On the subject of "ringing"

[mansr]

6 minutes ago, John_Atkinson said:

Incidentally, I have recently been examining the time-behavior of A/D converter antialiasing filters and have found just one which captures a band-limited impulse without any ringing before or after: the Listen filter on Ayre's QA-9 converter, designed by Charley Hansen and Ariel Brown.

Could you describe your "band-limited impulse" a bit more? How is it generated? What does the waveform look like?

[John_Atkinson]

13 minutes ago, mansr said:

Could you describe your "band-limited impulse" a bit more? How is it generated? What does the waveform look like?

 

In brief I created a "digital black " file sampled at 38kHz and drew the shape of the diagnostic waveform I needed with BIAS Peak's pencil tool. I then modified it  until I got the desired spectral content. To create an analog signal to feed to the ADCs under test, I decoded the signal with a DAC capable of handing 384k PCM data without downsampling. As the ADCs to be tested were all set to 96kHz sampling, the DAC's own ringing at Nyquist would be an octave above the ADC's output passband and would be rejected. Apologies but people will have to wait for a forthcoming article in Stereophile for more detail.

 

John Atkinson

Editor, Stereophile

[mansr]

4 minutes ago, John_Atkinson said:

In brief I created a "digital black " file sampled at 38kHz and drew the shape of the diagnostic waveform I needed with BIAS Peak's pencil tool. I then modified it  until I got the desired spectral content.

Well, that's rather vague.

 

4 minutes ago, John_Atkinson said:

To create an analog signal to feed to the ADCs under test, I decoded the signal with a DAC capable of handing 384k PCM data without downsampling. As the ADCs to be tested were all set to 96kHz sampling, the DAC's own ringing at Nyquist would be an octave above the ADC's output passband and would be rejected. Apologies but people will have to wait for a forthcoming article in Stereophile for more detail.

When you publish the results, will you include at least a plot of this "impulse" signal? If you made the file available for download, others could perform the same test on their equipment and get results meaningfully comparable to yours. Just something to consider.

[Ralf11]

15 hours ago, sandyk said:

results that can't be reasonably explained.

 

when coupled with a lack of methodology it is no surprise that people reject this

 

if you want to be believed you will need good strong evidence

 

unusual results require unusual proof -- and anything appearing to contravene known physical laws is indeed unusual

[mansr]

1 hour ago, Ralf11 said:

unusual results require unusual proof -- and anything appearing to contravene known physical laws is indeed unusual

New physics is pretty much confined to high-energy accelerators these days.

[buonassi]

4 hours ago, John_Atkinson said:

 

Incidentally, I have recently been examining the time-behavior of A/D converter antialiasing filters and have found just one which captures a band-limited impulse without any ringing before or after: the Listen filter on Ayre's QA-9 converter, designed by Charley Hansen and Ariel Brown.

 

John Atkinson

Editor, Stereophile

 

Thanks,  I should've noted that the ringing can also be a result of the encoding process as well as the decoding process.  

But isn't this ADC you mention achieving less ringing at the expense of high frequency loss?  I thought the ayer "listen" filters (at least in their DACs) were a super slow rolloff starting at about 15K?  It's the same for any DAC that uses the AKM4490 chip (note pic below).  

 

Presumably, when using both slow rolloff at encoding, then again at decoding, you're attenuating the top octave pretty aggressively....

 

 

[buonassi]

@John_Atkinson, Fiio X5 3rd gen, uses a filter called "super slow rolloff" that produces almost no ringing.  It must be proprietary to them because it's not listed on the AKM data sheet.  But I can clearly hear the high frequencies being attenuated when switching between this setting an a more traditional filter of the 4490.  Admittedly, this device's amp section really isn't up to snuff for the most critical of listening, but still the result of this special filter is very dry sound - eerily precise. But again, at a cost....  HF rolloff.

http://fiio.me/forum.php?mod=viewthread&tid=42106&page=1&extra=#pid117558

 

On a side note, I've somewhat given up on trying to eliminate ringing.  A minimum phase filter that dumps all the ripple energy after the transient is OK by me provided it's not excessive.  I've played around with iZotope SRC at length to find a good compromise between high frequency loss and post ringing - still avoiding aliasing at -96dbfs.  And of course, I never stray from the minimum phase region of the ringing parameter.  

 

Really it's the "pre" ringing that bothers me.  Post ringing has the effect of added harmonics on cymbals, giving them some more zing and sparkle, which can be overwhelming if too much post ringing is present.  A moderate amount really doesn't bother me.

 

 

 

[semente]

2 minutes ago, buonassi said:

@John_Atkinson, Fiio X5 3rd gen, uses a filter called "super slow rolloff" that produces almost no ringing.  It must be proprietary to them because it's not listed on the AKM data sheet.  But I can clearly hear the high frequencies being attenuated when switching between this setting an a more traditional filter of the 4490.  Admittedly, this device's amp section really isn't up to snuff for the most critical of listening, but still the result of this special filter is very dry sound - eerily precise. But again, at a cost....  HF rolloff.

http://fiio.me/forum.php?mod=viewthread&tid=42106&page=1&extra=#pid117558

 

On a side note, I've somewhat given up on trying to eliminate ringing.  A minimum phase filter that dumps all the ripple energy after the transient is OK by me provided it's not excessive.  I've played around with iZotope SRC at length to find a good compromise between high frequency loss and post ringing - still avoiding aliasing at -96dbfs.  And of course, I never stray from the minimum phase region of the ringing parameter.  

 

Really it's the "pre" ringing that bothers me.  Post ringing has the effect of added harmonics on cymbals, giving them some more zing and sparkle, which can be overwhelming if too much post ringing is present.  A moderate amount really doesn't bother me.

 

 

 

I've spent some time this evening playing with HQPLAYER's poly-sinc-xtr filters and I am having trouble deciding whether to choose minimum phase or linear phase. I don't like mp so much but when I switch back to lp something doesn't sound quite right.

I must try again tomorrow using poly-sinc-short filters but I like the transient "snappiness" of the xtr.

[John_Atkinson]

47 minutes ago, buonassi said:

But isn't this ADC you mention achieving less ringing at the expense of high frequency loss?  I thought the ayer "listen" filters (at least in their DACs) were a super slow rolloff starting at about 15K?

 

The high-frequency rolloff of the Ayre QA-9's Listen filter seems pretty benign. You can find my measurements at https://www.stereophile.com/content/ayre-acoustics-qa-9-usb-ad-converter-measurements

 

Below is fig.2  on that page, showing the frequency response at –1dBFS with the Listen filter,  analyzed in the digital domain, and data sampled at 192kHz (left channel blue, right red), 96kHz (left green, right gray), 48kHz (left cyan, right magenta)  The vertical scale is 1dB/div. The response at 20kHz is down by just a fraction of a dB at all 3 sample rates, though the rolloff is slow, which will potentially allow some aliasing at single Fs rates with some music.

 

John Atkinson

Editor, Stereophile

 

[buonassi]

8 minutes ago, John_Atkinson said:

The response at 20kHz is down by just a fraction of a dB at all 3 sample rates, though the rolloff is slow, which will potentially allow some aliasing with some music.

 

And therein lies the tradeoff: aliasing.  The "filtering" power of this seems very weak for red book material if we're talking about DACs and reconstruction filters for playback.  But this is ADC only we're discussing for the QA9, right?  Not SRC/downsampling from a digital master to redbook - where aliasing would matter most.  I'm too ignorant on ADC to understand why the slopes and bandpasses of the filters are essentially identical for all sampling rates.

 

Still, good article - I especially agree with your sentence on page 1 - 

 

"Given my extensive experience of both domestic and professional A/D converters, which has convinced me that the most critical process in digital recording is the initial analog/digital conversion—nothing downstream can put right whatever was done wrong in that conversion"
 

[Ralf11]

2 hours ago, mansr said:

New physics is pretty much confined to high-energy accelerators these days.

 

you forgot the interiors of theoreticians skulls - lotsa new physics rattlin' around in there

[buonassi]

16 hours ago, Fokus said:

 

That is a demonstration of equalisation with the filter transitions smack in the audible band. It has been known for three decades or so that under these conditions, given sufficiently steep transitions, linear phase pre-ringing can be audible.

 

But if you move the transition frequency up and up the audible effect vanishes.

 

Not sure it vanishes, but it's subdued for sure.  In terms of digital reconstruction filters on DACs, there isn't the same level of OVERT preringing, of course not.  It is very subtle, but it's there.  It can be measured, but that's really not the point.  If I'm inferring correctly, I think your assertion is that the human auditory system can't discern preringing of such low levels and therefore any preringing in digital filter design is of no consequence? 

[buonassi]

7 hours ago, John_Atkinson said:

 

As its frequency is at Nyquist, which is above almost everyone's HF cutoff frequency, you would think not. But when I discussed this with Karlheinz Brandenburg at an AES convention several years ago, he said that basically even if you can't hear the "ringing" as a tone, your brain could well be aware that something has happened when it starts and marks it as an acoustic "event." Then, when the peak is subsequently reached, that is marked as a spurious second "event," leading to confusion.

 

 

Does preringing only affect HFs near Nyquist?  The entire bandpass is affected I thought?  Preringing has less to do with the frequency domain, rather the time domain as I understand it, and all frequencies in bandpass are affected in the time domain regardless of Nyquist limit.  

 

I agree with your second sentence, which time domain theory would support.  Preringing doesn't exist naturally.  You can't "ramp up" to a transient.  That would necessitate breaking the laws of physics to somehow get a portion of the decay of a transient to your ears before the original transient takes place. When a linear phase transient-aligned filter is used, some information is pushed ahead of time during reconstruction, modeling precisely what I just described.  Unnatural. And your brain can hear it IMO.  

 

So, is unnatural worse?  No, it's just different and could be considered quite enjoyable by some people.  I like it with acoustic/orchestral music actually!  But for metal, with super fast drum fills, not a chance.  

[Spacehound]

7 hours ago, John_Atkinson said:

 

As its frequency is at Nyquist, which is above almost everyone's HF cutoff frequency, you would think not. But when I discussed this with Karlheinz Brandenburg at an AES convention several years ago, he said that basically even if you can't hear the "ringing" as a tone, your brain could well be aware that something has happened when it starts and marks it as an acoustic "event." Then, when the peak is subsequently reached, that is marked as a spurious second "event," leading to confusion.

 

On the subject of the time-symmetrical ringing of linear-phase digital filters, I assume Robert Watts would disagree; see my comment on his "million-tap" filter at https://www.stereophile.com/content/chords-million-tap-digital-filter

 

Incidentally, I have recently been examining the time-behavior of A/D converter antialiasing filters and have found just one which captures a band-limited impulse without any ringing before or after: the Listen filter on Ayre's QA-9 converter, designed by Charley Hansen and Ariel Brown.

 

John Atkinson

Editor, Stereophile

If pre-ringing  actually exists in 'reality' then the filter must be made out of the electrical equivalent of thiotimoline.

Failing that it is merely an artefact of the measurement methods.

 

There is only one guy on this entire forum that might be able to convince me otherwise.

[Spacehound]

40 minutes ago, buonassi said:

 

Does preringing only affect HFs near Nyquist? 

It doesn't happen at all if Nyquist/Shannon is obeyed.

 

'Bob' Stuart and his shills  deliberately  use meaningless measurement methods so it looks 'broken' and then try to sell us a magical fix.

It's why Meridian stuff often receives  poor  'sound quality'  reviews, and that's even before he dreamed up  MQA. 

[buonassi]

11 minutes ago, Spacehound said:

It doesn't happen at all if Nyquist/Shannon is obeyed.

 

are you sure you're not confusing aliasing/imaging with "ringing"?  Not trying to shut down the discussion, or insult you by any means.  It's me that has a lot to learn yet, especially about Nyquist and the "folding" that's involved with MQA.

[Spacehound]

28 minutes ago, buonassi said:

 

 

are you sure you're not confusing aliasing/imaging with "ringing"?  Not trying to shut down the discussion, or insult you by any means.  It's me that has a lot to learn yet, especially about Nyquist and the "folding" that's involved with MQA.

I don't think I am but I wouldn't stake my life on it. It's only  audio.

 

Pre-ringing  breaks cause and effect so it can only be an artefact,  or what you see if you use an invalid measurement method. Which I suspect they are doing deliberately. (Even John Atkinson,  said  the measurement was "illegal" but later backed off via  obfuscation in the hope we wouldn't notice. The magazines' contempt for their readers is endless and it's not just Stereophile.)

[buonassi]

25 minutes ago, Spacehound said:

I don't think I am but I wouldn't stake my life on it. It's only  audio, not Klingon babies

Pre-ringing  breaks cause and effect so it can only be an artefact,  or what you see if you use an invalid measurement method. Which I suspect they are doing deliberately.

The effects of preringing can be done away with just by using a higher sampling rate with a super slow rolloff filter applied after 20khz.  So a benefit of MQA as I see it would be a smaller file size.  But do we really need that with today's storage and transmission technology?

 

Man do I wish SACD or even 96khz/16 bit PCM would've become the standard and ousted redbook.  We wouldn't be having this debate today.

 

edit:  The effects of ringing both pre and post can be done away with.

[Spacehound]

1 hour ago, buonassi said:

It achieves absilThe effects of preringing can be done away with just by using a higher sampling rate with a super slow rolloff filter applied after 20khz.  So a benefit of MQA as I see it would be a smaller file size.  But do we really need that with today's storage and transmission technology?

 

Man do I wish SACD or even 96khz/16 bit PCM would've become the standard and ousted redbook.  We wouldn't be having this debate today.

 

edit:  The effects of ringing both pre and post can be done away with.

It varies by content but on average  an MQA file is no smaller than a regular FLAC file of the same 'nominal' resolution.

Though in practice, due to the need to make space for  the 'folding',  the resolution of a MQA file is less, 17 bits at the most rather than 24.  Nothing is free in the real world.

 

And as MQA data above 96K is totally fake, even including the noise,   with an MQA-claimed  192 file  they are using the doubled size for nothing that is 'true'. This won't matter audibly, but what's the point other than making your DAC say 192?

 

The  more you look the more you find MQA is a total scam from beginning to end. It achieves absolutely nothing and may often audibly degrade the sound. 

[Shadders]

9 hours ago, Spacehound said:

If pre-ringing  actually exists in 'reality' then the filter must be made out of the electrical equivalent of thiotimoline.

Failing that it is merely an artefact of the measurement methods.

 

There is only one guy on this entire forum that might be able to convince me otherwise.

Hi,

I am surprised that pre/post ringing has not been resolved or defined as an issue in the audio world.

 

How can such an issue be continually discussed as either it exists, or does not exist, and where and how it occurs ?.

 

If the AES was a professional organisation, then surely this aspect would have been well understood by the participants/members ? and published papers on the effect presented for the edification of all ?

 

Regards,

Shadders.

[psjug]

38 minutes ago, Shadders said:

Hi,

I am surprised that pre/post ringing has not been resolved or defined as an issue in the audio world.

 

How can such an issue be continually discussed as either it exists, or does not exist, and where and how it occurs ?.

 

If the AES was a professional organisation, then surely this aspect would have been well understood by the participants/members ? and published papers on the effect presented for the edification of all ?

 

Regards,

Shadders.

It has been shown good enough for me that there is no ringing.  By Archimago.  If ringing from linear phase reconstruction filters is a problem then let someone else do similar and show it.

[adamdea]

18 hours ago, John_Atkinson said:

 

In brief I created a "digital black " file sampled at 38kHz and drew the shape of the diagnostic waveform I needed with BIAS Peak's pencil tool. I then modified it  until I got the desired spectral content. To create an analog signal to feed to the ADCs under test, I decoded the signal with a DAC capable of handing 384k PCM data without downsampling. As the ADCs to be tested were all set to 96kHz sampling, the DAC's own ringing at Nyquist would be an octave above the ADC's output passband and would be rejected. Apologies but people will have to wait for a forthcoming article in Stereophile for more detail.

 

John Atkinson

Editor, Stereophile

I'm still baffled by this. What is the diagnostic waveform? It's an impulse only bandlimited. But band limited to what? 192Khz? So is that the same as a dirac put through a filter? If so what filter? Why not just use a real thing like a castanet?

[Shadders]

3 minutes ago, psjug said:

It has been shown good enough for me that there is no ringing.  By Archimago.  If ringing from linear phase reconstruction filters is a problem then let someone else do similar and show it.

Hi,

Yes, i agree - there is no ringing with audio signals energising a filter designed for audio. (not all the story, covers 99.99% of cases)

Regards,

Shadders.

[mansr]

9 hours ago, Spacehound said:

If pre-ringing  actually exists in 'reality' then the filter must be made out of the electrical equivalent of thiotimoline.

Failing that it is merely an artefact of the measurement methods.

 

There is only one guy on this entire forum that might be able to convince me otherwise.

First of all, pre-ringing isn't a well-defined term. In fact, it is never used outside of audio. Let us thus define it for the sake of this discussion as an impulse response having at least one negative excursion prior to the (positive) peak.

 

Symmetrical impulse response plots are often centred around time zero. On the face of it, this amounts to (part of) the response preceding the input. Such a system is termed non-causal and cannot be physically realised. We nevertheless use this representation because it is mathematically convenient. Now remember, we are dealing with a time-invariant system, which means a time-shift of the input results in a time-shift of the output. Thus, provided the impulse response is finite, we can make it causal simply by shifting it such that it becomes zero for all negative time values. The only effect of this is the addition of a constant delay to the output. Importantly, the frequency response is not affected.

 

In practice, a digital filter incorporating such a time-delay is trivially constructed. An all-analogue realisation is trickier, although it can be accomplished through the use of various delay elements.

 

Have I convinced you yet?

[adamdea]

18 hours ago, John_Atkinson said:

 

As its frequency is at Nyquist, which is above almost everyone's HF cutoff frequency, you would think not. But when I discussed this with Karlheinz Brandenburg at an AES convention several years ago, he said that basically even if you can't hear the "ringing" as a tone, your brain could well be aware that something has happened when it starts and marks it as an acoustic "event." Then, when the peak is subsequently reached, that is marked as a spurious second "event," leading to confusion.

It might in principle but is there any evidence of a process in the cochlea whereby nerves fire at inaudible frequencies (and very low levels)?. And is what about integration and masking? And how does this match what is known about the audibility of pre-ringing at audible frequencies.