16 versus 24 bit files

[StanleyB]

Does anyone hear a difference between 16 and 24 bit audio files? I happen to be in the possession of the 16 and 24 bit 44.1KHz version of the Norah Jones album Come Away With Me. I have listened to the 16 bit CD for years so I would consider myself to know it like the back of my hands. But the 24 bit version has made me rethink some of my ideas of high res files. I got the same album in 96KHz and 192KHz and think I can hear a difference between them and the 44.1KHz CD. But funnily enough the 24/44.1 seems to produce more info and imaging.
I'll be interested in any supporting or counter observations.

[Andrei]

Trouble is most 24 bit files are 48, 96, or 192 resolution. The only 24 bit 44.1 khz files I have the Beatles and Mumford & Sons. Unfortunately I do not have the 16 bit versions to do a comparison.
However I do like the quality of those Beatles and Mumford & Sons Albums, and tentatively put down some measure of that quality to the better range offered by the extra bits.

[lovejoy]

Years ago I was on a training course learning about Pioneer's very nice new DAT recorder which was capable of 96/24 and we were played a set of songs, the same music, same mastering, but all at different bit-depths and sample rates. We were all in agreement at the time that moving to 24-bit over 16-bit made a clearly discernible difference in terms of depth and subtleties, while increases in sample rate were marginal if at all discernible.

[NRG]

In theory increasing the bit depth should only improve the SNR but I've just recorded an LP track at 44.1Khz in 16 and 24 bit and compared the Spectrogram plots in Audacity. At first glance they look the same (they are not perfectly time aligned) but looking more closely there are some differences in the plots! What the cause is I have no clue! Its big so just the link:

http://s94.photobucket.com/user/_NRG_/media/other/ScreenShot2013-08-29at111323.png.html

[StanleyB]

These listening results that I have experienced have had me wondering whether the differences heard between 16/44.1 and 24/192 are actually due to the difference between 16 and 24 bit, rather than the difference between 44.1 and 192.

[Audioman]

It stands to reason that greater bit depth will increase resolution. Also increasing sampling frequency ensures there is no perceived filtering of musical frequencies around the limit of human hearing. Also higher sampling frequency removes digital artifacts well out of audible range. Hence the lower noise floor. I'm sure ultrasonic frequencies in music have some effect on the presentation of sounds we can hear though don't think that has been proved. I would think 24/48khz should be a minimum and doubt there is any real audible difference above 96Khz.

For some reason unfortunately EMI on Beatles and other stuff at the time of remastering thought it easier to work at 44.1, obviously with downconverting to CD resolution in mind. If they did these today rather than 5 or more years ago I'm sure the 24/96khz standard would be used. Note with the Beatles they transferred from analogue to 24/192 but did most of the EQ at 24/44.1. These became the approved masters and thus the vinyl was cut from the 24/44.1 files.

[Rothchild]

It's interesting isn't it, I'll have to have a play myself, but in theory the sample rate should deliver the more marked tonal difference because the slope of the filter doesn't need to be as aggressive with higher sample rates (notwithstanding that I'm still not sure how much 96kHz I want in my music!).

Certainly for recording I go for 24bit, but this is more because of the headroom it gives me rather than pure sonics (and if you're careful about gainstaging, room acoustics and noise you can make it very difficult to tell the difference).

On a related point I'm a bit puzzled about where some of these hi rez track come from, there's woefully little information on websites like HDTracks about how, for instance, they generated a 24/96 version of The Smiths 'Meat is Murder'- if someone had made the effort to go back to the master tapes and re-track it you'd think there'd be a bit more bumpf about the tape machine used for playback and the quality of the convertors etc used for capture, I see none. I fear that they're just upsampled? (I expect that this is less likely for a contemporary record like Norah Jones as it was probably made in 24/96 in the first place)

[StanleyB]

This is some info I gathered from http://en.wikipedia.org/wiki/Sampling_rate


Because of the Nyquist-Shannon theorem, sampling rates higher than about 50 kHz to 60 kHz cannot supply more usable information for human listeners. Early professional audio equipment manufacturers chose sampling rates in the region of 50 kHz for this reason. 88.2 kHz and 96 kHz are often used in modern professional audio equipment, along with 44.1 kHz and 48 kHz. Higher rates such as 192 kHz are prone to ultrasonic artifacts causing audible intermodulation distortion, and inaccurate sampling caused by too much speed. The Audio Engineering Society recommends 48 kHz sample rate for most applications but gives recognition to 44.1 kHz for Compact Disc and other consumer uses, 32 kHz for transmission-related application and 96 kHz for higher bandwidth or relaxed anti-aliasing filtering.

[DSJR]

The difference between 16 bit and 24 bit is 96db dynamic range and over 120db I understand. The ear has around 50db range on a sliding scale in comparison, so if the music is mastered properly and the DAC isn't doing silly things with its filtering? then there 'shouldn't' be any difference I believe.

Can you be sure that the hi-res recordings haven't been remastered, or done with less limiting? The dynamic range on most commercial pop/jazz/rock recordings isn't much more than 40db anyway, apart from long fadeouts, so I advise caution lest it's something else causing what you hear...

By the way, I think most of us are convinced on the excellent quality of a good Radio 3 broadcast, even when the digital feed is warmed up in the analogue transmission. The BBC has used THIRTEEN BIT PCM for decades now, the sacrifice being the noise floor, rather than 'resolution' by my reckoning..

Just a few thoughts, since the record industry has no scruples where an extra revenue-source is concerned. If hi-res sells, then they'll pander to it and make sure the sonics you hear via this form are different and perceived 'better' than the vanilla CD issue!

[Audioman]

The difference between 16 bit and 24 bit is 96db dynamic range and over 120db I understand. The ear has around 50db range on a sliding scale in comparison, so if the music is mastered properly and the DAC isn't doing silly things with its filtering? then there 'shouldn't' be any difference I believe.

Can you be sure that the hi-res recordings haven't been remastered, or done with less limiting? The dynamic range on most commercial pop/jazz/rock recordings isn't much more than 40db anyway, apart from long fadeouts, so I advise caution lest it's something else causing what you hear...

By the way, I think most of us are convinced on the excellent quality of a good Radio 3 broadcast, even when the digital feed is warmed up in the analogue transmission. The BBC has used THIRTEEN BIT PCM for decades now, the sacrifice being the noise floor, rather than 'resolution' by my reckoning..

Just a few thoughts, since the record industry has no scruples where an extra revenue-source is concerned. If hi-res sells, then they'll pander to it and make sure the sonics you hear via this form are different and perceived 'better' than the vanilla CD issue!

I suggest trying one of the King Crimson CD/DVDA releases. Essentially the same mastering for CD and DVD. Now report back if you think that 16/44.1 and 24/96 sound the same. There is quite a difference using a dedicated fairly upmarket CD player/transport and a Universal player for DVDA that is probably well short of what is possible with hi-res audio.

[StanleyB]

There are not many sites that have a reasonable enough info about the bitrate and sampling frequency of the various album releases. The one I tend to refer to is avaxhome.ws . Ignore the info from the ones that people have ripped. The ones to consult for info are the official digital releases. I wasn't aware that so many old vinyl cuts can now be had in 24 bit digital files.

[Richard Kimber]

24 bit 44.1 and 48 are IMHO not worth anything over normal 16bit CD quality. It's only when the sampling rate increases to 88.2 and above that I think you start to get a worthwhile improvement. I don't normally buy anything less than 88.2, except normal 16bit CDs.

- Richard.

[Stratmangler]

The problem with many of these threads about 24/192 stuff sounding better than 16/44.1 is that the files used are not from the same master.
All observations made are therefore invalid, and no conclusion can be sensibly arrived at.
Such comparisons are like comparing apples with oranges.

Try downsampling a 24/192 file (or 24/96 - whatever) down to 16/44.1, and then start comparing the two - it isn't difficult to do - you'll find that there's a lot less difference in sound quality between the two files.
And you'll at least be comparing apples with apples and oranges with oranges.

[WOStantonCS100]

The problem with many of these threads about 24/192 stuff sounding better than 16/44.1 is that the files used are not from the same master.
All observations made are therefore invalid, and no conclusion can be sensibly arrived at.
Such comparisons are like comparing apples with oranges.

For a "real world" comparison of resolutions, one needs to obtain/borrow a high res recorder with premium media (master tape dub ala Tape Project or mint condition 60s to 70s vinyl, for instance) and make your own copy to 16/44.1 and 24/192. Having done this many times, I can certainly say the comparisons, observations and conclusions are very telling. 24/192 is the closest I've heard to the analog input signal, distancing 16/44.1 by a large and noticeable margin, 24/44.1 slightly (but definitely not worth converting all my CDs / I'd rather use an upsampling DAC). I would imagine 5.6MHz DSD to be just as good, possibly better.


Try downsampling a 24/192 file (or 24/96 - whatever) down to 16/44.1, and then start comparing the two - it isn't difficult to do - you'll find that there's a lot less difference in sound quality between the two files.
And you'll at least be comparing apples with apples and oranges with oranges.

That doesn't make much sense to me. Downsampling high res removes data from the file and negates the gains, right off. I don't see the point in making a comparison after that; akin to dubbing a master tape to worn 8-track. What's the point? :scratch:

[Stratmangler]

For a "real world" comparison of resolutions, one needs to obtain/borrow a high res recorder with premium media (master tape dub ala Tape Project or mint condition 60s to 70s vinyl, for instance) and make your own copy to 16/44.1 and 24/192. Having done this many times, I can certainly say the comparisons, observations and conclusions are very telling. 24/192 is the closest I've heard to the analog input signal, distancing 16/44.1 by a large and noticeable margin, 24/44.1 slightly (but definitely not worth converting all my CDs / I'd rather use an upsampling DAC). I would imagine 5.6MHz DSD to be just as good, possibly better.





That doesn't make much sense to me. Downsampling high res removes data from the file and negates the gains, right off. I don't see the point in making a comparison after that; akin to dubbing a master tape to worn 8-track. What's the point? :scratch:

That's all well and good, but you don't think that a record company goes to the effort of recording at different resolutions for different formats, do you?
They don't, they have the studios number crunch things down.

As for upsampling, how can you add data that isn't in existance at the start point.
All that upsampling can do at best is add loads of redundant zeros.
The additional bits might fool the logic cicuit of a DAC into using a different output filter, but nothing more.

[StanleyB]

We seem to be drifting off a bit here. My question was with regards to 16 versus 24 bits. I have intentionally left out comparing sampling rates. Otherwise we have two variables to deal with, namely bitrate and sampling rate. It's difficult enough as it is for many to distinguish between the two. Putting both of them into the mix so to speak only helps yo muddle the waters.

What I have been able to hear is a larger dynamic impact with 24bit. I can also hear more clarity in low level signals. The most appropriate test for this is live sessions with people talking in the background. Murmurs in 16bits tend to turn to understandable speech with 24bits.

[lurcher]

Yep, I have found very similar things Stan. When I was messing around with passing signals through programmable logic on the way to DAC's it was easy to change the bit depth of the data on the way through. it was easy to hear the difference (with my kit in informal listening conditions) to at least 18/19 bits, beyond that I was guessing. Should be simple enough to take 24bit WAV files and filter them into 16,17,18 and so on bit versions without altering the sample rate. I have some code somewhere that I can stream a wav file through that would let that happen.Anyone got a 24 bit WAV file we could use as a test?

Though we would have to decide if we wanted to add dither that matched the bit depth.

[AlexM]

I'd agree that 24 bit files do seem to reveal better low-level detail, and I notice ambience cues and reverberation more distinctly. Not a HUGE difference, but noticable,

Nick/Stan, that is an interesting idea - you would ideally have different audio signals such as speech combined at differnent levels,say -40db, -60db, which are then truncated to different bit depths and then padded into a 24bit format to see how integlligability is affected by reducing the bit depth. Is this easily achievable?.

A controlled test of this sort would help weed out the influence of the mastering/production chain.

In terms of what I have noticed, I personally think you are into rapidly diminishing gains beyond 24/48. I have done some testing with downsampling with dither via DBPoweramp DSP, resampling a 24/192 recording from 2L to 24/96, 24/48, 24/44.1 and finally 16/44.1. No ABX or anything, but I couldn't tell the differnece between 24/192 and 24/96, 24/48 still sounded practically the same, 24/44.1 was a little worse and 16/44 was noticably different. Not a really big differnece between any of the 24 bit format.

Regards,
Alex

[realysm42]

Kind of on topic, I bought the Keith Greeninger DSD recording from Blue Coast recordings yesterday.

The first thing I noted was how well recorded it was; it was very "quiet", which actually translates into dynamic. Theres some music I wouldn't turn up past 20%, this comfortably gets up to 70%.

As for other SQ observations, it's very immediate and full, yet completely unforced in terms of presentation. Full of nuance.

I don't have any 24/192 music though, will explore more in future.

I do agree with Chris (Strat) there's an awful lot of apples getting compared to oranges here.

[Richard Kimber]

The problem with many of these threads about 24/192 stuff sounding better than 16/44.1 is that the files used are not from the same master.
All observations made are therefore invalid, and no conclusion can be sensibly arrived at.
Such comparisons are like comparing apples with oranges.



Well, not entirely. If you listen to enough hi-res recordings and enough CDs it's possible to form a judgement about whether there is a tendency for hi-res to sound better.

- Richard.

[Rothchild]

I want to try this, can someone direct me to a recording that was dead-cert made at 24 bit and where the same master was used to generate the cd and the hi-rez version?

Alternately, I have access to a reasonable recording facility does someone want to come up wiht a sensible experiment whereby I can capture the same recording / performance at both bit depths and I will happily post the results - I can't for now work out how to get my studio machine to record both 16 and 24 bit files at the same time....

[lurcher]

Nick/Stan, that is an interesting idea - you would ideally have different audio signals such as speech combined at differnent levels,say -40db, -60db, which are then truncated to different bit depths and then padded into a 24bit format to see how integlligability is affected by reducing the bit depth. Is this easily achievable?.

Don't see why not, though you could get a -40dB signal by taking a CD and adding 8 bits of zero (sign extended) to the MSB.

[User211]

In theory increasing the bit depth should only improve the SNR but I've just recorded an LP track at 44.1Khz in 16 and 24 bit and compared the Spectrogram plots in Audacity. At first glance they look the same (they are not perfectly time aligned) but looking more closely there are some differences in the plots! What the cause is I have no clue! Its big so just the link:

http://s94.photobucket.com/user/_NRG_/media/other/ScreenShot2013-08-29at111323.png.html

Try recording the same track again at either sampling rate. Does it look the same as the original? That'll tell you whether it is just variance in the analogue domain - or not. If not, then interesting:)

[NRG]

I understand where you are coming from and will try it but I cant think what would change so significantly in the analogue domain in such a short time. The Phono stage had warmed up, the deck was not stopped in between, the loading was the same and the tracks recorded within a minute of each other...

[User211]

I'd like to see the result Neal. But now things really have changed I guess - ambient temperature, warm up time etc.

So just do two at 16 or 24 bit one immediately after the other?

[Richard Kimber]

It seems to me that it depends what you are recording. I'm not convinced you'd notice much difference if it's just someone playing a guitar. On the other hand, if it's a large symphony orchestra playing Wagner, then I think you would.

- Richard

[Stratmangler]

Well, not entirely. If you listen to enough hi-res recordings and enough CDs it's possible to form a judgement about whether there is a tendency for hi-res to sound better.

- Richard.

It's also possible to have CD seem to have greater detail by making the overall level higher, thus making low level sounds become more apparent.
This is not the same as what's been taking place in the loudness wars, where dynamic limiting and compression have been misused.

The thing about comparing high resolution files to CD spec files is that often high resolution transfer is done some years after the CD release, so it isn't mastered from the same production transfer.
In between times the ADCs in the studio may have changed for something that is much better resolving.

[WOStantonCS100]

That's all well and good, but you don't think that a record company goes to the effort of recording at different resolutions for different formats, do you?
They don't, they have the studios number crunch things down.

Precisely my point. In order to get a real idea of how higher bit depth and/or sampling frequencies benefit what is heard, you have to sidestep the record companies most of the time. But, that doesn't mean that the technology doesn't have worth or doesn't do what it is designed to do. No need to throw the baby out with the bath water, so to speak. There are and will be companies (the majors will have to get the "ok" from the bean counters first) that will avail high res, properly.


As for upsampling, how can you add data that isn't in existance at the start point.
All that upsampling can do at best is add loads of redundant zeros.
The additional bits might fool the logic cicuit of a DAC into using a different output filter, but nothing more.

Again, that was my point. If the source is a 16/44.1 file, be it on a hdd or a cd-da, then it doesn't really matter if the increase in bit depth is done with software or hardware. So, in my opinion, do it on the fly and keep your CDs or your original file (save space and time) and let competent hardware deal with it.

[WOStantonCS100]

We seem to be drifting off a bit here. My question was with regards to 16 versus 24 bits. I have intentionally left out comparing sampling rates. Otherwise we have two variables to deal with, namely bitrate and sampling rate. It's difficult enough as it is for many to distinguish between the two. Putting both of them into the mix so to speak only helps yo muddle the waters.

What I have been able to hear is a larger dynamic impact with 24bit. I can also hear more clarity in low level signals. The most appropriate test for this is live sessions with people talking in the background. Murmurs in 16bits tend to turn to understandable speech with 24bits.

That infamous AoS thread drift...

Yes, I'd agree. Anytime one can increase the S/N ration, theoretically, it should render low level information more articulately, increase dynamic range. Wasn't that the purpose of going to 24 bit?

The thing is, aside from a computer program, I know of few DACS that only up the bit rate. Maybe I'm just ignorant on this (a distinct possibility). My experience is that most DACS that will increase the bit depth also want to up the sampling frequency at the same time. Ironically, my DV-RA1000HD will NOT copy a CD as 16/44.1 It extracts audio from a CD at 24/44.1 For all practical purposes it is a "24 bit" machine that is backward compatible so as to read a 16/44.1 disc. Obviously Tascam don't think 16 bit is a worthy format for a master recorder.

[Reffc]

The theoretical argument for 24 bit is fine and the technology widespread these days. In practice however, it's not a huge audible margin between the two if we're being honest. The main reason being that there's far greater divergence between recording (mastering) qualities of music than there is significant audible differences between the bit rates. As there is still no de-facto industry standard to address this, then the argument imho becomes somewhat pointless to a lesser extent. Just buy and enjoy music in either format. A well recorded 16 bit piece of music still offers a thoeretical 90dB dynamic range which is above human hearing perception anyway. Sampling rates at 44.1 are also adequate if properly implemented and it would be interesting to see how many people could pick out 88 or higher in blind listening tests. I have heard some superbly recorded high res files that would knock spots off most of my 16 bit CD recordings, but equally, I have some CD recordings that are impossible to distinguish between their hi-res counterparts. I did this comparison using the Astell and Kern A100 hi-res player loaded with hi-res versions of some of my 16 bit albums and tracks.

[StanleyB]

The theoretical argument for 24 bit is fine and the technology widespread these days. In practice however, it's not a huge audible margin between the two if we're being honest.
I used to be in that camp for sceptics as well for many years. But I have had to revisit my measuring and listening test in the last year or so in order to track down the reason for some things I had not paid attention to before. In my case it happened to be a couple of albums from Marvin Gaye that first caught my attention.

[Reffc]

I used to be in that camp for sceptics as well for many years. But I have had to revisit my measuring and listening test in the last year or so in order to track down the reason for some things I had not paid attention to before. In my case it happened to be a couple of albums from Marvin Gaye that first caught my attention.

I wouldn't really place myself in either camp nor do I really consider myself a sceptic. I accept the advances that are continually made in technology, but from my own perspective (listening tests), I never considered that the differences between well recorded versions of the same album were vast. They exist, but not in any night and day sort of sense. In the case where the same master was used, there's no discernible audible difference at all (unless we're fooling ourselves by suggestion which is a very common problem in hifi evaluation) but then why should there be? More to the point, why produce a 24/96 version of an original 16 bit master...upsampling as has been noted, doesn't give anything over the original file.

[Richard Kimber]

We seem to be drifting off a bit here. My question was with regards to 16 versus 24 bits. I have intentionally left out comparing sampling rates. Otherwise we have two variables to deal with, namely bitrate and sampling rate. It's difficult enough as it is for many to distinguish between the two. Putting both of them into the mix so to speak only helps yo muddle the waters.



But you can't sensibly separate them when considering sound quality. 44100 fits fine into 16 bits. 24/44100 won't sound any better, it's just a 'waste' of bits. You can't accurately represent a 192 sampled recording in 16 bits. So sample rate and not bit depth is the key to sound quality. Bits are just what you use to represent the audio at whatever sample rate you're using.

- Richard.

[StanleyB]

But you can't sensibly separate them when considering sound quality. 44100 fits fine into 16 bits. 24/44100 won't sound any better, it's just a 'waste' of bits. You can't accurately represent a 192 sampled recording in 16 bits. So sample rate and not bit depth is the key to sound quality. Bits are just what you use to represent the audio at whatever sample rate you're using.
Not so I am afraid. When the 1st CD players first came out, some of them were 14 bit machines. At the beginning of the 90's the bit rate conversion had reached 20bits. Let's remind ourselves that these machines were playing 16bit/44.1KHz dics at a higher bitrate, not a higher sampling rate.

[lurcher]

But you can't sensibly separate them when considering sound quality. 44100 fits fine into 16 bits. 24/44100 won't sound any better, it's just a 'waste' of bits. You can't accurately represent a 192 sampled recording in 16 bits. So sample rate and not bit depth is the key to sound quality. Bits are just what you use to represent the audio at whatever sample rate you're using.

- Richard.

Sorry, I just don’t understand what you are trying to say here. Bit size controls the resolution and so the available noise floor/dynamic range. Sample rate controls the highest frequency that can be recorded/reproduced. Both are important, but unconnected.

[WOStantonCS100]

...More to the point, why produce a 24/96 version of an original 16 bit master...upsampling as has been noted, doesn't give anything over the original file.

I agree. Might as well let a well-designed DAC do it as part of it's filtering scheme.

At the same time, I would certainly consider, and think it worth it, to obtain a 24/88.2 or higher res file straight from an analog master or a digital master of that same higher resolution.

[WOStantonCS100]

I used to be in that camp for sceptics as well for many years. But I have had to revisit my measuring and listening test in the last year or so in order to track down the reason for some things I had not paid attention to before. In my case it happened to be a couple of albums from Marvin Gaye that first caught my attention.

:)

[Andrei]

I want to drive from London to Glasgow in as quick as time possible. Will a car with a four liter engine be better than a car with a 3 liter engine?

But what about the tyres? are they the same?
No that's irrelevant; it is the skill of the driver.
Nah you are missing the point - it the design of the vehicle - low center of gravity will get you there faster - you can corner faster.
Wrong dude it is torque and turbocharging. You will get held up behind those trucks unless you can zip past them.
Makes no difference cause we have to keep to the speed limit.

Hi-fi is a subject with multiple variables. The op asked one variable.

[WOStantonCS100]

Hi-fi is a subject with multiple variables. The op asked one variable.

Is the red ball heavier than the green ball?

The op asked one question which will have varying answers based upon a multitude of possible criteria. :)

[drrd]

With very good AD and DA conversion I think not only is 16 bit a completely adequate final file resolution but 44.1K is also completely adequate bandwidth. I'll record at 24 bit to allow for processing files but once that's finished I hear no difference when cutting off the 8 LSBs. It makes sense, if 0dB is as loud as you want it then what's happening at the LSB with 16 bit is too quiet to be audible. Add another 8 bits below this and even if it's meaningful data, which it won't be down to the LSB even with SOTA gear, it's still inaudible. I've got plenty of storage and AD and DA gear that handles everything up to 192K and DSD, I just can't find a benefit to bandwidth or resolution that exceeds what I can hear.

[StanleyB]

With very good AD and DA conversion I think not only is 16 bit a completely adequate final file resolution but 44.1K is also completely adequate bandwidth. I'll record at 24 bit to allow for processing files but once that's finished I hear no difference when cutting off the 8 LSBs.
Not all of us have the same hearing sensitivity and frequency response. Many people have a far larger ability to hear very minute audio details that others cannot detect. So cutting off the 8 least significant bits arbitrarily is not the way to go.
One reason why oversampling of the CD data was first introduced (in the Marantz CD45 14Bit 4X oversampling CD player) was to attempt to recreate the missing bits that had been discarded at the A to D stage. But processing power at that time was not as powerful as it is today. So they had to use 14Bits instead of the full 16 bits of the CD format.

[drrd]

Oversampling doesn't recreate missing bits. Oversampling gives increased bandwidth so that noise-shaping can be performed. Noise-shaping increases the resolution of the converter but biased to the lower frequency range, at higher frequencies noise-shaping worsens the converter performance. My converters run at 2.82 MHz and are 5 bit, with noise-shaping they can give up to 24 bit resolution in the audible frequency range but increasing quantisation error with frequency will make the higher end far worse than that, hopefully though that's outside hearing range. A 14 bit converter will have better resolution if it's oversampled 4x and noise-shaped to push quantisation errors out to where they aren't so audible but it doesn't do any magic recreation of data.

If you have true 16 bit resolution for the whole range of 20Hz to 20KHz I would challenge anyone to listen to full-scale music and be able to hear extra bits. We are talking below -96dB at this point. You know 12 bit and even 8 bit digital sounds great if the music obscures the noise floor, they just have to be good bits for the whole audible frequency range, it's errors in AD quantisation and back again to analogue that the ear picks up on as sounding bad (see turntables as a good example of excellent linearity and poor S/N).

[StanleyB]

Oversampling doesn't recreate missing bits. Oversampling gives increased bandwidth so that noise-shaping can be performed.
That's what it is used for now. But in that period of the 80's designers were experimenting with all sorts of techniques to try to "recover" discarded data from the A to D process, or from lack of processing power to extract the full 16 bit format from a CD. Oversampling was tried out with three reasons in mind. One was to simulate missing bits, one was to raise the then very sharp filtering frequency, and the third one escapes me. Its effectiveness in recreating extra bits was of zero value, so that intention was abandoned.
Quite a few CD players were only 14 bit machines. One was even 12 bit, but I can't remember any more who released it.

[drrd]

I'm sure there was a little marketing mumbo-jumbo back in the day (eg. Pioneer Legato Link) and maybe ideas were being tried without an exact understanding of why they did or didn't work. Havent heard any suggestion that the Phillips 14 DAC or even later 1-bit 256xOS designs were doing anything radically different to modern noise-shaped Delta-Sigma. These days the techniques seem to be fairly clear, at least the basic stuff that converter makers are prepared to divulge. The aim is better resolution and better linearity but not pretending to make a silk purse from a sow's ear.

[StanleyB]

I'm sure there was a little marketing mumbo-jumbo back in the day (eg. Pioneer Legato Link) and maybe ideas were being tried without an exact understanding of why they did or didn't work.
I was one of the first engineers drafted in at the time (1983) from the repair industry to start testing and help the design engineers to come up with next generation solutions for CD playback. Philips and Sony were the first companies that needed my services, but others followed once I got known. This was well before any finished product was presented to any marketing team, and in many cases some machines and functions never made it to market. So what I am writing here is from an insider who was involved in the design of early CD equipment.

Some of the features I miss because they were not implemented or withdrawn after implementation are:
The track and CD album name on the display.
Memorizing of the tracks that were most frequently played on a CD, with an option to then play your favourite selection or the whole CD.
Selectable bitrate playback. This was to allow for suppression of low level noise on old analogue recordings that were converted for CD playback.

[drrd]

That must have been interesting work, comparing what's known now with then though do you think it's reasonable that converters could possibly just make up valid data (that is audible) or do we now just have better resolving and especially more linear converters? I occasionally bought some slightly duff gear over the last 30 years or so but there have been some stand-out good ones. Still the trend with time has been that the sound quality has just got better, the CDs from 20-30 years ago now sound great and well produced red-book sounds difficult to distinguish from hi-res. Coming back to my original comment though I'd stick with well produced 16 bit 44.1 as a final format as being just about spot on.

[Yomanze]

The first 16 bits are infinitesimally more important than the 16 vs. 24 bit, so it is more important to focus on the DAC type IMO. I would take 16 bits from a well-implemented multibit DAC than a delta/sigma, oversampling, upsampling DAC etc. anyday. In reality too, no analogue system can physically play anything near 24 bit, the noise floor of real life limits the actual resolution to a maximum of below 21 bits. Where 24 bit or 32 bit, maybe even 64 bit is very useful is in the studio due to the additional headroom during digital editing. EDIT: I should add that if you're listening at over 100db volumes in the listening position, aside from going deaf, you'd might then gain an advantage with the extra dynamic range beyond 16bits...

A little tin foil hat wearing person in me would think that companies pushing higher resolution may be inclined to work on these files and 'improve' them, or make them sound different...