The Audiophile Challenge

  On 4/16/2018 at 6:57 PM, mcbpete said:

Apparently in that Olveros piece it's 11x >20khz signals beating against a 1 Hz signal ( NOISEGATE 11 - http://www.stalk.net/paradigm/pd04.html ). Beat frequency is equal to the difference in frequency of two waves so the higher pitched oscilators are all being pulled down significantly by the 1Hz one rather than them all adding up to make frequencies in the 100+ Khz range.

 

 

Thanks, I remember them mentioning the specific frequencies when I studied this stuff in college but that was, shit, like 16 years ago at this point.  I knew it was sub and super-sonic but the only thing I'd read recently was an interview with Oliveros where she talked in passing about the bias tone being one of the sources, which is something they didn't mention in school.

I quite enjoyed the brief bit I heard, I love pure-tone synthesis stuff (like stuff by Eleh). I have to be in the right mood for it, but when I do I love the physicality of it

imd distortion, beats, combination tones, etc. do happen yeah, theoretically you can combine tones that are over the frequency range of human hearing generating new tones that are in the human hearing range but my question here is how does one produce such tones if every audio equipment out there has a frequency range of 20h-20khz? might be a dumb question but it's a genuine and honest one...

Edited April 16, 2018 by THIS IS MICHAEL JACKSON

  On 4/16/2018 at 6:48 PM, eugene said:

as i said, choose ANY track(s) you want and i'll cut you a lossless and lossy samples out of them for you to ABX.

 

  Quote

 

Regarding the psychoacoustics of lossy compression, we're talking about 25 year old technology based on 60+ year old research and aimed at getting acceptable sound quality at file sizes that are friendly to dial up networks, and it did that.  Even expecting it to be inaudible doesn't make sense. And again, the linear mathematical models it was based on are old and flawed. The time and frequency domain resolution of human hearing is far more sensitive than the mathematical models and lab research from the 1940s and 50s that people are still basing their assumptions about human hearing on could capture.

 

then logically it would be easy to ABX, yet there are almost no examples of people successfully abxing high bitrate mp3s and other more modern codecs. so maybe the codec programmers know what they're doing after all?

 

regarding the olveros example - imd distortion and aliasing artifacts (though less relevant here) do indeed exist, so what exactly is your point?

 

The Oliveros thing was a tangent.

 

All I know is in the ABX tests I've done, both self administered and administered by other people (although the only one I've had administered by another person was on ATRAC rather than mp3) there's a very characteristic softening of transients at frequency extremes in harmonically rich content (similar to the sound of a CD-R burned at high speed and played back on a CD player with poor error correction, which isn't really relevant today since everything has decent error correction now) that's easy to spot, and an overall loss of high frequencies, in lossy codecs in general.  And usually I actually prefer it to the uncompressed audio.

 

I've never really paid attention to the current state of lossy codec testing because for me mp3 is fine for sharing stuff and everything is available lossless now so for listening it doesn't really matter, get it lossless and encode your own mp3s for portable devices. I'd definitely be interested in seeing some results of properly administered ABX tests conducted recently by independent (not industry-funded) researchers.

 

I'll think of some tracks at work today and post a list tonight or tomorrow.

  On 4/16/2018 at 6:57 PM, mcbpete said:

Apparently in that Olveros piece it's 11x >20khz signals beating against a 1 Hz signal

:scratches head:

11 of the generators were set to operate above 20,000Hz...

the way you put it sounds like they're producing frequencies 11 times higher than 20kHz...

Soz by '11x' I meant '11 lots of', but yeah it does read the other way !

the test you can do for yourself in foobar is as properly administered as it gets. it does control for all other factors and calculates the probabilities for guessing. it's all that you really need to test lossy codecs, it's how it's done by codec devs themselves on hydrogen audio for example.

 

along with track names do post the time codes of desired parts, i think we'll stay in the clear legally if the samples are below 30 secs in such context.

...and the second of these tracks is appearing here for the first time. I of IV (1966) is a work of exceptional depth and penetration, using a tape delay system in combination with 12 sine tone square wave generators, an organ keyboard, a spring-type reverberation unit, two line amplifiers, two stereo...

 

as soon as i read that^^^ i knew that whole article/review was full of shit... sine tone square wave generators??? wtf is that? is it a sine wave generator or a square wave generator? :sighs:

Edited April 16, 2018 by THIS IS MICHAEL JACKSON

I'll read these posts later but I've got an unfolding family medical crisis (ironically it involves hearing) that just got more serious than I expected it to be, and I really have more important things to think about than tracks for an AB test on a message board. So if you want to upload some clips go for it, I'll AB them Wednesday, but after the last hour here I just don't have the energy to choose my own, and things won't be improving any time soon.

 

 

EDIT: how about this:

 

some tracks from Pauline Oliveros/Deep Listening Band's "The Readymade Boomerang" (I don't really care which), I think those would be really suitable for this kind of testing and I'm already pretty familiar with them.  Maybe some kind of slickly produced 80s pop-classical like Penguin Cafe Orchestra or Michael Nyman would be good, I couldn't really name a particular piece by either of them I'd choose, I'm not really a big listener of either but I've heard them enough over the years to know their general production style.  Maybe the soundtrack to The Cook, The Thief, His Wife and Her Lover would be a good choice, particularly the more sparse tracks that are mainly voice.

 

Most of my CDs have been packed away for years but I might be able to dig up the Oliveros album.  But not any time soon.

Edited April 16, 2018 by RSP

In the mean time here's an Ars Technica article that does a good job of summarizing the ramifications of the study I linked above, specifically relating them to the shortcomings of mp3 and other lossy algorithms that rely on assumptions about hearing derived from Fourier's research.

 

https://arstechnica.com/science/2013/02/human-hearing-beats-sounds-uncertainty-limit-makes-mp3s-sound-worse/

man, i'm not really interested in "in the meantime"s or whatever. you made a pretty ridiculous claim and since making it did everything to weasel out of actually proving it for real. the developers of lossy codecs already take into account the peculiarities human hearing, however it might work, simply by putting their encoded files to blind tests. so i don't understand what exactly are you trying to prove here by linking to that stuff.

 

for the third time, if you still want to stand by your claim that "the difference between 320kbps mp3 and CD audio is night and day." pick a track(s) that you think will tax the mp3 codec the most and i'll prepare the stuff for you to test. after all it was your claim that those tracks in the online test are not good enough for such tests, so naturally it's up to you to choose the right ones, otherwise you'll simply claim that that tracks i chose aren't appropriate as well.

Edited April 16, 2018 by eugene

the difference between 320kbps mp3 and CD audio is

Edited April 16, 2018 by THIS IS MICHAEL JACKSON

i still want that someone explains to me how can audio frequencies 5x 20kHz be recorded in reel to reel or whatever that composer used but nevermind....

As I say, the 1Hz tone brings the 11 20+ Khz signals down to audible range as a result of frequency beating (more info here, mainly the 'Beat Frequency' section - http://www.physicsclassroom.com/class/sound/Lesson-3/Interference-and-Beats )

  On 4/16/2018 at 10:39 PM, mcbpete said:

As I say, the 1Hz tone brings the 11 20+ Khz signals down to audible range as a result of frequency beating (more info here, mainly the 'Beat Frequency' section - http://www.physicsclassroom.com/class/sound/Lesson-3/Interference-and-Beats )

right, theoretical, on paper, maths... but how do you record them? if the frequency response of a reel to reel is aprox. 20hz to 20kHz... and if the speakers frequency response is also aprox. 20-20k and if microphones... same, etc etc... even if you try and do it today, in digital domain, you can generate these things with maxMSP, mathlab, c-sound, etc, but how do you actually play them or record them, that's what i would like to know... we've talked about the beat effect before... on the other thread... regarding binaural beats...

Edited April 16, 2018 by THIS IS MICHAEL JACKSON

You'd probably round 12 of these bad boys (or similar) up in parallel, into a single input of the tape recorder and the frequency beating would occur pre- the recording stage

 

 

The setup chain looking something like:

 

  On 4/16/2018 at 10:56 PM, mcbpete said:

You'd probably round 12 of these bad boys (or similar) in parallel, into a single input of the tape recorder and the frequency beating would occur pre- the recording stage

 

 

The setup chain looking something like:

 

cool, thanks for clearing my mind... :^)

she didn't even need the sub osc right? the beats produced by the over 20kHz osc's would fall in the audible range if desired... f₁=25kHz x f₂=30kHz would generate f₃=5kHz and f₄=55khz

Edited April 16, 2018 by THIS IS MICHAEL JACKSON

Yeah in my (admittedly limited) understanding of it even if you had just two sine waves as high above the human hearing of say 30Khz and 31Khz it would generate a beat frequency of 1Khz (the difference 'tween the two)

  On 4/16/2018 at 11:12 PM, mcbpete said:

Yeah in my (admittedly limited) understanding of it even if you had just two sine waves as high above the human hearing of say 30Khz and 31Khz it would generate a beat frequency of 1Khz (the difference 'tween the two)

yeah i'm sorry, my calculations where based on the ring modulation effect... it's all a big mess in my head lol, we have the inter-modulation distortion, the beat effect, the ring modulation, the combination tone, i guess they're all based in the same fundamental, sum and difference between 2 tones...

Edited April 16, 2018 by THIS IS MICHAEL JACKSON

Out of curiosity (as I didn't actually believe it myself, even though I was stating it as fact!) I created four sine waves using a 96Khz samplerate (my audio interface luckily goes up to 192Khz): 30Khz, 31Khz, 32Khz, and 33456Hz (I was feeling saucy!) and mixed them together

 

Obviously on their own was 'silence' as I had nothing that even remotely played back that frequency range audibly but when mixed together you get this irritating noise:

 

http://www.ilovecubus.co.uk/pete/superbell.mp3

that must be some kind of distortion, either dac's or the transducers' themselves or some kind of aliasing. i assume you see nothing besides the 4 sines on the software spectrum analyzer?

Edited April 16, 2018 by eugene

from what i gather the beat frequency will actually sound like tremolo effect, you won't magically get a tone at that frequency.

  On 4/16/2018 at 10:08 PM, eugene said:

man, i'm not really interested in "in the meantime"s or whatever. you made a pretty ridiculous claim and since making it did everything to weasel out of actually proving it for real. the developers of lossy codecs already take into account the peculiarities human hearing, however it might work, simply by putting their encoded files to blind tests. so i don't understand what exactly are you trying to prove here by linking to that stuff.

 

for the third time, if you still want to stand by your claim that "the difference between 320kbps mp3 and CD audio is night and day." pick a track(s) that you think will tax the mp3 codec the most and i'll prepare the stuff for you to test. after all it was your claim that those tracks in the online test are not good enough for such tests, so naturally it's up to you to choose the right ones, otherwise you'll simply claim that that tracks i chose aren't appropriate as well.

lol drag his ass

  On 4/16/2018 at 11:42 PM, eugene said:

that must be some kind of distortion

Yep, just realised that idiot boy here had all four samples normalised so when the four were added you ended up with that mess of heavily clipped sines leading to all those overtones. Looks proper pretty though

 

 

EDIT: Doing it 'properly' and you do indeed get creation of the exact pure sine waves at the difference of the other frequency you modulate it with. Physics/Maths wins !

  On 4/16/2018 at 11:52 PM, eugene said:

from what i gather the beat frequency will actually sound like tremolo effect, you won't magically get a tone at that frequency.

yes you do, you'll only hear a tremolo if the difference between the 2 freqs is below the audible range...

  On 4/17/2018 at 12:02 AM, mcbpete said:

EDIT: Doing it 'properly' and you do indeed get creation of the exact pure sine waves at the difference of the other frequency you modulate it with. Physics/Maths wins !

good to know it works! :thumbsup: what soft are you using?