PS3 SACD Playback

[The Seventh Taylor]

I'd point out also Seventh Taylor that the article you reference for London-Boy is actually included within the first post of the thread you had bumped, and indeed that thread was the original source for the spread of the story.

Almost. The article One referred to and kindly translated is .../mobile357.htm. The one I referred to is .../mobile356.htm. They are two different articles from a series of three.

Anyway, thanks for giving this its own thread. I'm not into necrophaelia but I'm new to this forum, found this thread through a search engine and felt urged to respond.

Now the discussion does seem to have gone in quite a different direction from here.

 

[randycat99]

So, tkf, do you, personally, understand what is being presented in your 2nd link? If you do have comprehension on that level, how did the faultiness in the 1st link not set off some serious alarms in your head?

Do you not see some areas in the material that could be faulty in the 2nd link? For instance, why do you absolutely need a high-quality clock that is accurate to within pico-seconds to be used in a system that resolves audio frequencies? That is about 1e6 times smaller than the time slice you could expect for even the shortest sampling event for an audio signal to occur on an SACD/DVD-A recording. None of this raises a question in your head?

 

[-tkf-]

So, tkf, do you, personally, understand what is being presented in your 2nd link? If you do have comprehension on that level, how did the faultiness in the 1st link not set off some serious alarms in your head?

Do you not see some areas in the material that could be faulty in the 2nd link? For instance, why do you absolutely need a high-quality clock that is accurate to within pico-seconds to be used in a system that resolves audio frequencies? That is about 1e6 times smaller than the time slice you could expect for even the shortest sampling event for an audio signal to occur on an SACD/DVD-A recording. None of this raises a question in your head?

Do you acknowledge there is a something called jitter or not?
Do you think jitter can be heared?

 

[randycat99]

Certainly, there is jitter. The vehicles by which it can manifest into a "sound" is the unsubstantiated part.

It seems to be thrown into the fray far too often by people who know very little about it, but profess to know. So do you understand the material in your 2nd link or not? Why is it that nanoseconds are of concern in the 1st link, but in the 2nd link, the concern moves to picoseconds? Maybe it doesn't matter how well the technology improves, because somebody will just reason that jitter on an even more infinitesimal scale prevents us from reaching audio nirvana?

Do you see the problems with the proposed setups in your 2nd citation? Do you realize the manner in which they are atypical to nearly all consumer implementations?

 

[Schaden]

I've been around the high end audio block a few times. Given a quality recording, which is paramount in hi-fi, the speakers and room acoustics determine the majority of the sound quality.

Analog sources have the most room for improvement. Night and day difference between low/medium/high end gear.

Digital sources are more subtle. You may think a bit is a bit. But I think even a novice can appreciate the improved instrumental focus, three-dimensionality, separation, and low-level detail resolution offered by the best digital equipment.

I challenge any of you to compare a $100 mass market cd player against a dCS or meridian DAC/transport combo. So long as the rest of the equipment is sufficiently revealing, there is a perceivable difference. In many cases, it's hard to justify the expense. Like someone alluded, diminishing returns kick in fast.

Much of hi-fi comes down to system synergy. Careful speaker placement and room treatments are critical. You won't find hi-fi gear in a Best Buy or Circuit City. But you don't have to spend a fortune either. Some of the finest equipment I've heard was DIY or made by low volume, internet direct, boutique companies.

I don't know how much better crimson book is to red book in a real world application. I'd want to compare formats, recorded from the same performance, mastered by the same person, reproduced on the same reference stereo, in a double blind test. That would be a feat !

 

[-tkf-]

Certainly, there is jitter. The vehicles by which it can manifest into a "sound" is the unsubstantiated part.

So it´s there but it cannot be heard?

That is all i wanted to know. As i said some think MP3 sounds just as the PCM some don´t. This is the classic Audio discussion it´s been going on for years and it hasn´t changed. The fun thing is, i used to be doing your part of the argument

 

[randycat99]

Jitter can exist in the transmission of a digital signal. That doesn't automatically mean there is a corresponding audio component once the conversion to analog is done. The nature of digital suggests that the signal can be increasingly corrupted by jitter, interference, noise, whatever, and the analog signal will be successfully reconstructed identical to the original as long as the signal is strong enough. At the point where the signal is not strong enough, you will get various blips in the analog signal as the digital signal yields uncorrectable errors. These blips or dropouts will be obvious as such.

You don't get this gradual "deterioration" of "soundstage", "tempo", or "analog smoothness", as is often claimed in myriad jitter scenarios. When someone starts speaking like that, you know to immediately to don your BS boots on.

 

[jonabbey]

So it´s there but it cannot be heard?

In the digital domain, yes. Much as you can't tell the difference in the result between adding a pair of 2's written in red crayon and adding a pair of 2's laser printed from a high end computer.. you get 4 in both cases. That's what the digital domain means.. you don't care about the analog nature of the medium which is getting the numeric information to you, you just care about whether you can read the numbers.

That is all i wanted to know. As i said some think MP3 sounds just as the PCM some don´t. This is the classic Audio discussion it´s been going on for years and it hasn´t changed. The fun thing is, i used to be doing your part of the argument

 

[phat]

Jitter can exist in the transmission of a digital signal. That doesn't automatically mean there is a corresponding audio component once the conversion to analog is done. The nature of digital suggests that the signal can be increasingly corrupted by jitter, interference, noise, whatever, and the analog signal will be successfully reconstructed identical to the original as long as the signal is strong enough. At the point where the signal is not strong enough, you will get various blips in the analog signal as the digital signal yields uncorrectable errors. These blips or dropouts will be obvious as such.

You don't get this gradual "deterioration" of "soundstage", "tempo", or "analog smoothness", as is often claimed in myriad jitter scenarios. When someone starts speaking like that, you know to immediately to don your BS boots on.

You're certainly correct that digital transmission is impervious to jitter as long as it's below a certain recovery threshold. Jitter at the DAC is what people worry about...

Jitter in the nanosecond range is audible for high frequency (~16 kHz) tones. Jitter in the picosecond range might not be audible for mono tones, but might affect perceived imaging of stereo audio. It's the high frequency components that most determine the so-called "sound stage", and they are the most affected by jitter.

Also, 44.1 kHz, 16-bit sampling is actually not all that good at capturing all the nuances that the human ear is sensitive to. Remember that phase gets implicitly quantized along with amplitude. A -60dBF tone for example is quantized into just 6 bits, and the phase quantization from that is signficant for a high frequency signal.

 

[randycat99]

Jitter in the nanosecond range is audible for high frequency (~16 kHz) tones.

We'll have to agree to disagree on this point. According to tkf's 2nd source, the most they could conjure from a deliberately induced jitter is a small spike in the spectrum at -130 dB. That's pretty much into the thermal noise floor of any consumer audio device known to man. So maybe you can hear this artifact 20 dB submerged under equipment noise with another 40-50 dB of ambient room noise on top of that, but I'll just have to maintain just a bit of skepticism over this claim.

Jitter in the picosecond range might not be audible for mono tones, but might affect perceived imaging of stereo audio. It's the high frequency components that most determine the so-called "sound stage", and they are the most affected by jitter.

Afaik, the perception of soundstage relies on phasing and amplitude occuring in the entire human auditory range above bass frequencies, not just high frequencies. If it is the case that treble is dominantly responsible for soundstage and jitter affects this treble, then the soundstage from a vinyl recording must be entirely nonexistent and utterly corrupted by media imperfection. It is there where wow and flutter occur at many orders of magnitude greater than any conceivable amount of picosecond jitter that happens in the digital world. Strangely, vinyl is somehow exempt or immune to this...must be more to this "soundstage" stuff than simple frequency dependent time delay effects, eh?

Also, 44.1 kHz, 16-bit sampling is actually not all that good at capturing all the nuances that the human ear is sensitive to. Remember that phase gets implicitly quantized along with amplitude. A -60dBF tone for example is quantized into just 6 bits, and the phase quantization from that is signficant for a high frequency signal.

Again, we'll have to agree to disagree on this point, as well. If you believe plain jane pcm is capturing only a subset of what the ear hears, then I suppose anything can be argued. The extension of this is how can we ever be sure any kind of audio capture mechanism will be sufficient, regardless of how high the bit depth and sampling rate? The more obvious culprit(s) in the chain which explains why you aren't hearing what you think you should be hearing is NOT the digital, but remains the mic'ing scheme and your speakers (and your own ears, for that matter). Beyond that, you have the mixing and processing that happened in the studio. In any/all of those stages, you will find things being done to the waveform to alter/adulterate/utterly transform the signal in frequency and phase that are many orders of magnitude more egregious than anything the digital could contribute, itself.

You may believe otherwise, but that's my 2 cts.

 

[phat]

We'll have to agree to disagree on this point. According to tkf's 2nd source, the most they could conjure from a deliberately induced jitter is a small spike in the spectrum at -130 dB. That's pretty much into the thermal noise floor of any consumer audio device known to man. So maybe you can hear this artifact 20 dB submerged under equipment noise with another 40-50 dB of ambient room noise on top of that, but I'll just have to maintain just a bit of skepticism over this claim.

The distribution of the induced jitter is also important. I haven't looked at tkf's 2nd source, but I recall an AES paper that says that 10ns jitter is audible for frequencies near the top end of human hearing.

Afaik, the perception of soundstage relies on phasing and amplitude occuring in the entire human auditory range above bass frequencies, not just high frequencies. If it is the case that treble is dominantly responsible for soundstage and jitter affects this treble, then the soundstage from a vinyl recording must be entirely nonexistent and utterly corrupted by media imperfection. It is there where wow and flutter occur at many orders of magnitude greater than any conceivable amount of picosecond jitter that happens in the digital world. Strangely, vinyl is somehow exempt or immune to this...must be more to this "soundstage" stuff than simple frequency dependent time delay effects, eh?

Wow and flutter are very low frequency phenomena, and probably do not affect the "transparency" of the soundstage. High frequency jitter, on the other hand, would perhaps have a more diffusive effect on the soundstage.

The thing is, stereo does not allow the soundstage to be reproduced accurately. It doesn't give realistic positional information. The soundstage one perceives is highly fictional. Wow and flutter could cause gradual drifting of the soundstage, while jitter could cause a "muddying" of the soundstage. It could be like seeing the bottom of a pond through slow long waves vs. lots of small ripples.

Again, we'll have to agree to disagree on this point, as well. If you believe plain jane pcm is capturing only a subset of what the ear hears, then I suppose anything can be argued. The extension of this is how can we ever be sure any kind of audio capture mechanism will be sufficient, regardless of how high the bit depth and sampling rate?

I'm not saying that sampling theory is all wrong. I just think that the metrics that people have been using to quantify the accuracy of sampling neglects the effect that quantization has on phase. Each sample is not really a point, but an error bar. Timing jitter is error in the x-axis, and bit depth affects error in the y-axis. Tones near the nyquist limit, that aren't at full dynamic range, and hence don't benefit from full bit depth, have a lot of phase information discarded that might be important.

 

[randycat99]

The distribution of the induced jitter is also important. I haven't looked at tkf's 2nd source, but I recall an AES paper that says that 10ns jitter is audible for frequencies near the top end of human hearing.

Essentially, you hear some "extra noise" in the extreme treble, if you hear it at all. Just a sidenote- if you are really averse to some extra noise in the extreme treble/ultrasonic range, then SACD is definitely not the right format for "soundstage".

Wow and flutter are very low frequency phenomena, and probably do not affect the "transparency" of the soundstage.

True, and you should also expect that if the timing is inconsistent at low frequencies, it only becomes even more so at higher frequencies. Essentially, the rudimentary design of an etched platter and mechanical turntable can barely eliminate speed/timing variations at low frequencies, thus it hasn't even a chance of keeping accurate timing at higher frequencies...yet we still have soundstage, right? I also doubt they were aligning the left and right audio tracks down to the nanosecond, let alone picosecond, as they carved the master disc. The technology for that sort of manipulation simply did not exist for the vinyl era. Even if it did, it would still be pointless, since the aforementioned imperfection of the media and mechanical playback scheme could never be accurate down to nanoseconds, anyway.

So I'm thinking the correlation between soundstage and jitter is looking pretty busted for a theory, unless anybody would like to go on the record that vinyl was inherently incapable of "soundstage" or any kind of redeeming stereo imaging effect, whatsoever.

The thing is, stereo does not allow the soundstage to be reproduced accurately. It doesn't give realistic positional information. The soundstage one perceives is highly fictional.

No doubt about that. The extension to that is that no one has yet invented the "perfect" mic arrangement to utterly capture an audio event in 3 dimensional space + time. It's all a compromise for the time being (though some very good results can still be attained). No amount of jitter or lack of it is going to make up for that. So I say, if something is to be blamed for not perfectly capturing a soundstage, at least put mic'ing techniques at the top of list. That's going to be many orders of magnitude more dominant an effect than any sort of jitter at the level of picoseconds.

Timing jitter is error in the x-axis, and bit depth affects error in the y-axis.

...and in both scenarios, scale is an important context. Would you sweat a time unit on the x-axis that is on the order of 1 million times smaller than the smallest division on your x-axis any more than you would sweat an amplitude unit on the y-axis that is on the order of 1 million times smaller than the value of your least significant bit? If this is a 16-bit system, this is essentially like attributing some sort of sound quality happening at a tiny value 20-bits beyond that. That's right- a 16-bit audio track actually needing 36-bits of resolution? Extreme, right? Certainly, that kind of extreme calls for skepticism. You may believe otherwise, but that's how the math seems to work out. We might as well give up now, because the sort of time precision in that realm might as well require a clock with time increments that could track the incremental motion of an electron in widths of 1 electron.

Tones near the nyquist limit, that aren't at full dynamic range, and hence don't benefit from full bit depth, have a lot of phase information discarded that might be important.

The only phase information that can be discarded in the process is the phase information belonging to frequencies pertaining to beyond the nyquist limit. If you want to store phase information down to the picosecond, you are essentially calling for a sampling bandwidth on the order of tera-Hz. I don't know if that is such a reasonable goal for the purpose of human hearing (which can barely break 20 kHz). At some point, you got to acknowledge that other parts of the audio chain are far more the bottleneck to fidelity than jitter at the picosecond level. Just the shear amount of phase anomalies that occur in a premium tweeter transducer element are like a meat grinder to the waveform compared to picoseconds of jitter happening in the digital domain.

 

[-tkf-]

That's going to be many orders of magnitude more dominant an effect than any sort of jitter at the level of picoseconds.

Picoseconds? i don´t think that is the case for consumer grade SPDIF interfaces.

 

[randycat99]

picoseconds, nanoseconds, microseconds...the imaging effect is still swamped by orders of magnitude by the fundamental arrangement and selection of microphones for the audio event. If that setup doesn't capture the quality of soundstage you anticipate for the recording, no amount of jitter or lack of it is going to make up for that. That was the context of my comment.

Looking at it from the other direction, if you are going to be concerned with the jitter in the digital chain, then you are making an assertion that all the matter of phase, timing, whatever happened absolutely perfectly in all the analog capture and manipulation stages that preceded it. The chances of that?...nill. Everything had some level of compromise before that sound even hit the digital domain. A bit of jitter isn't going to salvage/make/break the soundstage from that.

Then after that, the sound has to make it through your speakers (with meatgrinder level phase properties), through a room with imperfect acoustics, and the unique peculiarities of your own ears (which will have the phase needle spinning like a nose-diving plane ready for a crash). It won't be the same phase properties between left to right ear (or speakers due to manufacturing tolerances), so right there you can have a "soundstage bottleneck" on the order of hundreds of degrees of phase.

 

[Bohdy]

Looking at it from the other direction, if you are going to be concerned with the jitter in the digital chain, then you are making an assertion that all the matter of phase, timing, whatever happened absolutely perfectly in all the analog capture and manipulation stages that preceded it. The chances of that?...nill. Everything had some level of compromise before that sound even hit the digital domain. A bit of jitter isn't going to salvage/make/break the soundstage from that.

Then after that, the sound has to make it through your speakers (with meatgrinder level phase properties), through a room with imperfect acoustics, and the unique peculiarities of your own ears (which will have the phase needle spinning like a nose-diving plane ready for a crash). It won't be the same phase properties between left to right ear (or speakers due to manufacturing tolerances), so right there you can have a "soundstage bottleneck" on the order of hundreds of degrees of phase.

That sounds terrible! I'll never listen to recorded audio again!

 

[randycat99]

That's the thing- few people are really aware of the shear amount of phase and amplitude deviation that is introduced by ears and speakers as finite mechanical systems, but gladly the quality of sound we can achieve is still inspiring. It will be quite a day when we have all these issues sorted out to the point where mere picoseconds of jitter becomes the primary deviation to the ideal of "straight wire with gain".

 

[Colourless]

Here is an example:

The speed of sound in air ~= 340m/s

Lets say you are 30cm closer to the left speaker than to the right speaker.

In this case sound from the left speaker will get to you about 0.9 ms before the sound from the right speaker reaches you. That is a much more significant problem than jitter is and affects analog and digital just the same.

 

[randycat99]

...but how much delay is that in picoseconds? ...excellent point!

Looking at a different way, if the construction of the tweeters and front baffle were to be manufactured to a tolerance of 35 nanometers (assuming you had a magic laser pointer to enable you to perfectly align the speaker distances), that would still amount to 100 picoseconds of uncertainty. If there is any speaker manufacturer out there machining/assembling their tweeter parts down to any where close to a tolerance of 1000 nanometers, no mere mortal could afford one (let alone 2 to make a stereo pair). I don't know if the technology even exists to achieve something like this in a mass production scenario.

 

[-tkf-]

That's the thing- few people are really aware of the shear amount of phase and amplitude deviation that is introduced by ears and speakers as finite mechanical systems, but gladly the quality of sound we can achieve is still inspiring. It will be quite a day when we have all these issues sorted out to the point where mere picoseconds of jitter becomes the primary deviation to the ideal of "straight wire with gain".

Yeah it´s amazing that it´s even possible to hear music without throwing up.

 

[-tkf-]

Here is an example:

The speed of sound in air ~= 340m/s

Lets say you are 30cm closer to the left speaker than to the right speaker.

In this case sound from the left speaker will get to you about 0.9 ms before the sound from the right speaker reaches you. That is a much more significant problem than jitter is and affects analog and digital just the same.

I don´t think you will find any serious HiFi freaks setting 30 cm closer to one speaker than another. They spend days setting their room up and then the rest of their life tweaking it.