Reconstruction Filters made audible?

[castleofargh]

After don't feed the troll, here's a new forum advice. Don't feed the astral plane.

 

[gregorio]

you wanna tell me you dont hear the difference in the video? this is laughable at best

You wanna tell me the audible difference in the video does not comply with the conditions I stated?

every single instrument produces transients...

Near full scale at the filter cut off frequency, when it’s already been anti-alias filtered?

there probably also is a combination of sounds that mask preringing and *it depends* on how or if the preringing is audible....

There’s probably a combination of sounds that mask the pre-ringing that your own cited reference stated does not exist?

and while its less obvious, imo it still can make a audible difference

So you think this non-existent pre-ringing, which is probably masked, can still make an audible difference in your opinion? That’s truly beautiful, thank you so much! Lmao

i stated also preringing...

So your “one crucial thing” was actually two crucial things, neither of which even exist but are crucial anyway? Maybe we should make a list of “one crucial things”; bass piccolos, soprano tubas, unicorns, mermaids, sane audiophiles, etc., sounds like a lot of fun! You’re on some plane well beyond the astral or any other plane I’ve ever heard of. lol

After don't feed the troll, here's a new forum advice. Don't feed the astral plane.

What if they’re in some dimension beyond the astral plane? And, do I really have to stop feeding the troll, even though it’s so funny and entertaining?

G

 

[bigshot]

of course lets trough variables at it till no difference is discernible anymore

Then you agree that standard DACs should sound the same. Then the test should be to test the fancy audiophile filters against a standard DAC and see which ones sound the same, and which ones sound different.

these "high quality filters that degrade the sound" actually measure objectively better...

Then they should sound the same as an audibly transparent standard DAC. If something sounds different, it isn’t performing to spec

 

[Ghoostknight]

So you think this non-existent pre-ringing, which is probably masked, can still make an audible difference in your opinion? That’s truly beautiful, thank you so much! Lmao

just to clarify.... do you hear a difference in the linear phase demonstration video?
that doesnt sound "non-existent" to me and many others

Then you agree that standard DACs should sound the same. Then the test should be to test the fancy audiophile filters against a standard DAC and see which ones sound the same, and which ones sound different.

well GoldenSound did just that.... picking a filter that is similar to your usual dac filters and comparing it to a high quality one

and imo you got things backwards....

its not this -> " all dacs are transparent and the high quality sounds worse "
its this -> " there is no transparent to begin with and the higher quality one simply sounds better verified by measurements AND listening.."

 

[gregorio]

just to clarify.... do you hear a difference in the linear phase demonstration video?

Yes.

that doesnt sound "non-existent" to me and many others

Who is claiming that a very steep high pass linear phase filter in the bass region, acting on near full scale transients does not produce pre-ringing? I did not state that, Archimago did not state that, so who are you arguing with and how is your assertion therefore not a strawman argument? Don’t you know that a high pass filter in the bass region is different to a low pass filter in the ultrasonic region, that the output of a filter depends on its input?

Do you have anything other than fallacies and other planes of existence?

G

 

[VNandor]

on minimal phase filters , while the main attenuation happens at 20-22khz the phaseshift that occours can start as low as 12-14khz, might be a bit different on linear phase filters,

On minimum phase filters, the phase shift caused can be clearly measured at 12-14kHz. However, linear phase filters have a linear phase as the name would imply, and the passband is also a passband as the name suggests. As such, there is no difference between the sound in the passband of a filter like this. This should be common sense and the details (such as the exact cutoff of a particular anti-image filter) could be tested with a null test by any half-competent competent person, which perfectly explains why you think a linear phase anti-imaging filter with a cutoff over 20kHz could cause ringing but why you don't provide evidence.

Your video about ringing caused by a 20dB peak at 100Hz with plenty energy already there has nothing to do with an anti-imaging filter.
There is very little content to be filtered by the anti-image filter as the signal was already low-passed before. This is in stark contrast with the video you posted. In your case the difference is ~6dB concentrated at ~200Hz while the example provided with the anti-imaging filters the differences are about 0.1dB, with the difference being at mostly in the ultrasonic range (a fact that can be verified by anyone who wishes to do so). Can you see how claiming that an anti-imaging filter that's not even removing that much content in the first place won't cause preringing is different than saying a 20dB boost centered at 200Hz won't cause preringing with a snare drum? The video is a good example of preringing. It also has nothing to do with anti-imaging filters.

 

[bigshot]

well GoldenSound did just that.... picking a filter that is similar to your usual dac filters and comparing it to a high quality one

If he heard a difference, he should then compare them to other midrange DACs and figure out which of them sounds different than the rest and which one doesn’t. The one that doesn’t is the defective one because it’s highly unlikely that multiple DACs would be out of spec in the same way.

He’s assuming that his “high quality” DAC is better because in theory he thinks it should sound better. But that is expectation bias. The truth may be the opposite.

There’s only one way for a DAC to sound… accurate. There is no better than a proper DAC, only worse. It’s either right or it isn’t. Find the one that sounds different than the rest and you’ve found the defective one.

 

[eq1849]

Sure, but then with the limited compute power available in a typical DAC chip, you can't keep things flat under 20khz and actually achieve a correct Nyquist reconstruction,

This is a good point IMO.

There seems to be a common misconception the infinitely perfect math and reconstruction available in the Sampling Theorum is available in finite technology.

 

[bigshot]

There seems to be a common misconception the infinitely perfect math and reconstruction available in the Sampling Theorum is available in finite technology.

It's good enough to be perfect for the even more finite technology of human ears!

 

[Ghoostknight]

There is very little content to be filtered by the anti-image filter as the signal was already low-passed before.

lowpassed by what? do you refer to the analog lowpass on ADC's ?

It also has nothing to do with anti-imaging filters.

if i understand anti-aliasing filters correctly they are not more than a lowpass (aka some kind of EQ) themself with a different name to describe its purpose, anti-image also describes the same, so i dont think its that far off to compare linear phase EQ with "anti-aliasing filters" but of course its true, the lower the frequency the more obvious it gets but i believe the same behaviour you see/hear on low frequencys still exist on higher frequencys, timing is probably part of why it gets way less audible

i also just found out even "analog anti-aliasing filters" (some kind of lowpass) get used in specific applications

while the example provided with the anti-imaging filters the differences are about 0.1dB,

maybe the 0,1db are making the difference idk..

----

well we all saw a impulse response like this:
https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcQkw-esczQup-LLIEsYD5bHr8t_SKEFiL3wlw&s

So could someone clarify at which frequencys on a 20khz lowpass does the impulse response actually look like the above? at which frequency threshhold we can "definitely" say its just a peak in the impulse response without any preringing/postringing caused by the anti-aliasing filter? my guess would be this "alteration" of the impulse response goes way down into the audible range

----

also aliasing itself is definitely another point one should look at as these higher quality filters have better rejection at nyquist... can someone for sure say that aliasing artifacts at 18-20khz are lower than say -100db with standard dac filters?

 

[bigshot]

Inaudible. You are “hearing” expectation bias.

 

[gregorio]

This is a good point IMO.

Yes, clearly it’s a very good point; an audiophile marketing point that’s clearly “very good” because audiophiles such as yourself are being suckered by it. This allows audiophile DACs with more computing power or audiophile software products that use powerful computer CPUs to be marketed to audiophiles, who will then be happy to pay a premium for it. The best scams are the ones where those scammed never realise or believe they’ve been scammed! The actual facts are quite different different though and GoldenSound had no response when they were put to him, presumably because he knows the actual facts but chooses to push the marketing BS instead?

The actual facts are simple history, that anti-image filters “kept things flat under 20kHz and actually achieved a correct Nyquist reconstruction” with DAC chips in the 1990’s. So, obviously this “good point IMO” is actually complete BS unless you believe that DAC and other chips had more computing power in the 1990’s than they do today. And in case you weren’t aware, the world’s most powerful super computer in the late 1990’s (that cost ~$30m) had less computing power than a modern iPhone!

There seems to be a common misconception the infinitely perfect math and reconstruction available in the Sampling Theorum is available in finite technology.

Another great audiophile marketing lie! Sure, the companies that spent billions on developing and implementing digital audio, on chip R&D, on building chip manufacturing plants and on employing tens of thousands of the best scientists and engineers are all under “a common misconception” but some audiophile marketers (and those audiophiles suckered by them), most of whom have probably never even seen Shannon’s mathematical proofs, let alone understood them, have figured out and avoided this “common misconception”. Is it possible to even imagine a more extreme example of Dunning-Kruger?

What would even be the point of implementing “infinitely perfect math” when the output is an electrical (analogue) signal that according to the laws of physics can never be infinitely perfect (and that’s before we even start with acoustic sound or human hearing)? Of course though, those tens of thousands of scientists/engineers don’t know the laws of physics, or at least, don’t know them as well as a bunch of audiophiles.

Jeez, how is it possible not to be embarrassed spouting such BS and doing so in a science discussion forum of all places?

G

 

[gregorio]

lowpassed by what? do you refer to the analog lowpass on ADC's ?

He’s referring to the fact that the human voice, the vast majority of musical instruments and acoustic sounds produce far less content >20kHz to start with. For example, the highest pitched orchestral string instrument is the violin, yet only around 4% of it’s sound energy output is above 20kHz and that’s at just 1m, at an audience listening position it’s more like 1% or 0%. On top of that there is an analogue low pass (anti-alias) filter on ADCs, then there’s a low pass (decimation) filter also in ADCs and finally, there’s typically several to several dozen more anti-alias and anti-image filters applied during mixing and mastering, due to oversampling plugins, round trip conversions to analogue processing/effects units, or both.

i also just found out even "analog anti-aliasing filters" (some kind of lowpass) get used in specific applications

Analogue anti-aliasing filters are employed in all ADCs, along with subsequent decimation filters, which are also anti-alias filters.

So could someone clarify at which frequencys on a 20khz lowpass does the impulse response actually look like the above?

At NO frequencies “does the impulse response actually look like the above” in music recordings because music recordings do not contain such Dirac Impulses to start with and therefore cannot possibly have such a response! Pre and post ringing can occur but it requires certain conditions (typically, massively over compressed/limited transients) but it’s still relatively uncommon and still does NOT “actually look like the above”, it is far lower in level and what tiny level there is (if any) is concentrated at the filter cut off (stop band) frequency, which in the case of 44.1kHz would be around 22.05kHz. Hence why I stated that the track in question probably has roughly “-infinity dB” pre-ringing, because it does not appear to be massively over compressed.

can someone for sure say that aliasing artifacts at 18-20khz are lower than say -100db with standard dac filters?

Anyone who knows digital audio basics can “say for sure that aliasing artefacts at 18-20kHz are lower than say -100dB with standard DAC filters”, in fact the actual level “for sure” is -infinity dB because DAC filters cannot and do not create any aliasing artefacts! Alias artefacts can ONLY be produced during the ADC (or other re or down sampling) processes, not during the DAC (upsampling) process, hence why ADCs (and certain plugin processors) have anti-alias filters and DACs do not.

G

 

[VNandor]

lowpassed by what? do you refer to the analog lowpass on ADC's ?

Goldensound compared two different anti-imaging filters. The test used electronic music, presumably with no ADC involved apart from the samples used but I can't know that for sure. ADCs have gentle analog filters but employ steep digital low-pass filters by the way. I'm referring to all the (digital) filters involved in the production. Most likely every synth and non-linear effects use some. These are similar to anti-image filters. Very likely steep, linear-phase lowpass filters near half of the sampling rate.

if i understand anti-aliasing filters correctly they are not more than a lowpass (aka some kind of EQ) themself with a different name to describe its purpose, anti-image also describes the same, so i dont think its that far off to compare linear phase EQ with "anti-aliasing filters" but of course its true, the lower the frequency the more obvious it gets but i believe the same behaviour you see/hear on low frequencys still exist on higher frequencys, timing is probably part of why it gets way less audible

Yes, both anti-alias and anti-image filtering implies a steep linear-phase low pass filter set around half the nyquist frequency. Which is exactly why your example of a 20dB boost at 100Hz with a signal already containing plenty of that frequency already is actually not a good example of how music would react to an anti-image filter.

i believe the same behaviour you see/hear on low frequencys still exist on higher frequencys, timing is probably part of why it gets way less audible

It can exist and when it does, it can be demonstrated that it does. However, the filter will not produce ringing when there is nothing to filter, for example a filter with a 20kHz cutoff will not cause any ringing when the signal already cuts off at 10kHz. The reason people don't notice a high frequency cutoff under controlled conditions is because the filter do not change the sound at low frequencies and people don't hear the high frequencies so there is nothing to notice.

i also just found out even "analog anti-aliasing filters" (some kind of lowpass) get used in specific applications

Yes, as an example this how how most ADCs work. It uses a low order analog low pass filter that starts cutting off well above the audible frequencies. The filtering is still needed because despite the high frequency sampling, aliasing will still occur if there's enough signal above fs/2. The filter used for the downsampling is still the typical high order digital filter though. Is this something you don't know despite arguing about how sampling works for years at this point?

So could someone clarify at which frequencys on a 20khz lowpass does the impulse response actually look like the above? at which frequency threshhold we can "definitely" say its just a peak in the impulse response without any preringing/postringing caused by the anti-aliasing filter? my guess would be this "alteration" of the impulse response goes way down into the audible range

I can somewhat clarify it. Your guess is wrong, this "alteration" will not go down into the audible range if the cutoff is not set in the audible range. On top of that, if the signal doesn't contain a significant enough portion above the cutoff frequency, there will be no ringing regardless of where the cutoff frequency is. There's no fits all rule of when this definitely applies but generally it will not happen with typical music and typical filters and if a suspicion rises that it does, it can be demonstrated without any "guessing" or "believing".

As a quick example here's a guitar pluck "without" a lowpass and with a 10kHz brickwall filter lowpass done by audacity's linear phase EQ.
No ringing as one would expect.

I don't really understand why you are so unwilling to test something like this. Ringing lives rent-free in your mind for years and yet you never even bothered to test if it exists outside of your head or not. It is extremely unlikely to happen with well produced music and proper filters. This is a filter at 10(!)kHz, with a 100Hz or so transition band. And still no signs of ringing. And just because there might be one or ten example where there is some ringing it doesn't mean that there is always ringing happening just because you always "hear" it because in fact, preringing does not always happen it almost never happens.

 

[gregorio]

The test used electronic music, presumably with no ADC involved but I can't know that for sure.

Just in terms of a “by the way” and not wishing to dilute your point at all, but most likely there would be an ADC involved and very possibly more than one. Even since electronic music really took off as a mainstream genre (set of sub-genres) in the 1990’s and obviously relied almost entirely on digital technology; samplers, sequencers and eventually soft-synths and sample library software “units”, it was still standard practice to come “out of the box” (the DAW) into the analogue domain for certain processing, especially the addition of saturation/distortion. Even though modern saturation/distortion plugins are great bits of kit and far superior to those in the 1990’s, commercial studios still maintain a catalogue of vintage analogue units and even professional level project studios still usually have a rack with at least a couple/several analogue units. Unless it’s a “bedroom”/amateur production, the likelihood of all the mixing and mastering of a commercial electronic music release being entirely “in the box” is not very high. I can’t give you an accurate percentage but I’d be very surprised if it were as much as 50% and may well be significantly less than 20%.

G