On the subject of "ringing"

[Shadders]

Hi,

The initial posts and questions on the ringing in filters, and whether it affects can be heard, is open to debate. The following has been noted :

1. Ringing is due to the filters reaction to a transient event, where the transient event contains energy at the filter cut off frequency.

2. The transient event lasts for the duration of the number of taps in the filter (assuming FIR), hence a small number of taps, then a short effect, high number of taps, then a longer effect.

3. Analysis of the pre and post ringing shows that the peak energy of the ringing is at the filter cut off frequency (filters natural/transient response).

4. There is significant energy in the audible band, and although this energy is at a lower level, it is a burst of energy, and may therefore be audible.

5. The total energy at the filter cut off frequency is 0.5%, and current analysis shows energy in the audible band being 23%.

6. The intensity (amplitude) of the ringing is directly proportional to the amplitude of the energy at the filter cut off frequency.

7. If there is no energy at the cut off frequency in the signal passing through the filter, and the cut off frequency energy is added, this will generate ringing (filters natural/transient response), as this additional energy will be presented as a transient to the filter.
   
   To be confirmed : If the signal applied to the filter is at the filter cut off frequency, once the transient has decayed, then the filter is operating in its steady state condition. If the signal at the filter cut off frequency is increased in amplitude, then this may introduce ringing, based on the type of increase. A slow increase incurs no ringing, an abrupt increase may incur ringing.

8. Since the energy at the cut off frequency is, in general, small in an audio signal, then the ringing amplitude is small. The issue is that when an audio signal has energy at this frequency, appearing and disappearing (transitory – castanet example), it will cause ringing at the filter output. Although this ringing is low level energy, it contains energy in the audible bandwidth, and hence may be heard.
   
   Visual inspection and spectral analysis of the waveform will not readily show this, since the ringing is low level, the spectrum is wideband, despite the peak at the filter cut off frequency, and most importantly, it is transient in nature.

So, in summary, ringing does exist in audio replay based on the audio signal being reproduced, and the combination of sample rate and filter parameters. Whether it can be heard is dependent on the person listening, and the music being listened to, in addition to the aforementioned criteria.

 

Regards,

Shadders.

[psjug]

On the example with 5 tones and discontinuity - this is not band limited so I think not a good example at all to use for discussion of ringing.  Seems to me any calculations or thinking about ringing based on this example are best forgotten.

 

On the castanets file - I didn't realize the discussion had widened to include ringing in source files.  To be honest I am a little embarrassed that I had no idea that downsampling with modern tools would cause such a thing.  But the makers of SRC tools talk about it, and choosing filters fo rmore or less ringing, more or less aliasing. Clearly the issue is real, and the castanets spectrograms are convincing enough for me.  How better would you show it?  I would hope that engineers would make the right choices so that tracks don't have  T shaped spectrograms, but I never look at these so I have no idea.

 

On frustration in this discussion - those of us who are not in this business really don't need to get all worked up about all this.  If it gets to that point just click off the thread and pour a drink!

 

 

 

 

 

[Hifi Bob]

14 minutes ago, psjug said:

On the castanets file - I didn't realize the discussion had widened to include ringing in source files.  To be honest I am a little embarrassed that I had no idea that downsampling with modern tools would cause such a thing.

 

I wouldn’t worry too much: there was no ringing evident in the 96kHz original file and only when choosing what I deliberately described as an ‘insanely’ steep filter was the ringing significant in the 44.1k version.

[Shadders]

13 minutes ago, psjug said:

On the example with 5 tones and discontinuity - this is not band limited so I think not a good example at all to use for discussion of ringing.  Seems to me any calculations or thinking about ringing based on this example are best forgotten.

 

On the castanets file - I didn't realize the discussion had widened to include ringing in source files.  To be honest I am a little embarrassed that I had no idea that downsampling with modern tools would cause such a thing.  But the makers of SRC tools talk about it, and choosing filters fo rmore or less ringing, more or less aliasing. Clearly the issue is real, and the castanets spectrograms are convincing enough for me.  How better would you show it?  I would hope that engineers would make the right choices so that tracks don't have  T shaped spectrograms, but I never look at these so I have no idea.

 

On frustration in this discussion - those of us who are not in this business really don't need to get all worked up about all this.  If it gets to that point just click off the thread and pour a drink!

 

Hi,

The 5 sine waves file with discontinuity is a way of replicating the filter ringing. There is no difference - the issue is as stated by adamdea etc., is that the energy at the filter cut off frequency initiates the ringing (transient response) of the filter.

 

For the castanets, i varied the filter cut off to align with energy and no energy in the bandwidth, and the same effect. The sine wave allows for much easier extraction of the ringing.

 

Since the system is LTI, a discontinuous sine wave, or actual audio file are still valid.

 

The castanet file did not have ringing - as far as i can tell, but by modifying the filter cut off frequency, i could introduce ringing. So, do not be embarrassed - there is no downsampling as i am aware in this file.

 

On the general subject - yes, choosing the right filter is an issue for SRC etc., so maybe, since audio contains no information at 50kHz+, then we should all listen to 192kHz source material....

 

Regards,

Shadders.

[psjug]

2 minutes ago, Shadders said:

 

The castanet file did not have ringing - as far as i can tell, but by modifying the filter cut off frequency, i could introduce ringing. So, do not be embarrassed - there is no downsampling as i am aware in this file.

 

There is the high sample rate file, and then 3 downsampled files with more and less ringing.

[Shadders]

Just now, psjug said:

There is the high sample rate file, and then 3 downsampled files with more and less ringing.

Hi,

Apologies - i only used the high smaple rate file - i did not use the others. I also had an issue with the reading into Octave of the SoX files - so could not analyse those.

Regards,

Shadders.

[psjug]

I did not analyze either.  But if you look at the spectrogram closely, it is an animated gif that switches between the different filters.

[Jud]

On 2/10/2018 at 10:50 PM, Spacehound said:

There is no such thing as a 'positive' blind test and no 'correct' or 'expected' answer.

 

All the tests are is 'can you tell ANY difference'? It's not 'Can you tell THE difference'.

Thus it is 'binary'  (yes or no) so  100% objective as unless you CAN hear a difference you CAN'T have a 'preference'

 

No test requiring a verbal answer as to conscious, subjective perception should be referred to as  "100% objective."

 

But rather than jumping down this particular rabbit hole yet again, I'll be interested to read through the rest of the thread and learn something from folks who are far more informed than I am on the topic of digital audio filter behavior.

[Shadders]

Hi,

Ok- i saw this - and yes, it does indicate the ringing very well. The steeper the filter, then the large the number of taps which means the pre and post ringing last longer.

Here is the pre ringing for a sine wave with 500 taps.

Here is the same with 50 taps :

Regards,

Shadders.

[sdolezalek]

2 hours ago, Shadders said:

So, in summary, ringing does exist in audio replay based on the audio signal being reproduced, and the combination of sample rate and filter parameters. Whether it can be heard is dependent on the person listening, and the music being listened to, in addition to the aforementioned criteria.

 

Regards,

Shadders.

 

I think this appropriately summarizes 8 pages of posts that don't do very much to shed light on the subject. 

[adamdea]

4 hours ago, psjug said:

On the example with 5 tones and discontinuity - this is not band limited so I think not a good example at all to use for discussion of ringing.  Seems to me any calculations or thinking about ringing based on this example are best forgotten.

 

On the castanets file - I didn't realize the discussion had widened to include ringing in source files.  To be honest I am a little embarrassed that I had no idea that downsampling with modern tools would cause such a thing.  But the makers of SRC tools talk about it, and choosing filters fo rmore or less ringing, more or less aliasing. Clearly the issue is real, and the castanets spectrograms are convincing enough for me.  How better would you show it?  I would hope that engineers would make the right choices so that tracks don't have  T shaped spectrograms, but I never look at these so I have no idea.

 

On frustration in this discussion - those of us who are not in this business really don't need to get all worked up about all this.  If it gets to that point just click off the thread and pour a drink!

 

 

 

 

 

The reason to use castanets is actually that they are one of the few real music signals where you can see pre-ringing. In practical terms (somewhat submerged in the derailing of the thread) most music does not exhibit it for reasons which are actually quite clear when not submerged ie you need a transient with significant energy in the transition band. And anyway the ringing will be at that frequency (which you can't hear, even before masking effects are considered). 

Unfortunately this has been made to seem a lot more complicated than that; this will no doubt suit the purposes of the "there are more things under heaven and earth Horatio than are dreamt of in your philosophy" gang.

 

[buonassi]

being my 101th? no, that's not right, 101st post, I have taken the liberty of graduating myself from "freshman member" to to this..... promulgating which camp i'm in.  Damn do I feel assertive tonight.  

[Spacehound]

15 hours ago, Jud said:

 

No test requiring a verbal answer as to conscious, subjective perception should be referred to as  "100% objective."

 

But rather than jumping down this particular rabbit hole yet again, I'll be interested to read through the rest of the thread and learn something from folks who are far more informed than I am on the topic of digital audio filter behavior.

Sure it is.

We are doing a 'survey'  even if it's only of one person, and asking the listener whether  he can hear a difference or not.  That's binary.

What happens between his ears and his mouth is irrelevant.

 

Also what he says doesn't have to be true, true/false is binary too,  Nor does any overall result if there are lots of people. 

(I'm not having you  as a lawyer, you don't even know that things can be both objective and wrong )

[audiventory]

On 11.02.2018 at 8:50 AM, Spacehound said:

All the tests are is 'can you tell ANY difference'? It's not 'Can you tell THE difference'.

 

In context of ringing research, I'd ask hearing test participant "What is better?" (linear or "quasi-analog" [minimal phase filter]).

[Spacehound]

11 minutes ago, audiventory said:

 

In context of ringing research, I'd ask hearing test participant "What is better?" (linear or "quasi-analog" [minimal phase filter]).

That instantly changes what should have been an objective test  into a possibly  incomplete  list of individual, unrelated,  personal preferences, even if  there is no difference at all - you have prompted them  into thinking there is one.

 

So it  is totally worthless.

[audiventory]

1 hour ago, Spacehound said:

That instantly changes what should have been an objective test  into a possibly  incomplete  list of individual, unrelated,  personal preferences, even if  there is no difference at all - you have prompted them  into thinking there is one.

 

Big numbers (tens thousands) and clustering (tested samples, partisipant skills, etc.) allow to reduce uncontrolled individual biases.

 

I don't know what is practical benefits of researching "different or not". As example, linear and minimal phase filters sounds differently. What we can to do for sound improvement in a future audio units after the researching?

 

[Spacehound]

19 minutes ago, audiventory said:

 

Big numbers (tens thousands) and clustering (tested samples, partisipant skills, etc.) allow to reduce uncontrolled individual biases.

 

I don't know what is practical benefits of researching "different or not". As example, linear and minimal phase filters sounds differently. What we can to do for sound improvement in a future audio units after the researching?

 

Sure they can. But it has nothing whatsoever to do with my comments.

 

In practice? If you cant  hear any difference buy the lower cost box.

That's why most  Hifi manufacturers HATE blind tests and will come up with near endless utter nonsense to (fail to) convince sane people they don't work. It is why most US manufacturers (particulary)  have totally abandoned  the  realms of high fidelity and gone for 'impressive' instead.

[audiventory]

1 hour ago, Spacehound said:

In practice? If you cant  hear any difference buy the lower cost box.

 

As developer/manufacturer, I use available researching results (including blind and ABX tests) to improvement of subjective perceived sounding.

Me need to know what is sound better. Of course, knowledge that there is no difference is useful. But also need to know, that to do for sound perception improvement. It allow to move to target easier way.

 

Blind tests are too complex. It is very hard work, that demands serious knowledges about subject and measurements, performing accuracy, correct using of equipment and fixing subtlest details https://samplerateconverter.com/educational/hifi-blind-test

Also need to separate 2 direction of perceived sound improvement:

• high fidelity (low distortion level);
• sound enhancers (tube, vynil, tape, analog circuits, spatial improvements, it emulators).

I think, next jump in quality, expect us in spatial reproduction (sound hologram) https://samplerateconverter.com/content/where-limit-audio-quality

[Spacehound]

31 minutes ago, audiventory said:

 

As developer/manufacturer, I use available researching results (including blind and ABX tests) to improvement of subjective perceived sounding.

Me need to know what is sound better. Of course, knowledge that there is no difference is useful. But also need to know, that to do for sound perception improvement. It allow to move to target easier way.

 

Blind tests are too complex. It is very hard work, that demands serious knowledges about subject and measurements, performing accuracy, correct using of equipment and fixing subtlest details https://samplerateconverter.com/educational/hifi-blind-test

Also need to separate 2 direction of perceived sound improvement:

• high fidelity (low distortion level);
• sound enhancers (tube, vynil, tape, analog circuits, spatial improvements, it emulators).

I think, next jump in quality, expected us in spatial reproduction (sound hologram) https://samplerateconverter.com/content/where-limit-audio-quality

Nothing after the source can improve on it. They  either degrade it, leave it alone,  or alter it  to suit your personal preferences, which means it has left  the 'high fidelity' arena, which is what many of us pay for  So 'enhancers' don't exist.  I exempt room treatments from this though I don't use them.

 

Blind  tests can reveal price/performsnce ratios. Nothing else.

 

What do I do?

I will look at speaker reviews to see if they have  a flat frequency response/good phase performance. If not, I won't buy them.

Everything else I just buy what I take a 'fancy' to and can afford.  I almost never  listen before buying it and have not done so for at least ten years. It's the same with cars - it took me all of  15 minutes to make up mind, go to the Mercedes dealer, and buy one,  mostly because he's the nearest car dealer to where we live (UK).

[audiventory]

41 minutes ago, Spacehound said:

Nothing after the source can improve on it.

 

What is source? Acoustic wave before microphone?

 

41 minutes ago, Spacehound said:

Blind  tests can reveal price/performsnce ratios. Nothing else.

 

I'd looks wider to the matter. You forgot about development. It is almost impossibly to made home blind test.

 

41 minutes ago, Spacehound said:

I will look at speaker reviews to see if they have  a flat frequency response/good phase performance. If not, I won't buy them.

Everything else I just buy what I take a 'fancy' to and can afford.  I almost never  listen before buying it and have not done so for at least ten years.

 

Technically correct measurements have more than pair dimensions. Flat frequrency response say nothing about non-linear distortions.

 

There is need to scan linearity of in/out voltage for full frequency band including ultrasound area. And when you have these results, there is very complex analyzis stage with accounting of psychoacoustics.

 

 

41 minutes ago, Spacehound said:

I almost never  listen before buying it and have not done so for at least ten years. It's the same with cars - it took me all of  15 minutes to make up mind, go to the Mercedes dealer, and buy one,  mostly because he's the nearest car dealer to where we live (UK)

 

I more sensitive to car, that I ride But I can understand this approach in other things.

[Spacehound]

23 minutes ago, audiventory said:

 

What is source? Acoustic wave before microphone?

 

 

I'd looks wider to the matter. You forgot about development. It is almost impossibly to made home blind test.

 

 

Technically correct measurements have more than pair dimensions. Flat frequrency response say nothing about non-linear distortions.

 

There is need to scan linearity of in/out voltage for full frequency band including ultrasound area. And when you have these results, there is very complex analyzis stage with accounting of psychoacoustics.

 

 

 

I more sensitive to car, that I ride But I can understand this approach in other things.

No. It's the CD, file or whatever. They are not playing live in our rooms so that is all  most people have

 

Yes,  you cannot  easily  do good blind tests at home. 

 

I agree with the rest.

 

My car is an SLK 55, small but with a  5.5 litre V8.  It  does what I expected it to do and a lot more   

 

 

[audiventory]

1 hour ago, Spacehound said:

No. It's the CD, file or whatever. They are not playing live in our rooms so that is all  most people have

 

CD is product of sound enhancement, in general case. We can remaster it and give more "nice" sound Subjectively, of course.

 

1 hour ago, Spacehound said:

My car is an SLK 55, small but with a  5.5 litre V8.  It  does what I expected it to do and a lot more   

I like your choice!

[Jud]

11 hours ago, Spacehound said:

Sure it is.

We are doing a 'survey'  even if it's only of one person, and asking the listener whether  he can hear a difference or not.  That's binary.

What happens between his ears and his mouth is irrelevant.

 

Also what he says doesn't have to be true, true/false is binary too,  Nor does any overall result if there are lots of people. 

(I'm not having you  as a lawyer, you don't even know that things can be both objective and wrong )

 

Ah, the things you don't know about blind testing.

 

There's a lot in the scientific literature. If you're curious about the various complexities, you can start out looking at one of them by doing a search for the Iowa Gambling Task, and thinking about the implications for tests requiring conscious verbal responses. 

[Spacehound]

16 minutes ago, Jud said:

 

Ah, the things you don't know about blind testing.

 

There's a lot in the scientific literature. If you're curious about the various complexities, you can start out looking at one of them by doing a search for the Iowa Gambling Task, and thinking about the implications for tests requiring conscious verbal responses. 

.  

[buonassi]

 

14 hours ago, Spacehound said:

 

In practice? If you cant  hear any difference buy the lower cost box.

That's why most  Hifi manufacturers HATE blind tests and will come up with near endless utter nonsense to (fail to) convince sane people they don't work. It is why most US manufacturers (particulary)  have totally abandoned  the  realms of high fidelity and gone for 'impressive' instead.

 

well said.  I totally agree with your comment, and you know what?..... I can't help but ask myself this when i eval gear at home:

 

"If I didn't have them side by side, could I tell you which was which?  am I really missing any musical enjoyment by foregoing the more expensive unit"?

 

my answers are almost always "no and no" as I climb higher up the chain...... but.....

 

My point is..... I buy two of something when I make a purchase (ie headamp).  I later I sell the one I like less after courting them for a nice long time.   I find that differences are revealed as you became more "tuned in" to each unit's presentation and subtleties - and this takes time.  For me at least, once I hear something (like or dislike), my brain becomes trained and the affinity or aversion to what I'm hearing becomes much stronger.  Yes, "ringing" I'm looking at you.  I'm clearly forming a preference (subjective) and it's one that I can discern (again subjective).  What was first answered as "no and no" becomes "yes and yes" the longer of time you spend with those units.  

 

This hobby, or in my case, audiophile curse, is all about a few percent of subjective SQ improvement with each change.   And let's agree on this at least:  2% here and 3% there add up to a final chain that brings joy and satisfaction....for a little while at least.