Thread: A Search for Truth and Tonality, Part 2 ...

[fas42]

From http://www.whatsbestforum.com/showthread.php?6758-Do-we-use-our-ears&p=112513&viewfull=1#post112513 ...

For me the best description was the enthusiastic reaction of our administrator Steve Williams in Paul Stubblebine Mastering studio listening to a 1" master tape, as he posted in audiogon some time ago and I repeat from time to time with the help of google:

... but in my 38 years involved in this hobby I have just never heard anything as good as this ...NEVER. It was tantamount to being at the symphony and hearing it live. ...

Just to round off this current "rant", this is the sort of thing digital can do with ease. But, and it's a very big but, I would estimate that 99.9% of systems doing digital don't get anywhere near what they could do, for all the reasons mentioned by myself many, many times. Tape escapes a lot of the problems because the type of technology used to capture and retrieve the source material is far more resistant to "contamination" so to speak -- definitely not, because it's analogue, vs. digital ...

Frank

 

[NorthStar]

Frank, after serving your time, will you be able to reintegrate the society. :b

 

[jkeny]

Wrong, wrong, wrooong!!! What you hearing is introduced distortion, note this word, introduced, distortion occurring in the process of playback.

Are you denying that quantization errors exist in digital unless noise is "introduced" i.e dither? So can you tell me why this isn't a precarious solution to a natively flawed technology? Can you also tell me how using FFT methodology to recover the low-level signal from the introduced noise by averaging over many samples actually works for hearing?

I've spent 25 years plus working on this "problem", and I know it by heart, even though I haven't got all the technical answers for why it occurs. A dead, lifeless quality in low level detail when playing CDs is something I've always heard when listening to other people's systems, but one doesn't have to live with it, it can be fixed -- if I hadn't been able to "solve" it, I would have chucked the hifi setup out the window years ago, and I in fact did so, effectively, for about 10 or more years, because it got so frustrating ...

Yes I generally agree with you & believe that 16/44 digital is so difficult to get anywhere near working optimally that for all intents & purposes it is flawed!! I'm not sure if you offer any solutions other than generalisations abut getting rid of noise etc. all of which I agree with but it's implementation of this is nearly

But if people keep banging on, prattling, about needing more bits, less jitter, higher sampling rates, better dithering, you'll never, ever solve the problem in a wider context ! Never, ever ... it will just remain another, Never Ending Story ...

Frank

I'm agnostic about whether 16/44 is sufficient for optimal audio reproduction but I suspect that it isn't. I'm pretty sure that we need very low levels of deterministic jitter (in the 10s of picoseconds), I'm pretty sure that we need very low levels of noise in power supplies & grounds, I'm pretty sure we need very low levels of RFI & EMI system-wide, I'm pretty sure that I'm missing a bunch of other necessary criteria.

What's your actual implementation solution?

 

[fas42]

Frank, after serving your time, will you be able to reintegrate the society. :b

Bob, is that me "reintegrating with society" -- understanding the hierachy of suckee, and learning to love it -- or integrating society with a "better truth", hmmm ... ?

Actually, I'm not fussed: it's rather fascinating to see everyone settle back into their comfortable foxholes, totally assured of the veracity of their points of view -- if I went away and came back in 10 years there would be the same crew, same opinions, same arguments; not one iota of movement would have taken place, in any direction whatsoever: the comfortableness of it all takes precedence over everything ...

Frank

 

[fas42]

Are you denying that quantization errors exist in digital unless noise is "introduced" i.e dither? So can you tell me why this isn't a precarious solution to a natively flawed technology? Can you also tell me how using FFT methodology to recover the low-level signal from the introduced noise by averaging over many samples actually works for hearing?Yes I generally agree with you & believe that 16/44 digital is so difficult to get anywhere near working optimally that for all intents & purposes it is flawed!! I'm not sure if you offer any solutions other than generalisations abut getting rid of noise etc. all of which I agree with but it's implementation of this is nearly

Yes, there are quantization errors, but at a very low level! When I first really started playing with digital playback I made some major efforts to actually try and hear this happening, and it was a struggle. My best specimen is one of operatic highlights with Kiri Te Kanawa, which has been mastered at a very, very low level: the system has to run at maximum volume to sound any louder than a kitchen radio. And at the very start of the first track, super, super quiet that it is, running at maximum gain, with my ear next to the tweeter I can hear the digital hash happening, from poor dithering.

Yes, it's "easy" to hear this happening by running the signal through a pre-amp and upping the volume enormously, but this is a totally artificial situation! If you did the equivalent with a virgin R2R tape the hiss would be deafening, it would sound as if a massive thunderstorm was hammering your roof! You have to compare apples with apples ...

Not quite sure what you mean about FFT ...

Yes I generally agree with you & believe that 16/44 digital is so difficult to get anywhere near working optimally that for all intents & purposes it is flawed!! I'm not sure if you offer any solutions other than generalisations abut getting rid of noise etc. all of which I agree with but it's implementation of this is nearly

It's "flawed" because people keep focussing on the wrong areas to fix up on the replay side! If you listened now to one of the prized, collectors' special LPs from the Golden Era on a reasonable playback TT of the mid 60's would you swoon over the sound, or, think, hmmm, vinyl has got some issues ....

So, yes, it IS implementation that is the problem, but the industry still hasn't "got it", in the same way the makers of TTs, etc, in the 60's didn't understand what they had to do ...

Why I go on about the noise, distortion issue, is because that has been the solution for me! And when I see comments by owners of expensive CD players still complaining about things that bugged me in 1986 then I know that key progress has still not been made.

I'm agnostic about whether 16/44 is sufficient for optimal audio reproduction but I suspect that it isn't. I'm pretty sure that we need very low levels of deterministic jitter (in the 10s of picoseconds), I'm pretty sure that we need very low levels of noise in power supplies & grounds, I'm pretty sure we need very low levels of RFI & EMI system-wide, I'm pretty sure that I'm missing a bunch of other necessary criteria.

What's your actual implementation solution?

Yes, all those things need to be right, to a certain level of performance. But the "problem" with digital is that if just one of things, just one!, is not right then the sound is badly affected. And I mean, badly!

So, trite that it is, so trite it indeed sounds, the implementation solution is to fix every weakness, every problem area! And how do you know when you've knocked off every significant one? By the fact that the sound comes to life, and stays at that level! With digital, for me, either it sounds "right", meaning big, full, rich, enveloping; or it's miserable, yuckky, give it away: there's no inbetween ...

Frank

 

[NorthStar]

Bob, is that me "reintegrating with society" -- understanding the hierachy of suckee, and learning to love it -- or integrating society with a "better truth", hmmm ... ?

Actually, I'm not fussed: it's rather fascinating to see everyone settle back into their comfortable foxholes, totally assured of the veracity of their points of view -- if I went away and came back in 10 years there would be the same crew, same opinions, same arguments; not one iota of movement would have taken place, in any direction whatsoever: the comfortableness of it all takes precedence over everything ...

Frank

But in ten years from now, where will you be, in your 'teachings' and 'theories' and 'discoveries', and all that Jazz, Frank? :b

 

[NorthStar]

.... Not quite sure what you mean about FFT ...

Frank

Fast Fourier transform => http://en.wikipedia.org/wiki/Fast_Fourier_transform

 

[fas42]

But in ten years from now, where will you be, in your 'teachings' and 'theories' and 'discoveries', and all that Jazz, Frank? :b

Predicting the future, eh, Bob -- I'd better put on my clairvoyant's hat, I think. And, actually, I do do a reasonable job in that area ... talk about being versatile!!

Well, I would like to give the audio sleepyheads a good kick in the pants; but, as you can well and truly see, this is mighty difficult! I need to build up some momentum, not so easy for me any more -- I tried with the high power amp, then got sidetracked into the unravelling of compression thing. The latter is not really going to be a goer, as least with the technology of today, because what people are really complaining about is the current overprocessing of audio tracks, compression is just one part of that, and to really get somewhere with that greater problem is just too complex a task at present.

So, moved to the studio monitor project: easy, cheap to do, and gives me a chance to get a few heads turning: the current performance levels of the "good" names is barely there, so there is excellent scope for improving what people have to play with.

Down the track would like to assemble an integrated setup that could do, say, mid 130dB's peaks cleanly, Basspig territory, with good looks, for normal, domestic consumption.

As for "discoveries", I still don't fully understand the 'how' of the various interference and distortion mechanisms nobbling audio replay, so if I have the mental room and finances for it, I would like to assemble some solid test gear to help track down the full path of cause and effect ...


PS: I know what FFT is, just don't get what John's aiming at ...

Frank

 

[fas42]

From http://www.whatsbestforum.com/showt...re-information&p=112705&viewfull=1#post112705 ...

Actually, getting that detail isn't so hard, getting it at safe sound pressure levels is the real challenge

This is an interesting one, because it points at two challenges in the audio game: firstly, to be able to run the system up to and past realistic volume levels, and secondly to do that with effectively inaudible distortion. This is where decent studio monitors, for example, get the first need right: they can go up to the red line without any apparent distress. But, but, the sound they produce is so murky, irrespective of volume, that they need to be run at high loudnesses to pick where the problems are -- what Jack is eluding to.

And the reason for that is the murkiness: good ol' distortion, yet again! If you eliminate that, then it's trivially easy to recover detail; because it's always been there, but your ear/brain is having to fight its way through the grunge, of the layer of electronic noise to try and decipher it, and you mind just gives up, usually with fatigue and a possible headache. Simple methodology: minimal audible distortion equals maximal ease in registering detail, at all volumes.

Frank

 

[jkeny]

Yes, there are quantization errors, but at a very low level! When I first really started playing with digital playback I made some major efforts to actually try and hear this happening, and it was a struggle. My best specimen is one of operatic highlights with Kiri Te Kanawa, which has been mastered at a very, very low level: the system has to run at maximum volume to sound any louder than a kitchen radio. And at the very start of the first track, super, super quiet that it is, running at maximum gain, with my ear next to the tweeter I can hear the digital hash happening, from poor dithering.

Yes, it's "easy" to hear this happening by running the signal through a pre-amp and upping the volume enormously, but this is a totally artificial situation! If you did the equivalent with a virgin R2R tape the hiss would be deafening, it would sound as if a massive thunderstorm was hammering your roof! You have to compare apples with apples .
Not quite sure what you mean about FFT ...


It's "flawed" because people keep focussing on the wrong areas to fix up on the replay side! If you listened now to one of the prized, collectors' special LPs from the Golden Era on a reasonable playback TT of the mid 60's would you swoon over the sound, or, think, hmmm, vinyl has got some issues ....

So, yes, it IS implementation that is the problem, but the industry still hasn't "got it", in the same way the makers of TTs, etc, in the 60's didn't understand what they had to do ...

Why I go on about the noise, distortion issue, is because that has been the solution for me! And when I see comments by owners of expensive CD players still complaining about things that bugged me in 1986 then I know that key progress has still not been made.


Yes, all those things need to be right, to a certain level of performance. But the "problem" with digital is that if just one of things, just one!, is not right then the sound is badly affected. And I mean, badly!

So, trite that it is, so trite it indeed sounds, the implementation solution is to fix every weakness, every problem area! And how do you know when you've knocked off every significant one? By the fact that the sound comes to life, and stays at that level! With digital, for me, either it sounds "right", meaning big, full, rich, enveloping; or it's miserable, yuckky, give it away: there's no inbetween ...

Frank

I'm talking about the inherent weakness of digital audio - the fact that quieter sounds are coded using fewer bits, ending at the LSB which needs the support of dither to work. So I agree with you, if everything isn't exactly right & optimised then what suffers is the reproduction of these low level details (the bits towards the LSB end). Now what I'm saying & others are too is one approach to partly solving this issue is to relax some of the requirements needed for this reproduction of low level sounds by using higher resolution files - it would seem to be a sensible thing to do. There are also theoretical issues about whether higher resolution files are needed anyway to truly capture & reproduce audio optimally. So all-in-all it seems like it is a good idea. Just because you can hear a better reproduction of digital than most on your optimised system (which has taken you >20 years to achieve) does not mean that it is the only solution. Maybe by relaxing the requirements, more people will be able to achieve greater sound reproduction & it won't take them 20+ years to get there?

Now the other issues of noise & jitter still remain & have to be dealt with anyway! But I don't agree with your black & white assessment of digital audio - either it is perfect or it is crap - there are many gradations in between

 

[jkeny]

PS: I know what FFT is, just don't get what John's aiming at ...

Frank

Frank, what I'm getting at is that to accurately use the LSB dither is required. Dither is noise added to the signal. The only way to recover the signal from this noise is by averaging over many samples, whereby the non-randomness of the signal peaks above the randomness of the noise floor. This is the FFT method.

Now averaging is not a great way to hear music, I would contend. It would seem to me that there are many possible flaws with this approach such as the capture of the fast transients associated with the attack of instruments. If a transient is fast enough that it falls outside of the the averaging window of samples, then it will be missed or distorted from it's original waveform.

It might partly be at the heart of the explanation for the statements that you have quoted from the other threads about cymbals & tape hiss not sounding natural on digital reproduction!

 

[fas42]

I'm talking about the inherent weakness of digital audio - the fact that quieter sounds are coded using fewer bits, ending at the LSB which needs the support of dither to work.

John, this is noise we're talking about, analogue has exactly the same "problem": it's just that hiss and other artifacts are more acceptable to people's ears; hence you use dither to adjust the tonal quality of the digital quantisation errors to create the "nicest" noise. There's a reason that you hear master tape hiss on CDs -- because digital has far greater resolution then that glorious R2R signal. Try using your best tape recorder to actually capture audible quantisation error of a RB CD signal, matching maximum signal levels: anything audible will be drowned in a sea of hiss ...

So I agree with you, if everything isn't exactly right & optimised then what suffers is the reproduction of these low level details (the bits towards the LSB end). Now what I'm saying & others are too is one approach to partly solving this issue is to relax some of the requirements needed for this reproduction of low level sounds by using higher resolution files - it would seem to be a sensible thing to do. There are also theoretical issues about whether higher resolution files are needed anyway to truly capture & reproduce audio optimally. So all-in-all it seems like it is a good idea. Just because you can hear a better reproduction of digital than most on your optimised system (which has taken you >20 years to achieve) does not mean that it is the only solution. Maybe by relaxing the requirements, more people will be able to achieve greater sound reproduction & it won't take them 20+ years to get there

Yes, there are all sorts of ways of getting around the "problems", but much of it is overkill: a bit like having a 1000W amplifier which peaks at an output of 20W in the loudest passages when you listen to it. Which means that for one thing it's very expensive, in money and resource senses. From my own experiments I've determined that an MP3 file sounds dramatically better, played back over a cheap and nasty decoder, when you convert and resample to a 24/384 format. Why should this happen? Simple answer is, because the electronics, the PC processor and DAC chip, have far less work to do -- less stressed electronics = better sound. There's just a little downside: the music file is half a Gigabyte!!

There needs to be a better way -- the sort of thing I'm working on ...

Now the other issues of noise & jitter still remain & have to be dealt with anyway! But I don't agree with your black & white assessment of digital audio - either it is perfect or it is crap - there are many gradations in between

Unfortunately, for me, I don't find it so. I've heard "nice" digital, many times, and while that is acceptable to many, it is not for me, because large chunks of detail are stripped from it -- the ugliest woman can look reasonable if you blur the photo enough ...

I want to hear what is on the recording, not a heavily filtered version thereof, and with digital I have not heard or achieved a system capability where if all the detail IS coming through then it sounds pleasant even if not working at optimum. It may be possible, but I haven't experienced such yet ...

Frank

 

[jkeny]

John, this is noise we're talking about, analogue has exactly the same "problem":

No, it's not noise, it's the inability to accurately represent a waveform at a very low level using a few bits. try listening to a 3 bit digital recording & see if it's just noise or if it's distortion you hear!

it's just that hiss and other artifacts are more acceptable to people's ears; hence you use dither to adjust the tonal quality of the digital quantisation errors to create the "nicest" noise. There's a reason that you hear master tape hiss on CDs -- because digital has far greater resolution then that glorious R2R signal. Try using your best tape recorder to actually capture audible quantisation error of a RB CD signal, matching maximum signal levels: anything audible will be drowned in a sea of hiss ...

if it was just noise you would have a point but it's not just noise!!

Yes, there are all sorts of ways of getting around the "problems", but much of it is overkill: a bit like having a 1000W amplifier which peaks at an output of 20W in the loudest passages when you listen to it. Which means that for one thing it's very expensive, in money and resource senses. From my own experiments I've determined that an MP3 file sounds dramatically better, played back over a cheap and nasty decoder, when you convert and resample to a 24/384 format. Why should this happen? Simple answer is, because the electronics, the PC processor and DAC chip, have far less work to do -- less stressed electronics = better sound. There's just a little downside: the music file is half a Gigabyte!!

there's more than the filesize as a little downside - the sound is also a big downside!!

There needs to be a better way -- the sort of thing I'm working on ...

Unfortunately, for me, I don't find it so. I've heard "nice" digital, many times, and while that is acceptable to many, it is not for me, because large chunks of detail are stripped from it -- the ugliest woman can look reasonable if you blur the photo enough ...

I want to hear what is on the recording, not a heavily filtered version thereof, and with digital I have not heard or achieved a system capability where if all the detail IS coming through then it sounds pleasant even if not working at optimum. It may be possible, but I haven't experienced such yet ...

Frank

Why then do you upsample MP3? Your statements are logically incongruent if you are actually recommending upsampled MP3 as somehow acceptable!

 

[fas42]

Frank, what I'm getting at is that to accurately use the LSB dither is required. Dither is noise added to the signal. The only way to recover the signal from this noise is by averaging over many samples, whereby the non-randomness of the signal peaks above the randomness of the noise floor. This is the FFT method.

Now averaging is not a great way to hear music, I would contend. It would seem to me that there are many possible flaws with this approach such as the capture of the fast transients associated with the attack of instruments. If a transient is fast enough that it falls outside of the the averaging window of samples, then it will be missed or distorted from it's original waveform.

It might partly be at the heart of the explanation for the statements that you have quoted from the other threads about cymbals & tape hiss not sounding natural on digital reproduction!

This ability of digital to render accurately cymbal shimmer, fading smoothly to silence is one way to describe where the dilemma is. Take a particular CD, with high levels of cymbal detail: on one high end system that cymbal shimmer will simply not exist, it has been expunged from the soundscape by skillful filtering along the chain; on another it will sound like some vague, reasonably pleasant background hiss; and on yet another it will be the most aggressive, ear drum piercing slingshot of audible pain. Same CD, same track, totally, totally different audible sounds. Which is right? None, because all have distorted the sound in different ways: the real cymbal sound does exist on that CD, because I have heard it in all its glory ...

And it didn't require dithering to get it right: this shimmer is a very high level sound, and the fading to silence only requires the dither to be meaningful at the point where I can just hear it with my ear next to the tweeter ...

Frank

 

[jkeny]

This ability of digital to render cymbal shimmer accurately, and smoothly to silence is one way to describe where the dilemma is. Take a particular CD, with high levels of cymbal detail: on one high end system that cymbal shimmer will simply not exist, it has been expunged from the soundscape by skillful filtering along the chain; on another it will sound like some vague, reasonably pleasant background hiss; and on yet another it will be the most aggressive, ear drum piercing slingshot of audible pain. Same CD, same track, totally, totally different audible sounds. Which is right? None, because all have distorted the sound in different ways: the real cymbal sound does exist on that CD, because I have heard it in all its glory ...

And it didn't require dithering to get it right: this shimmer is a very high level sound, and the fading to silence only requires the dither to be meaningful at the point where I can just hear it with my ear next to the tweeter ...

Frank

Frank, it's getting tiring to hear that you have it all but never telling us the exact details of how you are achieving this. Can you do this so that I can judge where you are coming from? Maybe you have given these details before, I don't know?

 

[fas42]

Frank, it's getting tiring to hear that you have it all but never telling us the exact details of how you are achieving this. Can you do this so that I can judge where you are coming from? Maybe you have given these details before, I don't know?

Oh dear, this sounds familiar! Sorry, John, it's just not that simple: I used the expression of people thinking there was a Magic Bullet eons ago, and was roundly attacked for doing so ...

I'll give you an analogy: you have a space capsule that leaks air; if it leaks people inside die, so you have a real problem here. But the reason that it leaks is because all sorts of short cuts were taken making the capsule, the engineering was poor, the construction people couldn't be bothered doing a good job, the maintenance people don't think you have to check various things.

So what's the solution to making the space capsule work properly, not leak at all? Is there some mystical answer, some Great Secret? No, it requires perserverance, and the knowledge that the capsule can be made not to leak air, and the only answer is that every single hole has to be plugged. Every last one of them. If you leave one hole unplugged, then the people inside will still die, in spite of the fact that you put so much energy into plugging every other leak.

And that is the answer, pure and simple. People already know what the leaks are, but you have to take care of every single one of them ...

EDIT: In terms of what I have "done" to different pieces of gear it has been radically different over the years: I've obsessed about the quality of power supplies, the quality of connections, the level of vibration impacting various parts. Depending on how well the gear's been made, I worry about different things. The overall theme is to get the key electronics to function with as little stress, the least interference possible. How do I know when I've done enough? When the sound gets "as good" as I talk of ...

And to put it in a wider context: I'm looking at the internals of a well regarded studio monitor at the moment, and it's riddled with weaknesses. Barely acceptable power supply, poor connections all over the place, no attempts to prevent interference coming up the power cord. It's a wonder it works as well as it does ...

Frank

 

[jkeny]

Frank, it is not enough to talk in platitudes. You stated that

" the real cymbal sound does exist on that CD, because I have heard it in all its glory ..."

So it's time to specify what was the detail of that particular system that gave you this experience! I presume it was a real experience & not something made up so I'm assuming it was also a real playback system that you can describe in detail? I'm not expecting a magic bullet, just a description of a particular system!

 

[fas42]

Frank, it is not enough to talk in platitudes. You stated that So it's time to specify what was the detail of that particular system that gave you this experience! I presume it was a real experience & not something made up so I'm assuming it was also a real playback system that you can describe in detail? I'm not expecting a magic bullet, just a description of a particular system!

I've been working on a variety of systems over that 25 years or so. The first setup that gave me the good stuff was a Yamaha CDX-1100 battleship CD player, 1986 vintage, digital volume control directly driving a Perreaux 2150B, talking to B&W DM10s. The Yamaha, which I still have, driving the studio monitors, uses just single Burr Brown PCM56 chips for each channel: on maximum volume, low level material you can clearly hear these ancient multibit chips' less than 16 bit perfect linearity. But this was the setup that showed me what was possible, particularly on cymbals: what was holding me back then was the Perreaux's relatively weak power supply design, past a certain volume and the sound started to compress.

Since then I've gone through a variety of multi-bit, and delta sigma chipped CD players, non high end at all: the Yamaha was the best of the bunch, would probably cost $5,000 - $10,000 these days. The sound qualities have been consistent, something I look for. The current test setup is a throw away HT machine, Philips, quite well made compared to a lot of the current rubbish: it uses a cheap Cirrus Logic chip, better low level linearity than the Yamaha. I have noted that the delta sigma chips need to be heavily thrashed on turn on, quite a number of hours to get the treble working right. I'm also "handicapped" by the very ordinary speakers that are part of the HT, they also have to be driven hard a number of hours to free up the suspensions for best treble quality.

For more details of the current setup check this thread: http://www.whatsbestforum.com/showthread.php?2966-A-Search-for-Truth-and-Tonality ...

Frank

 

[jkeny]

Frank, What DAC were you feeding the Yamaha into? What digital volume control? I really can't see you achieving the sound that you claim you did with a digital vol control. What setting did you normally have the vol control ?

I don't see any clear evidence of your carefully looking after every detail of the system. In fact I would say from your descriptions, which have little to no evidence of any attention to detail, that your system is very ordinary.

Are you sure that your not exaggerating your experience? Who else has heard it that has posted their listening experiences?

 

[fas42]

Lots of talking about synergy ... this is all about adjusting flavours of distortion, to get the most pleasant combinations, a sort of blending of Scotch whisky type of thing ...

Lack of "synergy" will typically be because the type of interference generated by one component will more severely upset another: a simple example is that the spectrum of the nasty current spikes on the mains power being fed to the amplifier will play havoc with the power supply of the preamp, which will then happily spit extra levels of distortion further down the line. At least when the components come out of one manufacturing shop someone will hopefully have noticed that this can happen, and at least taken some measures to moderate the interaction ...

Frank