New Posts  All Forums:Forum Nav:

The Unix Philosophy of Audio

post #1 of 27
6/11/07 at 3:18pm
Thread Starter 
This post is mainly about how people approach audio, but I just wanted to talk about it because it is driving me crazy in a lot of threads, and I am wondering what other people think about it.

The Unix philosophy in computing is generally accomplish large tasks by stringing together many smaller ones. In computer terms this means you have a lot of small specialized utilities, and you pipe the output from one to another until the final output is what is desired. A good example of this is ripping and encoding CDs. You use something like EAC to handle the ripping job, something it is incredibly good at, and the output of the rip (i.e. the .wav's) is fed to your encoder, which is especially good at its job, psychoacoustic compression of audio.

In audio terms, let's say we have a CD we would like to listen to. Our chain looks something like this:

CDP -> DAC -> EQ -> Amp -> Headphones

We have six components in this chain, each suited to a very specific task. The CDP is dedicated to ripping the bits from the CD as accurately as possible. The DAC is responsible for converting these bits into the representative analogue stream. The EQ is responsible for shaping the resulting waveform in what ways we desire. The amp for amplifying the signal to the headphones specifications. The headphones are responsible for reproducing the resulting waveform. The cables are responsible for accurately transmitting the information from one stage to another.

I've always looked at audio this way, and I have a feeling that a lot of the "objectivists" on this forum the same way.

I've also come to realize that this way of looking at audio really isn't shared by most. Some general examples I've seen:

Quote:
Certain high end DACs and CDPs have synergy
Translation : Certain high end CDPs are defective. They are not accurately recreating the source bitstream.

Quote:
High end cable A has a much smoother sound, while high end cable B has a much tighter, aggressive sound.
Translation : High end cable A or high end cable B, possibly both, are defective. They are not reproducing the signal to a high degree of accuracy.

I could provide more examples, but you get the general idea. Given an unlimited amount of money, the ideal system would have completely "neutral" (I really hate this word, but from the post I think you understand the context) Headphones, DAC, CDP, Cables, and an Amp. The EQ would be the sole source of coloration.

In reality this is not possible, but we can come close. Headphones themselves for physical reasons are extremely difficult to achieve true neutrality, but we then have the EQ to compensate. For every other component though, there is no technical reason that they cannot be accurate to an absurdly high degree.

This philosophy also has the advantage in that it provides a good framework to test components. The digital end (CDP->DAC->) has a mathematically defined function we can measure against. The analogue side has a user defined set of functions we can test against. The greater the deviations, the worse our system is.

As I said at the start of this post though, this really doesn't seem to be the prevailing philosophy in audio. I've seen tons of stereophile reviews of cables with a long string of adjectives attached when under the UNIX philosophy we are looking for a single one. The use of an EQ seems completely foreign around here, instead trying to find perfect synergies between components that under the UNIX philosophy have an extremely well defined function.

So, well, there you have the UNIX philosophy of audio. What other philosophies do people ascribe to?
Reply
post #2 of 27
6/11/07 at 4:12pm
I subscribe to the 10,000 monkeys philosophy; try every combination available to me and stick with what I enjoy the most. When I want neutrality I reach for the earplugs.
Reply
post #3 of 27
6/11/07 at 5:04pm
Ditto. Your comments about cable and amp synergy are exactly what I think of when I hear talk about that stuff. Sure to some people some equipment 'sounds' better than others. But is it really accurate or does it simply just sound good to that particular person?

You gotta remember that with HIFI, things will start sounding the same. That's just how it works. You pay for stuff that will reproduce music faithfully and accurately. Just because something sounds different doesn't mean that it's better or necessarily accurate. Keep that in mind next time you shell out 10 grand for a cable.
Reply
post #4 of 27
6/11/07 at 5:15pm
Except that no one can agree on what is faithful and accurate. Furthermore, not everyone wants faithful and accurate. People want to like how it sounds, accurate or not. Some people say they seek faithfulness and accuracy when they have no idea what those things sound like. Limited measurements can also be deceiving since they do not reveal the whole picture. There is no clear standard for accuracy in many situations.
Reply
post #5 of 27
6/11/07 at 5:24pm
Thread Starter 
Quote:
Originally Posted by ooheadsoo View Post
Except that no one can agree on what is faithful and accurate.
I think this statement is a bit deceptive. Every component in that chain can be though of as a function on either a bitstream or a waveform. Compare what goes in and what comes out. If it doesn't match the function, then it is not accurate.

I also posted about this in another thread, but thanks to Nyquist's sampling theorem, for digital sources we have a hard error floor. It's not a very high one either, we have equipment that can measure waveforms to the necessary degree of accuracy.

There is another way too look at this statement though, which would make the philosophy a little clearer. If I had two cables, two power cords, etc., and they sounded different, by definition at least one of them is not accurate. Even if we assume there is no way to form a consensus about which is more accurate (a hypothesis I reject), you cannot deny that at least one is inaccurate.
Reply
post #6 of 27
6/11/07 at 6:02pm
One of my points was that most people don't care about what is accurate, especially in high end audio. People who buy cheap systems just want it to make sound. People who buy expensive systems usually want brand name recognition and good cosmetic appearance.

In the end, even the most ardent measurements freak will not listen to an accurate system if it sounds like crap. As a simplified example, you may say that speaker A measures very accurately flat in the amplitude domain, more accurately than speaker B. However there's still the phase angle/impedance, sensitivity, cabinet construction, off axis dispersion, power response, room response, linear and nonlinear distortion, and many more, some of which we may not have standardized measurements for. It depends on how deep you want to go. How heavily do you weight each measurement when you determine that speaker A is more accurate than speaker B? That's psychoacoustics, a fuzzier field than electrical engineering, I think you would agree. It's near impossible to objectively compare two different products to such a close degree. And in the end, someone may choose a speaker that would make me sick to my stomach to contemplate spending my hard earned money on. I wouldn't care if all the charts in the world claim that speaker A measures great if it sounds like crap to me. If this were not so, why are there still so many electrical engineers working on designing power amplifiers after 40 years of refinement, and I don't mean for specialized applications. I would think that the task would be academic by now.
Reply
post #7 of 27
6/11/07 at 6:37pm
Quote:
Originally Posted by Chu View Post
Given an unlimited amount of money, the ideal system would have completely "neutral" (I really hate this word, but from the post I think you understand the context) Headphones, DAC, CDP, Cables, and an Amp. The EQ would be the sole source of coloration.

In reality this is not possible, but we can come close. Headphones themselves for physical reasons are extremely difficult to achieve true neutrality, but we then have the EQ to compensate. For every other component though, there is no technical reason that they cannot be accurate to an absurdly high degree.
Doesn't work that way in the real world. Attempting to EQ a headphone usually results in more problems than it solves. Try to fix the frequency domain and chances are the time domain gets compromised, and phase anomolies are likely to be introduced. Ok well, let's fix the signal in the digital domain then, where we can manipulate the bits to our heart's content. Unfortunately, there's no free lunch here either, a very high-powered processor will be needed to juggle the bits and unless the algorithm design is really good, low level information will be lost and new anomolies may be introduced. Can it be done? Well, the jury is still out on this one, no commercially available unit is capable of transparently EQ'ing a bitstream, though some heavily modified units are starting to come close.

Amplification looks simple enough on paper, but again it's a non-trivial issue. Solidstate amps for various reasons tend to erase or flatten out low-level information and due to the nature of transistors & negative feedback they have an unnatural distortion spectra and residual noise-floor, both of which can shift around as the music signal changes. Tube amps have their own issues with frequency response, large amounts of distortion in poor designs (ie. most of them), and limited current or power output among other issues. Hybrids usually end up with the worst of both worlds. It is in my opinion, still impossible to design an "ideal amplifier", at this point we still have to make trade-offs based on what each of us perceives as the most important aspects be it frequency response, low-level linearity, impulse behaviour, or just plain musicalness, just as we do with headphones & speakers.

The real problem as I see it is this; we are trying to obtain "ideal sound" from an inherently flawed medium, that is, 16/44.1 digital, the compact disc. It takes far too much convoluted mathematical & processing gymnastics to pull decent sound off those stupid silver discs. Start with 24/96 or 24/192 digital, like a good DVD-A, and right off the bat we can throw out a lot of the convoluted interpolation guesswork and get a much more accurate signal. I'd also advocate the use of R2R DACs to eliminate all the questionable noise-shaping and filtering schemes necessitated by Delta-Sigma DACs, this would also incidentally eliminate a source of RF noise.

My philosophy is pretty simple, I want maximum linearity right down at the device level so I don't have to "fix" things up with negative feedback and/or overly complicated circuit designs. This I feel is the way to pass as much information through the system while screwing it up as little as possible. Make sure every last transistor, tube, capacitor, and so on is as linear and free from distortion as possible, then build the simplest circuit which will perform its function, no more, no less. Do it right and you'll have a system which is clean & neutral, and capable of exceedingly high levels of detail & resolution while preserving the proper tonal and spacial relationships between all the sounds.
Reply
post #8 of 27
6/11/07 at 6:56pm
Don't forget the recording itself. Oh what to do when you finally hear "the truth" and it sucks?
Reply
post #9 of 27
6/11/07 at 7:29pm
Switch to different headphones so as to adjust the sound such that it doesn't suck as much.
Reply
post #10 of 27
6/11/07 at 7:48pm
Oh but we aren't all designing power amplifiers.

Us electrical engineers have more important purposes in life. None of my colleagues has gone anywhere near a career involving audio amplifiers.

That being said, circuit theory is a science. It can be measured. The reason why amplifiers vary so much across the spectrum is because while you can estimate and model a component's behavior in a circuit, there's no way to 100% accurately predict what it will do. Sure you can get close, but models are models. They are close, but not by any means definitive. Even mathematical models of devices that cannot be solved by humans are not accurate enough to truly nail down the performance of a circuit in every circumstance.

Carrier concentrations vary with temperature, components have manufacturing tolerances, there's interference from external sources, less than ideal power supplies, amplifier input and output impedances vary, etc etc.

There are a lot of variables, and getting it right also includes a bit of luck. But while designing an amplifier may be an art, testing them is not. Like Chu said, it's either 1:1 with the original signal or it's not.

Unfortunately once the signal pops out of the speakers, you get into the realm of audio acoustics and things start to blur.
Reply
post #11 of 27
6/11/07 at 8:08pm
Quote:
Originally Posted by voxr3m View Post
Like Chu said, it's either 1:1 with the original signal or it's not.
Well, let's get it straight: It's never 1:1, is it?
Reply
post #12 of 27
6/11/07 at 8:15pm
Quote:
Originally Posted by ooheadsoo View Post
Well, let's get it straight: It's never 1:1, is it?
More or less. As long as the data stays in the digital domain it's possible to move it anywhere in the world onto any storage device and make infinite copies all with perfect 1:1 accuracy.

But as soon as we move it into the analog domain so we can hear it or see it, that perfection is forever gone.
Reply
post #13 of 27
6/11/07 at 8:30pm
Don't forget jitter in the digital domain. And that's just theoretical, we haven't gotten to the practical stuff, yet.
Reply
post #14 of 27
6/11/07 at 8:52pm
Thread Starter 
Quote:
Originally Posted by Roam View Post
no commercially available unit is capable of transparently EQ'ing a bitstream, though some heavily modified units are starting to come close.
This is an honest question, what does "transparently EQ'ing a bitstream" mean?
Reply
post #15 of 27
6/11/07 at 9:04pm
The computer analogies just don't work with audio. Computer processes start and end fully within the digital domain. Of course it's easy to glue components together that way - that's the advantage of digital. Audio ends up analog, not to mention that the starting point and intermediates are debatably not pure digital thanks to PCM & jitter.
Reply
New Posts  All Forums:Forum Nav:
  Return Home