maanantaina, maaliskuuta 12, 2018

Faktoja ja myyttejä

Toisaalla internetsissä, Arny Grueger, erittäin kovan luokan audioinsinööri ja hifiyhteisöjen vihaama objektivisti kirjoitti äskettäin aivan erinomaisen osuvan ja selkeän postauksen audiolaitteista ja niiden koetuista eroista ja sudenkuopista. Vaikka ukko on vähintäänkin kärkäs kieleltään ja snadisti kärttyisyyteen taipuvainen vanha jäärä, jonka kanssa itselläkin on ollut välillä napit vastakkain, ei miehen ymmärrystä ja tietämystä alan tiimoilta voi sivuuttaa. Ja vaikka tämä on suunnattu enemmänkin hifiyhteisöille, samat lainalaisuudet pätevät myös studiopuolelle: Luultavasti se behringerin satasen DAC on suunniteltu ja toteutettu huomattavasti paremmin kuin ylihypetetty, kymmenen kertaa kalliimpi hifilaite. Mikä on todella helppo tarkistaa spekseistä. Nykyään edullisetkin kamat ovat aivan todella hyviä, verrattuna esim todella kalliisiin pro-vehkeisiin, vielä 15 vuotta sitten. Eli musanteko ei pitäisi olla enää laitteista kiinni. Luvallansa kopion tämän erinomaisen selkeän ja osuvan tekstinsä tänne:
amphion2


Some comments on measurements and such:
All music is composed of a combination of pure sine waves at various frequencies and amplitudes. Therefore, all possible errors in reproducing and/or storing a signal fall into 1 of 4 categories because a signal is a single function (amplitude) of a single variable (time).
(1) Linear distortion commonly characterized using the terms or tests: bandwidth, frequency response, phase response. This changes the size of one or more of a signal's frequency components. Linear Distortion *does not* add new frequency components that were in the original signal.
(2) Nonlinear distortion commonly characterized using the terms or tests: THD, IM, Jitter, FM distortion. This changes the shape of one or more of a signal's frequency components. Nonlinear Distortion *does add* new frequency components *that are related* to those originally in the signal.
(3) Random noise commonly characterized using the terms or tests: SNR, Dynamic Range. This adds new random frequency components that are *not necessarily related* to those in the original signal.
(4) Interfering but non-random undesired signals commonly characterized using the terms or tests: Hum, EMI, RFI. This adds new non-random frequency components that were not necessarily in the original signal and *that are generally unrelated* to the ones that were there.
*Everything* that goes wrong with an audio signal fits into one of these categories, there can be none else, and they are all independent of each other. For example, frequency response variations can increase or decrease with no change to hum and noise. However, the same problem such as a bad part or a bad design can cause more than one of them at the same time.
Common related audiophile myths and answers:
• "Imperfect DA conversion and inadequate filtering could lead to an incomplete signal with little distortion and inaudible noise, but still incomplete."
Missing data in an audio signal whether analog or digital is generally very audible as an Interfering Signal. It most commonly sounds like clicks or pops. It can't be very subtle because when data goes missing suddenly and briefly, the signal goes from being there, to not being there, and the transition is like part of a square wave. The louder the signal, the larger the click or pop, and if the signal is a low-frequency sine wave, then the clicks and pops stand out particularly strongly.
• "If it is only loudness differences, it would be really easy for manufacturers to start a new kind of loudness-war to support sales."
The availability of more loudness is well known as such things as amplifier power and loudspeaker power handling capacity or efficiency. The influence of these things on sales is such that they are common specifications that are advertised for gear.
The audibility of differences in loudnesses is so commonly appreciated that just about every audio system has a volume control for changing it.
Unless you take special steps, any two pieces of audio gear you compare will gravitate naturally towards playing at a different loudness. If you try to match these differences using your ear, you will find it to be time-consuming and inaccurate.
"If you match levels by ear, you will hear the remaining level differences by ear" If you are serious about matching levels, you can do it with precision equal to hundreds or thousands of times less than the smallest change you can hear with just a PC with a cheap audio interface and freebie analytical software, or a cheap DVM and the ability to play simple digital files or CDs or DVDs you can burn for yourself or purchase inexpensively.
When level differences are small, they are not perceived as differences in level, but as subtle, almost indescribable audible differences. If you compare something to itself but with slight level differences, you will generally be able to distinguish the sound of the device from itself, just playing at a slightly different level and how silly is it to prefer something because its volume control is set a little bit differently?
• "It is entirely possible for equipment to measure exactly the same and not sound exactly the same."
Perhaps in 1950 or even sometimes in 1960. Audio measurements are not a fixed thing, but something that is constantly changing and improving, just like the rest of science.
The biggest innovation in audio measurements was the ready availability of computer-based measurements based on a mathematical analysis procedure called the FFT and it being married to computer graphics. This particular revolution was pretty well established in the late 1960s. Today you can take a $200 computer or cell phone and combined with free software, duplicate or exceed tests and measurements that required from one quarter to one-half of a million dollars in 1975.
• "All that is saying is that for that particular measurement, there is no perceptible difference. Real soundwaves in complex dynamic passages are much more complex than recorded measurements. Our hearing is even more complex than that."
All music is composed of a combination of pure sine waves at various frequencies and amplitudes. All of the 4 errors I listed above change the frequencies and/or amplitudes of those sine waves that make up all music. FFT analysis can recover the frequencies and amplitudes of even the most infinitesimal changes in frequencies and amplitudes. Today, we can readily discover and quantify those errors even when they are thousands of times less than what the ear/brain can reliably perceive. This is easy to demonstrate with simple, readily available tools if you know what you are doing.
• "You just can't reduce differences between components to comparing distortion numbers... There's a LOT more to it than that."
Finally, a non-myth. Please see my former comments about there being 4 possible kinds of errors in reproducing or storing audio signals.
Do any of these statements contain an amount of truth?
Generally No, as already stated above.
Could it be that it's not just loudness and bias that makes one audio component sound better than another one?
It could be.
However, it almost always is simple, trivial things like mismatching loudness, time delays between auditions, and using different music or different segments of the same piece of music in the comparison, or preconceived notions and biases that is why people perceive audible differences between reasonably good audio gear.
The proof of this assertion is that almost every time we make those issues go away with good experimental procedures, the reliable perceptions of audible differences also go away.
Please notice that without the existence of reliable audible differences, preferences are meaningless.
Do DAC designers tamper with EQ to make one pre pro's sound 'cleaner' than another one's (e.g. some companies include different filters with their DACs).
Yes. IME, one rather bad villain with this habit of making outlandish claims based on building gear with parts missing so as to create audible frequency response differences calls himself "Limpiazator" or some such.
Or is it all again part of the subjectivists' mythos and
Measuring frequency response is relatively easy and in general, I do it as part of setting up a listening experiment, as does any other worker with any pretensions of doing scientifically reasonable work.
Almost all modern audio gear can be reasonably easily adjusted for flat enough frequency response and low enough noise and distortion to be indistinguishable from other competitive gear, and also a simple short piece of audio cable. Proper procedures are almost always not ever used. Therefore the subjective opinions that you read most of the time on most audio forums are not worth the dynamite to blow your nose except as cultural artifacts.
Should I stop listening?
No, you should start engaging your brain when you listen and follow the example of say, Ethan Winer, just to mention a name that you might be familiar with. This is the opposite of following the example of say Michael Fremer, again to just mention a name you might be familiar with.

Ei kommentteja:

Lähetä kommentti