Stax Hi-Rez Revelation
Jul 18, 2002 at 1:51 PM Thread Starter Post #1 of 18

Matt

Are there any women on this board?
Joined
Jun 25, 2001
Posts
973
Likes
12
Until a few days ago, I was pretty much resigned to the fact that my Stax Classic System II didn't "do" digital. However, this verdict has seen a complete reversal in light of recent listening.

I have a DVD-A transport which I have been using for Redbook through an unmodded Art DI/O. While sifting around in some packed-away posessions the other day, I found a DVD-A sampler which I had, some months back, semi-pilfered from Best Buy.

The Stax system had always had this sort of "tension" with the digital source that I couldn't exactly put a finger on, but which made it substantially less pleasing to listen to than my vinyl rig, which sounded amazing (by the way, Stax and vinyl are made for each other).

Well, upon sticking in that Sampler disc (and making my way through the badly-recorded first track), I sat stunned as some big band played "Satin Doll"...right in my room! Now, the production values on this recording were good, but I have plenty of Redbook discs who's values are just as good. The element which placed the overall experience of this recording head-and-shoulders above the others was the incredibly relaxed-yet-extended treble ("relaxed" meaning all of that previously-mentioned perceptual "tension" just melted away). The timbral fidelity and, therefore, the sensation of reality, was tremendous. (The DI/O handled the 24/96 signal beautifully).

Another thing that shot out at me was the newly extended and refined dynamic range afforded by the 24 bit signal. There is a straightforward rock track which, along with it's real-sounding, crunchy distorted guitars, had this dynamic slam and a real sense of dynamic contrast that I've never heard from Redbook, ever. It was, again, very realistic.

The experience was very nearly vinyl-like.

It seems to me, then, that what the Stax's have a problem with is the inevitable signal degredation that 16/44.1 PCM imposes. I was thinking about this just last night: the highest range of just-born human baby hearing is approximately 20,000 Hz. The average adult caps off at about 14,000 - 15,000.

A 44.1/16 signal has as it's highest *representable* frequency 22,050 Hz. This "representation," though, is a severly broken-down sine wave, one turned into a sawtooth wave. Anyone who's heard the difference between a sine and a sawtooth wave knows what happens here, sonically. It's a change, a degredation, a digitally-imposed loss of fidelity.

However, since 22,050 Hz is clearly outside the range of human hearing, one might be apt to think "oh, well, that really means nothing." But does it?

What is one octave below 22,050 Hz? It's 11,025 Hz, a figure *well* within the regular range of human hearing. That's only *one* measly octave below, since, as indicated above, the frequencies of octaves increases exponentially. In musical terms, one octave ain't much. Most of what we hear in the mids and especially lows is very crowded-together, frequency-wise, and the gaps increase dramatically as you move up the musical scale.

So, if we have a single sample at the peak and single sample at the trough of the wave at 22,050 Hz, we have only double that for the 11,025 Hz signal, adding one sample half way between the two on the way down from the peak (the zero mark) and one half way back up to the peak (again, that darn zero mark). That is still a very boxy, very angular and very degraded signal and, again, well within most everyone's hearing range.

The Stax's are good at communicating whatever is sent to them, without glossing it over, so they give a full, psychologically-perceptible sense of that distortion that comes from 16/44.1.

Anyhow, I am really, really, really pleased and I now feel like I have gotten way more than my money's worth with the Classic System II.

- Sir Mister Matt
 
Jul 18, 2002 at 5:31 PM Post #2 of 18
Hi Matt (Sir Mister...)

With my headphones I have no problem at all with redbook CDs (although with my loudspeakers I wish the slightly «digital» overtone away...), even my electrostatics feel comfortable with them. Headphones are less critical in terms of the source characteristics than loudspeakers, IMO. And it may be the merit of my Theta DAC, too, which I appreciate a lot.

I still don't seriously know the sound of SACD and DVD-A, but you're certainly right with most of your objections against the CD format. In fact, the sampling rate is unsufficient for an adequate reproduction of high frequencies, as shown in the schematical graphs below:

sinuskurven.jpg


Just a bit more than 1 sample per half-wave at 20 kHz (averaged) and just a bit more than 1½ samples at 14 kHz are a really low resolution. The result is an amplitude modulation which reaches deep down into the audio frequency range. But common DACs don't reveal this behaviour. The reason: it's exactly this effect which is responsible for a treble roll-off when measuring the sine wave frequency response. To compensate it, an increase of the high frequencies is standard, followed by a sharp (anti-aliasing) low pass filter. This filter causes a ringing at around 21 kHz. Of course, 21 kHz are not perceivable. But the resonance bandwidth ranges into the audible frequency range, maybe down to 10 kHz. And those resonance effects can be perceived. It's funny: this filter resonance makes the amplitude modulation disappear – it just gets smeared! (Of course that's nothing that really contributes to a natural sound!) But not so in DACs which use «Spline» filters: those have a soft rolloff which make the frequency response decrease as of 10 kHz and reach –3 dB at 20 kHz. By the way: this is the real frequency response after D/A conversion without the treble compensation!

It is to say that the aforementioned amplitude modulation contains frequencies (fourier components) which are unharmonic to the reproduced tones. Maybe they're a further reason why the CD format has the reputation to sound «digitally»... So I look forward to the new audio formats, but am very undecided which one I should prefer. In addition, the musical bandwidth is still very narrow on both systems...

smily_headphones1.gif
JaZZ
 
Jul 18, 2002 at 5:57 PM Post #3 of 18
Your reasoning is wrong. Those graphs show sine waves that are linearly interpolated, which is going to do all sorts of bad things. The first graph at 21.5 kHz is especially wrong as you can't sample that frequency at 44.1 kHz without aliasing. Spline filters are not correct either. The only correct reconstruction filter is one that complements the original sampling method, which is point-sampling, and therefore requires sinc interpolation.

A sine wave near the Nyquist rate appears to be a sawtooth only if you use the wrong reconstruction filter on it. It will be a sine wave if you use a sinc interpolator (which corresponds to a brickwall low-pass filter).

--Andre
 
Jul 18, 2002 at 6:29 PM Post #4 of 18
Andre...

...your
reasoning is wrong! 21.5 kHz are definitely satisfying the Nyquist theorem (sampling frequency at least twice the reproduction frequency); the borderline is marked by 22.05 kHz (the half of 44.1 kHz). The graphs are defined as «schematic», they just link the sampling points, and of course they are meant to be low-pass-filtered subsequently (the german word «ungefiltert» means unfiltered), which means the «curves» would actually be curved (rounded) afterwards. The filtering is mentioned in my text, btw, and nevertheless a lot of «bad things» do really happen! As mentioned.

smily_headphones1.gif
JaZZ
 
Jul 18, 2002 at 9:10 PM Post #5 of 18
JaZZ, the graphs are completely misleading, for the reasons AndreYew pointed out. After applying a sinc filter, you would get a perfect reconstruction of the original signal. It's meaningless to show an interpolated version of the signal; this in no way reflects the final output of the DAC.

Ringing is an artifact of a particular low-pass filter implementation, not of the PCM sampling technique.
 
Jul 18, 2002 at 9:57 PM Post #6 of 18
MirandaX...

Quote:

«...the graphs are completely misleading...»


Maybe... of course they need interpretation.

Quote:

«After applying a sinc filter, you would get a perfect reconstruction of the original signal.»


I don't know what a «sinc filter» is – but either way the signal can't be adequately reconstructed – unless you are content with sine waves of arbitrary length or shape... Either the amplitude modulation is conserved (to be measured in DACs with Spline filter as Wadia) or the ringing will level it to a continuous sine curve (in common CD players) – at the price of a corrupt transient response.

Quote:

«Ringing is an artifact of a particular low-pass filter implementation, not of the PCM sampling technique.»


Yes, ringing only appears after the filtering – but filtering is already done before the recording, to prevent aliasing! So a part of the ringing can originate from the recording process. And the rest is emerging after DAC, during the indispensable (analog or digital) low-pass filtering in the CD player. Unless you use a Spline filter... but then you get the amplitude modulation...
eek.gif


Please explain how you figure the reconstruction from the given sampling points (assumed they're drawn accurately!) and/or an accurate high-frequency transient response provided the ringing is filter-inherent.

smily_headphones1.gif
JaZZ
 
Jul 18, 2002 at 10:32 PM Post #7 of 18
Jazz,

You are correct about 21.5 kHz. I don't know what I was thinking when I wrote that, as the Nyquist rate is at 22.05 kHz. I agree with MirandaX in that the graphs are very deceptive when linearly interpolated the way they are --- lots of people here have been confused by similar graphs appearing in Jan Meier's Analoguer article.

"I don't know what a «sinc filter» is – but either way the signal can't be adequately reconstructed – unless you are content with sine waves of arbitrary length or shape... Either the amplitude modulation is conserved (to be measured in DACs with Spline filter as Wadia) or the ringing will level it to a continuous sine curve (in common CD players) – at the price of a corrupt transient response."

What does this mean? And how is a transient corrupted that's meaningful in an audible way, and not only a visual way? The Wadia system is incapable of passing more than 1 sine wave without significant intermodulation distortion.

"Please explain how you figure the reconstruction from the given sampling points (assumed they're drawn accurately!) and/or an accurate high-frequency transient response provided the ringing is filter-inherent."

High frequency transient response exists only if you have a transient. The graphs you show were dealing with sine wave reconstruction.

--Andre
 
Jul 18, 2002 at 10:42 PM Post #8 of 18
Matt,
I had a similar experience with my Audio Technica W11R compared to Senn HD600/Clou Red. The W11R was sensitive to source. Vinyl sounded good and better than on the Senn. CD was another matter. The few HDCDs I have sounded more similar to vinyl but had a problem with treble on many CDs, but less on the best recordings.
I can't make any definitive conclusion about this without comparing with more components, things that I don't have now. It could be that these Japanese headphones are too detailed and over-emphasize deficiencies in the recording - some kind of interaction between deficiencies in the recording and the phone. Could also be that they are demanding on the source becasue they are more revealing. I also suspect that my old MOH headhone amp is not optimal in my configuration.
I also suspect the CD format. Some arguments for it are derived from mathematical theory. At the same time, deficits in common implementations are often acknowledged. I don't have the time to dig deep into this, mastering the theory and studying the implematations.
From a practical point of view, might it be simpler to make something really good with more bits and higher sampling rates?
 
Jul 18, 2002 at 11:30 PM Post #9 of 18
Andre...

Don't attach so much importance to my «linear interpolation» between the sampling points. The lines serve only for the clearness – to show the sample positions and amplitudes. They're not meant to demonstrate the final signal shape after D/A conversion. But with somewhat phantasy and intuition one can envision how the signal will look after low-pass filtering, i.e. rounding of the angles. And there's apparently no chance to retouch the amplitude modulation which results from the meagre sampling rate – 1-1½ samples per half-wave from 14 to 20 kHz.

But the ringing actually does it – though in an evil way. You wrote it's just visible, not audible... that's not true. The ringing also appears in the audible range: 14 or 16 kHz sine bursts show a clear ringing (with the signal frequency) – a result of the extended bandwidth of the (21 kHz) filter resonance. In the same way the amplitude modulation can be measured with DACs without treble compensation and resulting overly ringing. The Wadia's intermodulation tendency (? – never heard before...) may be a consecution of that, but its audible qualities speak for the concept anyway.

Quote:

«High frequency transient response exists only if you have a transient. The graphs you show were dealing with sine wave reconstruction.»


The context is that: it's easy for a DAC to «create» (not «reproduce»!) a sine wave – just sufficing resonant decay, and a proper sine wave is made! But the CD player's real job is to reproduce music – i.e. particularly transients!

smily_headphones1.gif
JaZZ
 
Jul 19, 2002 at 1:55 AM Post #10 of 18
...yeah, I see what you're saying.

I posted this because I had been resigned to the "fact" that the mid-end Stax's had that "metallic overlay" or "etch" in the treble. This latest experience has proven that simply wrong...or, if not completely wrong, very insignificant with hi-rez and I am extremely pleased about it.

I feel the Stax Classic System II represents an awesome, awesome value if you 1.) buy from EIFL Corporation and 2.) use it with a decent vinyl source and/or a decent DVD-A/SACD source. Excellent sound, very euphonic-yet-utterly-detailed in a way that surpasses Etys (and I love me some Etys!) I must say, though, that CD's definitely suck on them, because all of their flaws are laid right out in front of you, with no sugar-coating.

- Matt
 
Jul 19, 2002 at 2:53 AM Post #11 of 18
Matt,

I am close to your conclusion. It is amazing that we had the same experience on quite different components. I regard it as most probable that the CD format sucks! (to be simplistic). The new 24/96 DACs for CD, some with additional processing to enhance performance, also point to this.
But I want more experience before making a final conclusion.
 
Jul 19, 2002 at 3:00 AM Post #12 of 18
Quote:

Originally posted by Matt


I feel the Stax Classic System II represents an awesome, awesome value if you ... use it with a decent vinyl source and/or a decent DVD-A/SACD source. Excellent sound, very euphonic-yet-utterly-detailed ... I must say, though, that CD's definitely suck on them, because all of their flaws are laid right out in front of you, with no sugar-coating.

- Matt


I totally agree with you ... the problem with Stax electrostatics is that they are just too revealing and need a decent source to sound good. The electrets are not as revealing and are much more forgiving of poorer sources.

That's why I use my pair of Stax Electrets (SR-30) for bedtime listening connected to a JVC Combo (exclusively for CD). My other pair of Stax (Lambda Signature + T1 amp) is reserved for listening to vinyl and is connected to my tube pre-amp in the living room.

Just last Sunday I spent the evening listening to Proprius' Jazz at the Pawnshop (Vols. 1 and 2) and MFSL's The Power and the Majesty on my vinyl setup. Only after then did I realize how good these albums could (and should) sound.
 
Jul 19, 2002 at 5:51 PM Post #13 of 18
Jazz,

"Don't attach so much importance to my «linear interpolation» between the sampling points. The lines serve only for the clearness – to show the sample positions and amplitudes. They're not meant to demonstrate the final signal shape after D/A conversion. But with somewhat phantasy and intuition one can envision how the signal will look after low-pass filtering, i.e. rounding of the angles. And there's apparently no chance to retouch the amplitude modulation which results from the meagre sampling rate – 1-1½ samples per half-wave from 14 to 20 kHz."

The problem is that linear interpolation is the source of the amplitude modulation. If the signal were filtered with a standard brickwall filter, there would be no AM at all.

"But the ringing actually does it – though in an evil way. You wrote it's just visible, not audible... that's not true. The ringing also appears in the audible range: 14 or 16 kHz sine bursts show a clear ringing (with the signal frequency) – a result of the extended bandwidth of the (21 kHz) filter resonance."

But what is the frequency of this ringing?

"The context is that: it's easy for a DAC to «create» (not «reproduce»!) a sine wave – just sufficing resonant decay, and a proper sine wave is made! But the CD player's real job is to reproduce music – i.e. particularly transients! "

I don't think this is true. If I feed in a zero signal, I don't get any ringing. The DAC does not create a sine wave --- it recreates it. Perhaps the reason you dislike CD sound is something else.

--Andre
 
Jul 19, 2002 at 7:11 PM Post #14 of 18
...are saying may very well be perfectly correct, but there is, in my experience, a subjective but perceptually very quantifiable increase in sonic realism when one goes "hi-rez." It's definitely there.

- Matt
 
Jul 19, 2002 at 10:58 PM Post #15 of 18
(Kelly digs out his toldjaso hat.)

Matt
Remember when I first heard the Classic II and posted about how great I thought they were?

Remember how I said I wouldn't buy them yet because I wasn't happy with my source? (The same source you have now--the ART DI/O unmodded at the time.)

Remember how I told you that source was why I felt you weren't getting the most out of some of the headphones you were auditioning?

Ok, just checking.
smily_headphones1.gif
 

Users who are viewing this thread

Back
Top