Hi-res audio folk to introduce new rules and weed out impure noises High resolution (Hi-Res) audio is not as good it should be, causing an industry group to create new production guidelines to address what it says are "misperceptions" concerning what it takes to create the best recordings. The guidelines will come from the Producers and Engineers wing of the Recording Academy, the organisation …
This is all that needs to be said, and said much better than I ever could:
http://www.xiph.org/video/vid2.shtml
Oversample for the DAC, have 24bit depth in the studio before 10,000 rounds of autotune (other, useful, effects are available) but the human ear can't deal with more than CD quality - that's why it was defined.
The biggest issue is that all this high def audio will still be overdriven through 30p headphones (even if they are in a £300 case).
> but the human ear can't deal with more than CD quality - that's why it was defined.
To be historically accurate, the CD standard 44.1KHz was just the best they could do with late-70s technology... the original Sony PCM digital systems recorded to video tape, and 44.1Khz was the highest sample rate they could accommodate.
[As an aside: early CD players were a truly remarkable feat of engineering! Processing data from the CD (at rate of over 4Mbits/s) with 1980 era LSI chips was quite something...]
The main problem with the 44.1kHz sample rate is that it requires *very* steep reconstruction filters. It is basically impossible to design such a filter without significant passband ripples and/or time-domain problems. Bumping the sample-rate up (to at least 48kHz) allows for better behaved filters:)
The second problem that the sensitivity of human ears varies by frequency, and in the region of greatest sensitivity (3-4kHz) the apparent noise floor of a (non-noise shaped) dithered 16bit signal *can* be heard. Using a noise-shaping dither can improve things, but with a 44.1kHz sample-rate there really isn't enough frequency headroom available...
Of course, 24/192 is complete overkill!
In fact, a properly engineered 16/96 system, using pre-emphasis and a noise-shaped dither, should easily exceed any realistic requirements.
While it may be overkill, as an engineer, I like the idea of capturing the reproducing the highest quality that can done.
Its like owning a mechanical watch. Sure I have a phone that syncs to a time source so its accurate enough, but to think about what it takes to design and build a watch with 500 moving parts and fits on your wrist? Now that's a combination of art and function that I can appreciate.
Is there something wrong in striving to do the best you can do or have we become a society of where 'good enough' is the best we can do?
"Its like owning a mechanical watch. Sure I have a phone that syncs to a time source so its accurate enough, but to think about what it takes to design and build a watch with 500 moving parts and fits on your wrist? Now that's a combination of art and function that I can appreciate.
Is there something wrong in striving to do the best you can do or have we become a society of where 'good enough' is the best we can do?"
The best timekeeping device would be the digital one that regularly synchronizes to the NTP pool, possibly with a backup to plug time from GPS signals. The mechanical timekeeping device, while nostalgic, is not better in any way.
You are confusing "better" with "requires more effort". Very protestant of you.
"The main problem with the 44.1kHz sample rate is that it requires *very* steep reconstruction filters. It is basically impossible to design such a filter without significant passband ripples and/or time-domain problems. Bumping the sample-rate up (to at least 48kHz) allows for better behaved filters:)"
There's no need to bump up the recorded sample rate to deal with this very real engineering concern: all you need to do is fake a higher sampling rate by interpolating your 44.1kHz samples. A single interpolation between real data points will allow you to feed your DAC at 88.2kHz - beating 48kHz sampling by a big margin.
44.1kHz sampling really is enough to record the full human-audible bandwidth from the point of view of almost all people and almost all repro gear. Recording at a higher rate isn't necessary; engineering nouse such as interpolation to produce a higher data rate into the DAC deals with the low pass filter issues Ed-H mentions.
Homeopathy starts out rationally enough, taking a slightly wider view of human health. A view that acknowledges that there might be a connection between body and mind, perhaps via a hypothesized structure some call 'the spinal column'.
But then, suddenly, it veers over the line into the criminal insanity of dilution, the memory of water, and similar idiotic concepts.
High End Audio starts out rationally enough, 'if good is good then better is better'. Anyone seeing the steady progress up to the invention of CDs might expect that the progress would continue to even better technologies.
But then, suddenly, it veers over the line into the criminal insanity of directional speaker wires, over priced power cords, gold plated fuses and similar idiotic nonsense.
There must be a word or term for this pattern. How about 'Good, Better, Best, Crazy'?
Well, actually, it is important to have those sounds. I'm not an audiophile by any means, in fact I actively try to debunk the bollocks the audiophools punt about (previous comments include dipping audiophools in liquid nitrogen to lower the noise they emit) but in that particular instance they're right, supersonic and subsonic frequencies can be important.
It's basic theory (and practice), if you mix two fundamental frequencies together you will have an output that contains the two fundamentals *and* the additive and the subtractive frequencies of both.
So even though the fundamental frequency of such a sound is outside your hearing range, it will heterodyne with other frequencies and produce within your hearing range, they contribute to a fuller, more rounded sound.
For an example of how harmonics can make something sound different, grab a copy of audacity and have it generate a sine wave and a square wave at 440Hz. Mute one track and listen to the other.
Then swap tracks.
The square wave track will sound 'smoother', more rich, a fuller sound. performing a fourier transform on the square wave will reveal that it's rich in harmonics.
A little light reading, might want to take a pencil, notepad and calculator:
http://mathworld.wolfram.com/FourierSeriesSquareWave.html
It's basic theory and practice, yes... but in spite of this, the human ear can only hear those products which fall in the 20-20kHz (or thereabouts, depending on age) frequency band.
If an inaudibly high frequency is heterodying with a lower frequency (whether the lower frequency is within the audible range or not) the only product which will be audible is the difference frequency - the sum will be even further out of the range of hearing.
And guess what - those products are all adequately represented within the basic 'CD quality' signal.
A suggestion: take your harmonics experiment, but set the sample rate as high as you can - 96 or 192ks/s, for example. Now compare a sine and a square wave at 440Hz, and, as you state, you will easily hear a difference since the square wave has a number of harmonics within the audible range.
Repeat the experiment at 4400Hz; you *should* still be able to hear the difference; there is still one harmonic in range, at 13200Hz. Now try it at 8Khz instead, and unless your hearing is exceptional, you will be unable to hear any difference - only the fundamental frequency is within your hearing range: the third harmonic is at 24kHz. And yet it's clearly present in the signal you're listening to.
The comment above about steep filters is on the mark, which is why at least 48kHz sampling rate is needed. Buyt frequency response is not the issue. It is the bit representation of the samples. 16 bits leaves complicates orchestral music sounding like a muddle, and puts graininess in the strings. 24 bits is totally adequate. That said, some CDs with bit--mastering which supposedly gives an effective 19 bits, sound quite good.
@Hey Nonny Nonny Mouse and all
I'll try that. Might be worth trying Audacity's low pass filter on the square wave track and progressively reducing the cut off frequency until your (individual) perception of the 'richness' of the square wave changes notecibly. Make a note of the cut off frequency you end up with. It might be lower than you imagine. You can play with pure data (puredata.info) as well to build up waveforms by superimposing harmonics (fourier synthesis).
I've stopped worrying about audio quality above CD or 196kb/s mp3 these days: hypersonic starts around 10 - 12kHz for me.
Random thoughts: visual processing starts in the retina and optic nerve, and our visual model of what is before us is synthesised in a couple of regions of the brain at quite a high level above the raw signal from the rods and cones. Damage to the various components of the visual cortex has subtile effects (Oliver Sacks wrote about aspects of that in one of his books).
Perceptual psychologists (the experimental ones not the boutique ones) have developed a concept of 'categorical hearing' - a musician trained in the Western tradition will tend to bin sine waves into pitch classes even when they are quite a few (hundred) cents off an actual pitch. Non-trained ears can recognise the sine waves as different at a finer level. BUT talk to a pianist who works with a string quartet (e.g. playing a piano quintet) ... the quartet musicians will drift into perfect fifth intervals instead of the equal intonation interval if they rehearse without the pianist for any length of time...
My direction of travel here is that 'music' may actually be made in our minds rather than existing as a sound field in a room, so you only need to produce *enough* of the sound field to trick the mind into making the music... mumble mumble
Coat icon: its time I took a walk. Mine's the one with the penny whistle in the pocket.
keithpeter wrote:
I've stopped worrying about audio quality above CD or 196kb/s mp3 these days: hypersonic starts around 10 - 12kHz for me.
My wife is a audiologist and tested me, I can hear nothing above 8K now.
HCD, CD or telephone hold music, it all sounds the same to me.
Sorry, you're misapplying the idea. Heterodyne frequencies come about when you multiply the two frequencies together, like in stages in radio equipment that concert to IF frequencies or modulate an AM radio signal. They should never occur at any measurable level in an audio mixer, which is an adder, not a multiplier.
Hey Nonny Nonny Mouse offered: "It's basic theory (and practice), if you mix two fundamental frequencies together you will have an output that contains the two fundamentals *and* the additive and the subtractive frequencies of both."
Such mixing *requires* a non-linearity.
In radio receivers, this function is called a "mixer stage". In the preceding RF low noise amplifier stage, the designer works long hours to ensure that the LNA is as linear as possible. Presumably the audio circuit designers are equally careful.
Apparently, it is. Back when i were moving faders for a living, I had the opportunity for a chat with a bloke from Klark Technik (they make rather good graphic equalisers, amongst other things) and while most of the conversation disappeared some way over my head, he made the point that they engineered their kit to take account of harmonics outside the threshold of human hearing simply because it seemed to make a difference. These weren't mental 'put your speakers inside a pyramid' audiophiles, these were people building robust kit that was going to be put into flight racks and sent around the world.
Something, something psychoacoustics, if I recall correctly.
In the engineering world, this is referred to as "The Emperor's New Clothes" syndrome. People who claim to be able to hear differences like these are, quite simply deluding themselves into hearing something that they can't perceive, because they know - intellectually - that the signal is different.
Do exactly these same tests under truly blind conditions, and they won't hear a thing. Nor will they be able to discern a system that reproduces these sounds from one that doesn't.
No it's perfect to go alongside my new super-wide-band video format - I capture UV which you can't see and I need an audio format that you can't hear to go with it.
Obviously you'll need to wear sunscreen when you watch it but that's nothing for a dedicated audio-cine-phile.
I've started to record music directly off my record player and encode in flac and there are albums where I have both the cd and record (even some where I've downloaded the mp3s version as well) and it's amazing what a difference it can make on a well setup system (1970s amp, wharfdale diamond speakers and a numark ttx to hold up against my 2 year old trying to play with it, external sound card with phono in/out and cd deck) some radiohead albums sound massively different on vinyl and I know the mclusky re-releases got remastered and it shows.
It's worth the effort if you live your music.
@ Sgt_Oddball
It is possible of course that the Vinyl and CD versions were from different masters, so they will sound different, because they are different performances.
Tape and vinyl both have (and have always had) inherent limitations, which are far more noticeable than a 16 bit limit or a 22 kHz limit ever will be - so yes, I agree, they will sound different.
Now which you prefer is probably based partly on what format you *first* or "most significantly" heard the music from - or even which master you heard most...
They sound different because the older systems severely distort the signal. They simply distort it in ways that you like. Take a look at the output on a frequency analyzer, or any other objective measurement tool, and you'll find that the older systems absolutely suck in terms of fidelity.
Quite a few modern amplifiers even acknowledge this shortcoming when they include "soft clipping" circuits that mimic the non-linear rolloff exhibited by tube amplifiers. This is a gross distortion of the signal, but quite a few listeners find it pleasing.
Some people think the current overuse of vocoders is also pleasant. That doesn't mean that Madonna is a good singer.
You are mistaking damage from lousy remastering by a moron who was given some fancy studio gear to play with during his intern year for difference from "audiophile sound".
Grab something that was remastered properly under paranoiac original (or at least some of them) artist control like the reissue of the Moody Blues albums and compare.
Replying to myself (I know - bad sport).
A lot of the damage done during bad remastering today is deliberate because marketing insists on making the sound mp3/modern friendly. There are a couple of pretty good studies on this. The summary is that:
1. Quiet transitions are gone - think of Pink Floyd. It would never get past marketing in this day and age. Everything should go off the median volume just a bit and be LOUD. VERY LOUD. EXTREMELY LOUD. As LOUD as the encoding in iTunes will allow.
2. Bass. Beats is not alone here - making the bass go BOOM is a must.
You can see that most remasters surrender to marketing and make it sound "modern" by comparing the tracks (remastered and original) on an analyzer - it is bloody obvious. _THIS_ has nothing to do with audiophilia. It has to do with idiocy and lack of artistic control over the remastering.
For example - I bought a remastered Scorpions album this week and I could not recognize it. It was vandalized to the point where it was barely recognizable.
Well put. Going for high resolution audio is pointless as long as the music industry continues to insist upon destroying the quality of music during mastering. A high resolution recording that has been smashed to bits during mastering will sound far, far worse than a CD quality recording that has been mastered to sound good instead of loud.
The first thing the industry should do is ditch this stupid practice (so many online streaming services auto-level the volume anyway - making the whole thing pointless). Once they done that and are mixing/mastering music with the intention of achieving maximum sound quality, then maybe high-def audio formats may be something that can sell.
Until then, they're just effectively trying to sell high-resolution photos of a turd. Doesn't matter how sharp and detailed they are, the result is still a turd at the end of the day
That's right. Proper mastering is key. Modern recordings of popular music are mastered to be played on a noisy environments, like automobiles. They are significantly compressed and there is no dynamic range. With that in mind, properly mastered recordings converted to 192 kb/s mp3 sound really. Its not the mp3 that's at fault.
So you're calling Steve Albini a moron? That's bold... Same goes for elbow (who also control vinyl releases).
Not every band jumps on the band wagon because it's hip. I do have other releases that aren't as good (blue's parklife being one), I've also got a few radiohead albums (such as King of limbs and kid a) which pulls out more detail from the music and I've lost count howmany times I've lost myself in kid a and I'd still pick the vinyl over the cd copy I've got..
I can definitely hear the difference on my FLAC 24/96 wireless stream to Qualcomm Allplay ( the only multiroom that supports lossless wireless HD audio along with a DAC that can do it justice).
Official releases like Led Zeppelin II sound amazing, even unofficial releases like a vinyl rip of Queen II at 24/96 sounds fantastic.
Put a CD of those albums on, its noticably dull.
But a double-blind study is hard to argue with. Take a high-res audio source, then feed it optionally through back-to-back A/D and D/A convertors at 16-bit 44.1kHz, and see if you can tell the difference. Answer: you can't.
http://drewdaniels.com/audible.pdf
Just because something says SACD or DVD-A on the label doesn't mean it was mastered as high-res audio.
Back in the early days of vinyl, I know for a fact that some CD masters were recorded from vinyl copies.
Record companies tend to be cheap and nasty, and the likelihood they'd consistently make a special effort to get the very best from high-res media isn't high.
At the very least I'd have wanted to see some word length analysis to check that the supposed high-res content was actually there in the first place.
A more realistic test would be to record high-res audio with a clean signal path to master recorder, and then run that through the A/D/A system.
As someone who has spent a lot of time listening to converters, I find it amazing that it's apparently impossible to hear the effects of a mid-price A/D/A converter at all, never mind the source. Unless you're in the professional bracket (PrismSound, etc) most converters really don't sound that transparent. And I know from experience that the difference between a 24-bit master and a downsampled 16-bit master is absolutely and reliably audible.
@TheOtherHobbs - And I know from experience that the difference between a 24-bit master and a downsampled 16-bit master is absolutely and reliably audible.
There are only two ways for this to be the case:
- You are a bat.
- You're not doing it properly double blind.
OK, that's not fair - there is a third way - the 16bit master was mucked around with too much - adding dither noise at each step - that's why we record (ok, also to allow us to be cautious with gain settings) and master at 24bit.
Once mastered there is no reason not to down-sample to 16bit/44k, and every reason *to* do so.
To make it double blind you really need to make sure that your audio signals are at the same level to a ridiculous accuracy - because as salespeople know "louder is better". No really - even stupidly low differences (like 0.2dB, which is not consciously audible btw) show a distinct preference for the louder source.
"you're a bat"
Um, not necessarily. A 24-bit system has a much lower noise floor (or alternatively, a higher headroom) and therefore a greater dynamic range than a 16 bit signal.
As a rule of thumb, allow 6dB/bit for overall dynamic range. Then subtract (at the recording stage) 12dB for headroom an 11dB for quantisation noise - so a 16 bit system, irrespective of bit rate, will have a practical signal to noise ratio of 73dB and a dynamic range of the same order. That's not significantly improved over a 1980s broadcast tape machine, as it happens...
The extra eight bits in the 24-bit system allow an equivalent 48dB improvement in noise or headroom - easily audible provided (a) that your recording front end is both sensitive enough and quiet enough to be effective on close-to-silent signals, and (b) your listening equipment and environment is equally quiet and isolated enough to be able to hear very quiet signals. This is unlikely to be the case outside a professional recording studio.
@142 - 12dB because that's about the minimum you can get away with; 18dB is safer.
@nijam - the S/N ratio gets worse as the signal amplitude decreases because the quantisation noise is constant; 11dB was the average used in the BBC (might have changed since I left).
What the? NO!
That is absolutely incorrect.
You can *and should* use a properly designed digital system all the way up to just under 0dBFS. (allowing for intersample peaks of about 1dB or so).
There is no sonic advantage to having peaks 18dB below that. None. At all. Just huge disadvantages.
Unless you've got the most appallingly designed analogue components in your system/converters.
You may misunderstand my context, 142... for a final output, surely, use all the bits you can (but be aware of the nastiness of compression-for-loudness).
When I say 18dB headroom, I would expect the nominal zero dB signal to be 18dB below 0dBFS - but I would expect that 18dB to be used by the signal.
In, say, a live studio environment you will find that live voice and music both have level extremes which are generally unrehearsed and, trust me, the last thing you want is digital clipping. In analogue systems the approach was to assume 12dB headroom since an analogue mixer is usually a bit kinder as the signal approaches clipping. However, even there, you are shoving a 775mV signal through amplifiers with at least +/-24 supplies.
In a music recording studio (as opposed to, say, a live concert) you would, I assume, have much greater control over individual levels. Nonetheless, you would generally want a sound mixer with at least 24 bit internal representation, even with a 16 bit input and output.
However, it's some years since I left the BBC so while I have a lot of experience with good studio technique, practice may well have changed since I left. And of course, music recording is a different kettle of fish from live studio work.
@ Neil Barnes
The reason we run at -18dBFS as our "new zero" is for recording headroom - there is no reason to leave it there when we export the finished article.
So that's 12 bits you can add back on.
The 11 bits for dither noise - I rather suspect that is not true.
On the basis that a guassian dither will allow you to hear a tone at well underneath a single bit of amplitude, albeit with some noise, I struggle to believe that a sane dither would *lose* 11dB of resolution.
There are *extremely* good reasons to oversample as part of the DAC (it makes the analogue side filters much easier for instance) and very good reasons to use higher bit depths (it provides an excellent noise floor and headroom when recording) - but for the product, CD quality is way past most of our hearing capacity...
"...some CD masters were recorded from vinyl copies."
Meanwhile, I've watched the 'How it's made' episode on the resurgence of vinyl LPs. As the engineer was preparing to cut the master, a key step was inserting the CD into the cutter machine.
4 A NOTE ON HIGH-RESOLUTION
RECORDINGS
Though our tests failed to substantiate the claimed ad-
vantages of high-resolution encoding for two-channel au-
dio, one trend became obvious very quickly and held up
throughout our testing: virtually all of the SACD and
DVD-A recordings sounded better than most CDs—
sometimes much better.
It's too bad but the original LP masters of older music are already compression limited and it takes "remastering" directly from tape to digital to fix that. All you need is for the A/D convertor to be beyond reproach, not better media.
If music was recorded when the vinyl LP was the primary playback method, the base frequencies were limited on the LP pressing master so the recording would fit on the LP. Only the master tape was at full dynamic range. Then there were the RIAA equalization curves applied during preamplification for playback.
A normal CD will capture all the detail and resolution of the old master tape which is why the latest Led Zeppelin cd's sound better than the LP (and different). Using high resolution CD's and "lossless" formats with a dirty source just gives you a really sharp but noisy CD.
This. One thousand times this.
I own recordings from the 1970s, 40 years ago!, with sound quality 100 times better than most of what's been produced in the last 10 years. This whole "hi def" thing is yet another ploy to get you to re-purchase what you already own all over again. You can re-format crap sound all day. It's still crap sound.
To the record companies: Get good audio engineers back in the studio. Listen to well-produced albums from 40 years ago, hang your head in shame at the low-quality mush you've been churning out, and FIX IT!
They knew how to produce great sounding recordings 40 years ago, They still do, and hint, it has nothing to do with "hi def".
"They knew how to produce great sounding recordings 40 years ago, They still do, and hint, it has nothing to do with "hi def"."
Actually, they knew how to record high quality even before that...take a listen to Buddy Holly and the track "True Love Ways"....which was recorded back in the late 1950's, nearly 60 years ago.
On a good quality hi-fi system (that can show dynamic range), the sheer openness and quality of what they did in ONE take, and with all the musicians in the room at the same time, was truly breathtaking. And, truth be told, you don't even have to like Buddy Holly to hear how good this track is.
The existing resolution/quality is perfectly adequate.
The quality of the content, not its reproduction, is so much more of an issue - it also cannot be solved by technical evolution. It is the creative, artistic aspect of the work. It requires different (i.e. artistic) skills, and ability. It also doesn't sell new hardware based on new technology - equipment from the last couple of decades is perfectly good enough.
If you are really looking at supposedly higher quality reproduction, you'll likely want a completely acoustically isolated and perfect listening room. You'll probably have to treat it as a clean room as any dust particles will play merry hell with the whole affair. And you'll spend your energy trying to discern whether it was a Northern or Central line tube that you think you could hear for 1.34 seconds after the 3 cymbal crash, rather than whether it was a jolly fine piece of music or spoken word or whatever.
Good original content, produced well, is what the consumers want. Not fancy technology.
When I finally bought our first 1080p TV, during the initial 'critical viewing' phase that goes with a newly purchased TV, I noticed quite a bit of diffraction on some of the high contrast scenes. It was very annoying, and it took me a few days to find the cause.
After I trimmed my eyelashes a bit, the picture quality snapped into sharp, diffraction-free focus.
There's a lesson in there somewhere.
If consumers don't have equipment (headphones/speakers, DAC) that can reproduce lossless material of good provenance with a reasonable degree of fidelity then it really doesn't matter how good the recording is or what bit depth/sampling rate it's recorded at. So let's get them to at least CD quality first, before worrying about anything else. I'm positive that even the best recordings are let down by shitty postprocessing work, lossy encodings, cheap DACs/resamplers and headphones.
Then we worry about whether or not HRA is of any use to anyone. Personally I'm of the belief that you can never have too much information, and yes, beauty is in the ear of the listener--so some people may simply never notice, honestly, and that doesn't have to be a bad thing.
@Sebby - Personally I'm of the belief that you can never have too much information.
Can I interest you in my new TV standard - it doesn't just deal with the visible spectrum, but goes all the way to X-Rays...
There is a point at which information is sufficient. At the point where I can perfectly reproduce the full spectrum of visible/audio frequencies with a dynamic range to span 'barely detectable' to 'painful' then I can only add:
- Pain
- Information that is instantly discarded, since the receiver (the eyeball or ear) simply can not deal with it
It's not easy to cover important technicalities with brevity and clarity, but you managed it. Nevertheless, prejudice and ignorance is hard to do battle with.
There will also be people who say 192kHz sampling is better than 96kHz, and look for more. There will be dealers who are financially dependent on doing things like selling £100 analogue cables for transferring CD audio from player to amp.
"Want to hear the difference? No problem. Use your ears..... See? You get what you pay for. SO much better than using a cheap digital cable. That'll be £100, please. So happy to have a discerning customer."
Prejudice and ignorance are really easy to deal with - Double Blind tests. No fucker can tell the difference - you dont even get the placebo effect that might come with homoeopathy. Its all marketing woo.
Work out the Doppler effect from the warm air rising from your body while sitting on the sofa and then follow the link to buy some of my special 'air stop' foam at £200 a cubic meter, or alternatively my £10,000 room conditioning tape which you string across the room to prevent air flow.
> the Nyquist-Shannon sampling theorem proves that a bandwidth-limited wave can be perfectly captured and reconstructed to be as smooth as the original.
Are you saying the music is a truly periodic signal, or that it is sampled in time from minus infinity to plus infinity or something similar? If not, how do you apply the theorem to get *perfect* reconstruction for finite non-periodic signals?
Of course, perfect reconstruction is not actually needed in practice. But that is a different matter and one has to go and do the hard work... Waving the sampling theorem in front of people is just bullshit.
@DN4
A bandwidth limited wave is not necessarily periodic - we show periodic functions (whether sine waves or square waves) as they are easy to demonstrate (you don't need to do time synced overlays).
But a bandwidth limited waveform is purely represented by an appropriate sample rate. Whether it is finite in time or not. The FFT of the waveform would be harder to generate, but we're not trying to.
> Are you saying the music is a truly periodic signal, or that it is sampled in time from minus infinity to plus infinity or something similar? If not, how do you apply the theorem to get *perfect* reconstruction for finite non-periodic signals?
As I recollect, the periodicity only gives uniqueness of decomposition. That uniqueness may perhaps have some practical benefit, but any decomposition would still be correctly re-composed. As for the infinite duration issue, you can easily show that you get the same results (at least so far as human hearing is concerned) as by feeding an arbitrary duration of silence into the sampling process before and after the recording.
So actually it isn't bullshit in practice.
pssst? hypersonic refers to much greater than the speed of sound (and is a nebulous term. supersonic is faster-than-sound, hypersonic is faster-than-supersonic, typically somewhere over mach 5). ULTRAsonic refers to high frequencies above the range of hearing.
the REAL problem with 44.1k sample rate is that at the nyquist limit (sample rate/2) there's ZERO phase information. high frequency sounds rely on phase relationship to maintain stereo directionality. in fact, a 22Khz sound thats at the wrong phase relative to the samples will be nulled out, and a 21.95Khz signal will pulsate as it goes in and out of phase with the sampling (this is similar to the moire effect of almost parallel lines). even at considerably lower and quite audible frequencies like 10Khz the amount of phase error introduced by sampling is significant and seriously messes with the stereo 'staging'.
@John Geek - what makes you think that any sampled signal is recorded with frequencies up to the Nyquist limit? That's why there are brick-wall filters on the input and a reconstruction filter on the output (and no, let's not go into the mess that a poorly designed filter can cause...)
Oversampling at very high rates has one obvious advantage: it makes the necessary pre- and post-sampling filters much easier to design and build.
We were doing good up until your post. Only one (just 1 !) typo in Neil's post, but obviously a typo because he had the rest correct. Every other post had 100% accuracy in the correct case of "kHz" and "dB". It was a joy.
And then you came along... THREE "Khz"!
Sigh.
Getting higher quality sound than the common herd is a sort of religion with some people. You will no more convince them that their chosen route to heaven isn't valid than you will convince religious fundamentalists of the same thing.
Maybe that's because music can have a direct line to the emotions, but, like sex, the effects aren't under conscious control. So people need a way to get in the right frame of mind. Concentrating on the finer points of sound quality forces you to pay attention to the actual sounds, and avoids distracting thoughts. And a feeling of inner satisfaction at being able to hear subtleties that escape most people probably helps as well.
The exact means doesn't matter so long as it achieves the desired result. In the mid 1950s a friend of my father's explained to me how listening to Chopin on his acoustic gramophone was a far superior experience to the new LPs. Perhaps it was the slower wow of the newer medium that bothered him. But I think it's more likely that, having invested money and emotion in a superior wind-up gramophone, music on an ordinary LP record player just didn't turn him on.
My friends Father owns a few high end horn gramophones. In the 90's he made quite a bit of money by renting them to a specialist CD company who were recording old 78 records. Surely the thing that your friends Father found pleasing was the fact that it was using a horn rather than speakers?
Even if "high def" audio recordings are any real improvement on CD, you're not going to notice the difference unless your reproduction gear is up to the job.
Hardly anyone has audio reproduction gear that counts as "top notch hi-fi" by the standards of the 1970s, let alone kit that's better than that. So for most people, even if "high def" recordings do provide a potentially detectable improvement, it's not going to be apparent on playback.
So what's it all about, really?
The article's point about the poor quality of many CDs is a very good one. If these "high def" audio types really wanted us to have access to decent quality music, they'd be concentrating on producing well mastered CDs which weren't mucked up with dynamic range compression and other nasties.
But of course that's not what "high def" audio is about, is it? It's really all about selling a dream and making money from people who can't tell the difference between impressive looking specs and genuine high quality.
A couple or four other points.
Oversampling to increase the apparent sampling rate from 44.1kHz CD standard has a real engineering benefit: it means the DAC designer can easily come up with circuitry that gives greater rejection of unwanted frequencies with less distortion of all sorts, thus producing a higher quality analogue signal to send to your amp and speakers. But interpolation of a 44.1kHz signal to higher frequencies will give you this benefit without the need for higher sampling rates in the recording - and you can buy stand alone DACs to do just this at moderate cost (I've got one so I can play music from a PC HDD on my proper stereo).
The original article's claim that "Increasing the sampling rate extends the frequency response" is true, but isn't a real advantage since 1) most audio repro gear can't reproduce higher than the 22kHz theoretical max of CD and 2) most human ears can't hear such frequencies, which is why audio repro gear mostly doesn't bother with them. (my ears are nearly 50 years old: when they were much younger, they could hear up to about 22kHz when tested. They can't hear much above 16kHz these days.)
The advantage of 24 bit audio over 16 bit audio is that you've got greater dynamic range available, which is genuinely very useful in the recording and mixing stages.
From the point of view of "final consumer recording" even if your ears are able to spot the difference, you'd need one hell of a listening environment to take advantage of any such improvement - not just top notch repro gear, but nigh on perfectly insulated from noise, ideal echo characteristics, and so on.
From my experience, the reason some CDs sound worse than the "equivalent" LP is because of dreadful mastering of the CD. I sometimes get the idea that the record company has hired the cheapest chimp they could find to crank out CD masters as quickly as possible while paying the proverbial peanuts, and to hell with the actual music.
CDs have astronomically better dynamic range than LPs, which also means better noise figures even before you consider dust and wear on LPs. LPs can just about match CD frequency range, until they get worn - wear which increases noise and distortion. Technically, there's no excuse for a CD to sound worse than an LP. If you've got LPs which sound better than the equivalent CDs, the problem is probably bad CD mastering (or maybe you need a better CD player).
I sent one CD back to Amazon after a single listen: a recently remastered compilation by the original artists which was so awfully done the music sounded like it was bad cover versions by a band which had no idea what the music was supposed to sound like.
Sometimes, you can get a CD which is remixed to be a huge improvement on the original LP. It might not be to everyone's taste, but Roger Glover's 25th Anniversary Edition remix of Deep Purple's 1972 "Machine Head" is a case in point.
And finally, given that the youth of today mostly seem happy with lossily compressed files downloaded and played back on their laptops or tablets (or phones - shudder!), surely just striving to achieve CD quality would be an improvement for most?
There are three types of music:
1. Studio-recorded, multi-tracked, mixed, autotuned, normalised (i.e. compressed dynamic range to make it VERY LOUD) pop. No point in making that hi-res, since it's all been through a load of computers first and bears no resemblance to original vocals or instruments. I remember students preferred the "fizz" of 128k MP3* to the original CD as that was the sound they were used to!
2. Live rock/pop. Few of the pros can sing or play as well as on their records (the X Factor is unmitigated drivel but at least they have to be able to sing live) and you have crowd noise/stage buzzes so no point in hi-res.
3. Classical recordings. There might be a case for hi-res but it's a small market.
*I can hear that 128k MP3 is a bit crackly but 256k or high quality VBR, with any reasonable encoder, is as good as CD for me.
> bears no resemblance to original vocals or instruments
So!? Artistically, **we don't want them to**. Otherwise, we'd just stick a single binaural mic in the rehearsal room and release that, and save everyone a fuck load of 120 hour work weeks. There is detail there. An awful lot. In automated delays, very subtle instument stacks and incidental layers, very, very nice artificial reverbs (much nicer than real rooms), very very crude reverbs, distortions, and tons of other artistic sonic choices - there's some gorgeous imaging techniques in play these days too.
Much of which gets lost at 128kbps MP3.
I think I;ve read before about people who say they have 'golden ears' and can tell the difference between very subtle increases in dynamic range and sampling rate.
I however have 'lead' ears, which means I've never really been able to tell the difference between rubbish and amazing, or any grades in between. Which saves me money, as I can happily continue on my merry way with my £10 bluetooth earphones and MP3 files and get exactly the same experience as if I had spent £10k on some crazy super kit.
Ignorance may not be bliss, but it's a hell of a lot cheaper.
"It would be a brave person who said that HRA has no audible benefits whatsoever but..."
I am that brave person. Your home equipment can't deal with higher frequencies, neither the digital part nor the analogue part of your signal chain. If you throw a perfect HRA signal at your equipment you are likely to hear a difference but that difference will be mostly signal interference from your amplifier. The amp is simply designed to work really well for a certain band of frequencies...
Your speakers can't output these high frequencies either (and not your headphones either) and if the signal makes it to the speakers (or headphones) you'll either get interference or muddiness. You would need one or two extra tweeters specialised for higher frequencies to make HRA worth it.
Completely possible, but nobody has this equipment. I've never seen such equipment, at least.
Studios use high sampling rates and bitrates because they are working with multiple signals and suffer less signal degradation when mixing if high resolution audio is used there. After it's all been mixed together the hypothetical benefits are very, very small. And the real world benefits are nil.
Most off the shelf mass market hi-fi may not be able to reproduce higher frequencies, however it is quite easy to source tweeters that have a good frequency response up to 40 KHz which I have tested and used in the lab (admittedly not for the purposes of reproducing music).
Similarly whilst many PA power amplifiers 'boast' that they filter everything above 22 KMz it Yamaha at least make ones that are reasonably flat up to 40 Khz and have useful output almost up to 100 KHz.
Whilst you may not be able to 'hear' above ~20 KHz it is still possible for many people to percieve something well above this. My boss can tell when our air-coupled ultrasonic transducers are operating even from a distance, and they are typically operating above 160 KHz. with a high-enough burst reppetition rate I can 'hear' if they are working when close up.
There is a clear difference between my £400 hi-fi and my friend's £40,000 system, but I have better things to spend my money on. Also I have some CDs that on his system are terrible to listen to becasue of the quality of the recording, which sound fine on my system.
> Your speakers can't output these high frequencies either (and not your headphones either)
Not all the way up to 192k's nyquist of 96k, of course, but almost every tweeter and audio cable will give response way over 20kHz, though the linearity suffers.
That we can't hear those frequecies is a different point, obviously.
Hard to trust studio engineers when it comes to sound quality as half of them have been deafened on the job. Just listen to the awful speakers* they monitor on and the painful levels. As a friend who sells audio stuff said recently, most engineers have reliability as their number one priority when choosing gear -- fidelity somewhat less. Worse, recent artists seem to actually seek what they call a lo-fi sound -- e.g. albums by Doves.
* is that a JBL in the picture ? Nuff said.
OK, to be fair - a lot of recording studios employ cheap-ass speakers, simply because they're trying to ensure that their product will sound OK on the craptastic systems so many people listen too. You'll see those same guys listening through $6 earbuds, too, for the same reason.
The rest of your post is spot on, however. There is no difference these "high def" systems bring to the table that any human can actually hear.
You shouldn't be so sure of yourself, the science understanding hearing is a an evolving one, and is made incredibly more complex than just wave propagations by the brain. Psychoacoustics is in its infancy, and it is still awaiting for its first studies with large population samples.
But It is a fact for example that most young people can discern higher frequency than older ones, and that one looses frequency sensitivity as one becomes older. It is a also a fact that many people physically feel low frequencies, which may or may not enhance their hearing experience.
In the same way, do not assume that because you are unable to tell two colour aparts other people can't (hint: true tetracromats).
Also please don't put gold-cable hi-fi fanatics and professionals in the same bag.
These are the same people who've claimed that gold-plated connectors sound better, or promote "oxygen-free" copper for its "superior" signal transmission qualities.
I'd love to have any of these self-proclaimed "engineer" submit to a truly blind, random listening test, and see if they can simply pick out their "high def" playback from an ordinary CD. When a similar test was done to compare high and low end stereo systems (after much reluctance, when the blind nature of the trial was revealed) a similar gaggle of audio "experts" wound up picking out a cheap, ~$100 K-Mart utility system over a staggeringly expensive, multi-thousand dollar one. More than once.
The Myth of the Golden Ear has simply found a new bottle of snake oil to hawk.
Information & technology. As IT professionals, we ought in principle to be in favour of capturing and storing more information. It baffles me that so many of us seem keen to pay good money for music formats that throw information away. I would venture to suggest that those in favour of crappy equipment and lossy compression are people who don't listen actively to music. You never hear these luddite arguments about video - "Oh 405 lines black and white analogue TV is fine. Nobody needs colour and HD." If a bunch of activists are promoting the highest possible standards of audio recording and mastering, that's good for everybody. For those who actually can, or only claim to, hear the difference, quality is maintained. It doesn't stop you downsampling it and playing it through earbuds if you want. So it gives freedom of choice. And that's a good thing.
You can't hear the difference. There's no sensible information being "lost" by ordinary CD standards, and no discernible information is being retained by "high def" systems.
As IT professionals, we ought to be appalled that such psuedo-science is being peddled in the public marketplace, and have a duty to expose it as the snake oil it is.
> You can't hear the difference.
Wanna bet? Or, would you care to back that assertion up with reference to any rigorously conducted double blind tests?
But, even if you can't hear the difference, imagine that, in future, your audio software can zoom in on, say, the first violinist in the orchestra, so that you can pick out clearly what she's playing. You'll be very glad that you have a recording that has enough detail and a low enough noise floor.
That's one of the points of recording stuff - it's for the future, not just for today.
Those blind tests have been debunked for flaws like choosing music that's been badly mastered. And bad mastering is one of the evils these HD proponents would like to do away with - which will benefit everybody, even when played back at 44.1kHz, 16-bit.
I'll ask again, what's wrong with archiving music with the highest possible quality? We're upping the quality with which we capture digital images and video all the time. Storage costs are dropping all the time. And HD audio doesn't even involve killing kittens. But, oh no, we have to get all grumpy and stick with our shellac 78s.
@GasppardWinkler
Archive away - but you still don't need to archive more sound than is audible. CSI "enhance" will still only be possible in hollywood.
The fundamental difference between audio and video is that we can change the spatial resolution of an image (moving or static) by moving it relative to our eyeballs.
When I buy/download/stream music why should I transfer many times more data than I need?
The difference between High Res and CD is audible by many people with normal hearing.
The problem is, you need a good setup for this.
You can hear the highest res audio on crap iPhone headset or even crappier more expensive Beats B.S. and you won't tell the difference from a lower res source.
Similarly if your recording sucks because you have no idea how to record, mix or master, or you are recording horrible instruments or voices, no amount of high res audio will save you.
Conversely, with a good Hi-Fi / Pro installation, calibrated and rightly set up for the audience, as well as a not too crappy room (sound-wise), and of course good recordings, I beg to differ, many people will hear the difference.
For final playback purposes, Dan Lavry (who makes arguably the best converters going) argues that for a person with a normal ear (ignoring people who can hear 24khz sine waves) any audible improvement at higher sample rates is the DAC's manufacturer being unscrupulous and hyping above 10kHz or so, in order to make people feel they're getting extra detail. You can get the same sonic improvement at a lower sampling frequency by just applying a subtle EQ boost.
And you're not going to hear the bit depth difference (on playback) unless someone's been careless with the headroom, left out the dithering, or applied a bunch more digital processing to a 16bit source before playback, or it's being played loud enough to cause instant hearing damage. Granted it's not that unlikely that one of those scenarios is in play, but you did say "properly configured".
----This all assumes they're the same master source at different sr/bd. If they're different sources, then of course all bets are off.
@GaspardWinckler
The reasons for higher rate sampling and higher bit depths is for the recording studio, and for the post processing work.
For final playback - no point at all. For the steps before (and during) the thousands of processing steps which are applied to each track, before those tracks are grouped, down mixed, then processed, grouped and downmixed again, and again, and again....
After all of that you are grateful for the extra bit depth you used initially, the noise floor is still inaudible.
The higher sample rates are primarily used to reduce the cost of analogue filters (required to produce the bandlimited signal for encoding) by allowing a secondary application of a digital filter before downsampling (these filters are *much* cheaper). I imagine there is a minor lower latency benefit for digital processing internally in a live mixer as well...
I challenged my "self appointed audio expert" friends to try if they could decide which was a record or a HiFi Dolby cassette recording of the same record, played back to back. None could.
I have never been to a recording studio to listen to the "original" and thus have never had any means of comparing a recording to this "original" sound. The originals for me are whatever I hear the recording on. For all I know I might not like the "original" sound. I know for sure that some pieces of music I really like I do not like the live version of.
I have Linn Akurate gear capable of 24/192 playback. But I don't buy hi-res anymore. Two reasons: Firstly I was becoming slowly convinced that my favourite recordings were not always 24/96 or higher and secondly because I came across this article which is so obviously based on solid scientific knowledge that it is impossible to ignore:
http://xiph.org/~xiphmont/demo/neil-young.html
Once you get to the point about TVs that omit x-rays not giving you a better picture and also that emitting stuff above the audible range may actually create intermodulation artefacts in the audible range (why should your tweeters maintain linearity at that kind of frequency?) then you realize that hi-res audio is a con and that well mastered, uncompressed, 16/44 CD audio is more than good enough.
I'm just waiting for the first "Well you must be bloody deaf then" comment
Which is the typical retort of an audiophile who's just had his snake oil system slighted by accepted and proven science.
I'm a hobby producer, and yes there is a lot of advantage to using hires audio throughout the production chain. It gives you headroom and it gives you the ability to apply multiple effects without significant quantisation errors. However for the end product delivery there's no point in doing anything more than 16/44.1.
The primary reason is that the vast majority of people will listen to the music on transducers that can't even follow a 16/44.1 accurately,
Even those with truely accurate transducers, if such things exist, will not benefit from anything higher than the CD format as the extra information is simply not audable.
Hires audio is simply the music industry's way of digging into your wallet again for age old stuff that you've already got.
“lossless audio that is capable of reproducing the full range of sound from recordings that have been mastered from better than CD quality music sources.”
This is an utterly empty, meaningless statement. I'll just note that CDs themselves meet this nebulous non-definition.
I'll be sure to put this equipment on my rack right next to the oxygen-free free, directional speaker cables with gold-plated connectors I bought for $200 per inch.
Full disclosure - I love listening to, and discovering, music. I have a moderately good vinyl hi-fi system plus I also own a Pono player and all my digital audio files are in uncompressed, lossless AIFF.
My take on this is simple... if you can hear the difference and are happy to pay for it then go ahead and enjoy.
A simple test... can you hear the difference between a FM / DAB radio broadcast and a CD? More interesting, what about a CD and a vinyl record? If you think vinyl sounds better then you will likely hear and enjoy the difference with HRA. If not, don't worry... it's subtle.
On a similar note, can you see the difference between DVD (480P) and "Full HD" (1080P)? What about "Full HD" and 4K? Personally I cannot... so I will continue to enjoy 1080P and will use the money I save to buy snake oil for my vinyl music and unidirectional USB cables for my digital audio ;-)
Forgot to mention that Pono also has a "revealer" option. You can take any of your own existing HRA music and the Pono Desktop app will convert it into 5 different "qualities"... MP3 (Amazon), AAC (iTunes), 44.1/16 (CD), 96/24 and 192/24. You can then play the song and switch dynamically between each quality setting to see if you can hear the difference for yourself (http://www.whathifi.com/news/ponorevealer-lets-you-compare-aac-mp3-and-high-res-tracks).
The human ear is not a perfectly linear device. As a result, percussive sounds can be affected audibly by frequency components above the limit for human hearing of steady tones - because they will raise the maximum total amplitude of a transient, causing it to encounter additional distortion, among other reasons.
I'm a bit late to the party, but here's how Sony advertises its technology: https://plus.google.com/+JensGrivolla/posts/1vC6GcLJRov
They compare their HRA to MP3s that are artificially cut off at around 10kHz. To me that's a clear case of false advertising, possibly fraud, but then IANAL...
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
