Why is digital audio so complicated? Where did it all go wrong?

[Hifi_swlon]

So, I've been thinking about this quite a bit over the past months, following topics all over the place on various forums etc - including here obviously -  as well as from my own experimentation.

Lets just get this out there first - I'm not sure how factually correct any of this, it just one persons opinion - but as far as I can tell, CD was invented about 30 years ago(?) and it bought digital audio to the masses, but many still argue to this day that it takes an ultra-expensive player to make CD sound 'good', and that's still miles from being as good as vinyl, or at least from sounding 'realistic' or 'pleasing' (read natural I suppose).  Computer based audio followed, and it seems like only an uber expensive computer based setup can sound as good as an uber expensive CD player, and in most cases the sounds is still only just believable as being 'real' and 'natural' in many cases, although some say high-end digital is now equal or better than vinyl, so it seems it can be done.

In my mind, the theory's simple - digital data is sent from A and received at B, where it is converted to analogue for output.  If dropouts aren't occurring, then we assume the data arrived unharmed, but that's not quite good enough apparently.  (OK, there are a lot of things in the chain like DACs, amps and speakers but lets ignore those for now and assume they're perfect.).  The bit I'm talking about is this getting the digital data from source to destination, and the related problems that are bounded around.  Jitter, EM interference, Signal Integrity, clock bending, power supply noise, ground noise, galvanic isolation, atomic clocking, digital cable boundary effects, cable impedance, reflections, the list goes on I'm sure - but all seem to contribute in some way towards making this specific audio version of digital, somehow prone to lots of issues, and there is no length you can't go to extract a bit more just by making this digital data 'better'.

Now, we've been to the moon, and are planning to go to Mars, we're starting to get a much deeper understanding of the universe and how it was formed, we've made radical advances in medicine and science - many of us carry computers in our pockets that are more powerful than the early supercomputers.  There's a whole swathe of other things too many to mention here obviously.  (OK so we're not so smart because we've also f'ed up a lot of the planet and caused untold suffering at the same time…but technology-wise we seem to know what we're doing).

So what went wrong with digital audio?  Why has something that in this day and age should be so easy, turned out to be so complicated?  Are the problems all real, or  part real, or is the actual problem just not really understood? Or are there a lot of people taking advantage of the fact that its hard for humans to make decisions about what they hear so are deliberately misinforming or making it more complicated than it needs to be? Or a combination of?

I've done a fair bit of tweaking - albeit nothing compared to some - and think I'm open-minded, but often at the end of it I start to question myself, and wonder whether those that say it's just the brain playing tricks might be right (at least in part).

Be keen to hear what others think….

[NickB]

So, I've been thinking about this quite a bit over the past months, following topics all over the place on various forums etc - including here obviously -  as well as from my own experimentation.

Lets just get this out there first - I'm not sure how factually correct any of this, it just one persons opinion - but as far as I can tell, CD was invented about 30 years ago(?) and it bought digital audio to the masses, but many still argue to this day that it takes an ultra-expensive player to make CD sound 'good', and that's still miles from being as good as vinyl, or at least from sounding 'realistic' or 'pleasing' (read natural I suppose).  Computer based audio followed, and it seems like only an uber expensive computer based setup can sound as good as an uber expensive CD player, and in most cases the sounds is still only just believable as being 'real' and 'natural' in many cases, although some say high-end digital is now equal or better than vinyl, so it seems it can be done.

In my mind, the theory's simple - digital data is sent from A and received at B, where it is converted to analogue for output.  If dropouts aren't occurring, then we assume the data arrived unharmed, but that's not quite good enough apparently.  (OK, there are a lot of things in the chain like DACs, amps and speakers but lets ignore those for now and assume they're perfect.).  The bit I'm talking about is this getting the digital data from source to destination, and the related problems that are bounded around.  Jitter, EM interference, Signal Integrity, clock bending, power supply noise, ground noise, galvanic isolation, atomic clocking, digital cable boundary effects, cable impedance, reflections, the list goes on I'm sure - but all seem to contribute in some way towards making this specific audio version of digital, somehow prone to lots of issues, and there is no length you can't go to extract a bit more just by making this digital data 'better'.

Now, we've been to the moon, and are planning to go to Mars, we're starting to get a much deeper understanding of the universe and how it was formed, we've made radical advances in medicine and science - many of us carry computers in our pockets that are more powerful than the early supercomputers.  There's a whole swathe of other things too many to mention here obviously.  (OK so we're not so smart because we've also f'ed up a lot of the planet and caused untold suffering at the same time…but technology-wise we seem to know what we're doing).

So what went wrong with digital audio?  Why has something that in this day and age should be so easy, turned out to be so complicated?  Are the problems all real, or  part real, or is the actual problem just not really understood? Or are there a lot of people taking advantage of the fact that its hard for humans to make decisions about what they hear so are deliberately misinforming or making it more complicated than it needs to be? Or a combination of?

I've done a fair bit of tweaking - albeit nothing compared to some - and think I'm open-minded, but often at the end of it I start to question myself, and wonder whether those that say it's just the brain playing tricks might be right (at least in part).

Be keen to hear what others think….

Like most things in life it only gets complicated when someone can make lots of money by confusing you. A bit like cables, power supplies, different file types, DVDa, DSD etc etc.

Set up a Raspberry Pi with Hifiberry+ run coax or optical in to Devialet connect to it with Roon and enjoy.

Don't listen to me though I am just an old cynic 

[SwissBear]

Be keen to hear what others think….

Three ideas, which could help each of us reach an appropriate answer to your interesting question:
- in the Audiophile hobby, we are not aiming at any improvement which has any objective value, but we are in a search for aesthetic perfection ; aesthetics has no limit whatsoever...
- the search for aesthetic perfection in 'Audiophilie' is a recurring process : the better you listen, and the better you discern what you expect, the more you expect from your system...
- in human activities which are in relation to arts, there is no limitation of financial means ; there will always be people who have enough money to ask for more, be it in paying for the best painters or the best architects to give pride as it happened during renaissance, or to build the finest cars, or the finest audio systems, for each and everyone of us feel happy with the what we can buy.

So it does not have anything to do like researching the best computer to break-out the human genome and help find treatment for 'orphan' diseases. It's just researching the most appropriate way to restitute sounds, to a variety of people who do not share the same expectations, the same aesthetic quest, and who do not have the same amounts of money to spend...

So, coming back on earth, we have the right not to be fooled by people who use our quest for audiophile perfection to ruin us selling us overpriced equipment. But we have the duty to respect each of us' desire to feel unique having found the system which suits the best his/her audiophile quest...

[f1eng]

It depends on whether you are judging whether something is an accurate reproduction of the original or just whether it sounds nice to you (which is probably the only thing that really matters)
If you were not the original recording engineer you can not know whether the reproduction is accurate, since even if you are present at the performance as it is being recorded your ears are not where the microphones were positioned, and that makes a BIG difference.

I have been making recordings since the mid 1960s and none of the analogue recorders I used produced an output indistinguishable from the microphone feed.
Every digital recorder I have used was closer to the microphone feed than even the very best hugely expensive analogue recorders.

I worked on record players in the mid 1970s. The lowest distortion pickup cartridge I saw was 2% distortion, and was mono. High quality stereo pickup cartridges often have well over 5% distortion in the mid band and way more at high frequencies. Record players also pick up airborne and structure borne vibration which they add to the cartridge signal. Luckily, the distortion is euphonic and the pickup is a bit like added reverb, which sounds nice.

So, as somebody who has been frequently in a position to compare the microphone feed with the output of the recorder, and worked in the LP business at its height, I can assure you that digital recordings are potentially audibly completely accurate as long as the correct antialiasing filter is used and the ADC is never driven to clipping.
No analogue system is, but luckily (and it is luck not engineering) the shortcomings of analogue recorders, the LP manufacturing process and record players are almost all euphonic and nice sounding.

So good LPs sound nice but they are certainly not and accurate reproduction of a recording and never could be.

If a digital recording doesn't sound nice either the recording was done wrong (ie clipping or no anti aliasing filter) or the microphone choice, or its position (both have a huge effect on the sound) is not to the listeners taste because what comes out of an properly engineered DAC will nowadays be audibly indistinguishable from what went into the original ADC.

Now if a bit of added colour is what somebody likes then LP is a good choice, though the two non-euphonic shortcomings of LPs, speed variation and noise, annoy me a  lot.
To get coloured sound from digital is more difficult. Non-linear DACs, leaving out the re-construction filter and eccentrically engineered analogue stages can do it though

CD standard digital is not quite capable of recording the whole human loudness range from the threshold of hearing to the threshold of pain, but in reality, a quiet domestic room is 30dB, not 0dB so in a domestic situation CD is capable of reproducing everything a human can hear, from the deepest bass to the highest treble, and from a sound just not quite loud enough to hear in a domestic room to the threshold of pain. More than tho other parts of most domestic hifi systems, in fact.

If a high res recording sounds nicer it will almost certainly be because it is a better recording, not because of anything added by "high resolution"-ness.
High res is useful for recording headroom and unexpected peaks but not necessary for playback, although some re-sampling calculations are not audibly transparent so there is a risk in making the conversion.

For me the way digital has "gone wrong" is in its latest guise as streamed files.
Unlike CD and LP there are multiple standards, rather than one, mp3, aac (mp4), flac, alac, wav and so on. Track labelling has not been standardised in any versatile way either IME so I find it irritatingly hopeless for classical music.

Another thing The quality of recordings varies massively. I have fantastic sounding CDs of all ages and terrible ones. I have fabulous sounding LPs and some dire ones too.
IME the difference in sound quality between mediums (and frequently hifi equipment) is less in magnitude than one often encounters between recordings.

So, in summary, after 50 years experience recording, designing record players and being a hifi enthusiast I know that CD can produce an audibly transparent reproduction and LPs can not but that does not mean the recording itself sounds nice.

Nowadays I just listen to music and don't bother much about equipment or medium, CD, LP or streamed file. I sometimes even listen to original master tapes, reel-to-reel, cassette and DAT.

Record players are good for enthusiasts who like playing around with equipment as much or more than listening to music, and I know a few of them...

[Borgen]

This reminds me of a blind test report I read from a UK hifi magazines on amplifiers some 30 years ago. They conducted a blind test of around 8-10 amplifiers towards an audience and there was one amp that won by quite some margin. Interestingly, they showed the measured frequency response of the amplifiers as part of the report. Most amplifiers had relatively flat respons over the 10Hz-20kHz band except one, which basically was ok in the 1kHz-10kHz band and was swinging a lot outside this band (yes it was the one the test group preferred). It seems to me that people like a bit of imperfection and I guess this is what the LP gives you.

...or is it that the digital audio with DAC sampling gives a lot of high frequency noise that you can avoid with a LP playing into an old fashioned pre-amp/ power-amp combo??

[f1eng]

This reminds me of a blind test report I read from a UK hifi magazines on amplifiers some 30 years ago. They conducted a blind test of around 8-10 amplifiers towards an audience and there was one amp that won by quite some margin. Interestingly, they showed the measured frequency response of the amplifiers as part of the report. Most amplifiers had relatively flat respons over the 10Hz-20kHz band except one, which basically was ok in the 1kHz-10kHz band and was swinging a lot outside this band (yes it was the one the test group preferred). It seems to me that people like a bit of imperfection and I guess this is what the LP gives you.

...or is it that the digital audio with DAC sampling gives a lot of high frequency noise that you can avoid with a LP playing into an old fashioned pre-amp/ power-amp combo??

Digital gives measurably less high frequency noise than LP. It is the case that LP is incapable of recording treble levels as high as CD (if they are on the recording) because of cutter limitations. Tape recorders are the same. One can get a flat frequency response at -20dB but nowhere near in the treble at 0dB so it is certainly the case that LP is incapable of being as loud in the top octave as CD, but frankly very few recordings have very high levels at very high frequency so this limit is rarely a problem on real music.

[Krisp]

It depends on whether you are judging whether something is an accurate reproduction of the original or just whether it sounds nice to you (which is probably the only thing that really matters)
If you were not the original recording engineer you can not know whether the reproduction is accurate, since even if you are present at the performance as it is being recorded your ears are not where the microphones were positioned, and that makes a BIG difference.
...

100 % acknowledge with everything you wrote.

What else went very wrong is the mastering. The loudness war ruins a lot of music esp. pop and rock music. I would vote for a floating point type of encoding. Here the compression rate could for instance be defined in the playback device (with high compression in car stereo, portable player or smartphone) and the max. resolution would be available for audiophile playback.



Cheers,

Krisp

[Silvertouran]

I remember The Gadgetshow did a blind test along the lines of CD/Vinyl/MP3. And yes you've guessed it. They preferred the mp3. Personally I think if it sounds "nice" then thats all that really matters. A lot of people get hung up on values and figures but that's only half the picture. Over the last 30 years i have come to the conclusion that the biggest influence to the sound isnt all the Techy stuff ,but that speakers make the biggest difference to the sound ,including positioning of course.

[Hifi_swlon]

Very interesting thoughts so far…..

I was actually thinking more specifically about the digital data itself, and it's 'quality', getting it from A>B and successfully converted in the DAC, rather than the wider question of subjectivity. If an expensive CD transport and an expensive computer server manage to create a more 'realistic' sound due to how they deliver the data, then there has to be something to it unless people are imagining things completely.

Oh, and I forgot another popular one to add to the list right now from the MQA marketing blurb, which is the 'time smeared' data problem that apparently exists in the audio files we buy, and isn't fixable regardless of how perfectly we transport the data to its destination. Could this actually be the real answer we're looking for, or just another part-fix, or a non-issue just to confuse us further…

I also saw an interesting video clip (I can't remember what it was exactly) but I think it was Stereophile magazine versus scientists/engineers or researchers, and the 'opposition' to Stereophile basically said that in all their testing and experiments over the past years, they've never seen people be able to identify different amplifiers that measured the same and were level matched, and that, in fact, its the human bias based on seeing the 'box' that decides how it sounds. I would love to take part i something like that, to see if a raspberry Pi DAC/Amp versus a Naim Statement or whatever could be picked out blind, I find it fascinating. Anyway, drifted off my own thread!

[SwissBear]

I would love to take part i something like that, to see if a raspberry Pi DAC/Amp versus a Naim Statement or whatever could be picked out blind, I find it fascinating.  Anyway, drifted off my own thread!

Another interesting bias is the reliability of our memories in these fields : https://www.youtube.com/watch?v=XgRlrBl-7Yg