John Atkinson: Yes, MQA IS Elegant...

[vl]

On 8/26/2018 at 2:56 PM, John_Atkinson said:

 

Because there is no spectral content above the new Nyquist frequency of 22.05kHz, due to the SRC's high-order low-pass filter. (Actually, spectral content above 22.05kHz analyzed in the digital domain after resampling to 96kHz, lies at approximately -127dB ref. the peak pulse level.) It is thus a "legal" 44.1kHz signal.

 

John Atkinson

Editor, Stereophile

 

The definition of a "legal" signal deserves some discussion.

 

From Wikipedia we have the definition of the Nyquist-Shannon sampling theory:

 

If a function x(t) contains no frequencies higher than B hertz, it is completely determined by giving its ordinates at a series of points spaced 1/(2B) seconds apart.

 

There is also a modern clarification in a following paragraph:

 

Modern statements of the theorem are sometimes careful to explicitly state that x(t) must contain no sinusoidal component at exactly frequency B, or that B must be strictly less than ½ the sample rate. 

 

Both statements are technically correct.  The sampling theory implies that when a signal with limited bandwidth B is sampled at 1/(2B) second intervals,  all the information in the original signal is contained in the sampled signal.  This sampling theorem does not however tell us how to retrieve the original information, without unwanted byproducts.

 

In the case of the sampling rate being exactly 2x of the bandwidth, an unwanted byproduct is the ringing excitation of the anti aliasing or reconstruction filter.  If the sampling rate is more than 2x the bandwidth of the signal, ringing of the anti aliasing or reconstruction will not occur.  

 

Anytime we limit the bandwidth of the of a signal by LP filtering it, we harm its transient property.  If we use a gentle filter, there will be less ringing.  If we use a brick wall filter, the ringing will be more concentrated at the "wall".  We can choose the characteristics of our filters.  If we do not go for "audiophile" specifications, we can use a gentle LP anti aliasing (AA) filter that may be a couple dB down at 20 KHz.  This will minimize the concentrated ringing at the brick wall frequency.  This is a choice of the designer of the ADC or the recording engineer.

 

In the case of the DAC, the first D to A operation usually occurs in software in the upsampler.  The reconstruction filter is first applied there.  If the signal is "legal" in the sense that the AA filter creates a condition of oversampling in sampling rate, The DAC will not reconstruct any ringing near the audio band in the upsampler.  If the "legal" signal excites the brick wall filter in the ADC, the DAC will reconstruct and reproduce it.  Downstream the DAC will produce more ringing far above the audio band and it will not be audible.  

 

My message is that if we choose the AA LP filter wisely, we can probably band limit a musical signal to fit into a CD with very good (but not perfect) results.  A brick wall AA LP filter should probably be used with care, with good understanding of the spectrum of the music being recording.  Jokingly speaking if we have musical instrument or human voices that are like impulses, we probably do not want to use a brick wall AA filter.  

 

If we do not want to compromise the transient of the music, than we should not band limit it at the ADC.  Just use a high enough sampling rate.  

 

Finally a word on MQA.  As a compression scheme, and a lossy one at that, it is quite elegant.  I can also say the same for our good friend MP3.  While MQA tries to provide large bandwidth in a small bucket and it appears to be quite good in doing so, it may have some unwanted byproducts that deserves more attention.  Since it uses minimum phase filters, how does the phase distortion affection the sound, compared to linear phase filters?  Since the MQA filters are gentle, leaky filters, there will be alising, the effect of that may be quite audible.  

 

One effective challenge for MQA is to compare an MQA encoded track with a 24/192 PCM track.  MQA wins if both tracks sound the same.  If the MQA track sounds different, even if it is better, that may be a sign of byproducts.

 

vl

[vl]

8 minutes ago, Rt66indierock said:

 

Very little doubt MQA changes things. I asked Bob at the LAAS in 2017 about plugins for DAWs so we could tell what MQA was changing in studios. The conversation continued at the AES convention later that year. 

I agree.  That is to be expected.  With the use of "leaky" filters that allow aliasing and minimum phase filters that have significant phase distortion, the sound has to be different.  

vl

[vl]

23 minutes ago, Archimago said:

 

I'm sure audiophiles are "all ears" to know if there's some super duper DSP system in the MQA process that actually does something to make music sound better or in the mastering process...

 

 

It appears that MQA's leaky filters and their associated impulse responses are the only remaining technical smoke screens MQA and its proponents are repeating over and over again in their apparent publicity effort to influence those who are less technically informed.  

 

To eliminate the brick wall filter ringing one can choose to use a non brick wall anti aliasing that has enough attenuation at the Nyquist frequency.  MQA is not required.  Compare the sound from the Sony Classical CDs to the classical music CDs from Decca and EMI.  The upper two octaves are usually far closer to live sound in the Sony while the other two brands have their characteristic house sound that is often bright and harsh.  CDs with its modest sampling rate and bit depth can sound very good, when the format is used properly.  MQA is not needed.

 

 

[vl]

4 hours ago, Jud said:

 

Keeping things accurate: "Leaky" is bad.  Minimum phase filtering is widely used in audio, including by nearly all room equalization software.

 

I agree that many room EQ SW use minimum phase filters.  The better ones, like Dirac Live, improves on the impulse response of the system.  This means the phase distortion of its filters is low.

[vl]

25 minutes ago, Jud said:

 

No. It very much depends on what is meant by "improves impulse response."  It could very likely mean any ringing energy is moved to after the impulse, which is simply another way of saying the filter is minimum phase. Any minimum phase filter will do this, not just those used by Dirac. Go to http://src.infinitewave.ca/ and have a look at the impulse response of any minimum phase filter there. 

 

I should clarify.  I was referring to the band limited impulse response of my Dirac Live system, which is limited by my preference to 17 KHz on the high end and 30 Hz on the low end.  The attached screen shot shows the impulse response of the audio system including the speakers.  The left impulse is before Dirac Live correction.  Please note the slight dispersion (spread out) of the impulse in the first mS after the impulse.  The right impulse is with Dirac Live correction.  This dispersion is gone and the pulse is taller.  This indicates better phase behavior after correction.  

 

The Dirac Live unit, a miniDSP model, operates at a sampling rate of 96 KHz.  So its filter ringing is not visible in this band limited presentation.  It uses a blend of FIR and IIR filters.

 

[vl]

16 minutes ago, Jud said:

 

I do have speakers with linear phase crossovers, and I fancy I might be able to hear the difference.  I've even passed a couple of (literally two) self-administered blind tests.  But I'm not confident I could pass additional blind tests. 

 

My main speakers are KEF LS50s.  I do not know what crossovers they have.  Looking at the impulse response of the system before Dirac Live correction, you can see that they are quite time coherent.  I crossover the LS50s to two woofers at 200 Hz using linear phase FIR filters running at 96 KHz.  You can see from the right impulse that the corrected impulse response is well behaved.  

 

Most two way monitors do not achieve this level of time coherence.  I find with these monitors, it is hard to tell a minimum phase or linear phase reconstruction filter apart.  With my system I can easily tell.  The linear phase filter has better transient response, is sharper and has more details.  The minimum phase filter smears.

[vl]

2 minutes ago, Jud said:

 

That's interesting, since one might think the opposite - it's one of the reasons minimum phase filters are used, to avoid pre-ringing that is purported to "smear" transient response.

 

The minimum phase filter has phase distortion, which may affect the perceived transient response.  The ringing of the filter at its Nyquist frequency should be inaudible to most people, even at the 44.1 K sampling rate.  If this ringing is from the AA filter at the ADC and if it is perfectly rendered by the reconstruction filter in the upsampler of the DAC, it should be inaudible, pre ringing or not.  If the reconstruction filter does not have enough computation power, this ringing may be reconstructed with some distortion.  Distortion means non linearity.  Non linearity means intermodulation, which may fall into the audio band and become audible. 

[vl]

13 minutes ago, Jud said:

 

Here - a linear phase and a minimum phase filter.  See how there's no pre-ringing to possibly smear the initial transient in the minimum phase filter?

 

 

 

We were talking about two different things.  Your impulse response is that of the anti aliasing or reconstruction filter.  The ringing frequency is at the Nyquist frequency, which is 22.05 KHz for CDs.  This single frequency ringing is above the audio band.  I have yet to understand how such inaudible ringing can lead to transient degradation of audio signals that are audible.

 

My impulse response is that of the entire audio system, included the speakers, limited to a bandwidth of 30 Hz to 17 KHz by my design.  It is a band limited impulse.  Since it is in the audio band, it represents what we hear.  In my system the Dirac Live DSP is done at 96 KHz, so its filter ringing, at 48 KHz, should be inaudible.

[vl]

23 minutes ago, Em2016 said:

 

Per Jussi @Miskajust one page back:

 

"looked from two extremes, both Chord talking about transient accuracy with extremely long filters and MQA talking about transient accuracy with extremely short filters are both right in a way, but only looking at things from one point of view while ignoring others. As usual in life, truth is somewhere between the extremes..."

 

https://www.computeraudiophile.com/forums/topic/49609-john-atkinson-yes-mqa-is-elegant/?do=findComment&comment=866897

 

 

There is a distinction.  Long filters (meaning very steep) are very good for the reconstruction of information contained in the digitized signal.  Short filters are very good for letting transients pass through them, but they are very poor for anti aliasing or reconstruction, as these leaky filters allow aliased signals to reach the audio components downstream, and the ears of the listeners.

[vl]

47 minutes ago, crenca said:

 

 

Most of my listening (must be >90% now) is HP's, so no crossover (though I tend to use software crossfeed - a complicating factor?).  I am liking the sound of upper frequencies like cymbal, brass, etc. now that I am using  linear or "zero" phase most of the time, they seem to bring these out in a more accurate way, whereas there is a "smear", or softness when I use minimum phase.  I also think strings seem more subtly accurate as well. By upsampling everything to DXD and using the quality filters in HQPlayer, I am probably getting just about the best out of linear phase.  All this is andoctotal only...

 

I experience the same with my LCD2 headphones.  A good pair of HPs usually have better amplitude and phase characteristics than a two way monitor of standard design.  With Dirac Live my speakers and HPs have very similar sonic characteristics.  

 

Good upsampling goes beyond multiplying the sampling rate.  The better upsamplers use polynomial curve fitting, probably for both upping the sampling rate and extrapolating 16 bits to 24 bits.  There is some art involved.  Most upsamplers, regardless of cost, sound a little sterile.  I have come across two upsamplers that are noticeably more musical.  They come in the Auralic Vega and the Cambridge Audio DACMagic Plus.  The latter is about 10% the price of the former but it gets 90% of the sonic quality when playing CDs.  The Vega upsamples to 1500 MHz and the DACMagic Plus to 384.  The higher upsampling help suppress the aliased byproducts in D to A conversion.  A very powerful DSP is needed to do a good job.

[vl]

22 minutes ago, Jud said:

 

Yeah, I can't tell with headphones (though admittedly I haven't tried a lot).  The differences I hear or imagine I hear with speakers have to do with soundstaging and location within the soundstage, and that's totally different with my headphones.

 

Yes.  HPs have different imaging and sound staging than speakers in a normal room, as the room tends to impart its characteristics on the sound.  However the HPs and speakers can sound similar in frequency response, transient, details, reverb, etc.

[vl]

13 hours ago, mcgillroy said:

Slightly OT but can you demonstrate this? Which CDs to compare and do you have any graphs too?

 

Here are two Sony Classical CDs that are representative of the Sony Classical CD sound.  As for Decca just pick any CD recording of the San Francisco Symphony Orchestra.

 

Wagner, Zubin Mehta, SK 45749

Hilary Hahn, Zinman, Beethoven violin concerto, SK 60584

 

The Sony Classical CDs sound like a microphone feed, but not the Decca or EMI ones.  The Sony CDs sound good on any CD player.  The Decca and EMI ones sound good only on the best digital playback systems.  I believe Sony chose a non brick wall AA filter for these recordings.  

[vl]

7 hours ago, Shadders said:

Hi,

From the miniDSP sweb site :

REW uses IIR filters, while Dirac Live uses mixed-phase filtering - in effect, a combination of IIR and FIR filters. FIR filters are more powerful than IIR filters, but more expensive to implement. See the app note FIR vs IIR filtering for information on the differences between these two different types of digital filter.

(https://www.minidsp.com/applications/digital-room-correction/dirac-live-vs-rew)

 

I have not studied this, but one aspect that immediately occurs is that to implement room correction in regards to amplitude you will be splitting up the audio band into many smaller bands - so the gain of each band can be modified to achieve an overall flat response.

 

The band pass filters will always have out of band energy applied to them - and this will always be in the hearing frequencies (20Hz to 20kHz).

 

Therefore filters with ringing are a bad idea - you will need filters with no ringing.

 

Someone else may be able to confirm this.

 

Regards,

Shadders.

 

There are several issues at hand.  One is FIR vs IIR.  One is minimum phase vs linear phase.  Another one is filter ringing or not.

 

FIR takes more computation resources to implement and sounds significantly better than IIR.  I did a demo to my friends FIR and IIR crossover filters and they immediately heard the difference.  Most EQ devices and most miniDSP devices use the inferior IIR filters.  I was lucky to have found a miniDSP product that supports FIR at the 96k sampling rate.

 

Linear phase filters have constant group delay.  Sounds of all frequencies are delayed by a given amount of time.  Minimum phase filters have variable group delay.  Their latency is not fixed.  It is frequency dependent.  If we want to recover the information captured in a digital recording without amplitude and phase distortion, linear phase reconstruction filters should be used.

 

All brick wall filters ring, if they are excited at their Nyquist frequency.  If you do not want them to ring, simply put an effective AA filter in to limit the bandwidth to less than the Nyquist frequency.  These are all under our control.

[vl]

7 hours ago, Rt66indierock said:

 

I got to disagree with you as would Rob Watts and others. Preringing is necessary to reconstruct transits properly. 

 

I concur with this statement.

[vl]

7 hours ago, Jud said:

 

As you noted, there is disagreement.  Some folks (including people not from the Meridian/MQA school) who actually do have some education and training in filter design use minimum phase filters because they think pre-ringing makes transients sound incorrect (since pre-ringing doesn't occur in nature).  Other folks agree with you and Rob Watts.  I hope to figure out a little more regarding what folks are disagreeing about by asking questions. 

 

Two points here.

 

1. Can a min phase filter do a perfect reconstruction job?  Look at this example.  I have a microphone feed with an upper bandwidth of 40 KHz.  I make a recording at 24/96.  Note that since the signal is band limited below 48 KHz, there will not be filter ringing.  Now I use a min phase filter to reconstruct the signal.  What I get will be distorted in phase.  If I use a lin phase filter for reconstruction, I shall not get distortion in amplitude and phase in the reconstructed signal.

 

2. If a recording engineer or producer is ill informed enough to use a brickwall filter to band limit a signal, he will capture the ringing of the filter.  In this case using min phase filters may make this brickwall limited signal sound less bad.  Personally if I come across such a recording (one should be able to tell ringing at 22.05 KHz on a spectral display) I shall just say goodbye to it.

[vl]

1 hour ago, mansr said:

If the impulse response is the same, they sound the same.

 

I agree.  Do FIR and IIR filters have the same impulse response?

[vl]

1 minute ago, mansr said:

Any response that decays to zero can be implemented as a FIR filter. It may be possible to achieve the same response more efficiently using an IIR structure. Categorically stating that FIR sounds better than IIR is a mistake.

 

Thanks for the explanation.  Let me clarify that the same 4th order 200 Hz LP and HP crossovers created by rePhase at a 96K sampling rate demonstrated that the FIR version sounded better - in transients and details.

[vl]

3 hours ago, John_Atkinson said:

 

As I have said before, spectral analysis shows that the "band-limited impulse" I used has no content above 22.05kHz. However, as I showed in the article, it does have sinc-function ringing present at 22.05kHz.The inference to be drawn is that every musical transient in a CD master will be accompanied by sinc-function ringing at Nyquist, either from the original A/D converter's anti-aliasing filter (if the recording was made at 44.1kHz), or from the sample-rate converter's low-pass filter used to create the master from 2Fs or 4Fs files. It seems incontrovertible, therefore that that ringing will excite the playback DAC's reconstruction filter, which will impose its own ringing on musical transients.

Read more at https://www.stereophile.com/content/zen-art-ad-conversion

 

John Atkinson

Editor, Stereophile

 

 

Please help me understand.  

 

In the real world why would a recording engineer or music producer apply an anti aliasing filter that is brick wall steep at the Nyquist frequency?  This is guaranteed to excite the ringing of the anti aliasing filter during the A to D process and the ringing of the reconstruction filter during the D to A process.  Wouldn't the use of a gentler, non brick wall filter to reduce the signal bandwidth to LESS THAN the Nyquist frequency make more sense?  This will avoid excitation of the filters at the Nyquest frequency.

 

Assuming the recording engineer uses a brick wall filter to truncate the bandwidth of the incoming analog music signal at the Nyquist frequency and the music content has energy at the Nyquist frequency, the resulting filter ringing will add its energy to the signal at the Nyquist frequency only, not at other frequencies.  In the case of the CD, does adding a 22.05 KHz ringing to the signal produce an audible effect to the music?

[vl]

1 hour ago, Miska said:

Tricky parts are many modern pop/rock recordings that have been recorded with modern ADCs or mastering tools with half-band filters. Where the content systematically reaches exactly Nyquist frequency at high levels (with associated aliasing products in the top frequencies)...

 

If they are dumb enough to do that, they deserve second rate sound quality.

[vl]

9 minutes ago, mansr said:

Isn't that the main conceit of MQA, at least as sold to consumers? I'm only pointing out that eliminating ringing of reconstruction filters can be done entirely at the production end. Nobody needs MQA for that.

 

I agree.  Even if the ringing is imparted into the recording during the A to D process, we as listeners have choices during playback.

 

1. Leave the ringing there if you believe it is above the audio range of most people therefore not audible.

2. In the playback system apply a moderately steep filter, starting at say 18 KHz, with a healthy attenuation at 22.05 KHz.  That will get rid of the ringing that we cannot hear.

 

I have such a filter in my Direc Live processor, at 17 KHz.  The reason I put it there is so many CDs are badly produced.  They have high distortion in the last octave.  Some CDs, like from Sony Classical, are always clear all the way to the very top end.  This filter does not affect music much but it makes the bad CDs more listenable.

[vl]

12 minutes ago, mansr said:

They typically use whatever filter their ADC provides. Here's the step (5 ns rise time) response of a TI PCM4220:

 

Any filter with a cut-off below 50 kHz or so will "ring" a little since some musical instruments (mainly percussion) extend that far. If you want no ringing whatsoever, you must record at 176.4 kHz or higher (if we're sticking to the usual rates). At 96 kHz the music can occasionally extend beyond the Nyquist frequency, and thus any filter will necessarily encounter some energy at its cut-off frequency.

 

Did you see my article? Filtering to slightly below Nyquist during production avoids ringing of reconstruction filters entirely.

 

A low-pass filter does not add anything to the signal. The "ringing" results when higher frequencies are taken away. It was there all along, just hidden.

 

Thanks for the explanation.  I am in agreement with you on all counts.  I should have written - Is the presence of a 22.05 KHz ringing audible in music?

[vl]

3 minutes ago, fas42 said:

 

Ummm, this encapsulates where some people have topsy turvy thinking - no CDs are "badly produced" and "have high distortion in the last octave". What some masterings do is emphasis the weaknesses in the playback chain - the problems all exist on the consumer's side, because the manufacturers of the playback gear haven't done enough to reduce the typical digital artifacts, which cause the people listening to decide that they have a "bad CD".

 

It may take much exposure to a well sorted playback rig, easily presenting these 'difficult' recordings in their full glory, for many to understand what's going on, unfortunately.

 

 

I also agree with what you wrote.  We were making different observations of the same thing.  Some CDs are mastered to sound good on just about any player.  Unfortunately many can sound good only on an excellent playback system.  I want to call the mastering engineers who produce universally good sounding CDs good mastering engineers.

 

Relating to our topics on minimum and linear phase filters, it took me years to arrive at a digital playback system that can play Decca and EMI CDs with good sonic quality.  The magic is in the box that contains the DAC chip but it is not the DAC chip that gives the better sound.  It is the upsampler.  In my case I found a DSP with high computation power using a "long" (steep) linear phase filter is needed to handle the "bad" CDs.

 

 

[vl]

6 hours ago, John_Atkinson said:

 

Please read the thread. I was responding to the assertion made by several posters, that the ringing of the DAC's reconstruction filter was due to the down-sampled file having spectral content above 22.05kHz. It didn't.

 

John Atkinson

Editor, Stereophile

 

Given a brick wall filter set at the Nyquist frequency of the sampling rate being used, signal energy ABOVE and AT the Nyquist frequency will excite the filter ringing.  To avoid ringing the signal should be band limited to BELOW the Nyquist frequency.  Having no signal energy above the Nyquist frequency is not sufficient to prevent ringing in the filter.  To me a "legal" signal for a digital system is one that is band limited to LESS THAN the Nyquist frequency, assuming the anti aliasing and reconstruction filters are set at the Nyquist frequency.

[vl]

2 hours ago, mansr said:

Not above, only at.

 

Thanks for the correction.  I should rephrase what I wanted to say.  Given a brick wall filter set at the Nyquist frequency of the sampling rate being used, a signal with energy bandwidth ABOVE or AT the Nyquist frequency will both excite filter ringing.

 

I would like to add my perspective for JA that from the sampling stand point a "legal" signal is one that is less than or equal to the Nyquist frequency in bandwidth.  Nyquist and Shannon in their theorem did not say what filters should be used for anti aliasing and reconstruction.

 

From a digital audio implementation, 1/2 fs brick wall filters are often used for anti aliasing and reconstruction.  In this scenario I would like to suggest that a "legal" signal is one with bandwidth less than (and not equal to) the filter bandwidth, as it is probably not the intention of the recording engineer or producer to willfully impart filter ringing to the recorded music.

[vl]

17 minutes ago, mansr said:

No, that's still not right. Signal energy above the filter cut-off, in the stop band, does not cause "ringing."

 

I understand and agree with what you wrote.  What I meant to say was given a signal with energy bandwidth at or above 1/2 fs, filter ringing can be excited.  The signal with energy bandwidth above 1/2 fs contains energy at 1/2 fs.