Reconstruction Filters made audible?

[VNandor]

I was commenting re. videos in general; I wasn't singling out goldensound's videos, the opposite actually: I mentioned his videos are one of the better ones in terms of proving some level of topic index at least.

In general, information density in videos is simply too low for me, I prefer written media. That is not the same as telling those who make videos "to f##k off with their dumb videos", as you put it.

If I write some documentation and someone says they don't want to read it, please make a video for me (it has happened, seems to be a trend), I would still point them to the written documentation for the specifics they were after, but I wouldn't consider them rude nor would I feel offended.

bigshot specifically refused to watch the video recommended by goldensound. He very much singled it out.
To me, it's clear he doesn't refuse watching the video because he lacks the time or because he prefers text. He refuses to acknowledge the video because he prefers to not get his views on what's potentially audible be challenged.

I see the context to this is already provided. I want to add it's not specific just to this particular case, the same happened with the klippel THD tests and with various lossy compression tests. Whenever evidence is provided that something is audible under controlled conditions, he tries his hardest to dismiss it either by refusing the acknowledge it just like here, accusing people of cheating, or saying that the tests results are irrelevant to everyone, just because it's irrelevant to him. Sometimes you have to know the context so you can read between the lines.

 

[gregorio]

But imagine that you put in hours of effort to craft a compelling argument for a situation exactly like this just so some randoms can effectively tell you to f##k off with your dumb video because there is no way I am going to watch that. Rude, isn't it?

I did watch it personally and it is an interesting watch. I’m having a lot of trouble just blindly accepting the results though. According to his own FFT data, the difference starts at around -98dB. Assuming a reasonable peak listening level of 80dBSPL, the difference would therefore be at roughly -18dBSPL. If that difference were smack in the middle of the critical band, still it would be an order of magnitude or so better than anyone has ever demonstrated, even with a young child’s hearing in a world class anechoic chamber. So this result would already be fantastic/unbelievable, however, the difference isn’t in the critical band, it’s at about 20-20.5kHz, which makes it another order of magnitude or two more fantastic/unbelievable. Additionally, he claims one of his identifying differences is according to “how big” it sounds, the spatial information/echoes but “bigness” doesn’t correlate with high treble let alone ultrasonic content, it correlates with high bass/mid freqs and spatial information/echoes usually has no content above about 12kHz let alone audible content at 20kHz and again is typically much lower in the freq spectrum <8kHz.

I don’t know what’s going on, the bigness/spatial info would indicate it’s something far lower in the spectrum he’s picking up on and therefore IMD from his amp or cans might be a good candidate, although there are others. He may have made some mistake aligning, saving or playing the files, it maybe a consequence of testing using a NOS DAC, it maybe he’s whacking the volume up and hearing the difference between TDPF dither and noise-shaped dither or the whole thing might be a scam. It would take some extremely robust independent examination for me personally to believe he really is able to apparently achieve something so far beyond the well established human thresholds.

G

 

[Ghoostknight]

bigshot specifically refused to watch the video recommended by goldensound. He very much singled it out.
To me, it's clear he doesn't refuse watching the video because he lacks the time or because he prefers text. He refuses to acknowledge the video because he prefers to not get his views on what's potentially audible be challenged.

I see the context to this is already provided. I want to add it's not specific just to this particular case, the same happened with the klippel THD tests and with various lossy compression tests. Whenever evidence is provided that something is audible under controlled conditions, he tries his hardest to dismiss it either by refusing the acknowledge it just like here, accusing people of cheating, or saying that the tests results are irrelevant to everyone, just because it's irrelevant to him. Sometimes you have to know the context so you can read between the lines.

its not just bigshot, tho he seems to be most "ignorant"

well what i figured out (and hence i refused to post in the sound science forum for a while) is that its literally about believe, its like trying to convince a christian god doesnt exist, no point in doing that...

 

[Ghoostknight]

I did watch it personally and it is an interesting watch. I’m having a lot of trouble just blindly accepting the results though. According to his own FFT data, the difference starts at around -98dB. Assuming a reasonable peak listening level of 80dBSPL, the difference would therefore be at roughly -18dBSPL. If that difference were smack in the middle of the critical band, still it would be an order of magnitude or so better than anyone has ever demonstrated, even with a young child’s hearing in a world class anechoic chamber. So this result would already be fantastic/unbelievable, however, the difference isn’t in the critical band, it’s at about 20-20.5kHz, which makes it another order of magnitude or two more fantastic/unbelievable. Additionally, he claims one of his identifying differences is according to “how big” it sounds, the spatial information/echoes but “bigness” doesn’t correlate with high treble let alone ultrasonic content, it correlates with high bass/mid freqs and spatial information/echoes usually has no content above about 12kHz let alone audible content at 20kHz and again is typically much lower in the freq spectrum <8kHz.

I don’t know what’s going on, the bigness/spatial info would indicate it’s something far lower in the spectrum he’s picking up on and therefore IMD from his amp or cans might be a good candidate, although there are others. He may have made some mistake aligning, saving or playing the files, it maybe a consequence of testing using a NOS DAC, it maybe he’s whacking the volume up and hearing the difference between TDPF dither and noise-shaped dither or the whole thing might be a scam. It would take some extremely robust independent examination for me personally to believe he really is able to apparently achieve something so far beyond the well established human thresholds.

G

you forget one crucial thing.... phaseshift and/or preringing

you purely go by frequency response which is the wrong way to look at it imo as you are probably right, that a -98db difference isnt likely to be audible "as is"

tho, one can enlighten me at what db level preringing exists, atleast with phaseshift im fairly certain that people can hear it if they know what to listen for and it has nothing todo with frequencyresponse
if i go by the usual animated impulse response graphs usually preringing is like 1/5 of the impulse, not sure if this is true to reality

Also @GoldenSound one would need to clarify if he is "actually" hearing a 20khz sinewave.... or if he just passed the test because of audible preringing/phaseshift
personally i can still hear a 15khz sinewave but its already in the rolloff region (flat to about 14khz) and i still can hear differences between filters...
also i barely heared of persons above 20 years that hear that high

 

[VNandor]

I don’t know what’s going on, the bigness/spatial info would indicate it’s something far lower in the spectrum he’s picking up on and therefore IMD from his amp or cans might be a good candidate, although there are others. He may have made some mistake aligning, saving or playing the files, it maybe a consequence of testing using a NOS DAC, it maybe he’s whacking the volume up and hearing the difference between TDPF dither and noise-shaped dither or the whole thing might be a scam. It would take some extremely robust independent examination for me personally to believe he really is able to apparently achieve something so far beyond the well established human thresholds.

I don't think he made a mistake when editing the files, it looks correct to me. I certainly have no evidence he messed up there. He specifically pointed out he did not use any dither when exporting the files. If there is dither, it could only come from the original track. I think using a NOS DAC is valid when testing the differences of these kind of anti-image filters. The filters tested are there to massively reduce the high frequencies. Normally, I wouldn't think the levels are high enough to cause even a measureable IMD but given the circumstances it might be something to focus on.

According to his own FFT data, the difference starts at around -98dB. Assuming a reasonable peak listening level of 80dBSPL, the difference would therefore be at roughly -18dBSPL. If that difference were smack in the middle of the critical band, still it would be an order of magnitude or so better than anyone has ever demonstrated, even with a young child’s hearing in a world class anechoic chamber. So this result would already be fantastic/unbelievable, however, the difference isn’t in the critical band, it’s at about 20-20.5kHz, which makes it another order of magnitude or two more fantastic/unbelievable.

At midway through the tests, he switches to listening to the transients specifically instead of the spaciousness of the song as a whole. Maybe when he listened to the song as a whole, there were only short sections that sounded different (and after the switch he definitely focused on short sections only). If one could accurately determine which short time window he actually used to differentiate between the samples, I'm "sure" the difference in the restricted part would be higher.

Additionally, he claims one of his identifying differences is according to “how big” it sounds, the spatial information/echoes but “bigness” doesn’t correlate with high treble let alone ultrasonic content, it correlates with high bass/mid freqs and spatial information/echoes usually has no content above about 12kHz let alone audible content at 20kHz and again is typically much lower in the freq spectrum <8kHz.

Is it really so unlikely that some change in sound comes across as spaciousness even though the change is not in crosstalk or reverb or something else that's commonly linked to the feeling of spaciousness? It's established the perception of sound don't necessarily have to line up with the actual changes in sound. As an example, a small change in volume would not come across as a small change in volume, people would use a ton of vague descriptions to describe the change.


The peak differences were around -40dBFS and at extremely high frequencies which make the result of the test fantastical. I understand the need for more further testing for a claim so outlandish. I would have loved to see a recording of the DAC or even the amp output during the test as that would not have been hard to organize and could remove an other layer of potential pitfalls. Something harder to do would be testing the headphones specifically focused on high frequencies and distortions creating frequencies below the fundamental. I think at some point he did a more thorough hearing test than just blasting sine waves to his ear and it turned out he still hears frequencies above 20kHz so that part could already be somewhat covered.

I'm sitting in the extraordinary claims require extraordinary evidence camp so I guess I'll never convince myself this is a one in a million situation and not just some unrecognized mistake despite a competently done test. I would be very interested in a follow up video which is way more than I could say about the typical youtube """testing""" videos (such as the one originally posted) but I don't think I'll ever get one. The views and to some extent the reactions here on the already existing one indicates there's just no real interest in it so there is not point in making it.

 

[gregorio]

well what i figured out (and hence i refused to post in the sound science forum for a while) is that its literally about believe, its like trying to convince a christian god doesnt exist, no point in doing that...

You figured out something incorrectly then. And, it’s absolutely nothing like trying to convince a Christian there’s no god. We have a large body of reliable, repeatable evidence which conforms to the scientific method and therefore results in well established science, Christians have none at all. To demonstrate that well established science is incorrect will require some very robust, reliable, repeatable evidence also conforming to the scientific method of course. This is all summed up in the old adage “extraordinary claims require extraordinary evidence” and while GoldenSound’s video appear to provide more reliable evidence than the typical audiophile impressions, still a YouTube video by someone with vested interests is a long way short of extraordinary evidence.

you forget one crucial thing.... phaseshift and/or preringing

Given a typical fast roll-off filter as he suggests, it will always be a linear phase filter, therefore no phase shift and so it’s totally valid to ignore phase. And as for preringing:

tho, one can enlighten me at what db level preringing exists

Given the music he used for testing didn’t contain any Dirac impulses and didn’t appear smashed to death with a limiter and therefore somewhat similar to a Dirac impulse then the level of pre-ringing would be roughly -infinity dB, IE. None at all. I’d have to see the audio file to be certain but if there is any at all, it would be miniscule.

G

 

[gregorio]

Seems I was writing my response as you were posting yours.

I don't think he made a mistake when editing the files, it looks correct to me. I certainly have no evidence he messed up there.

Same here, it looks correct to me and I have no evidence to the contrary. The only thing I’ve got is a couple of experiences when such editing of files looked equally as correct to me but there was an inadvertent error. I noticed when he was editing there was part of the waveform that looked like a glitch, the sort of basic error you get when not using a zero crossing editing point, although it was on both files. Even though it looks tiny on a digital waveform, it can actually cause a full scale overload, equally it could be inaudible, no way to tell without listening to it.

He specifically pointed out he did not use any dither when exporting the files.

He stated he didn’t add any when saving the files but that the “typical filter” he compared used TPDF dither while the super steep filter used noise shaped dither.

Normally, I wouldn't think the levels are high enough to cause even a measureable IMD but given the circumstances it might be something to focus on.

It does seem somewhat unlikely I agree but as you say, given the far less plausible conclusion …

Is it really so unlikely that some change in sound comes across as spaciousness even though the change is not in crosstalk or reverb or something else that's commonly linked to the feeling of spaciousness?

It’s not just some change in sound though, it’s specifically a change in ultrasonic sound, if we take the claim at face value. He actually stated how “big”, which would indicate way lower in the spectrum (in the hundreds rather than high thousands of hertz), although I suppose “airiness” could be interpreted as spaciousness and that does occur in the treble (roughly the 13-16kHz range).

I would be very interested in a follow up video which is way more than I could say about the typical youtube """testing""" videos …

He’s British and there are quite a few British universities who’d love a subject with seemingly superhuman hearing abilities for a nice juicy paper. That’s what I’d like to see, scientific verification without a vested interest. I’d be more than surprised if that ever happens as it’s pretty much a no win situation for him. Either it’s demonstrated to be some mistake, he has normal hearing and can’t discern what he’s claiming or it’s demonstrated that he differs from almost all other humans and can discern what he claims, which would make his assertions on audibility true but inapplicable to all or virtually all other audiophiles.

G

 

[VNandor]

He stated he didn’t add any when saving the files but that the “typical filter” he compared used TPDF dither while the super steep filter used noise shaped dither.

I completely missed that detail.

I noticed when he was editing there was part of the waveform that looked like a glitch, the sort of basic error you get when not using a zero crossing editing point, although it was on both files. Even though it looks tiny on a digital waveform, it can actually cause a full scale overload, equally it could be inaudible, no way to tell without listening to it.

I'm fairly sure that is also present on the unedited track. The original track is full of these transients that look a lot like an editing error, especially at the zoom level he used.

He’s British and there are quite a few British universities who’d love a subject with seemingly superhuman hearing abilities for a nice juicy paper. That’s what I’d like to see, scientific verification without a vested interest. I’d be more than surprised if that ever happens as it’s pretty much a no win situation for him.

I really have no idea how this kind of research gets done. Isn't this something he would get compensated for?

 

[gregorio]

I completely missed that detail.

Yep, it’s certainly a possibility. While below audibility at reasonable listening levels, it’s not hard to differentiate at levels 20dB or so higher, although not so much using the sections he appeared to use for differentiation.

I'm fairly sure that is also present on the unedited track. The original track is full of these transients that look a lot like an editing error, especially at the zoom level he used.

I don’t know exactly what zoom level he’s at but the only waveforms I’ve ever seen like that were always glitches. I didn’t notice any in the audio being streamed though.

I really have no idea how this kind of research gets done. Isn't this something he would get compensated for?

Typically a researcher becomes aware of someone demonstrating some unusual ability, either from that person contacting them directly, someone else contacting on behalf of the person or indirectly through another researcher, university or institution. For example an audiologist mentioning to a contact about some particularly sensitive patient and word getting back to the research who then asks the audiologist see if their patient is OK with being contacted by a researcher. This last way, word of mouth, is probably the most common way.

And yes, he would get compensated, there is typically an “honorarium” even for just being one of many ordinary subjects in a study, for a special individual case, it would certainly be full expenses plus a reasonable honorarium. Don’t ask me exactly what, it can vary enormously, depending on various factors.

G

 

[bigshot]

I’d like to see him pass a listening test between two midrange NOS DACs without selectable filters. I’m sure you can find “high quality” filters that degrade the sound. I don’t think you can tell the difference between DACs with the proper filter. It isn’t hard to find one of those.

 

[Ghoostknight]

We have a large body of reliable, repeatable evidence which conforms to the scientific method and therefore results in well established science,

yea which leads to your "believe window", nothing wrong with that per say and no point in going on a science vs religion ramble
it was a metaphor, imo an accurate one for either subjectivist or objectivist

Given a typical fast roll-off filter as he suggests, it will always be a linear phase filter, therefore no phase shift and so it’s totally valid to ignore phase. And as for preringing:

i just included phaseshift since its a option with some reconstruction filters, tho i doublechecked, goldensound was using linear phase filters

Given the music he used for testing didn’t contain any Dirac impulses and didn’t appear smashed to death with a limiter and therefore somewhat similar to a Dirac impulse then the level of pre-ringing would be roughly -infinity dB, IE. None at all. I’d have to see the audio file to be certain but if there is any at all, it would be miniscule.

sounds unrealistic

a highpass filter in the bassregion in FIR mode is definitely audible, most engineers will tell you that and there are enough comparison videos on youtube to check for yourself

while a lowpass at 20khz is definitely less audible than the above highpass im fairly certain its nowhere near "-infinity dB"
found some measurements but as usual, no real db scale, but also something like 1/5 preringing of the signal
http://archimago.blogspot.com/2018/02/musingsmeasurements-on-blurring-and-why.html

I’d like to see him pass a listening test between two midrange NOS DACs without selectable filters. I’m sure you can find “high quality” filters that degrade the sound. I don’t think you can tell the difference between DACs with the proper filter.

of course lets trough variables at it till no difference is discernible anymore
these "high quality filters that degrade the sound" actually measure objectively better...

imo you got things way wrong here..... "dacs with proper filters" have the same crappy filters and therefore sound similar to one another

 

[VNandor]

sounds unrealistic

That's especially rich coming from you. I'm sure according to you, the crystal sitting in your grounding box opening up the soundstage is somehow a very realistic and reasonable and not insane at all take!

Unfortunately, for any kind of "ringing" to occur there must be a sharp spectral cutoff going on. The music is already low passed just above 20kHz so even the peak levels above 20kHz is well below 40dB compared to the average level of the rest. Even when focusing only on the short window of the transient parts. It's not "according to me" or I something "I believe" it is something that could be verified by anyone having a spectrum analyzer and knowing how to use it which I imagine excludes you so you either "believe it or not".

 

[Ghoostknight]

I'm sure according to you, the crystal sitting in your grounding box opening up the soundstage is somehow a very realistic and reasonable and not insane at all take!

well, my 432Hz tuning work "just sitting in the room" either hit by sound or air/wind producing a tone under the room noisefloor which isnt audible as a tone still makes a audible difference, just to throw something out there

Unfortunately, for any kind of "ringing" to occur there must be a sharp spectral cutoff going on. The music is already low passed just above 20kHz so even the peak levels above 20kHz is well below 40dB compared to the average level of the rest. Even when focusing only on the short window of the transient parts. It's not "according to me" or I something "I believe" it is something that could be verified by anyone having a spectrum analyzer and knowing how to use it which I imagine excludes you so you either "believe it or not".

yea, i dont, since i actually listened to an additional 16-22khz lowpass FIR filter, which others with an analyzer may have not

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, but since the rolloffs usually already "start" at 17-18khz in miniscule amounts i still believe that some ringing occours in the audible range BEFORE actual big time attenuation hits

--------

on audibility of linear phase filters:

someone in the comments put it quite right:

Linear phase mode (pre-ringing) - WOOP! WOOP!
Zero Latency mode - PAM! PAM!

Now after listening to this you might can imagine how it would sound on higher frequency transients, its literally smearing of transients imo
the "woop woop" sound i imagine comes from low bass hitting before and after the actual transient

 

[gregorio]

yea which leads to your "believe window", nothing wrong with that per say and no point in going on a science vs religion ramble
it was a metaphor, imo an accurate one for either subjectivist or objectivist

Sure, you could put it in terms of a “belief window” from the point of view that one could choose not to believe the established science. For example, you could choose to believe that 1+1≠2 and some people apparently believe the earth is flat, etc. However, a metaphor or comparison to religion is about as fallacious as it gets because one of course needs to be irrational not to believe in science.

i just included phaseshift since its a option with some reconstruction filters, tho i doublechecked, goldensound was using linear phase filters

Right, so your statement that we’re “forgetting one crucial thing, phase shift …”, is obviously not crucial, in fact it doesn’t even exist!

a highpass filter in the bassregion in FIR mode is definitely audible, most engineers will tell you that and there are enough comparison videos on youtube to check for yourself

Are these “most engineers” kids in their bedrooms or audiophiles on YouTube just calling themselves engineers? All the engineers I know actually have some education in sound/music engineering and will tell you that: 1. A high pass filter in the bass region is not a low pass filter in the ultrasonic region. 2. That even a HP linear phase filter in the bass region is definitely NOT audible unless certain conditions are met, EG. A very steep (high pole) filter AND a transient with a very high level.

while a lowpass at 20khz is definitely less audible than the above highpass im fairly certain its nowhere near "-infinity dB"
found some measurements but as usual, no real db scale, but also something like 1/5 preringing of the signal
http://archimago.blogspot.com/2018/02/musingsmeasurements-on-blurring-and-why.html

Well done, thanks for providing evidence that contracts your assertion and supports mine. Though why you would prove yourself wrong and effectively argue with yourself is a mystery! Your linked article states of ringing/pre-ringing: “Let's not spend much time on this. As I've already discussed recently …” going to that second link provides the following conclusion:
“The key here is to remember that within a properly bandwidth limited signal where all the frequencies are below Nyquist, a linear phase FIR filter actually does not create ringing regardless of the impulse response appearance.” - This raises two obvious points, firstly, as the “filter actually does not cause ringing” then how can the level of ringing be anything other than “-infinity dB”, you think maybe there’s still some dB amount of ringing when no ringing is created? And secondly, as mentioned above, even your own cited reference disagrees with you!

"dacs with proper filters" have the same crappy filters and therefore sound similar to one another

How is a “proper filter” a “crappy filter”? If it’s a “crappy filter” then by definition it is not a proper filter!

G

 

[Ghoostknight]

2. That even a HP linear phase filter in the bass region is definitely NOT audible unless certain conditions are met, EG. A very steep (high pole) filter AND a transient with a very high level.

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

every single instrument produces transients... sometimes better recorded and sometimes not... so you will end up with transients hitting the filter IF YOU USE IT.... basicly a useless argument for PLAYBACK

there probably also is a combination of sounds that mask preringing and *it depends* on how or if the preringing is audible.... my point is with the right material/songs you can hear a difference and if the music you are listening to is somewhat diverse then you will hit songs where it makes a difference...

1. A high pass filter in the bass region is not a low pass filter in the ultrasonic region.

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

Well done, thanks for providing evidence that contracts your assertion and supports mine. Though why you would prove yourself wrong and effectively argue with yourself is a mystery! Your linked article states of ringing/pre-ringing: “Let's not spend much time on this. As I've already discussed recently …” going to that second link provides the following conclusion:
“The key here is to remember that within a properly bandwidth limited signal where all the frequencies are below Nyquist, a linear phase FIR filter actually does not create ringing regardless of the impulse response appearance.” - This raises two obvious points, firstly, as the “filter actually does not cause ringing” then how can the level of ringing be anything other than “-infinity dB”, you think maybe there’s still some dB amount of ringing when no ringing is created? And secondly, as mentioned above, even your own cited reference disagrees with you!

yea i just wanted to show the measurements...

Right, so your statement that we’re “forgetting one crucial thing, phase shift …”, is obviously not crucial, in fact it doesn’t even exist!

i stated also preringing...

Sure, you could put it in terms of a “belief window” from the point of view that one could choose not to believe the established science. For example, you could choose to believe that 1+1≠2 and some people apparently believe the earth is flat, etc. However, a metaphor or comparison to religion is about as fallacious as it gets because one of course needs to be irrational not to believe in science.

you are missing the point... (or just search sentences you can blabber around some BS....)