HiRes FOMO - Am I really missing anything?

[vintologi]

One other point to make about Googlebot's test is that two other people got statistically significant results (p< 0.05) when using completely different systems.

It's not as if it's an isolated test, either. I found a dozen similar tests on HA before I stopped looking

I don't doubt that some people are able to differentiate between sampling rates in ABX tests (at least 44.1 vs higher).

Next step is finding out what the cause of this is. Problem is that it's very difficult for most people to hear any difference and you also need speakers and other equipment able to properly reproduce the ultrasonics.

I would guess that with good speakers less IMD is requires for it to make an audible difference but obviously it's still going to be extremily difficult to pick up on that and it's even harder to get 16/16 like googlebot did. Googlebot did however also reduce the bit-rate to 16 when downsampling which is suboptimal due to it introducing another potential way for audible difference between tracks (which might be easier to pick up with ELAC speakers).

 

[Keith_W]

Does anyone listen to music?

I should hope not. Speakers are for measurements and test signals only. I would NEVER defile my speaker by shoving random oscillations through it. The very thought of it makes me feel dirty.

 

[vintologi]

I personally very much do not trust music distributors to downsample properly (i honestly trust members of the 2 big music torrent sites more for that) and it can also be the case that the 24-192 version has a better master than the 16-48 version.

Another advantage with 24-192 is that you can probably hear a clear difference if you play it at 1/4 the speed.

I should try to find some more 24-192 to 24-88 guitar albums.

Let me know if you find some albums with suitable spectrals for testing the impact of ultrasonics, in most cases i will be able to download it myself so no need for you to share files (except in very rare exceptions).

 

[voodooless]

play it at 1/4 the speed.

Is that a thing people do?

 

[vintologi]

Is that a thing people do?

Never heard of someone actually doing that.

And most 24-192 releases has basically all garbage 50Khz. I only found some albums with good spectrals past 50Khz thanks to a compilation someone posted elsewhere.

 

[john61ct]

Obviously the same source content is required for both A & B, storage space is NOT the issue, it's multichannel transmission and DSP processing where costs triple or 5x more.

So start with a given 96+kHz uncompressed FLAC, as well mastered as can be found.

Use quality gear commonly used by respected studios, even if legacy compared to the more recent AoIP networking gear.

And TRANSMIT the signal from the digital source device, "A" being via one ADAT lightpipe channel, "B" SMUX'd over two, to as high a quality DAC as you like.

Do normal humans perceive - AUDIBLE SQ - difference between those digital transports, using normal speakers?

Personally, if the only issues are ultrasonics and IMD, and special transducers are needed for just SOME listeners to MAYBE BARELY hear the differences...

 

[vintologi]

Obviously the same source content is required for both A & B, storage space is NOT the issue, it's multichannel transmission and DSP processing where costs triple or 5x more.

So start with a given 96+kHz uncompressed FLAC, as well mastered as can be found.

Of course what we are interested in is testing the impact of sampling rate specifically so the best method is converting down to 24-48 and then up again.

It's not enough that the mastering is good. You also need to check that the spectrals themselves are good.

Do normal humans perceive - AUDIBLE SQ - difference between those digital transports, using normal speakers?

Personally, if the only issues are ultrasonics and IMD, and special transducers are needed for just SOME listeners to MAYBE BARELY hear the differences...

Obviously you need speakers that extend past 20Khz to take advantage of the wider frequency response offered by higher sampling rate than 44.1 KS/s. There are however cheap speakers that extend past 30Khz.

But even with high end speakers extending to 50Khz it seems like less than 1% can actually tell a difference between say 24-192 and 24-48 even with cherrypicked music specifically for that task.

 

[john61ct]

Obviously you need speakers that extend past 20Khz to take advantage of the wider frequency response offered by higher sampling rate than 44.1 KS/s.

Yes if that's the only topic area where a difference may lie, it does not interest me personally.

I thought there were other, more generally relevant areas of "quality" being claimed.

 

[vintologi]

I thought there were other, more generally relevant areas of "quality" being claimed.

The wider frequency response does also allow for better impulse response but since the difference is all in the ultrasonics it's difficult to actually hear.

At lower bit-depths higher sampling rates can be useful for improving the signal to noise ratio (especially when noise-shaping is used) but 24-44.1 shouldn't have any audible quantization noise to begin with.

Higher sampling rates does make it easier to construct the filter to not cause audible differences below 20Khz. With SoX (and anything using it) the difference below 20Khz is tiny due to how steep the filter is but if a less steep filter is used (such as due to the dac being suboptimal) you might be able to pick up on a small difference from that.

And even if your dac happens to have suboptimal filters you can simply have your computer resample it something higher.

You do also need a filter for downsampling but here again with SoX there is hardly going to be any difference below 20Khz.

 

[john61ct]

DACs are a solved category, not expensive to get SOTA.

I'm not sure why Sox is relevant to recreational home audio playback?

I have no need for manipulating / converting my files, and no plans to require a PC as player, maybe RPi for passthrough DSP but likely using Camilla / HLC with native Linux tools.

 

[vintologi]

DACs are a solved category, not expensive to get SOTA.

I'm not sure why Sox is relevant? I have no need for format conversion, and no plans to require a PC as player, maybe RPi for passthrough DSP but likely using Camilla / HLC with native Linux tools.

It depends on where you are getting your files from because you are going to depend on others downsampling to say 24-48 for you.

 

[john61ct]

It depends on where you are getting your files from because you are going to depend on others downsampling to say 24-48 for you.

Sorry if I was not clear enough above.

I am not concerned about storage space, my NAS is 90% empty my cloud backup has 36TB free.

Also with streaming, high res creates no problem.

It is only within my own system that these silly high "standards" raise costs exponentially – ADAT channels, DA/AD conversion, DSP crossovers especially tap counts for bass frequencies, etc.

$100-200 devices as opposed to a grand or more! without any real benefits...

 

[vintologi]

Sorry if I was not clear enough above.

I am not concerned about storage space, my NAS is 90% empty my cloud backup has 36TB free.

Also with streaming, high res creates no problem.

It is only within my own system that these silly high "standards" raise costs exponentially – ADAT channels, DA/AD conversion, DSP crossovers especially tap counts for bass frequencies, etc.

$100-200 devices as opposed to a grand or more! without any real benefits...

You could always do some tests blind yourself and compare.

High resolution audio

CD quality allows you to reproduce frequencies below 22.05 Khz at a a signal to noise ratio of up to 96 dB, this is actually pretty good considering CD was format was released in 1982. It's actually very difficult to fully utilize the Signal to Noise ratio CD offers even today The topping...

vintologi.com

You are probably not going to be able to differentiate them without training but good to confirm just to be sure.

Note that i personally do not actually use any DSP so i just grab the highest sampling rate available mainly for the placebo benefits.

 

[Peterinvan]

From Google AI:

my hearing above 7 KHZ is declining, what can I perceive above that?
Even though your hearing above 7 kHz is dropping, you can still perceive very high sounds as changes in overall loudness, soft air movement, or subtle shifts in sharp tones. You might also notice hidden high-frequency patterns when they mix with lower sounds you can still hear well.

How High Frequencies Feel and Sound

• • Combined tones: High frequencies blend with lower sounds. They add brightness and crispness to speech and music.
  • Volume shifts: You notice when a high sound starts or stops. You hear the change in total loudness, even if you miss the exact pitch.
  • Spatial clues: Very high sounds help you tell where a noise comes from. Your ears use them to find sounds beside or behind you. [1]
  • 
    
  • You can perceive harmonics above 7 kHz, even with declining hearing in that range. Most musical instruments and human voices naturally produce multiple overtones that extend far beyond 7 kHz. [1, 2]
    Your brain can perceive these high-frequency harmonics through a few specific auditory phenomena:
    
    Missing Fundamental Effect
    • Pitch reconstruction: Your brain calculates pitch by looking at the spacing between lower harmonics.
    • Perceived presence: If you hear harmonics at 2 kHz, 3 kHz, and 4 kHz, your brain automatically knows what the higher harmonics (like 8 kHz) should sound like.
    • Sound fullness: You still perceive the tone as rich and complete, because your brain fills in the missing high-frequency pieces. [1, 2, 3, 4]
      • Accessible frequencies: These new tones fall into your healthy hearing range under 7 kHz, allowing you to perceive the presence of the higher harmonics.
    • 
      Acoustic Brightness and Texture
      • Crisp consonants: High harmonics of speech sounds like "s," "f," and "sh" change how clean the words feel.
      • Instrument timbre: Harmonics above 7 kHz give cymbals, violins, and acoustic guitars their "air" and "sparkle."
      • Dullness detection: If you completely remove the harmonics above 7 kHz using an audio equalizer, you will instantly notice the sound becomes muffled or dark. [1, 2, 3]
    • 
      
    • Bone and body conduction bypass your middle ear completely, allowing you to perceive extremely high frequencies—even into the ultrasonic range—by vibrating your skull and directly stimulating your inner ear. In fact, while airborne hearing stops around 20 kHz, bone conduction can allow humans to sense frequencies up to 50 kHz to 120 kHz. [1, 2, 3]
      
    • However, because your hearing above 7 kHz is declining due to inner ear changes, bone conduction interacts with your auditory system in very specific ways:
      
      Bypassing the "Middle Man"
      • Direct fluid stimulation: Air-conducted sound must travel through your ear canal and shake your eardrum. Bone conduction skips these parts. It sends vibrations
      • Mechanical force: Even if the high-frequency hair cells in your inner ear are declining, intense physical vibrations from bone conduction can force the entire cochlear structure to vibrate, creating a sensation of sound. [1, 2, 3, 4]
    • 
      
    • 
      
    • The Catch for Declining Hearing
      
      • Shared receptor cells: Bone conduction is excellent for bypassing outer or middle ear blockages (like fluid or a damaged eardrum). However, the vibrations still rely on the inner ear's hair cells to send signals to the brain. [1, 2, 3, 4]
      • Reduced sensitivity: Because your specific decline is likely sensorineural (related to the hair cells responsible for frequencies above 7 kHz), bone-conducted sounds in that exact range will still feel much quieter or require significantly more power to perceive compared to lower pitches. [1, 2, 3, 4]

 

[welwynnick]

Sorry if I was not clear enough above.
I am not concerned about storage space, my NAS is 90% empty my cloud backup has 36TB free.
Also with streaming, high res creates no problem.
It is only within my own system that these silly high "standards" raise costs exponentially – ADAT channels, DA/AD conversion, DSP crossovers especially tap counts for bass frequencies, etc.
$100-200 devices as opposed to a grand or more! without any real benefits...

This is one of my points about high resolution audio all along.
The storage and distribution of 16/44.1 audio was a challenge in the 20th century, but not in the 21st.
CRT TVs and fax machines are no longer the cutting edge of technology.

 

[somebodyelse]

I'm not sure why Sox is relevant to recreational home audio playback?

LMS uses it to change sample rate, bit depth or encoding when the endpoint doesn't support one or all of those in the file being streamed. I think there was a plugin that used it to do EQ too. It may be used for similar things by other player/streaming software.

 

[Kal Rubinson]

LMS uses it to change sample rate, bit depth or encoding

So does Jriver.

 

[john61ct]

LMS uses it to change sample rate, bit depth or encoding

Aha! So automatically included, no setup required?

 

[john61ct]

The storage and distribution of 16/44.1 audio was a challenge in the 20th century, but not in the 21st.

nor 24-bit.

But for us poors, ADAT connections and DSP at 96kHz or higher costs more than it's worth.

 

[welwynnick]

But for us poors, ADAT connections and DSP at 96kHz or higher costs more than it's worth.

Trinnov use $100 CPUs for digital room correction at up to 192kHz, so it can't be that difficult. Granted, Trinnov are expensive, but an ordinary PC can do DRC.
There are lots of software tools, some of them filter designers, and some of them convolvers, and some of them both, but they're all capable and fair value:

• Acourate - AudioVero
• Audiolense Convolver Download – juicehifi.com
• Cavern
• Comparison of convolution engines | Audio Science Review (ASR) Forum
• Dirac - Digital sound improvement that transforms your HiFi experience
• DRC: Digital Room Correction
• Focus Fidelity - Digital Room Correction
• Hang Loose Convolver (HLC) - Accurate Sound
• HQPlayer Desktop – Signalyst
• JRiver Media Center software
• juicehifi.com
• LinFIR Software - Dem Audio
• PGGB - Offline remastering
• rePhase - Official Site - Free FIR filtering tool
• REW - Room EQ Wizard Room Acoustics Software
• Umami Audio - Perfect Soup MIMO Speaker Calibration