, it will in almost all cases not "sound bad" but it will either not play at all, skip or exhibit clicking noises. So it usually is not really a gradual thing (like with a scratchy LP) but a defective disc is very often useless. Of course there is some gradual process going on but one will only hear faults when the disc is basically "gone". And in large collections of several thousands CDs there are discs one only listens to once every ten years or so...
That's why the lifetime is a very relevant question for both older collections and used discs.

[Andante]

This is a thing that can be easily misunderstood. This is complicated, but I try to explain this. The sample rate 44100 Hz and dynamic depth of 16 bits are pretty much the minimum in order to have "compromised full quality" in late 70's/80's technology. Today digital technology is much more advanced and 44100 Hz/16 bit audio is not as compromised it used to be decades ago.

Reconstruction filter
DACs must reconstruct the original signal from the samples stored on the disc. By the way, theoretically this can be done perfectly within the limitations of dynamic range, something a lot of people don't understand. So, theoretically we can have the original signal plus noise at quantization level + dither noise in order to avoid granulation. The only requirement is that we don't have frequencies above 20000 Hz (theoretically half of the sampling frequency = nyquist frequency = 22050 Hz). Anyway, DACs must reconstruct the original signal from the samples and this involves reconstruction filtering. This filter must be VERY deep between the narrow frequency band 20000 - 22050 Hz in order to block frequencies above nyquist frequency and pass frequencies below 20000 Hz. Back in the 80's this was tricky. Oversampling makes things easier. Two times oversampling means our new sampling frequency is 88200 Hz and the new nyquist frequency is 44100 Hz meaning the reconstruction filter can fall down between 20000 Hz and 44100 Hz.

Back in the days it would have been easier to reach great sound quality with a bit higher sampling frequency so in that sense 44100 Hz was a compromise. Today, digital technology has developped so much we can use all the potential of the 44100 Hz/16 bit audio and luckily it is just good enough for human hearing. Higher sampling rates (96 kHz, 192 kHz...) and bit depths (24) just makes many think CD is "lacking". What CD is lacking is multichannel audio and that's it.

That's what I thought  no..... honestly you left me behind in the first sentence.   

[Parsifal]

That's what I thought  no..... honestly you left me behind in the first sentence.   

Try this. CD Data data effectively says.

At t = 0.000025 sec the signal is 2.3232 volts
at t = 0.000050 sec the signal is 2.5654 volts
at t = 0.000075 sec the signal is 2.6523 volts
at t = 0.000100 sec the signal is 2.7645 volts
at t = 0.000125 sec the signal is 2.8543 volts
...

In it simplest implementation the recorder measures the signal voltage ~40,000 times per second and the player reproduces the signal by outputting those same voltages, updating 40,000 times per second. The question is, what is the voltage between those samples? If the player was too literal the output signal would look like a staircase, jumping to the next output voltage 40,000 times per second. That is a very bad representation of the original waveform. What a decent CD player must do is generate a smooth waveform between the discrete sample points.

The original CD players generated the smooth waveform mostly with analog circuitry. The output was designed to have a sluggish response that could not make the sharp steps. Modern CD players make use of digital filtering and oversampling. They add extra samples between the real samples to make the waveform artificially smooth.

[71 dB]

That's what I thought  no..... honestly you left me behind in the first sentence.   

Thanks for your honesty!  0:)

It takes effort to get into digital audio. There's a lot of old myths concerning CD and some of them are even less true now than a few decade ago.

Try this. CD Data data effectively says.

At t = 0.000025 sec the signal is 2.3232 volts
at t = 0.000050 sec the signal is 2.5654 volts
at t = 0.000075 sec the signal is 2.6523 volts
at t = 0.000100 sec the signal is 2.7645 volts
at t = 0.000125 sec the signal is 2.8543 volts
...

This is a nice example. The more truthful representation would be:

t = 0.000000000 ms, y = 0001000100100011₂ = 4387
t = 0.022675737 ms, y = 0001000011001000₂ = 4296
t = 0.045351474 ms, y = 0001000001011111₂ = 4191
t = 0.068027211 ms, y = 0000111111100010₂ = 4066
t = 0.090702948 ms, y = 0000111110011010₂ = 3994
....

Here y is the sample value (-32767...+32768) and the green zeros and ones are the bits. Typically the analog output voltage of a CD-player for maximum sample value +32768 is about 2 volts.

In it simplest implementation the recorder measures the signal voltage ~40,000 times per second and the player reproduces the signal by outputting those same voltages, updating 40,000 times per second. The question is, what is the voltage between those samples?

This is one thing about digital audio that many have difficulties with. Digital audio deals with bandlimited signals, because the sampling theorem demands it. As crazy as it may sound the mathematical truth is we don't need to care about what happens between sample points. The signal will behave "properly", becauce it is bandlimited and we have samples taken frequently enough. It's like taking notes of the position of a train every second. The acceleration of a train is slow enough to ensure the train doesn't make "sudden moves" within one second.

If the player was too literal the output signal would look like a staircase, jumping to the next output voltage 40,000 times per second. That is a very bad representation of the original waveform. What a decent CD player must do is generate a smooth waveform between the discrete sample points.

The "staircase" is a mathematical concept and far from reality. Staircase would have infinite bandwith which is impossible in real world. It's very unfortunate many people have got a wrong impression about digital audio because of that staircase illustration. Even if a CD play produced a staircase signal, we would have pre amp, power amp, loudspeakers/headphones and our ears bandlimiting the sound extensively. In fact, the main reason for the use of a reconstruction filter is to avoid electrical interferences caused by high frequency electric radiation.

The original CD players generated the smooth waveform mostly with analog circuitry. The output was designed to have a sluggish response that could not make the sharp steps. Modern CD players make use of digital filtering and oversampling. They add extra samples between the real samples to make the waveform artificially smooth.

There is no need to do anything artificial. We can do things that are 100 % faithful to the original information. For example we can double the sample rate by putting zeros between every sample. This is called zero padding. Sample stream 4387, 4296, 4191, 4066, 3994, ... becomes
4387, 0, 4296, 0, 4191, 0, 4066, 0, 3994, 0,... Then we bandlimit the stream to 0-20 kHz and we have a smooth stream at sampling frequency 88.2 kHz. We added zeros meaning the information was not altered, only the sample rate was doubled. To preserve the original amplitude of the signal, we simply double the sample values before zero padding (8774, 8592, 8382, 8132, 7988,...) which means we have to go to higher bit depths (at least 17 bits) to avoid overflow. To make the sampling rate 4 times higher (176.4 kHz), we add 3 zeros between every original sample and so on...

The frequency response of a filter (analog or digital) is round depending on the order. The higher order filter the steeper the response can be. We can also make a filter steeper allowing some ripple, but for high quality audio ripple must be kept small. The steeper a filter is, the more it "rings", which is harmful for sound quality. Increasing the sample rate before reconstruction filter, we can use "relaxed" low order filters and have less ringing and ripple.

[Andante]

Thanks scarpia and 71db,   it still takes a bit of digesting, in the end I must admit I am satisfied with my CDs they give excellent reproduction through my old Hi Fi set up. I think a solid state system such as a super usb stick would be space saving and perhaps a longer life providing they could display the music details on a hand held remote