High resolution

When Sony/Philips released the Compact Disk early 80’s , we were impressed.
650 MB on such a small platter!
Today the CD is very much like a 5¼ floppy disk, very big and a low storage capacity.

DVD stores 4.7 GB and a tiny micro SD card 256 GB.

1 TB micro SD cards are expected to be available soon.


Given the technology at that time, Sony/Philips had to find a compromise between duration and sound quality. For a long time Philips thought 14 bits would do (the first DACs by Philips were 14 bits. Using some noise shaping they squeezed 96 dB out of it).

In the end, they settled for 2-channel audio with 16-bit word length and a sample rate of 44.1 kHz.

 

16 bits allow for a dynamic range of 16*6= 96 dB and as Shannon/Nyquist tells us, a maximum frequency of 22.05 kHz (44.1/2).
Pretty good values by the way, especially when compared with vinyl.

 

Our audible range is 30 – 20.000 Hz (when young). Sampling at 40 kHz is in principle sufficient but there need to be some additional room for the filtering (transition band).

That’s why CD audio uses a slightly higher sampling rate of 44.1 kHz.
This allows for covering the entire audible range and using a very steep filter (brick wall) right after the 20 kHz to go down as fast as possible before ½ fs (fs=sample rate) is reached.

 

Recordings are often made with a greater bit depth (24 bits is almost common the last decade) and a higher sample rate.
The moment this recording is made available on CD or in the same format as a download, it has to be re-sampled to 16/44 otherwise it won’t play.

 

Hence it has to be filtered to remove all frequencies above 22kHz and dithered as the truncation of the bit depth otherwise produces audible artifacts.

 

Likewise if one records at 44.1 kHz.
The input must be bandlimited, no frequencies above ½ fs allowed.

Obvious you can’t get rid of this steep filter.

 

The moment one switches to other media, there is no reason to cling to the Rebook (CD) format.

DVD often contains 24 bit/ 96 kHz audio.
Downloads can be in any sample rate.
The only requirement is that your media player/audio device can play it.

Technically there is no reason to cling to this late 70's Redbook audio standard.

Bit depth

Almost all hires is 24 bits

A 24 bits recording has a dynamic range of 24*6= 144 dB
Sounds impressive compared with CD’s 96 dB.

In practice the noise floor of your gear e.g. -120 dBFS will be the limiting factor.


As Redbook (audio CD) is a very popular format, a recording made with 24 bits has to be truncated to 16 bits (otherwise it is not Redbook and won’t play on a CD player).
Inherent to this truncation is the quantitation noise.
Each and everybody in the audio industry applies dither to de-correlated this noise from the audio. That is the paradox of dither, improving sound quality by adding noise!
Obvious each and everybody also implicitly agree that signals at -96 dBFS are audible, otherwise dithering wouldn’t make sense at all.

 

The danger of 24 bits

Since 16-bit already provides enough dynamic range to deafen you, 24-bit provides enough to literally kill you.

This is a common misconception often mentioned on audio fora.
This would be true if a 24 bit file plays substantially louder that a 16 bit one but it doesn't.


The maximum output of a DAC is 0 dBFS, be it a 16 bit or a 24 bit audio file.
Hence a 16 bit file format goes down to -96 dBFS and a 24 bit down to -144dBFS.
A 24 bit will resolve the (tiny) details below -96 dBFS
A 24 bit file doesn’t play louder, it can play softer!

 

If a DAC can go down to -120 dBFS and you apply substantial amplification, say a SPL of 110 at 1 m, you can make all those tiny details below -96 dBFS audible.
But you do have to play fff loud.

Sample rate

Higres has sample rates like 88, 96, 176, 192 or higher.

Divide the sample rate by 2 (1/2 fs in the lingo) and you have the highest possible frequency in the recording.

 

The standard argument against highres is that our hearing stops somewhere.
20 kHz when young and the older we get, the lower this upper threshold.
We can’t hear above the upper threshold of our hearing. Period.

It sounds like buying recordings containing information above the upper threshold of our hearing is like paying a premium for some emperor’s new clothes.

 

A more relevant question is if a signal contains frequencies above the threshold of our hearing, can this have an influence on what is happening in the audible range?

This can indeed be the case.

Down sampling

Down sampling might introduce artifacts due to the filtering.

Uli Brüggemann did a very nice experiment.
Upsample the original 44.1 to 176 and back to 44.1

 

Original signal
Secret Rabbit Code upsampling to 176
Secret Rabbit Code downsampling from 176 to 44.1

Brick wall

You get an alias regardless the sample rate. This alias starts at ½ fs.
In case of 44 kHz this is at 22 so pretty close to the upper limit of our hearing.
You need a very steep filter (brick wall) to preserve both the audible range and filter out the aliases.
In case of 192 the alias is at 96.
Again we need a filter but we might decide to use a less steep one and starts a bit earlier because there aint much musical live at 96 kHz.
This make the filtering less intrusive.

Drivers

All drivers has an operating range. If you feed it signal above this range it will breakup.
Our tweeters are in general protected by a high pass filter, not a band pass filter.
If the upper limit of what our tweeters can reproduce is below the  highest frequencies in the recording, we indeed feed signals above the breakup point of this driver.
This might alter the sound including the one in the audible range.

IMD

The IMD (inter modulation distortion) generated by the content above 20 kHz might map into the audible range.

 

As out systems are non-linear, playing 2 test signals e.g. 19 and 20 kHz will result in intermodulation distortion, a kind of “harmonics” left and right of thes 2 signals.
Hence the IMD of 2 signals above the upper threshold of out hearing can result in IMD inside the audible range.

 

example of IMD

Time resolution

A common misconception is that the time resolution of CD audio is insufficient.
As the sample rate is 44.1 kHz, the time between 2 samples is approximately 23 microseconds.
This is above the threshold of our hearing, as we can resolve difference is time down to 15 microseconds.

 

The  Shannon-Nyquist theorem proves that we can fully reconstruct an analog signal as long as there are no frequencies above ½ fs in the incoming analog stream.
Nothing is lost below 22 kHz.
The step between two samples has nothing to do with time resolution, it is simply another expression of the sample rate.

A nice visualization can be found here: Time Resolution in Digital Audio - Christiaan Luther

 

Can we conclude that the time resolution of CD-audio is perfect?
No, it isn’t.
With 16 bits and a theoretical upper limit of 22050 Hz the best possible value is:

t min =1/(2π×22050 Hz×1×(2 16 1))=110 ps

as 1 pico is 0.000001 of a micro second, this way below the threshold of our hearing.
At lower frequencies e.g. 1 kHz signal at -20 dB the result is 24 ns.
Again way below the threshold of our hearing.

Mans Rullgard has more: Time Resolution Of Digital Audio

Conclusion

Obvious, high-resolution recordings are no snake oil.
They do offer a bigger dynamic and frequency range and are less demanding on filters.

Technically spoken, they do trump the Redbook format.

The question is if all these technical benefits actually translate into a better sound quality.

Mastering

A recording available on CD and as a highres download looks like the ideal one to do a comparison between the 2 formats.
Sometimes the differences are striking.
The highres sounding much better, less dynamic compression and no clipping.
Don’t mistake this for a property of the format.
The Redbook format is very popular hence management has it say. It should stand out (hence louder than the rest) , sounds good not only in the room but also in the car (hence dynamic compression), etc.
The highres format is less in demand, only audiophools are willing to pay the premium.
Hence good mastering preserving the quality of the recording is still allowed.
Highres can sound better than the equivalent on CD because it is (still) management free!

Can we hear the difference?

Often no audible differences between CD audio and higher resolutions are reported on the internet. Just as no audible differences between CD audio and high bitrate MP3 is often reported.

As 1+1=2 some conclude that there isn’t a difference between MP3 and hi-res audio!

 

On audiophile forums, you will find enthusiastic claims that high-resolution sound better.  
The question of course is better than what?
Often hi-res recordings are recent recordings or recent remasters.
Recording and mastering technique has improved over the decades.
One might mistake improvements in recording technique for improvements in the format.


Recording a hi-res download with the same recording on CD sounds like a fair comparison but it can be a bit tricky.
There is no guarantee that the source is the same.
Even if it is, it might be different masters.

 

The best way to compare is to have a true high-resolution track and down-sampled it to redbook format (16/44).

Compare them in a unsighted test.

You can find a couple of them on the internet.

 

AVS/AIX High-Resolution Audio Test: Ready, Set, Go! - Scott Wilkinson

AVS/AIX High-Resolution Audio Test: Take 2 - Scott Wilkinson

24-Bit vs. 16-Bit Audio Test - Part II: RESULTS & CONCLUSIONS - Archimago's musings

 

Often the discussions on the fora turns in to a flare.

A nice example: Conclusive "Proof" that higher resolution audio sounds different.

 

Most of the time these listening test concludes that most of us can't hear the difference.

According to Pras and Guastavino trained listeners can hear the difference in a ABX test.

 

It is currently common practice for sound engineers to record digital music using high-resolution formats, and then down sample the files to 44.1kHz for commercial release. This study aims at investigating whether listeners can perceive differences between musical files recorded at 44.1kHz and 88.2kHz with the same analog chain and type of AD-converter. Sixteen expert listeners were asked to compare 3 versions (44.1kHz, 88.2kHz and the 88.2kHz version down-sampled to 44.1kHz) of 5 musical excerpts in a blind ABX task. Overall, participants were able to discriminate between files recorded at 88.2kHz and their 44.1kHz down-sampled version. Furthermore, for the orchestral excerpt, they were able to discriminate between files recorded at 88.2kHz and files recorded at 44.1kHz.

Scam

You can find reports like this one:

I am interested in everyone's experience with the quality of various HD Tracks labels in hi rez. I have had excellent results with Reference Recordings, Dorian, Chesky and 2L. But I was very disappointed with the BIS recording of Osmo Vanska's Beethoven symphonies 2 & 7. I had seen rave reviews of the SACD version, but when I downloaded it in 24/88.2 from HD Tracks, the sound quality was nothing special. My RBCD version by Gunther Wand on RCA has better SQ
http://www.audioasylum.com/cgi/t.mpl?f=pcaudio&m=72939

Somebody listening to hi-res and is missing the hi-res sound.
Others chime in; ask for a sample and it turns out that this recording doesn’t contain any signal above 22 kHz.

One of those examples where you pay a premium for a hi-res track and get up sampled CD audio in return.
Unfortunately this is not a single incident.

http://www.audioasylum.com/cgi/t.mpl?f=pcaudio&m=74993
http://www.audioasylum.com/cgi/t.mpl?f=pcaudio&m=72939
http://www.audioasylum.com/cgi/t.mpl?f=pcaudio&m=81291

 

As long as companies selling hi-res don’t tell you what the resolution is of the source used, you run the risk of being scammed.
These incidents also proof that some can hear the difference between CD and hi-res.

Listening with your eyes

There are a trick to check if a recording is true hi-res.

An obvious one is the frequency spectrum.

If the spectrum it is cut off at 21kHz the source is probably 44.1 (CD)

 

A nice overview of a couple of recordings with spectrograms by Fujak.

Reference
  1. Listening with your eyes - Bruce Brown
  2. Hires Audio - soundcheck's - audio@vise
  3. HD-Produktionen auf dem Prüfstand - Fujak
  4. The tradeoff of 192 kHz sampling - Mikael Vest and Peter Scheelke - Digital Audio Denmark
  5. Sampling Rate Discrimination: 44.1 kHz vs. 88.2 kHz - Pras, Amandine; Guastavino, Catherine. Affiliation: McGill University, Montreal, Quebec, Canada
  6. AVS/AIX High-Resolution Audio Test: Ready, Set, Go! - Scott Wilkinson
  7. AVS/AIX High-Resolution Audio Test: Take 2 - Scott Wilkinson
  8. Conclusive "Proof" that higher resolution audio sounds different - Amirm
  9. 24-Bit vs. 16-Bit Audio Test - Part II: RESULTS & CONCLUSIONS - Archimago's musings
  10. Audio Distortion Measurements - Brüel & Kjær

  11. Does the nano iDSD measure: Good, Bad or Ugly?
  12. The great audio myth: why you don't need that 32-bit DAC - Robert Triggs
  13. Time Resolution in Digital Audio - Christiaan Luther
  14. Time Resolution Of Digital Audio -  Troll audio