[Digital Enigma]

Hey I’m just happy the thread is back on track

 

[bigshot]

Can you provide an example of such artifact, before and after it is fixed?

Adding to that question... and what kind of artifacts? It certainly can't fix compression artifacting or clipping.

 

[gregorio]

Apodizing filters can solve the artifacts caused by poor recordings.

As @danadam requested, please provide some examples of poor recordings where the audible artefacts are solved by an apodizing filter, and along the same lines:

The hearing sense will feel that the sound is far away, and some Apodizing filters will feel that the high frequency is slightly dark;

What apodizing filters that are not a pathologically bad/broken design (Eg. That deliberately reduce fidelity within the adult audible freq range) will make “the hearing sense feel that the sound is far away” and “slightly dark”?

G

 

[KMann]

Could you please go into a bit more detail on what you mean by "an illegal signal" and "other illegal content". Thanks.

From what I understand, the Nyquist-Shannon sampling theorem simply states that the sample rate must be at least twice the bandwidth of the desired signal to avoid aliasing. So the 22.05kHz bandwidth limit is merely a function of having the sample rate used for 'CD quality' digital audio set at 44.1kHz. Now if we were talking about 'Hi-Res' digital audio with a 192kHz sample rate, then its Nyquist-defined bandwidth would be from 0Hz-96kHz.

An impulse or a square wave or a step function are some examples of 'illegal' signals because their analog equivalent will have an infinite bandwidth (it is a sudden instantaneous change with infinite slope). If they had ever truly passed an ADC (even a mediocre one at that), you would still see the ADC filter's response instead (a combined response of both its analog antialiasing filter and its digital filters) as some pre and post ringing and the extent of the ringing will be dependent on the ADC. Illegal because, you should never be able to see an impulse or a step signal in music that was recorded and passed through an ADC.

There are exceptions to this, during mastering, the signals can get clipped, or in cases of badly mastered tracks (for example victims of loudness wars), you can see severe nonlinearity (peaks are just saturated and flat). These are artifacts introduced after they have been recorded, these are not true band limited signals either. So, when these 'illegal' signals pass through another low-pass filter, ringing is expected (due to Gibbs phenomenon) when there is jump discontinuity. However, the frequency of the ringing is closer to the cutoff frequency of the filter and are often times are just outside the audible band. So, the audibility of such ringing is questionable.

 

[KMann]

Can you provide an example of such artifact, before and after it is fixed?

A case can be made for an apodizing filter where the filter cuts off at 20kHz and opening up the possibility of reducing any aliased frequencies that folded back due to the ADC's filters not being good enough. So, objectively it could be argued that an apodizing filter can 'correct' for some of the ADC errors. The flip side to this is, given it is above the audible range, will it matter subjectively? If the roll-off starts any earlier than 20kHz, then it also opens up the possibility of reducing the highs.

Adding to that question... and what kind of artifacts? It certainly can't fix compression artifacting or clipping.

it depends on how the filter is designed. An apodizing filter that is not abrupt (i.e., no sharp edge), but at the same time has high attenuation, can 'round' the edges where clipping occurs without much appearance of ringing. This would be easy to show and is aesthetically more pleasing to look at. When it comes to audibility, the fact still remains, this is still past the audible frequency range, so will it matter?

An apodizing filter can sound different than a non apodizing filter, but I don't think one can attribute these differences to what they do to frequencies beyond what we can hear.

 

[bigshot]

OK. It seems we are talking entirely in theory here. You haven't actually heard it correct artifacts.

flip side to this is, given it is above the audible range, will it matter subjectively? If the roll-off starts any earlier than 20kHz, then it also opens up the possibility of reducing the highs.

The answer to your first question is, it won't matter at all. And the answer to the second sentence is oversampling.

 

[KMann]

OK. It seems we are talking entirely in theory here. You haven't actually heard it correct artifacts.

In theory, does it correct artifacts? yes, it can correct some artifacts. Have I heard it correct artifacts? I can't hear past 16kHz.

And the answer to the second sentence is oversampling.

What question was this whose answer is oversampling?

PS: Both my questions were meant to be rhetorical.

 

[bigshot]

I'm interested in information that can be used to make music sound better in the home. I understand that other people have fun with"what ifs" and discussions about theoretical or inaudible sound. It's just tough to discern the difference between stuff that matters and stuff that doesn't sometimes.

 

[71 dB]

I'm interested in information that can be used to make music sound better in the home.

- Improve room acoustics
- Fine tune speaker placement and "toe in" angle
- Optimise listening point
- Upgrade speakers if necessary

That's about it. If I forgot something, feel free to add.

 

[bigshot]

That seems like a good list to me! Oddly those subjects are rarely discussed at Head-Fi.

I’d add judicious equalization and signal processing to your list.

 

[71 dB]

That seems like a good list to me! Oddly those subjects are rarely discussed at Head-Fi.

Expensive DACs and snake oil cables have a shimmering mysticism to them for many that can feel "sexier" than installing acoustic panels on your walls and moving your speakers inch by inch to perfect the soundstage. Also, there is little doubt for these methods of working, so where is the debate?

I’d add judicious equalization and signal processing to your list.

Yeah, that is a good add. Thanks!

 

[KMann]

I'm interested in information that can be used to make music sound better in the home. I understand that other people have fun with"what ifs" and discussions about theoretical or inaudible sound. It's just tough to discern the difference between stuff that matters and stuff that doesn't sometimes.

I was just helping keep the 'eye on the ball' so to speak. It is easy to loose focus when there are varying opinions and many of them can be correct too. Theory helps understand what is happening to the signals, how we perceive it is another story.

For example if a process improves a reconstructed signal in the audible range, is it inaudible? And how will one know if the process in fact improved the accuracy of reconstructed signal? A frequency domain analysis will not provide the answer because there are infinite ways where a signal can be different in time domain but be exactly the same in frequency domain. An extreme example is an impulse, an impulse is when all the frequencies arrive at the same instant in time, but you get white noise when all those frequencies are spread apart. Digital filters do not affect frequency response alone, they do affect how the different frequencies pass through them and the delay each frequency goes through.

There is a lot of discussion on frequency domain as it is the easiest thing to measure, but we have not learnt to analyze the time domain signals as much and gain a deeper understanding of what matters there too. For example in headphones, it is very easy to look at a frequency response and judge the tonal characteristics and even use Harman curve as way to know how a headphone may be liked by a majority. If the headphone had very low distortion, then it would be relatively easy to EQ it to the Harman curve. If two different headphones are EQ'd to the same curve and both had very low distortion (below the audibility threshold of what is accepted), will that mean no one can tell the two headphones apart in a blind test (assume level matching, etc)? I do not think so, because, based on the transducer design and ear cup design etc., the time domain characteristics will be very different. One may be able to look at the waterfall plots and tell that they are different as it does show the behavior, but even there it will be hard to judge other factors such as depth and width of soundstage or resolution.

 

[bigshot]

Headphones have degrees of openess and closedness and they fit on different people’s heads differently. They also may have different dynamics and handle bass differently. These can affect how they sound. But you could EQ them to match your target curve and other than those things, it would probably be close enough to make you happy.

An oversampling DAC should be able to give you transparent sound within the audible range. And the kind of impulse you’re talking about probably doesn’t exist in recorded music, so it doesn’t matter.

 

[VNandor]

A frequency domain analysis will not provide the answer because there are infinite ways where a signal can be different in time domain but be exactly the same in frequency domain. An extreme example is an impulse, an impulse is when all the frequencies arrive at the same instant in time, but you get white noise when all those frequencies are spread apart.

It is also an incorrect example because if you transform a dirac impulse to the frequency domain you get all the phases lined up while if you transform noise to the frequency domain the phase is going to be random. Any function that describes a phyisical quantity changing over time can be transformed back and forth between the time and frequency domain without any ambiguity.

 

[KMann]

Digital filters do not affect frequency response alone, they do affect how the different frequencies pass through them and the delay each frequency goes through.

if you transform a dirac impulse to the frequency domain you get all the phases lined up while if you transform noise to the frequency domain the phase is going to be random. Any function that describes a phyisical quantity changing over time can be transformed back and forth between the time and frequency domain without any ambiguity.

On the contrary, that is exactly my point, but most of the analysis and measurements are done to highlight the frequency response. In theory yes you can go back and forth an infinite number of times, but for measurements, a lot of information is lost as FFTs are done with fixed window sizes and applying averaging and choice of the windowing functions and duration of the window affects how the frequency response looks and the phase response becomes useless at this point as it is time averaged and there is no way to just go back from there.