The new forums will be named Coin Return (based on the most recent vote)! You can check on the status and timeline of the transition to the new forums here.
The Guiding Principles and New Rules document is now in effect.
Digital Audio and New Encoding?
ArcSyn
Registered User regular
Registered User regular
Here's a question for some audiophiles:
Correct me where I'm wrong, but this idea just hit me one day.
Many audiophiles consider analog recordings and storage better than digital because it loses no quality from the actual sound of the instrument, where digital recording requires some sort of loss in the form of bit rates.
Here's my understanding:
An analog wave for sound is a smooth curve where a digital form of that wave is actually a number of steps (per second) and that number is determined by the bit rate.
So essentially, all data between one step and the next is lost and could change the sound heard. Which is why some audiophiles prefer records vs. CDs.
Here's where my idea kicked in:
Take a program for images like Photoshop vs. Illistrator.
Photoshop is our digital format where it pixellates the image (many small dots determined by a dpi) and the data between those dots is essentially lost. If you enlarge the image after it has been created, it is possible to see these dots (blockiness in the lines).
Illistrator can use mathematical equations to determine the wave form and curve and angle of lines so that when an image is enlarged or shrunk the viewer does not notice any loss of image quality. This allows an image to be enlarged from a thumbnail to a billboard and not be "blocky".
Back to audio:
With rising storage solutions allowing more and more space (as I assume it would be needed), would it not be possible to digitally store music as a mathematical equation, preserving the essence of the true wave form and curve so you would not actually lose the audio between samplings?
Essentially allowing an infinite bitrate, though it would not even be a bitrate, but a determination of the wave through it's mathematical equal.
Has this already been thought of, or even already implemented?
Am I crazy?
Correct me where I'm wrong, but this idea just hit me one day.
Many audiophiles consider analog recordings and storage better than digital because it loses no quality from the actual sound of the instrument, where digital recording requires some sort of loss in the form of bit rates.
Here's my understanding:
An analog wave for sound is a smooth curve where a digital form of that wave is actually a number of steps (per second) and that number is determined by the bit rate.
So essentially, all data between one step and the next is lost and could change the sound heard. Which is why some audiophiles prefer records vs. CDs.
Here's where my idea kicked in:
Take a program for images like Photoshop vs. Illistrator.
Photoshop is our digital format where it pixellates the image (many small dots determined by a dpi) and the data between those dots is essentially lost. If you enlarge the image after it has been created, it is possible to see these dots (blockiness in the lines).
Illistrator can use mathematical equations to determine the wave form and curve and angle of lines so that when an image is enlarged or shrunk the viewer does not notice any loss of image quality. This allows an image to be enlarged from a thumbnail to a billboard and not be "blocky".
Back to audio:
With rising storage solutions allowing more and more space (as I assume it would be needed), would it not be possible to digitally store music as a mathematical equation, preserving the essence of the true wave form and curve so you would not actually lose the audio between samplings?
Essentially allowing an infinite bitrate, though it would not even be a bitrate, but a determination of the wave through it's mathematical equal.
Has this already been thought of, or even already implemented?
Am I crazy?
ArcSyn on
0
Posts
to supplement terrorbyte's insightful post, lossless methods of recording digital audio have been around for a while. they do take up massive amounts of disk space. but as hard drives increase in size, yes, people are slowly but surely moving to the larger formats for audio. mp3s won't die out for a while though.
technically speaking, digital audio is already a "mathematical formula" of the actual audio. but it's a bit different than vector graphics.
steam | Dokkan: 868846562
you guys are missing what he's talking about
lossless audio still has a bitrate and a samplerate. it's still "pixels"
standard CD's are 44100 samples per second.
the closest to analog we have come so far is SACD, at over 2 million samples per second and it's 1-bit instead of 16-bit (I could explain that but that's a side point). they are still samples and it's not an algorithm like he is asking for.
I would like this "vector based audio instead of pixels" idea as well, but honestly it's not going to see a lot of corporate backing when it seems the masses think that 128kbps mp3s out of their ibuds sound awesome.
so we must wait for some small group of audiogeeks to have a breakthrough.
2. It is possible to digitally model an analog signal to an amazingly detailed degree. Such that you can describe a waveform with better resolution than the waveform itself.
3. Your system would not work because analog-to-digital sampling systems are by nature discrete, i.e. they have a finite amount of detail they can record.
4. 32-bit float PCM audio allows for neigh infinite dynamic range and at high enough sample rates can model audio no human can possibly hear. A lot of mastering anymore is done in this format and stored as such.
Here's an image to help you out.
Yes and no. Let's suppose you could broadly generalised graphics into two categories:
1.) Images which can be vectorised (corporate logos, flow charts, diagrams...).
2.) Images which cannot be vectorised (photographs, paintings...). These are sampled/scanned in.
I suppose you can broadly generalise digital music into the same two categories:
1.) MIDIs (and other similar formats) - from what little I know, they describe how the music should be played. I like to think of them as sheet music in digital form. They'll describe which instrument plays what note and what time and all that. This is similar to vector based formats, which describes how to, say, draw a line from point a to point b using what colour.
2.) MP3s (and other associated formats) - I guess these are the equivalent of paintings/photos in that they're sampled forms of analogue data.
So, I guess the format already exists, but just like vector graphics can't describe everything I see in real life, MIDI sound can't quite duplicate everything you can hear.
it was hard for me to even read past #2...
(i'm using pixels instead of samples because the differences between a vector and a pixel is exactly what the OP is asking for in audio)
so you think you can arrange a series of pixels to look like a curve, and then expect me to believe that they are not still just a series of pixels when you zoom back in? that they magically dissapear at some point and form a perfect curve?
as for 3. of course it would not work with current A/D sampling equipment because it is still based on sampling.
as for being unable to vectorize a painting.. well it's only because
a. noone has found/looked for an algorithm that does this kind of process
b. technology is not fast enough yet to even think about researching an algorithm of this scale.
when this all comes to audio though, it's mostly
C. the status quo is now 128-192kbps lossy files downloaded from online music stores. most consumers don't care/have never heard what the extra detail of hi-def audio can sound like.
so this will be a niche interest for lifetimes. ah well. <slight sarcasm> kids these days...with their 'hip hop' music.. :P </sarcasm>
As has been mentioned, the best use is for pure digital creations, rather than conversions.
this has been thought of millions of times by audio engineers who have already created formats of SACD and DVD audio..
your avatar....
just. oh. my. god.
that creeped me out so much more than it should have.
As for the OP, it doesnt make a shit of difference if you are using desktop stereo speakers anyways, as they will be the limiting factor for audio anyways.
Which means any developments in storage and recording must be coupled with speaker playback, which to my knowledge, is far behind.
It makes me wonder what made ArcSyn think of this topic, as it is basically the first question that everyone thinks of when they have to actually be concerned about the quality of something being recorded.
Are you doing a project involving some recording, ArcSyn?
Try not to hurt yourself while reading up on the Nyquist-Shannon sampling theorem. When you get done with that you can come back and appologize.
To summarize an answer to your dimwitted reply to my second point, as long as the sampling rate is sufficient to model the signal's Nyquist frequency then it is indeed possible to perfectly reconstruct the original signal mathematically.
if you would actually read your own link you'd see that "A signal that is bandlimited is constrained in terms of how rapidly it changes in time, and therefore how much detail it can convey, in between discrete instants of time. The sampling theorem means that the uniformly spaced discrete samples are a complete representation of the signal if this bandwidth is less than half the sampling rate."
so the standard format for audio today is hacking off any information that's greater than 22050 hz. it still isn't perfect. it's just 'good enough'
that's the whole point I've been trying to make.
Where would you get hi-def audio? Are you talking CD-quality, or something better?
I'm talking cd-quality/analog that's well recorded (so much stuff is compressed to shit these days) and listened to through gear that's better than ibuds.
and no I'm not talking idiotic multi-thousand dollar audiophile gear, you can improve the quality of your listening experience by a large margin for less than the cost of a 2nd ipod.
At the moment however I get more utility out of the convenience of lossy MP3s than I do out of high definition audio. That doesn't mean I don't occasionally pop in a DVD-A disc, however.... the market for "HD Audio" is very slim. When you can get audio of that quality into MP3-like file sizes... then we'll be talking.
But pretty much everyone has listened to CDs. So are you saying that most people have not heard music of that quality from high quality gear?
I'm just curious -- I like music, and if you're saying that there is a difference between CD quality and some other quality, could you give an example so I can hunt it down and give it a listen? And if you're saying that it's the equipment, I probably won't upgrade right now, but I'm still interested.
Get a decent digital amplifier, a cd player (heck any dvd player will work) with an optical digital output, halfway decent pair of speakers, and good speaker wiring.... you can do all that for surprisingly cheap. You can build top of the line speaker wiring out of ethernet cables, for instance.
While all this can be done for cheaper than the cost of an iPod, it's a lot easier to buy the iPod.
yeah that's basically what I'm talking about, it's the equipment.
I'm not sold on SACD/DVD-A either though. I've tried them both and most of the time if you sit down and listen, the difference between 2ch sacd and 2ch CD is placebo. they're really just platforms for 5.1 audio
Bandwidth really doesn't have a lot to do with what you're talking about though I think. You're talking about interpolating frequencies within specified bandwidths already set... I think. I'm no expert on audio-formats, but I'm pretty sure that this "interpolation" is taken care of by adding Quantization noise, which smooths out the resulting data. I'm mostly guessing at that from what little I've read on Wikipedia though.
All digital quantization techniques are going to be lossy compared to reality, whether they're CD, SACD, or MP3. It's just a question of how much loss is acceptable to people. Based on the MP3s available on the internet, I'd guess that they're willing to put up with a lot. They're not stupid. They're used to hearing everyday sounds, and know what good sounds like. They just aren't that picky about their stereos.
Stop, take a deep breath, and realize you are out of your fucking depth Donny.
A CD has a sampling rate of 44.1KHz and a sample size of 16 bits. That means a CD purchased today can accurately reconstruct sounds up to 22.05KHz with a dynamic range accuracy of about five decimal places. That means a CD available right now can play just about any audio frequency a human being can distinguish. Not only can we not distinguish sounds above about 22KHz but most audio equipment can't reproduce the signal even if it had to.
The point you have been trying and failing to make is PCM sampling is somehow inadequate for capturing audio. I tried to show you that you were not only wrong but impressively wrong. You're now trying to reframe your original dimwitted argument as being about sampling rates and not sampling methods. You do not understand what you are trying to talk about.
As an aside, there is an argument to be made that CD's are just "good enough" but not great. This is true, the sampling rates and sizes on CDs were chosen by the record industry as to not be too good but acceptable to almost everybody. Considering CDs were head and shoulders above the quality of cassettes and LPs they were readily accepted once their price dropped to consumer levels. The only people that are really bothered by CD audio quality are classical music aficionados and self-deluding audiophiles. Even though you can't hear frequencies above aboue 22KHz there's still quite a bit of acoustical energy produced by some instruments, violins and such especially. While this cannot be heard directly is can be sensed just like super low frequencies may not be distinguishable but are quite obviously felt. This is where DVD-A and SACD come in as they offer better audio reproduction than CDs. For almost everyone however they're complete overkill and will likely never be mainstream.
Also MP3/AAC compression does not change the output frequency of the audio end product. They will spit out pretty much the same frequencies as the original signal, there's nothing preventing them from doing so save for artifacts of compression. A 44KHz MP3 has the same capability to reproduce a 22KHz signal as a 44KHz WAV. Whether a 22KHz signal actually survives the encoding from the original waveform to the compressed version is based on a large number of factors. However there's nothing fundamental about psychoacoustic modeling that prevents such a compressed file to be unable to reproduct any particular supported frequency.
I can't stop you from reading between the lines, but here's the only few lines that matter.
I'm done here...the OP came to ask if a 'vector audio' digital format existed and I think it's been well explained as to why it doesn't.