MQA is Vaporware

[mansr]

5 minutes ago, Jud said:

Yep, I realize this.  Just that as mentioned I've seen (by more folks than ESS) "time domain artifacts" used to describe ringing, and "frequency domain artifacts" used to describe aliasing/imaging.  Thinking of time and frequency domain performance as having to be balanced also (for me at least) helps get at the concept of conjugate variables, where optimization of one necessarily means less optimization of the other.

I see what you're getting at, but that's just not how it works. There is no such thing as time domain performance or frequency domain performance. The time and frequency domains are merely different ways of looking at the same thing, the signal.

 

I think much confusion stems from the fact that the impulse response, while fully characterising a system (its Fourier transform is the frequency response), isn't very intuitive to look at. It's tempting to think that a more impulse-like appearance (steep slopes, little ringing, etc) is better when in actuality it is less accurate.

[crenca]

8 minutes ago, mansr said:

I see what you're getting at, but that's just not how it works. There is no such thing as time domain performance or frequency domain performance. The time and frequency domains are merely different ways of looking at the same thing, the signal.

 

I think much confusion stems from the fact that the impulse response, while fully characterising a system (its Fourier transform is the frequency response), isn't very intuitive to look at. It's tempting to think that a more impulse-like appearance (steep slopes, little ringing, etc) is better when in actuality it is less accurate.

 

Masnr,

 

I think I am following you but could it not be argued that when you compare two signals - such as a (largely hypothetical/abstract) "original signal" (i.e. the live performance) with a signal that exists somewhere along the playback chain (e.g.  the analogue output of a DAC or an amp) that the difference between the two can be characterized in the way that is often done and to which Jud is pointing to?

 

Whatever the answer, this aspect is hard to wrap the mind around - particularly for those of us who are not steeped in engineering.  It is obvious to me that it also is something that is often unintentionally or intentionally  obscured and used by all sorts in "the industry" in a kind of pseudo-technical language in an attempt to sell their product...

[Charles Hansen]

5 hours ago, mansr said:

It's not that simple. An ideal sinc filter has a perfect cut-off with no ringing or aliasing/imaging. Unfortunately, it is also infinitely long and thus unpractical.

 

Hi Mansr,

 

Actually that is not correct. An ideal sinc filter would be infinitely long and thus have infinitely long pre- and post-ringing.

 

EDIT: That is precisely why one must make tradeoffs between time-domain performance and frequency domain performance. An ideal sinc filter would have perfect frequency response, but terrible impulse response.

 

The only way around the trade off is to raise the sampling rate to the point that the tradeoffs are negligible. In the case of audio, a factor of 10x (200kHz) is sufficient, and192 kHz is close enough to that to get essentially perfect performance in both domains.

 

Hope this helps,

Charles Hansen

[fung0]

20 hours ago, Charles Hansen said:

Hello Fung0,

 

While I agree with your very well articulated premises for most publication. I am fairly certain there are exception to the above quoted sentence.

 

I know of at least one US print magazine and one US webzine that allegedly only review equipment if you advertise. (They may occasionally break their own rule, so as not to make it too obvious.) Advertisers are allegedly guaranteed good reviews, and the more valuable the ad contract (size and frequency of placements, along with contract duration) allegedly the better the reviews will be. One publication allegedly took a loudspeaker company from start-up mode to major player  within 3 years, almost single-handed. One publication allegedly will sell cover shots to the highest bidder. One publication allegedly  will write lengthy positive reviews in exchange for non-financial incentives such as all-expenses-paid luxury vacation (which can be easily disguised as "travel expenses/reimbursements").

 

As you correctly note, the product being pimped must meet a certain level of performance (at least in the writer's mind) or else it would be too obvious and the reviewer/publication would lose credibility.. You will never see a mediocre product promoted in this way. But "sweetening the pot" can result in reviews that are more praiseworthy than would they would otherwise receive.

 

It's also well known that some manufacturers allegedly will not submit products for review to certain magazines. I think even Magnepan publicly acknowledges that they will not submit review samples to Stereophile, due to the fact that they do not measure ""well" under JA's loudspeaker test protocol. That is just one example and one reason.

 

Best, regards,

Charles Hansen

 

I've certainly seen occasional extremes of collusion between publishers and vendors, as you describe. But I think you'd agree that, in general, influence is more subtle. We know there are 'paid shills' out there. But 'useful idiots' are far more numerous.

 

Unfortunately, few publications these days truly support their writers in pushing for high standards of accuracy and honesty. (Doing proper research is expensive, for a start.) Even fewer publications make any pretense of fairly presenting all facets of every issue. About the best you can do is find one whose slant you can live with.

 

I've definitely encountered manufacturers who avoid bad reviews in the way you describe regarding Magnepan. Once you're on their 'black list,' you're lucky to get a polite answer to an email. You certainly never get interview access to company execs, and you needn't dream about getting product for review. The companies that have treated me that way have generally been ones selling products that were greatly over-valued, and vulnerable to having their bubble punctured.

[Charles Hansen]

6 minutes ago, fung0 said:

I've certainly seen occasional extremes of collusion between publishers and vendors, as you describe. But I think you'd agree that, in general, influence is more subtle. We know there are 'paid shills' out there. But 'useful idiots' are far more numerous.

 

Hello Fung0,

 

Actually, unfortunately I have to disagree. There are only two audio print magazines in the US, and 50% is more than "occasional" in my book. There are far more online publications, and there I would agree with you - one known example and one or two suspected examples out of many dozens of websites is more in line with what I would call "occasionally".

 

EDIT: I agree 100% with all of your other points, I find them to be extremely insightful. To have the wisdom for self-reflection like that is rare, and I commend you for sharing your insights with the world in general. It will even have the honest reviewers and publications thinking twice about what they are doing.

 

Cheers,

Charles Hansen

[rickca]

21 hours ago, Charles Hansen said:

I know of at least one US print magazine and one US webzine that allegedly only review equipment if you advertise. (They may occasionally break their own rule, so as not to make it too obvious.) Advertisers are allegedly guaranteed good reviews, and the more valuable the ad contract (size and frequency of placements, along with contract duration) allegedly the better the reviews will be. One publication allegedly took a loudspeaker company from start-up mode to major player  within 3 years, almost single-handed. One publication allegedly will sell cover shots to the highest bidder. One publication allegedly  will write lengthy positive reviews in exchange for non-financial incentives such as all-expenses-paid luxury vacation (which can be easily disguised as "travel expenses/reimbursements").

The MQA business development team must love these publications.

[Charles Hansen]

2 minutes ago, rickca said:

The MQA business development team must love these publications.

 

Yes, it makes part of their job very straightforward. And they are not stupid people at all - quite the opposite, they are extremely clever. Any digital engineer will admit that the basic idea of "folding" the low amplitude high-frequency information under the LSBs of the baseband data is a clever idea. While one may have legitimate disagreements about the audibility of the choices made by MQA, it doesn't detract from the fact that they are clever people. Clever enough to manipulate the perceptions of the (to quote from the MQA annual business report) "key opinion makers in the music industry and journalists". Read the full report from last year at:
 

https://beta.companieshouse.gov.uk/company/09123512/filing-history/MzE3MzU2NDUyM2FkaXF6a2N4/document?format=pdf&download=0

 

The easily bought-off webzines don't apparently register among MQA's "key opinion makers", hence the need for unfair and/or misleading "comparisons".

[fung0]

16 minutes ago, Charles Hansen said:

Hello Fung0,

 

Actually, unfortunately I have to disagree. There are only two audio print magazines in the US, and 50% is more than "occasional" in my book. There are far more online publications, and there I would agree with you - one known example and one or two suspected examples out of many dozens of websites is more in line with what I would call "occasionally".

 

EDIT: I agree 100% with all of your other points, I find them to be extremely insightful. To have the wisdom for self-reflection like that is rare, and I commend you for sharing your insights with the world in general. It will even have the honest reviewers and publications thinking twice about what they are doing.

 

Cheers,

Charles Hansen

 

Sounds like the last survivors in print are being forced to make increasingly horrible compromises...

 

Thanks for your kind EDIT. I think about this stuff a lot - the whole problem of disseminating useful information to an appallingly gullible public. We have endless information a few clicks away, but most people seem no have no ability (or inclination) to discriminate between genuine insight and total hogwash.

[Charles Hansen]

33 minutes ago, fung0 said:

Sounds like the last survivors in print are being forced to make increasingly horrible compromises...

 

Thanks for your kind EDIT. I think about this stuff a lot - the whole problem of disseminating useful information to an appallingly gullible public. We have endless information a few clicks away, but most people seem no have no ability (or inclination) to discriminate between genuine insight and total hogwash.

 

Hello Fung0,

 

On the first point I would say "choose" rather than "forced". Nobody that I know of is holding a gun to their head saying, "Take this money or I'll kill you."

 

On the second point, I think it boils down to personality types. One common measure is the Myers-Briggs personality inventory. Two of the quadrants comprise about 20% of the population, and it seems that those are the ones that question "authority". The other 80% are content to just go along with things, as long as they are "comfortable" in a physical sense. It doesn't apply to just audio, but to politics, religion, or any other type of belief system.

 

Cheers,

Charles Hansen

[Fokus]

6 hours ago, mansr said:

I think much confusion stems from the fact that the impulse response, while fully characterising a system, isn't very intuitive to look at.

 

The problem is that to many the impulse response looks very intuitive to look at. And as a result, many do look at it.

 

[Fokus]

2 hours ago, Charles Hansen said:

Actually that is not correct. An ideal sinc filter would be infinitely long and thus have infinitely long pre- and post-ringing.

 

Mans probably meant that, given a band-limited signal, the replay Sinc filter does not ring. Which is correct.

But it sidesteps the issue of how to obtain a band-limited signal at the production stage.

 

Quote

 

An ideal sinc filter would have perfect frequency response, but terrible impulse response.

 

 

Only if you are looking at the signal.

 

I repeat it once more: ringing due to excessive filter steepness is perfectly audible, but only if the ringing frequency falls within the subject's audible range, and the ringing envelope exceeds the envelope of the relevant cochlear filter for that frequency.

 

For most adults and CD rate or higher it is a non-issue.

 

[Fokus]

2 hours ago, Charles Hansen said:

Any digital engineer will admit that the basic idea of "folding" the low amplitude high-frequency information under the LSBs of the baseband data is a clever idea.

 

And not invented by anyone related to MQA:

 

http://www.aes.org/e-lib/browse.cfm?elib=10276

[Charles Hansen]

14 minutes ago, Fokus said:

 

1) Mans probably meant that, given a band-limited signal, the replay Sinc filter does not ring. Which is correct.

But it sidesteps the issue of how to obtain a band-limited signal at the production stage.

 

2) Only if you are looking at the signal.

 

3) I repeat it once more: ringing due to excessive filter steepness is perfectly audible, but only if the ringing frequency falls within the subject's audible range, and the ringing envelope exceeds the envelope of the relevant cochlear filter for that frequency.

 

4) For most adults and CD rate or higher it is a non-issue.

 

Hi Fokos,

 

I took the liberty of numbering your points for easy reply:

 

1) Exactly. The easiest way is to sample at (say) 10x higher than the highest frequency of interest. Then no filtering is needed (at least in the case of audio. But since the CD standard was set as soon as it was barely achievable for a commercial enterprise, the problem is that there are many musical instruments with musical energy above 20kHz. One solution would be to omit the anti-aliasing and reconstruction filters altogether and live withe the aliases. Plenty of people own "non-oversamping DACs and that demonstrate that strict adherence to sampling theory is not required for musically satisfying results. Think about all of the good times people had in the 60s and 70s listening to their favorite songs on the AM radio with a maximum bandwidth of only 5kHz. How important are all the numbers anyway. Even FM radio was limited to only 15kHz. Numbers and specifications have little relationship to the actual listening experience.

 

2) It depends on the system and the listener. I can hear differences in blind testing, as can many others. However, in general you are correct. It requires training and/or sensitivity to hear these differences.

 

3) See #2.

 

4) See #3.

 

4 minutes ago, Fokus said:

And not invented by anyone related to MQA:

 

http://www.aes.org/e-lib/browse.cfm?elib=10276

 

Nice find! Thanks for the link. This shows that nothing in MQA is new. Not the "folding", not the minimum-phase slow rolloff filters - just the hype and the attempt to extort money through licensing and royalties. I guess they will patent anything these days.

 

Cheers,

Charles Hansen

[Fokus]

1 hour ago, Charles Hansen said:

Nice find! Thanks for the link. This shows that nothing in MQA is new.

 

0) Not a find at all. It is clearly stated in one of the MQA patents. People discussing MQA should start reading these patents.

 

As for your replies to 2)-3)-4): sure you can hear things. But was all else scrupulously kept equal? Not easy, that.

[Charles Hansen]

2 hours ago, Fokus said:

As for your replies to 2)-3)-4): sure you can hear things. But was all else scrupulously kept equal? Not easy, that.

 

Hi Fokos,

 

My degree is in physics, and I've done real research in real labs on real projects. I am familiar with correct protocols and methods. My ability to hear many of these things only came with many decades of experience. Since this has been my job since 1985, I have more experience than most. As a manufacturer it is much easier to control the constants and ensure that everything is equal. For example when listening to different digital filters, all I have to do is mute the preamp, load new coefficients into the FPGA, and unmute the preamp. The DUT had a small toggle switch on the rear, to select between two sets of cofficients so we would load two sets of coefficients at once. I cannot reach the toggle switch from my wheelchair, so the other engineer would flip the switch for me. Nearly all of the tests were done blind. Sometimes we would load two new sets, but usually I would pick a winner and replace just the loser. I did this for 6 hours a day every day for 3 or 4 months and must have listened to many hundreds of sets of different filter coefficients, looking for patterns and trends. For example in every case (using mostly Redbook sources - high res sources were extremely rare back in 2009 - just a hand ful of DVD-Vs and DVD-As which were used to double-check) I found that regardless of the other filter parameters that minimum phase sounded better than linear phase. This was just one of eight or ten different filter parameters tested.

 

Obviously this type of thing is not available to a typical audiophile. All of the listening tests we perform are done with equal rigor.

 

Hope this helps,

Charles Hansen

[Fokus]

That's not what I meant. Ringing gets loaded with many sins, but many comparisons change at least two factors at the same time. When MP wins, is it because of the eliminated pre-ringing (which exists only at Fs/2), or because of the phase distortion (which extends down into the audible band)? When a less steep filter wins, is it because of the reduced ringing, because of the drooping treble, or because of the increased levels of imaging?

 

[mansr]

5 hours ago, Fokus said:

Mans probably meant that, given a band-limited signal, the replay Sinc filter does not ring. Which is correct.

That is exactly what I meant. An impulse is not band-limited.

 

4 hours ago, Charles Hansen said:

The easiest way is to sample at (say) 10x higher than the highest frequency of interest. Then no filtering is needed (at least in the case of audio).

No, regardless of the highest interesting frequency (20 kHz for audio), you need to sample at more than twice the highest present frequency. If you sample at 200 kHz without anti-alias filtering, and the input has a strong component at 399 kHz, it will alias to 1 kHz and ruin your recording. In practice, microphones have very little response beyond 50 kHz (most drop of well before that), so you can get away with sampling at 192 kHz, or even 96 kHz, using little or no filtering. Or you could if such ADCs existed. Audio ADCs are typically sigma-delta designs with a sample rate of 5.6 MHz or more followed by a digital downsampling filter.

 

4 hours ago, Charles Hansen said:

But since the CD standard was set as soon as it was barely achievable for a commercial enterprise, the problem is that there are many musical instruments with musical energy above 20kHz. One solution would be to omit the anti-aliasing and reconstruction filters altogether and live with the the aliases. Plenty of people own "non-oversampling DACs and that demonstrate that strict adherence to sampling theory is not required for musically satisfying results.

For playback, yes. The image frequencies are above 20 kHz and thus inaudible. Omitting the filters when recording would result in ultrasonic frequencies being aliased into the audible range. On playback, letting the image frequencies go unchecked is also a bad idea, even if they are inaudible. Firstly, their presence can cause audible intermodulation products in amps and speakers, and secondly, burdening the amp and speakers with frequencies you don't want and can't hear can impact their performance at the frequencies you do want. A little inaudible ringing above 20 kHz seems rather harmless by comparison.

[mansr]

7 hours ago, Charles Hansen said:

Any digital engineer will admit that the basic idea of "folding" the low amplitude high-frequency information under the LSBs of the baseband data is a clever idea.

I'd say it's not really all that clever. The only reason for such gymnastics is to send the encoded stream to the DAC over a standard PCM interface rather than simply decode it in software. The obvious reason to insist on this is to extract royalties from DAC manufacturers.

[Charles Hansen]

8 hours ago, Fokus said:

That's not what I meant. Ringing gets loaded with many sins, but many comparisons change at least two factors at the same time. When MP wins, is it because of the eliminated pre-ringing (which exists only at Fs/2), or because of the phase distortion (which extends down into the audible band)? When a less steep filter wins, is it because of the reduced ringing, because of the drooping treble, or because of the increased levels of imaging?

 

 

Both of your examples contradict yourself, as you maintain that any ringing (or aliasing close to Nyquist) is inaudible. If we ignore the internal self-inconsistency, your question amounts to "How many angels can dance on the head of a pin?" The difficulty is that all the parameters you note are inextricably linked at a given sampling frequency. That is why I noted that the only way around the problem is to increase the sampling frequency up to around 200 kHz or more.

 

Hope this helps,

Charles Hansen

[Charles Hansen]

7 hours ago, mansr said:

1) That is exactly what I meant. An impulse is not band-limited.

 

2) No, regardless of the highest interesting frequency (20 kHz for audio), you need to sample at more than twice the highest present frequency. If you sample at 200 kHz without anti-alias filtering, and the input has a strong component at 399 kHz, it will alias to 1 kHz and ruin your recording. In practice, microphones have very little response beyond 50 kHz (most drop of well before that), so you can get away with sampling at 192 kHz, or even 96 kHz, using little or no filtering. Or you could if such ADCs existed. Audio ADCs are typically sigma-delta designs with a sample rate of 5.6 MHz or more followed by a digital downsampling filter.

 

3) For playback, yes. The image frequencies are above 20 kHz and thus inaudible. Omitting the filters when recording would result in ultrasonic frequencies being aliased into the audible range. On playback, letting the image frequencies go unchecked is also a bad idea, even if they are inaudible. Firstly, their presence can cause audible intermodulation products in amps and speakers, and secondly, burdening the amp and speakers with frequencies you don't want and can't hear can impact their performance at the frequencies you do want. A little inaudible ringing above 20 kHz seems rather harmless by comparison.

 

1) Real music made with real instrument recorded with real microphones are not band-limited to 22.05kHz either.

 

2) I am talking about the real world, not some imaginary one where one deliberately injects a high-frequency, out-of-band signal into the audio signal. Also, it is easily possible to downsample delta-sigma ADC's using filters that don't ring and instead emulate the bandwidth restrictions of a purely analog chain.

 

3) There is no difference in aliasing between recording and playback. Almost always half-band filters are used because they only require half the number of taps and therefore are half the price. Half-band filters are always down only -6dB at the corner frequency (Nyquist), and when used as anti-aliasing filters will typically let through some spectral content between Nyquist and about 24kHz (for CD). When played back with no reconstruction filter, the subsequent aliasing will be greater than about 20kHz.

 

The problem is that putting a steep, sharp filter into the chain causes audible problems. Putting two steep, sharp filters in the chain compounds the audibility of the problems.

 

On the other hand if no anti-aliasing filter is used, then the aliasing will extend to lower frequencies. If we assume that MQA music spectral distributions of music are accurate (they are actually not as they do not include any percussion instruments - just one of the many ways that MQA goes out of their way to mislead) then the musical spectrum is down -50dBFS at Nyquist and falls at roughly --40dB/octave above that. Aliasing at 10kHz would be about -72dBFS, so even a CD-rate system with no filters would have a very small amount of aliasing below 10kHz. One would have to listen to such a system to see which was the more objectionable problem - the insertion of two steep, sharp filters in the chain, or the presence of low leves of aliasing in the top octave. Many have found that eliminating just one of the filters (reconstruction) makes for an audible improvement. I know of no experiments where the antialiasing filter has also been eliminated.

 

Thanks,

Charles Hansen

[mansr]

1 hour ago, Charles Hansen said:

1) Real music made with real instrument recorded with real microphones are not band-limited to 22.05kHz either.

True. The signal on a CD, however, is by definition. At this stage we can only take what we're given at face value and reconstruct the corresponding analogue signal as best we can.

 

1 hour ago, Charles Hansen said:

2) I am talking about the real world, not some imaginary one where one deliberately injects a high-frequency, out-of-band signal into the audio signal. Also, it is easily possible to downsample delta-sigma ADC's using filters that don't ring and instead emulate the bandwidth restrictions of a purely analog chain.

Sure, if the modulator noise is well separated from the highest audio frequency a filter with a transition band in the gap will have no ill effects. Some high-end recording interfaces achieve this, but they are not commonly used. Many recordings, even from audiophile labels, show rising noise levels already at 35 kHz.

 

1 hour ago, Charles Hansen said:

3) There is no difference in aliasing between recording and playback.

Huh? Aliasing during recording creates unwanted content within the audible range. In most cases, they are too low in amplitude to be heard beneath the real signal, but they are there nonetheless. Imaging during playback creates unwanted content above the audible range. It would be harmless if not for its potential to wreak havoc with downstream components.

 

1 hour ago, Charles Hansen said:

Almost always half-band filters are used because they only require half the number of taps and therefore are half the price. Half-band filters are always down only -6dB at the corner frequency (Nyquist), and when used as anti-aliasing filters will typically let through some spectral content between Nyquist and about 24kHz (for CD). When played back with no reconstruction filter, the subsequent aliasing will be greater than about 20kHz.

Aliases of 24 kHz only reach down to 20 kHz, which most of us can't hear anyway.

 

1 hour ago, Charles Hansen said:

The problem is that putting a steep, sharp filter into the chain causes audible problems. Putting two steep, sharp filters in the chain compounds the audibility of the problems.

True. The question is how much of a problem this actually is.

 

1 hour ago, Charles Hansen said:

On the other hand if no anti-aliasing filter is used, then the aliasing will extend to lower frequencies. If we assume that MQA music spectral distributions of music are accurate (they are actually not as they do not include any percussion instruments - just one of the many ways that MQA goes out of their way to mislead) then the musical spectrum is down -50dBFS at Nyquist and falls at roughly --40dB/octave above that. Aliasing at 10kHz would be about -72dBFS, so even a CD-rate system with no filters would have a very small amount of aliasing below 10kHz. One would have to listen to such a system to see which was the more objectionable problem - the insertion of two steep, sharp filters in the chain, or the presence of low leves of aliasing in the top octave. Many have found that eliminating just one of the filters (reconstruction) makes for an audible improvement. I know of no experiments where the antialiasing filter has also been eliminated.

Conducting such a test should be easy enough.

[Jud]

12 hours ago, Charles Hansen said:

For example in every case (using mostly Redbook sources - high res sources were extremely rare back in 2009 - just a hand ful of DVD-Vs and DVD-As which were used to double-check) I found that regardless of the other filter parameters that minimum phase sounded better than linear phase. This was just one of eight or ten different filter parameters tested.

 

Hi Charles - This raises a couple of topics of interest to me.

 

I find that in my system, with the Vandersteens, I tend to favor linear phase filters.  Perhaps 25+ years of owning Vandys (yes, I would have strongly considered Avalons had I been looking at that price range ) has gotten me accustomed to a particular type of imaging that comes with linear phase throughout the chain.  Or maybe I'm just accustomed to something in linear phase filters (having owned one of the early Theta DACs for quite a while).

 

You, as you say, have long experience listening to filters.  And I very much liked the QB-9 (listening at a friend's) and Pono Player (which I still own) which I assume both resulted from your listening process.

 

Now humans are very very bad at storing sounds in memory for more than a few seconds, so any time someone does listening comparisons, that's a rather fraught sort of thing.  We might well ask what really is being compared.

 

But one thing humans are really, really good at is pattern matching. We build up recognition of these patterns with long experience and/or training.  So we can legitimately say we may be able to recognize familiar patterns in sound.

 

When I recognize a familiar pattern of sound that for me includes linear phase filters, and for you includes minimum phase filters, as what sounds "right" to us, are we recognizing accuracy, some inborn preference, or simply what fits a familiar pattern?

 

And when listeners hear MQA or something else they think is new and great, are they recognizing greater accuracy, some inborn preference, or something that breaks out of the familiar pattern of what they've heard before?

[Charles Hansen]

Hi Mansr,

 

I think we're pretty much saying the same thing. The only area of potential disagreement is that you are implicitly saying that any noise above ~20kHz is inaudible with continuous sine waves and is therefore of zero consequence to human hearing. I disagree with that, based on my own direct experience.

 

At one time our top-of-the-line power amplifier was based completely on FETs - both J and lateral MOS devices. It had a  -3dB point of 100kHz. We made a new amplifier that had an output stage with an emitter-follower triple using complementary BJTs. (This is the so-called "T-Circuit" developed by Bart Locanthi and used by JBL and many others since.) When we listened to the first prototype (with a -3dB point of 170kHz, it sounded "shut in", "closed down", and "rolled off". By adjusting the compensation capacitors that keep the circuit stable with capacitive loads (unnecessary with lateral MOSFETs) such that the -3dB point was raised to 250kHz, the amp again sounded as open and extended as the previous amp did with only 100kHz bandwidth.

 

I have no explanation for this except to note that just because the ear/brain cannot hear continuous sine waves above ~20kHz does not mean that transients with waveforms steeper than that are not detectable by the ear/brain:

 

https://evolutionnews.org/2013/02/human_hearing_o/

 

Another area of disagreement is the audibility of aliasing. Clearly the higher frequency aliasing produced by only filtering one side of the record/replay chain should by itself produce less audible artifacts. But that leads to the necessity of using steep filters with sharp corners, which in my experience definitely produce more sonic damage that does high-frequency aliasing noise that is far below the background noise level of nearly all listening environments. I don't think there is any way to "prove" either standpoint is correct. There is such a wide variety of capability in listening capabilities due to years of intensive training (similar to skills of playing a musical instrument due to years of intensive training, that there will never be one "correct" answer. The best case would be to have a situation that applied to a particular percentage of the population. Otherwise we are left with the situation we currently have where many can hear the differences between (say) cables and many cannot.

 

But to insist that I cannot hear something just because you cannot also hear it is as silly as the doctor to tell me I am not in pain simply because he is not in pain. Or for me to tell a world class violinist that it is impossible to play Tchaikovsky's violin concerto simply because I cannot.

 

Hope this helps,

Charles Hansen

[Charles Hansen]

3 minutes ago, Jud said:

 

Hi Charles - This raises a couple of topics of interest to me.

 

I find that in my system, with the Vandersteens, I tend to favor linear phase filters.  Perhaps 25+ years of owning Vandys (yes, I would have strongly considered Avalons had I been looking at that price range ) has gotten me accustomed to a particular type of imaging that comes with linear phase throughout the chain.  Or maybe I'm just accustomed to something in linear phase filters (having owned one of the early Theta DACs for quite a while).

 

You, as you say, have long experience listening to filters.  And I very much liked the QB-9 (listening at a friend's) and Pono Player (which I still own) which I assume both resulted from your listening process.

 

Now humans are very very bad at storing sounds in memory for more than a few seconds, so any time someone does listening comparisons, that's a rather fraught sort of thing.  We might well ask what really is being compared.

 

But one thing humans are really, really good at is pattern matching. We build up recognition of these patterns with long experience and/or training.  So we can legitimately say we may be able to recognize familiar patterns in sound.

 

When I recognize a familiar pattern of sound that for me includes linear phase filters, and for you includes minimum phase filters, as what sounds "right" to us, are we recognizing accuracy, some inborn preference, or simply what fits a familiar pattern?

 

And when listeners hear MQA or something else they think is new and great, are they recognizing greater accuracy, some inborn preference, or something that breaks out of the familiar pattern of what they've heard before?

 

Hi Jud,

 

In loudspeakers, everything is a tradeoff. It is not like the case of digital audio where one can escape the tradeoffs by going to a high-enough sample rate. There are several drawbacks to the use of first-order crossovers as used by Vandersteen. His very latest Model 7 overcomes one of them and achieves pistonic motion from the drivers, which I consider to be a "first order" problem in loudspeakers, meaning that non-pistonic motion of the diaphragms adds horrible amounts of noise to the music signal that is far worse than anything that aliasing might cause.

 

Even when he fixed that problem (at a very high price - much higher than the Avalons I designed), there is still the problem that the drivers are 90º out-of-phase at the crossover point. That means (assuming the tweeter is above the woofer) that there will be an excess of energy at the crossover point that is aimed downward, typically toward the listener's feet. Depending on the crossover frequencies used, that will result in changes to the tonal balance in the room which are quite audible. The only way around that problem is to either used a D'Appolito configuration (as did the Dunlavy's and Duntech's) or a coaxial solution (not used by anybody that I know of with a first-order crossover).

 

What I have found is that in an odd way that the Vandersteens make better listening tools for an electronics manufacturer than a "better" speaker such as an Avalon or the TADs with beryllium cones that operate pistonically. Specifically the Vandys only "sing" when everything is just right. In contrast, when the electronics are really good, the sound will be amazing with a pistonic speaker - to the point where one may overlook a minor flaw that would be exposed by the Vandersteens. I know because we have had all kinds of reference speakers in our sound room - Avalon Eclipses for about 6 years, Vandy 3 for about two years, Wilson W/P 7 for about 2 years, and now TAD Ref Ones for about the last 8 years. Plus another half-dozen that lasted only a few weeks or months each.

 

I have a different philosophy than most people. Specifically I believe that "system matching" is a bad idea. If you have dull speakers, you can achieve an "improvement" by adding bright speaker wires - for a few weeks. Then you realize that two wrongs do not make a right and now you have two problems with your system. I have always found that one approach is more "correct" than another approach, and not just that I "like" it better.

 

The reason that I believe this is simple. When I hear systems that are built like a house of cards by carefully matching the colorations of the various components, it can sound absolutely spectacular with certain recordings or even certain kinds of music. But when you put a different recording or a different kind of music, it just falls apart.

 

In contrast a system composed of "correct" (accurate, but not the way that term is normally used by audiophiles), one can play any recording of any type of music and forget about the recording and just be swept away by the music. If you have a Pono Player, I think you probably know what I am talking about, especially when used with balanced headphones.

 

Hope this helps,

Charles Hansen

[Jud]

@Charles Hansen - Completely agree regarding "system matching."

 

Something that interests me greatly right now, that I touched on briefly in my last comment, is how we each form our notions of what makes good or accurate sound.

 

About the TADs - I've heard TADs, as well as other speakers designed by Andrew Jones, and think he is a very smart man.