High Resolution Audio Controversy on Wikipedia

[KozmoNaut]

That's fair. there are conflicting explanations on this subject. Should I instead reference subjects' responses in “The audibility of typical digital audio Filters in a high-fidelity playback system ” Convention Paper, Presented at the 137th AES Convention 2014, the paper you referenced in your post "High Resolution Audio: Does It Matter"? At least we can agree that for "some reason" there is an audible difference between high resolution and filtered content, although it is not entirely understood.

The conclusions in that paper are highly contested, for instance on this very forum in the MQA thread.

I'll summarize:

1. Across all 160 trials, the aggregate result is is 56.25% correct in identifying "hi-res" vs. downsampled. This is hardly statistically significant.

2. When low-passing the signal without quantizing to 16 bits, the result was statistically insignificant. So they spun this against the digital filters, instead of investigating further.

3. They used clearly non-optimal rectangular dither, which is known to produce intersample modulation distortion, which can certainly be audible. The fix is to use a proper dithering algorithm instead, such as triangular dither which is included in just about every piece of audio software worth mentioning. The fix is not to declare CD-quality PCM audio as "bad" and insist on "hi-res" audio.

4. The difference between filters is balancing on the edge of statistical significance (talking about the 16-bit samples that the participants could identify). It also points towards the opposite of their original hypothesis. They thought a sharper low-pass filter would be more audible, but it turned out to be less audible. They spin this into something about temporal smearing and ringing in a wider frequency range, which seems like post-rationalization.

5. They erroneously think that the gap between samples imposes a limit on the time-domain resolution, leading to "grainy" or "digital" sound. Obviously this is complete nonsense, as anyone familiar with digital sampling will tell you. It's just a repeat of the "stairsteps" myth.

All the paper shows is that if you use bad dithering or no dithering at all when downsampled, some people can (with great difficulty) identify some audible flaws under optimal conditions. That's not really particularly exciting or groundbreaking.

Considering the amount of effort spent on trying to prove any kind of audible benefit to "hi-res" audio, the conclusions reached are rather damning for that entire segment of the industry.

 

[Eirikur]

All the paper shows is that if you use bad dithering or no dithering at all when downsampled, some people can (with great difficulty) identify some audible flaws under optimal conditions. That's not really particularly exciting or groundbreaking.

Considering the amount of effort spent on trying to prove any kind of audible benefit to "hi-res" audio, the conclusions reached are rather damning for that entire segment of the industry.

For such a test, wouldn't it make more sense to up-sample the 44.1/16 ?
Or use a DAC that can do this internally with great accuracy?

 

[KozmoNaut]

For such a test, wouldn't it make more sense to up-sample the 44.1/16 ?
Or use a DAC that can do this internally with great accuracy?

If you want to compare with "hi-res" audio, you need to start with the highest resolution files, similar to if you want to compare lossless with lossy codecs.

Of course, it then also makes sense to upsample the downsampled files, so everything is played back at the exact same bit depth and sample rate, to avoid any DAC issues with different samplerates.

 

[JJB70]

a greater bit depth means I can crank up my music to disgusting volumes without a song sounding 'rough' or grainy, and that was something I could easily try out for myself. Also, since I do listen to music fairly loud it was quite easy to notice the lower noise floor of my higher res content once I had quiet enough gear.

Listen at disgusting volumes and the problem will soon go away once your hearing is degraded enough not to notice audio quality. Shouldn't take too long

 

[Ron Texas]

https://www.audiostream.com/content/fear-24-bit-planet-high-resolution-audio-0

 

[anmpr1]

As is the case with most controversial topics, you have to take articles at the On-Line Encyclopedia of General Know-it-All, aka Wikipedia, with a certain amount of reservation, if not skepticism. Unless you know who is writing, and their backgrounds and biases... only then can you make a wholly informed determination about the goodness of what you are being told.

 

[KozmoNaut]

As is the case with most controversial topics, you have to take articles at the On-Line Encyclopedia of General Know-it-All, aka Wikipedia, with a certain amount of reservation, if not skepticism. Unless you know who is writing, and their backgrounds and biases... only then can you make a wholly informed determination about the goodness of what you are being told.

Looking through the discussion on the Talk page is a handy indicator.

 

[CDMC]

One other thing to keep in mind is even were people able to distinguish between Redbook and Hi Res, it doesn't mean high res sounds better. Interestingly when Archimago had people do blind comparisons between high bitrate MP3 and Redbook of the people that could distinguish between the two, there was preference for MP3.

At the end of the day the real issue with recordings isn't Redbook v. hi res, it is the utter crap of the mixes and high levels of compression to satisfy the artists, managers and loudness. I would gladly listen to a well mixed version of an album in 320 kbps MP3 or 256 kbps AAC over a mediocre version in 192/24.

 

[Fluffy]

Interestingly when Archimago had people do blind comparisons between high bitrate MP3 and Redbook of the people that could distinguish between the two, there was preference for MP3.

Totally true. It happened to me as well – I listened to mostly MP3 (probably low bitrate also) when I was younger, and at some point I tried to listen to albums I knew very well in lossless FLAC. In some cases, I thought it sounded harsher and unpleasant and actually preferred to MP3 version. There is a simple explanation – low bitrate mp3 cuts off frequencies above a certain point that depends on the bitrate, usually at 16khz. I was used to listen to music without that high treble present, and when I heard it for the first time, it was like a piercing nail scratching sound was added on top of the music.

It took me a while but eventually I got used to listen to music with the whole spectrum playing, so now when I hear mp3 with something like 192k bit rate or lower, I notice the missing treble and it sounds worse.

I think a similar thing happened back in the day of the first CDs, when people that were used to analog recordings and gear that had more roll off in the highs, suddenly heard them as clear as they could hear the lows, and it sounded harsh.

 

[anmpr1]

I think a similar thing happened back in the day of the first CDs, when people that were used to analog recordings and gear that had more roll off in the highs, suddenly heard them as clear as they could hear the lows, and it sounded harsh.

I think some early CDs were just recorded/mastered poorly. It would not have surprised me if some transfers had the high's boosted, just because they could. Some of my early CDs had aggressive high ends. A good recording is the biggest determinant of sound quality.

 

[CDMC]

I think a similar thing happened back in the day of the first CDs, when people that were used to analog recordings and gear that had more roll off in the highs, suddenly heard them as clear as they could hear the lows, and it sounded harsh.

I am old enough to remember when CDs came out and hearing the early ones. I think it was a combination of things:

1) There was this amazing silence, even though earlier CD playback devices were "only" 14 bit (remember we either used vinyl with clicks and pops, or cassette with hiss). With analog recordings, you would often set volume based on the background noise, with no such background noise, you would crank it up with a CD because there was no reference and you wanted to hear that stuff that you never heard in the tape or vinyl version.

2) Yep, there was brightness, in part because all of a sudden there was a source flat out to 20khz. Tape rolled off far before that and vinyl degrades in the high frequencies the more it is played.

3) Early CD players were harsh sounding while also often sounding veiled. We didn't know a lot. Brick wall filters caused horrible phase shift. Jitter was a problem. As we started using oversampling and learning about jitter and controlling it, playback devices improved.

4) In many cases, downstream components now revealed their shortcomings, as the source was no longer the limitation.

 

[KozmoNaut]

Don't forget that people had to learn how to master for digital with flat frequency response, rather than tape and vinyl. A master recording that was appropriately boosted in the high frequencies to account for analog media would sound overly bright on digital formats.

And in some cases, the vinyl master would be put out on CD by mistake, with the RIAA curve.

 

[PaulD]

Don't forget that people had to learn how to master for digital with flat frequency response, rather than tape and vinyl. A master recording that was appropriately boosted in the high frequencies to account for analog media would sound overly bright on digital formats.

And in some cases, the vinyl master would be put out on CD by mistake, with the RIAA curve.

I was a junior in a studio at the end of the 80s when we had to change mid-production from 48-track analog (2 x MCI 24-tracks machines) to digital (2 x Sony 3324s) because of the 3M(?) tape shedding issue on the 2" tapes... I remember the main engineer saying, "It's amazing how the cymbals and top end sound the same after a week of mixing with digital." Yes, with analog you used to boost the top end to very roughly compensate for the HF loss due to repeated tape passes. So early digital recordings were often harsh, because the engineers were used to working with analog and needed to retrain for digital.

 

[amirm]

1. Across all 160 trials, the aggregate result is is 56.25% correct in identifying "hi-res" vs. downsampled. This is hardly statistically significant.

Oh, it is statistically significant. See: https://www.audiosciencereview.com/forum/index.php?threads/statistics-of-abx-testing.170/

 

[Wombat]

I think some early CDs were just recorded/mastered poorly. It would not have surprised me if some transfers had the high's boosted, just because they could. Some of my early CDs had aggressive high ends. A good recording is the biggest determinant of sound quality.

Some '80s CDs require de-emphasis to get normal treble frequency response. If played on a player that does not provide for this(more likely with some modern players) or one that does not automatically turn this feature on then the CD will probably sound 'bright'.

CD de-emphasis.

Incomplete list of pre-emphasised CDs

I have read that EAC will 'flag' pre-emphasised CDs.

* The Cambridge Azur 840C does not de-emphasise. Reference: Post #51, here - https://forums.stevehoffman.tv/threads/list-of-cd-players-that-can-decode-pre-emphasis.276489/page-3

A player with this

(or the white on black version) should de-emphasise if it is from an established reputable brand.
Be careful with hi-end and boutique products as they may hold, as Cambridge does with the 840C, that the de-emphasising process interferes with the up-sampling process. The Cambridge Azur 840C incorrectly shows the RedBook logo.
The Azur 851C appears to be the same in this regard.

 

[amirm]

3. They used clearly non-optimal rectangular dither, which is known to produce intersample modulation distortion, which can certainly be audible.

They explained that in their test the dither would not be audible. I provided the quote in the MQA thread.

 

[amirm]

All the paper shows is that if you use bad dithering or no dithering at all when downsampled, some people can (with great difficulty) identify some audible flaws under optimal conditions. That's not really particularly exciting or groundbreaking.

Considering that people think the rightmost bits of 24-bit content is so random as to be self-dithering, the findings cannot be put aside using that argument.

Even if it is just the effects of dither, that too would be significant and would point to the need to buy the original master and not let anyone decide what dither to use to down convert it to 16 bits.

 

[KozmoNaut]

Oh, it is statistically significant. See: https://www.audiosciencereview.com/forum/index.php?threads/statistics-of-abx-testing.170/

56% is hardly a "night and day difference", as some claim it would be.

They explained that in their test the dither would not be audible. I provided the quote in the MQA thread.

They made a claim that it would likely be inaudible. And they strongly recommend triangular dither in some of their other writings, so why even choose rectangular for this test? It's not a "typical" filter as they claim, and it does introduce anomalies.

The test samples they used were also bordering on pathologically dynamic, rather extreme even when compared to good musical recordings.

They gamed the setup to get the result they wanted.

 

[JJB70]

To assess the value of any technical paper you have to read it in full, understand the underpinning assumptions and any test methodology and have sufficient knowledge and understanding to be able to read it and develop an informed opinion on the contents. The fact that something exists in a paper does not make it correct.

 

[Kal Rubinson]

56% is hardly a "night and day difference", as some claim it would be.

56% with statistical significance says that a majority of subjects hear a real difference. It does not deal with the magnitude/