24-bit/192kHz is pointless?

[usernaim250]

Just saw this thread and haven't read through all of it but in case Chris (cs) is still here, do you have cites on the audible dynamic range of 30 dB? I have access to JAES, JASA, Proc. IRE, JSMPE/SMPTE, etc.

 

Thanks much if you can point me in the right direction.

 

--Eric

 

[SoundAdventurer]

You´ve all forgotten the Voodoo factor here!

 

When hi-frequency sound vibrates together with lower frequency sounds they together produce sounds, and audible sounds too. That´s why we respond to higher frequencies than normal hearing range...

 

Sampling in 192KHz is way better than 96KHz, just because of the Voodoo factor. I would gladly take a blind test anytime, and I would pass. It´s easy. Try to picture it like this, like the sky would open, like the roof went away, and there was more air and easier to breathe. Like moving from a 3D world to a 5D world.

 

I look forward to 384KHz too

 

[audioAl]

I enjoy music at any bit if it's sonically sound, great musicians are just great sounding. Bill Frisell sounds good today.

 

[sandyk]

I didn't read through the whole thread, (so shoot me !) but I was amazed at the number of people who can't seem to hear much (if any!) difference between 16/44.1 and 24/192.

This makes me seriously wonder about the quality of the sound files stored on the PCs/MACs/ Servers of many members, and their playback software as well as their playback equipment.

I am 70 years of age. with industrial type hearing damage after working in the technical side of Australian Telephone Exchanges for a little over 44 years.

I can very clearly hear the differences between a well recorded CD, and the 24/96 or 24/192 ripped .wav files from a good quality DVD-A. The difference to me , is like winding the clock back 30 years with my hearing.

I have also uploaded many comparison files to other forum members in Rock Grotto, as well as an upload for members here.

Almost without exception, the higher resolution .wav files are commented upon as sounding great.

The people that I upload these comparison files to,do however have a minimum of a good DAC,(usually via SPDIF) and a high quality headphone amplifier, or a better than average stereo system. Several have the Twisted Pear "Buffalo" DAC and Class A headphone amplifiers, and headphone amplifiers.

I can assure you that the difference between 44.1 and 24/96 recordings such as "Fleetwood Mac-Rumours" is absolutely stunning with the right storage and playback equipment. Too many of you seem to think that high quality audio is wasted on older people due to natural hearing deterioration. Nothing could be further from the truth.

Carly Simon's 24/192 recording of "You're So Vain" is light years ahead in every respect to the normal CD version.

I just wish that I had been able to obtain more high quality 24/96 and 24/192 DVD-As before they were discontinued. However, I suspect some DVD-As may have been upsampled, and only marginally better than the normal CD version.

 

SandyK

 

[ampman]

This is a great discussion. I have two DAC's 16/44 and 24/96. I am amazed at times I listen to the same music and the 16/44 sounds better than 24/96. At other times the higher resolution sounds better. I guess it depends on the original quality of the recording and my mood at the time.

 

[shakebb]

i'm new to cas... after reading so many information here... i'm just curious (and sorry for asking perhaps a dumb question) if anyone has tried comparing a devilsound usb dac with a weiss dac2/minerva playing the same music at 16/44.1? the reason i'm asking is that the majority of my collection will still be music i ripped from my cds...

 

i've got a macbook pro already and wanna get a dac to start enjoying my music again...

 

tks guys!

 

[audiozorro]

IMO native high rez music files, such as 24/88.2 or higher, should almost always sound superior to those files downsampled to lower resolutions, like 16/44.1. Similarly native 16/44.1 may often sound superior to those files upsampled to higher resolutions. I happen to believe the apparent sonic superiority is often more of a function of the DAC than the resolution of the music file itself.

 

I am not convinced that a DAC that handles multiple resolutions, handles each one to the same absolute level of fidelity. Others have discussed different DAC chips or digital filters in more technical details elsewhere. It is just my opinion that a fixed NOS DAC has an inherent advantage in sounding superior to a DAC that handles multiple bit and sampling rates. Just like the best RDBK CD player or the rare SACD only player, has an inherent advantage in sounding better than a universal player that handles multiple formats. Of course implementation is everything and will trump theory everyday.

 

Against the preceding backdrop we have music available in many different bit and sampling rates. It is not practical to have a dedicated DAC for each. My ideal DAC would be affordable and have excellent playback capabilities of 16/44.1, 24/88.2, 24/96, 24/176.4, 24/192, DSD and DXD files. Since my ideal DAC does not exist and I probably couldn’t afford it anyway, I have multiple DACs. Computer audio is just in its infancy and the digital landscape is changing relatively quickly, so I sometimes feel like a canoeist paddling (upstream or downstream?) to digital audio nirvana.

 

But 24-bit/192kHz is not pointless so long as there are music files in that format. One of many things computer audio does for me is the possibility of getting studio master music files at their native resolution, without the downsampling, compression or other deleterious processing effects. That is, the studio master files exist and that’s what I want at a reasonable price. In the old analog days, the gold standard for music files was the studio master audio tapes or the first generation of direct to disc LPs. I prefer to have a choice instead of having someone else decide that they will only downsample their studio masters to 16/44.1 or lossy formats such as MP3 for the mass market.

 

 

[mtspark]

Having studied auditory neurobiology extensively and being an electronics engineer, I'm in no doubt that the shift to 24-bit/192kHz is beneficial, not so much because you're going to notice a huge difference in sound quality like the transition from LP to CD, but mainly because it should result in a clearer perception of spatial positioning within a recording. ie. The finer details should be improved.

 

Early stages of the auditory system have amazing temporal resolution; a figure of 10us was mentioned in one paper. This results in the ear processing 100kHz signals! So I believe that 192kHz is actually about right. The neurons can phase lock precisely to the incoming stimulus up to a maximum frequency of around 5kHz. This phase locking ends abruptly when this information is translated into a place code or spatial map within the mid-brain.

 

So what I'm saying is that with such a huge bandwidth implied by 192kHz sampling, we don't actually care about hearing range per say, we are more concerned with temporal resolution at lower frequencies - up to 5kHz. This will enable you to accurately place sounds in your brains model of 3D space.

 

However, for this to work, the temporal accuracy of the recording is going to have to be utterly and faithfully preserved throughout the entire audio chain. A tall order I would say, especially at the loudspeaker end. But a careful headphone listener should in theory be able to notice the difference.

 

To the guy somewhere in this thread that said time dispertion would be a problem with two cascaded brickwall filters in modern sigma-delta DAC's, I don't think that's the case. These filters will be FIR types and have linear phase and therefore there won't be any time dispertion no matter how many you cascade. Just more delay

 

Cheers,

Mark.

 

 

 

[idiot_savant]

are you talking about being able to resolve to within 10us between channels?

If so, 44.1k is, surprisingly more than adequate. If not, I'll be quiet...

 

your friendly neighbourhood idiot

 

[sandyk]

"However, for this to work, the temporal accuracy of the recording is going to have to be utterly and faithfully preserved throughout the entire audio chain. A tall order I would say, especially at the loudspeaker end. But a careful headphone listener should in theory be able to notice the difference."

 

Hi Mark

I may be 70, but boy, do I notice this difference with the 24/192 rip of the track "You're So Vain" from the Carly Simon DVD-A "No Secrets"!!! Another well recorded DVD-A that shows off the benefits of higher resolution, is the 24/96 "Fleetwood Mac-Rumours" It is very obvious via headphones.

 

SandyK

 

[warrensomebody]

44.1kHz = 22.7us between samples, so if people can distinguish 10us difference, you would need 100kHz to represent that as an audio signal.

 

[idiot_savant]

is that 44.1k can only represent 22uS or so phase between channels.

This isn't the case.....

Remember, there is no such thing as a perfect impulse in the real world - for real-world signals, 44.1k can represent inter-channel phase of a few hundred picoseconds,

 

 

your friendly neighbourhood idiot

 

[Bob Stern]

Mark (mtspark): "Early stages of the auditory system have amazing temporal resolution; a figure of 10us was mentioned in one paper. This results in the ear processing 100kHz signals! So I believe that 192kHz is actually about right."

 

Even if it's true that the ear can respond to spatial cues up to 100 KHz, what is the advantage of recording at a sample rate of 192 KHz instead of 96 KHz if the recording microphone has negligible output between 48 and 96 HKz? Have you ever seen measurements of a microphone's frequency response from 48 to 96 KHz? (I'm not being coy; I haven't seen such measurements myself.)

 

idiotsavant: "for real-world signals, 44.1k can represent inter-channel phase of a few hundred picoseconds"

 

Very interesting! I can see how that would work by imagining the A/D conversion of two sine wave signals with a time offset shorter than the sampling interval. The time offset results in a difference in amplitude at each sampling point, allowing the sine waves to be reconstructed with the correct time offset by the D/A.

 

[idiot_savant]

According to shannon/nyquist, the time and frequency domains are irrevocably linked - so, more (real) bits can equal better temporal resolution between channels ( all else being equal ), as counter-intuitive as it sounds....

 

 

your friendly neighbourhood idiot

 

 

[idiot_savant]

Frequency response isn't all there is to it...

 

higher sample rates let you get away with awful filtering, or at the very least a wider choice of filters....

 

 

your friendly neighbourhood idiot

 

PS I'm sorry if previous post seemed patronising, but lots of people get hung up on the 22.7us thing...

 

[warrensomebody]

When I mentioned 100kHz, I simply meant that this would be the sampling rate required to represent 2 waveforms offset by 10us. This has nothing to do with impulse response (the sharpness of the edges of the signal). The frequencies of the signals themselves would of course be much lower -- at least half that frequency as per the sampling theorem. This also has nothing to do with the slope of the filters required to remove the carrier frequency and the phase shift induced by them. It would be interesting to determine what the sampling rate would need to be to guarantee that the filter-induced shift was less than 10us across the audio band. Or conversely, what's the maximum phase shift incurred by a 192kHz sampling rate filtered by linear or minimal phase filters?

 

[idiot_savant]

to represent a 10uS phase shift.... honestly, 44.1k will do it just fine!

 

I'll try and knock something up sometime today to try and show this,

 

 

your friendly neighbourhood idiot

 

[MartinColloms]

Great set of responses........

 

A few observations from me based on listening to pro and consumer material and products......

 

Higher rates are certainly good for filter sound quality, and all being equal and even if you can't directly hear the higher rates, they truly increase resolution. Rate and bits both work together defining resolution.

Now for a given rate, say 88.2 it is quite easy to hear properly dithered truncation of a solo piano of a nominal 24 bit recording to 22bits. You could consider that 176-192, 20 bit plus would be a good reference, knowing of course that we are generally nowhere near this in practice even though it is a good standard to work at for mastering and production processing.

I agree with many contributors that integer rate conversion , up sampling etc is always better than non integer and asynchronous methods.

The fact that you can hear such subtle differences doesn't mean that you have to have the higher standard to enjoy music.

Very good recordings via top replay mechs can be very good indeed on 44.1/16. It all depends on how well it is done and there are several design routes and combinations thereof to get good digital replay.

 

I am familiar with high level replay, yet play and enjoy my many red book recordings, which still clearly show even more subtle quality differences for amplifiers and cables.

 

One swift does not a summer make and it is surprising to see the larger number of speculative assumptions made about products and technologies based on very cursory experimental data , often a single product sample of the technology under debate.

 

We are still learning about digital sound, even after 30 years, and many design and technology decisions are being made for us by chip designers who may not know what high quality sound reproduction is.

 

Martin Colloms Tech Ed. HIFICRITIC Stereo Magazine

 

 

[fmak]

It is well accepted by audiophiles that hirez does sound better. It is not just the frequency content, but the moving up in the frequency range of the brick wall and subsequent analog filter that are important.

 

[wgscott]

Sorry for the necropost, but this looks like one of the most interesting threads I have seen here.

 

I've been trying to understand digital music from the point of view of optical reconstruction, which is something that I know a little bit about, so it is quite interesting to see some of the same ideas, but in a different context.

 

Three stick out as important:

 

1. Shannon/Nyquist sampling theory -- you can fully represent a periodic function of a given frequency by sampling twice that frequency. Everything I do to image molecular structures relies on the veracity of this.

 

2. Fourier truncation artifacts. Human physiology places an upper limit on audible frequency between 15 and 20 kHz depending upon age. So sampling at 40 kHz should be adequate. Yet many people claim there is a degraded signal if higher sampling rates are not used. The only explanation that seems consistent is that the higher sampling frequencies help to minimize Fourier truncation effects. The optical analogy is kind of interesting. What's kind of interesting is you can fill in the missing higher-resolution data with artificial junk, and it is still much better behaved ...

 

3. Apodization. Smearing the data gives it a more rounded appearance, but the payoff is a much more well-behaved image reconstruction, eliminating many imaging artifacts. It is also a common technique used in optical spectroscopy.

 

[sturgl]

B&W's Society of Sound has an excellent (and easy to read) article covering the evolution of digital. Some quality attention was given to 16/44.1 PCM and higher bit-depths / sampling frequencies. I highly recommend it:

 

http://blog.bowers-wilkins.com/lab/sound-quality-lab/the-definitive-guide-to-24-bit-flac/

 

 

[ted_b]

analogies to help folks understand bit depth and resolution. I keep bristling at those who simply equate 96k or 176k sample rates with the same frequency response! It's more than that; it's the increased resolution along the WHOLE sine wave, not just the upper reaches. To me, going 24 bit is paramount (and he explains it's mainly due to the huge increase in dynamic range, echoed by the arguments of folks like Mark Waldrep) but sample rate is also important and quite discernible (a native 24/192 file's sound quality is usually quite discernible vs a downrez of the same to 24/48). My $.02....YMMV.

 

[tresaino]

agree this is a great thread, lots of learning, thanks

 

[wgscott]

I keep bristling at those who simply equate 96k or 176k sample rates with the same frequency response! It's more than that; it's the increased resolution along the WHOLE sine wave, not just the upper reaches.

 

A sine wave is of a single, discrete frequency. Any arbitrary sound can be decomposed into an infinite sum of sine waves, each with a characteristic frequency, amplitude and phase. When these are summed together, you get the sound. The question is what happens when you only include sine waves up to a particular frequency. Human hearing goes up to 20kHz, max. The Nyquist/Shannon sampling theorem tells us that if we want to sample the 20kHz sine wave, we need a sampling frequency of 40kHz in order to completely describe that wave.

 

If you sample at 96kHz, you can fully reconstruct ALL of the sine waves from 0kHz through 45kHz. If human hearing only goes up to 20kHz, the question becomes what contribution, if any, does sampling at a frequency higher than 40kHz make to the content of the sound you can actually hear.

 

So it isn't a simple matter of "equating ... sample rates with the same frequency response." But it is important to recognize what sampling frequency is and how it relates to reconstructing the sound.

 

 

 

 

 

[quadmethyl]

Thank God someone got to the science before I saw this article. Equating 192khz to its upper threshhold of describing a wave isn't accurate at all. It means the curve/wave is described at a resolution of 192,000 times a second meaning that if you divide 18k upper harmonics in only about 2.25 pieces/samples (44,100 slices) you get very bad sounding cymbal crashes vs. (192,000 slices). For the purpose of this arguement slices = samples. The same way 24 bit word lengths sound much better describing dynamic range then 16 bit. People thought 1080p HD wouldn't make a difference and now everyone is oooh'ing and ahhh'ing it and wanting their own HD LCD flatscreen. This justification goes for the entire frequency range and not just the upper end. Any musical instrument sound is going to have tons of upper harmonics that make it musical and are sometimes subtle in the way they augment the original fundamental. Now that I've beaten the dead horse...