Taiko Audio SGM Extreme : the Crème de la Crème

[oldmustang]

Measurements of Quads done a long time ago by Hi Fi Choice (UK) and more recently in Stereophile show that Quads are indeed very linear as most loudspeakers go. Their distortion at moderate volumes is around (0.1%), which is about -70 dB. So, you get about 12 bits of linearity. But only at moderate volumes. Above 90 dB or so, Quads will start to distort before shutting down. Klipschorns and La Scalas are quite linear as well with about 0.1% distortion at higher volumes. Notice both of these are far from resolving even a 16-bit Redbook CD recording, let alone 24-bit high res recordings. A speaker that can resolve all 16 bits of a CD is science fiction still. As regards 24-bit high res, that’s a pipe dream. As for 32 bits? You’d need to maintain linearity down to -192 dB. Even if you spent a billion dollars, I doubt you could manage that in any hifi component. Maybe in a multibillion dollar physics lab with superconducting wires and temperatures close to absolute zero Kelvin. In the real world with what we can put in an actual living room? Not a chance in hell.

You seem to be inferring that because speakers have such high levels of distortion that they serve as some kind of brick wall filter that keeps us from hearing effects that lay below their linear resolution, measured in bits.

I think we can demonstrate that is not true for any number of audio parameters that occur at extremely low levels, both analog and digital and yet are audible via our imperfect speakers -- extremely low levels of odd order harmonics vs. similarly low levels of even order harmonics, correlated noise vs uncorrelated noise, jitter effects measured in picoseconds -- all things that immediately come to mind. There are many, many more.

In fact, I would hazard a guess that Emile routinely addresses factors in his designs that occur below the level you postulate as un-hearable.

Steve Z

 

[kennyb123]

In fact, I would hazard a guess that Emile routinely addresses factors in his designs that occur below the level you postulate as un-hearable.

His mistake is failing to consider alternate explanations around why bit depth beyond 16 bit can benefit audio reproduction.

 

[godofwealth]

You seem to be inferring that because speakers have such high levels of distortion that they serve as some kind of brick wall filter that keeps us from hearing effects that lay below their linear resolution, measured in bits.

I think we can demonstrate that is not true for any number of audio parameters that occur at extremely low levels, both analog and digital and yet are audible via our imperfect speakers -- extremely low levels of odd order harmonics vs. similarly low levels of even order harmonics, correlated noise vs uncorrelated noise, jitter effects measured in picoseconds -- all things that immediately come to mind. There are many, many more.

In fact, I would hazard a guess that Emile routinely addresses factors in his designs that occur below the level you postulate as un-hearable.

Steve Z

Absolutely I am, but I’m making a far stronger statement that might shock you, but it’s an experiment that any of you can try in your listening room. Engineers and scientists like to abstract complex systems by drawing boxes around them and looking at the “transfer function“ of the overall system. You can do this at the level of a single component — a resistor, a capacitor, a transistor, a vacuum tube — or a Hi-Fi component — a DAC or a preamp or a speaker or, in fact, the entire reproduction chain from the original album in whatever format to the sound it produces in your listening room. In each case, we ask a basic question: how linear is this transfer function?

So,, here’s the experiment any of you can do. Pick your favorite album, choose whatever components you want — analog or digital, solid state or tube amplification, dynamic speaker or stat or horn speaker etc. Now, record the actual sound your speaker makes in your listening room when playing back this album.

Now comes the important part. The second time around, play back the recording of the sound your speaker made of the original album, once again through the same reproduction chain. Repeat a dozen or more times. What do you think you’ll end up with?

To show you how nonlinear speakers and rooms are, we need a comparison. For example, if you take a dozen amplifiers (or preamplifiers), and daisy chain them, but level matched each amplifier in the chain to ensure that the chain of 12 amplifiers produced the same volume as a single amplifier. Quad, the British company did this test many years ago, showing that no listener could tell apart one of their Quad 606 amplifiers from a dozen of them hooked up together. The point here is not to critique this test, it’s to make a contrast with our original experiment. Amplifiers are very linear in comparison to speakers and rooms.

So, what happens if you do the original experiment? Well, amazingly enough, after a dozen repetitions of recording your speakers playing back an album, and playing the last recorded sound each time, you get ABSOLUTE NOISE! The signal to noise ratio is 0 dB! It does not matter if you played a jazz album, a classical album, a rock album, a choral piece or an opera. A dozen repetitions through the entire system representing your entire reproduction chain, you get complete noise at the end. Why? Because the room and speakers are so nonlinear, they totally dominate. You get what sounds like FM noise and no music whatsoever. Contrast that with the amplifier experiment where listeners could not tell apart one Quad 606 amplifier from 12!

The late Amat Gopal Bose, a brilliant scientist and engineer at the Massachusetts Institute of Technology, who taught a legendary class on acoustics at MIT for several decades did this experiment for his students every year. Each time, the students were completely slack jawed listening to complete noise coming out after a dozen or more repetitions. It’s useful to keep this experiment in mind. Compared to all the other components, the room and the speakers are most nonlinear ones by several orders of magnitude (like many millions of times more nonlinear). .

I will not enter into a debate about 16-bit or 24-bit or 32-bit. I’ll just say plenty of reviewers and contributors to this forum think vinyl sounds better than digital. That should alone tell you a lot. Speakers and rooms are so inherently nonlinear and noisy, it dominates. everything else. If you want to truly advance audio reproduction, you might want to focus on the most nonlinear components in the chain.

 

[RUIN*ish]

Absolutely I am, but I’m making a far stronger statement that might shock you, but it’s an experiment that any of you can try in your listening room. Engineers and scientists like to abstract complex systems by drawing boxes around them and looking at the “transfer function“ of the overall system. You can do this at the level of a single component — a resistor, a capacitor, a transistor, a vacuum tube — or a Hi-Fi component — a DAC or a preamp or a speaker or, in fact, the entire reproduction chain from the original album in whatever format to the sound it produces in your listening room. In each case, we ask a basic question: how linear is this transfer function?

So,, here’s the experiment any of you can do. Pick your favorite album, choose whatever components you want — analog or digital, solid state or tube amplification, dynamic speaker or stat or horn speaker etc. Now, record the actual sound your speaker makes in your listening room when playing back this album.

Now comes the important part. The second time around, play back the recording of the sound your speaker made of the original album, once again through the same reproduction chain. Repeat a dozen or more times. What do you think you’ll end up with?

To show you how nonlinear speakers and rooms are, we need a comparison. For example, if you take a dozen amplifiers (or preamplifiers), and daisy chain them, but level matched each amplifier in the chain to ensure that the chain of 12 amplifiers produced the same volume as a single amplifier. Quad, the British company did this test many years ago, showing that no listener could tell apart one of their Quad 606 amplifiers from a dozen of them hooked up together. The point here is not to critique this test, it’s to make a contrast with our original experiment. Amplifiers are very linear in comparison to speakers and rooms.

So, what happens if you do the original experiment? Well, amazingly enough, after a dozen repetitions of recording your speakers playing back an album, and playing the last recorded sound each time, you get ABSOLUTE NOISE! The signal to noise ratio is 0 dB! It does not matter if you played a jazz album, a classical album, a rock album, a choral piece or an opera. A dozen repetitions through the entire system representing your entire reproduction chain, you get complete noise at the end. Why? Because the room and speakers are so nonlinear, they totally dominate. You get what sounds like FM noise and no music whatsoever. Contrast that with the amplifier experiment where listeners could not tell apart one Quad 606 amplifier from 12!

The late Amat Gopal Bose, a brilliant scientist and engineer at the Massachusetts Institute of Technology, who taught a legendary class on acoustics at MIT for several decades did this experiment for his students every year. Each time, the students were completely slack jawed listening to complete noise coming out after a dozen or more repetitions. It’s useful to keep this experiment in mind. Compared to all the other components, the room and the speakers are most nonlinear ones by several orders of magnitude (like many millions of times more nonlinear). .

I will not enter into a debate about 16-bit or 24-bit or 32-bit. I’ll just say plenty of reviewers and contributors to this forum think vinyl sounds better than digital. That should alone tell you a lot. Speakers and rooms are so inherently nonlinear and noisy, it dominates. everything else. If you want to truly advance audio reproduction, you might want to focus on the most nonlinear components in the chain.

In my experience, if you don't take care or the room and mains it's an uphill battle. No matter how good your kit is..

 

[Bobvin]

Measurements of Quads done a long time ago by Hi Fi Choice (UK) and more recently in Stereophile show that Quads are indeed very linear as most loudspeakers go. Their distortion at moderate volumes is around (0.1%), which is about -70 dB. So, you get about 12 bits of linearity. But only at moderate volumes. Above 90 dB or so, Quads will start to distort before shutting down. Klipschorns and La Scalas are quite linear as well with about 0.1% distortion at higher volumes. Notice both of these are far from resolving even a 16-bit Redbook CD recording, let alone 24-bit high res recordings. A speaker that can resolve all 16 bits of a CD is science fiction still. As regards 24-bit high res, that’s a pipe dream. As for 32 bits? You’d need to maintain linearity down to -192 dB. Even if you spent a billion dollars, I doubt you could manage that in any hifi component. Maybe in a multibillion dollar physics lab with superconducting wires and temperatures close to absolute zero Kelvin. In the real world with what we can put in an actual living room? Not a chance in hell.

How exactly is this relevant in the Taiko thread?

 

[wil]

How exactly is this relevant in the Taiko thread?

It’s interesting though. Perhaps @godofwealth can start a thread?

 

[Blackmorec]

Absolutely I am, but I’m making a far stronger statement that might shock you, but it’s an experiment that any of you can try in your listening room. Engineers and scientists like to abstract complex systems by drawing boxes around them and looking at the “transfer function“ of the overall system. You can do this at the level of a single component — a resistor, a capacitor, a transistor, a vacuum tube — or a Hi-Fi component — a DAC or a preamp or a speaker or, in fact, the entire reproduction chain from the original album in whatever format to the sound it produces in your listening room. In each case, we ask a basic question: how linear is this transfer function?

So,, here’s the experiment any of you can do. Pick your favorite album, choose whatever components you want — analog or digital, solid state or tube amplification, dynamic speaker or stat or horn speaker etc. Now, record the actual sound your speaker makes in your listening room when playing back this album.

Now comes the important part. The second time around, play back the recording of the sound your speaker made of the original album, once again through the same reproduction chain. Repeat a dozen or more times. What do you think you’ll end up with?

To show you how nonlinear speakers and rooms are, we need a comparison. For example, if you take a dozen amplifiers (or preamplifiers), and daisy chain them, but level matched each amplifier in the chain to ensure that the chain of 12 amplifiers produced the same volume as a single amplifier. Quad, the British company did this test many years ago, showing that no listener could tell apart one of their Quad 606 amplifiers from a dozen of them hooked up together. The point here is not to critique this test, it’s to make a contrast with our original experiment. Amplifiers are very linear in comparison to speakers and rooms.

So, what happens if you do the original experiment? Well, amazingly enough, after a dozen repetitions of recording your speakers playing back an album, and playing the last recorded sound each time, you get ABSOLUTE NOISE! The signal to noise ratio is 0 dB! It does not matter if you played a jazz album, a classical album, a rock album, a choral piece or an opera. A dozen repetitions through the entire system representing your entire reproduction chain, you get complete noise at the end. Why? Because the room and speakers are so nonlinear, they totally dominate. You get what sounds like FM noise and no music whatsoever. Contrast that with the amplifier experiment where listeners could not tell apart one Quad 606 amplifier from 12!

The late Amat Gopal Bose, a brilliant scientist and engineer at the Massachusetts Institute of Technology, who taught a legendary class on acoustics at MIT for several decades did this experiment for his students every year. Each time, the students were completely slack jawed listening to complete noise coming out after a dozen or more repetitions. It’s useful to keep this experiment in mind. Compared to all the other components, the room and the speakers are most nonlinear ones by several orders of magnitude (like many millions of times more nonlinear). .

I will not enter into a debate about 16-bit or 24-bit or 32-bit. I’ll just say plenty of reviewers and contributors to this forum think vinyl sounds better than digital. That should alone tell you a lot. Speakers and rooms are so inherently nonlinear and noisy, it dominates. everything else. If you want to truly advance audio reproduction, you might want to focus on the most nonlinear components in the chain.

Just to mention; most recordings are normalised to account for the reverb of an average room during replay. Adding a further 11 sets of room reflections in series is simply adding 11 sets of noise. I don’t think we need an experiment to prove that. Reflections of reflections of reflections of reflections of reflections of…………You get the picture. One set of reflections are required to sound good. 11 further sets not so much.

The biggest differences between amps and loudspeakers in the above experiment are almost entirely down to the physical differences between the way voltages and soundwaves behave. I think the late Amat Gopal Bose was taking advantage of his student’s ignorance.
I‘m going to make an unproven guess and say that I believe that if I conducted the same experiment in a perfectly treated room with perfectly linear loudspeakers I‘d still get the exact same result because I’m simply distorting the signal in a compounding way i.e the very first reflection is ok and desirable….its how sound waves behave. But when I replay that first signal, the first room reflection is replayed, and now there’s a second room reflection, added to the first thats already there. The addition of that second reflection is recorded as a distortion, which is to say bucket loads of wrong signal which now compounds another 10 times.

Now back to Taiko and all the good stuff going on.

 

[kennyb123]

I will not enter into a debate about 16-bit or 24-bit or 32-bit. I’ll just say plenty of reviewers and contributors to this forum think vinyl sounds better than digital. That should alone tell you a lot. Speakers and rooms are so inherently nonlinear and noisy, it dominates. everything else. If you want to truly advance audio reproduction, you might want to focus on the most nonlinear components in the chain.

It tells me that one can reach any conclusion they want if their arguments are based on logical fallacies. Just because our speakers are our most flawed component doesn’t mean that they aren’t capable of revealing a ton about our music. Of all the threads to suggest otherwise, this seems the oddest one to do it in.

 

[godofwealth]

How exactly is this relevant in the Taiko thread?

Good point. It’s relevant in reminding everyone that as good as the Taiko is, it can’t fix nonlinearities in rooms or speakers, unless they decide to include some form of DSP or room correction as part of their software. Some servers do include room correction, e.g., Lyngdorf, and incidentally, also pulse width digital amplification It’s probably not hard to support both of these on the Taiko since it has enough compute power. I’m not sure either of these is in their roadmap. But I hope they do include it. One has to think synergistically about the entire reproduction chain, not just the first step. We now have the technology to fix at least some of the huge room induced nonlinearities and speaker errors. Might as well include that as an option.

 

[godofwealth]

It tells me that one can reach any conclusion they want if their arguments are based on logical fallacies. Just because our speakers are our most flawed component doesn’t mean that they aren’t capable of revealing a ton about our music. Of all the threads to suggest otherwise, this seems the oddest one to do it in.

Sure, that’s a fair argument. But if the average speaker has distortion around -50dB, how much further down in resolution do you think you can hear? Another 60 dB into the noise? 100 dB into the noise? To fully resolve 24 bits, we need -144 dB. That’s 94 dB down in volume from where the average speaker is producing noise. Is human hearing that good at resolving signal buried in noise? All I’m doing is asking the question. It’s a thought experiment.

 

[jbrrp1]

At this point, please start your own thread to indulge in this with those that might wish to. You really have more than belabored it on this one where it isn't all that germane.

 

[godofwealth]

Agreed. Apologies for taking up the bandwidth in the Taiko thread!

 

[kennyb123]

Sure, that’s a fair argument. But if the average speaker has distortion around -50dB, how much further down in resolution do you think you can hear? Another 60 dB into the noise? 100 dB into the noise? To fully resolve 24 bits, we need -144 dB. That’s 94 dB down in volume from where the average speaker is producing noise. Is human hearing that good at resolving signal buried in noise? All I’m doing is asking the question. It’s a thought experiment.

That's a straw man argument as no one is saying the sound quality benefits from 24 bits (vs 16 bits) come from being able to fully resolve 24 bits. This isn't the place to have this conversation so that's all that I will say.

 

[bonzo75]

How exactly is this relevant in the Taiko thread?

GoW keeps writing passionate irrelevant essays in various threads with lots of theory and fallacies

 

[BaconBrain]

@Taiko Audio A brief couple of questions from the DIY Community:

- Would Taiko be willing to publish their recommended BIOS Settings for the Asus WS C621E Sage Mainboard? I feel that it would be a great guideline for individuals (like myself) who want to build their own Extreme replica.

- What is the approximate timeline for making the Taiko USB Card available to non-Extreme Owners?

Thanks in advance!

 

[Kingsrule]

@Taiko Audio A brief couple of questions from the DIY Community:

- Would Taiko be willing to publish their recommended BIOS Settings for the Asus WS C621E Sage Mainboard? I feel that it would be a great guideline for individuals (like myself) who want to build their own Extreme replica.

- What is the approximate timeline for making the Taiko USB Card available to non-Extreme Owners?

Thanks in advance!

That's big balls...
Really?? You want Taiko to give up their info so you can build a replica Extreme??

 

[Jeffy]

@Taiko Audio A brief couple of questions from the DIY Community:

- Would Taiko be willing to publish their recommended BIOS Settings for the Asus WS C621E Sage Mainboard? I feel that it would be a great guideline for individuals (like myself) who want to build their own Extreme replica.

- What is the approximate timeline for making the Taiko USB Card available to non-Extreme Owners?

Thanks in advance!

Yes and will Nordost send out free Oden 2 cables?

 

[DW101]

Good point. It’s relevant in reminding everyone that as good as the Taiko is, it can’t fix nonlinearities in rooms or speakers, unless they decide to include some form of DSP or room correction as part of their software. Some servers do include room correction, e.g., Lyngdorf, and incidentally, also pulse width digital amplification It’s probably not hard to support both of these on the Taiko since it has enough compute power. I’m not sure either of these is in their roadmap. But I hope they do include it. One has to think synergistically about the entire reproduction chain, not just the first step. We now have the technology to fix at least some of the huge room induced nonlinearities and speaker errors. Might as well include that as an option.

I understand the desire the treat this holistically, but my own experience of room correction is that I got a far better result treating the room through physical means than via DSP software.

I had a small listening room and got some (large-ish) B&W 800 D3 loudspeakers and wanted to be able to control any potential excess bass in the room, so I bought and used Dirac room correction software. It did indeed correct the frequency response of the transfer function from music signal to microphone measurements at the listening position pretty well perfectly. However ultimately I found although it was a very flat measured frequency response it also had a '2D' flat stereo image and an unengaging musical presentation.

I then went back to basics and treated the room, no DSP correction, and ended up with good 3D imaging and musical engagement, with a less perfect but pretty good measured frequency response. I concluded that real-time DSP room correction might be a good idea for 'consumer' type systems, but not the type of high-end audiophile equipment most people on this forum appear to have.

I could be proved wrong, but based on that experience, and the fact that the additional real-time processing the Extreme would have to do to apply the correction via digital filtering would seem to be a 180 degree about face from what I understand Taiko's consistent approach of having a highly capable compute platform that is being asked to do decreasing amounts of CPU operations, resulting in corresponding sound quality uplifts that I've heard first in TAS and then XDMS, I personally wouldn't be too interested in using any potential Extreme-based software room correction capabilities.

(As probably irrelevant background information I also have some professional experience with similar DSP techniques having implemented active noise reduction on military equipment).

 

[BaconBrain]

That's big balls...
Really?? You want Taiko to give up their info so you can build a replica Extreme??

No offense intended.

Taiko does make a lot of it’s products available to the DIY community and if Taiko does not want to reveal any of it’s intellectual property then I certainly understand and respect that. I was only looking for a recommendation to be used as a guideline.

 

[ticktock]

No offense intended.

Taiko does make a lot of it’s products available to the DIY community and if Taiko does not want to reveal any of it’s intellectual property then I certainly understand and respect that. I was only looking for a recommendation to be used as a guideline.

As I recall, Taiko has been quite clear that the proprietary USB card will not be made available to the broader DIY community. I believe that's true of the XDMS playback software too. Much earlier, I had considered the DIY route but eventually concluded it would not get me close enough to a genuine Extreme to make the dollar savings worthwhile.