heywood on 8/9/2015 at 17:22

---

I did just that, built a fanless Linux music server that I controlled from a laptop. This was 7-8 years ago. Then in 2010 I decided to shed a lot of audio stuff in preparation for moving to Aus, and shrink down to just one source, an integrated amp, and a pair of bookshelf speakers. All my data, files/records, and media (music, photos, and movies) were put on a NAS so I wouldn't have to ship them. So the source had to draw from the NAS. The source I ended up getting was a Linn Majik DS, and it easily bettered my home-brew music server + Weiss DAC2. DLNA control points were still a bit clunky back then, but with the way things are now I would never go back to that kind of solution. Unless your hobby is tinkering with computers.

If I didn't mind being locked into one vendor, I might have tried Sonos + external DAC. Or LMS + Squeezebox + external DAC. But I like the interoperability that DLNA provides. We've got control point apps on all our phones and tablets and a household full of DLNA capable devices (Rokus, music streamers, wireless speaker, BD player, etc.).

Muzman on 9/9/2015 at 07:51

---

Not to be Mr argumentative-pants or anything, but the discussion is often fun. I throw these articles into the audiophilia discussion, which may be familiar to people already.

(https://people.xiph.org/~xiphmont/demo/neil-young.html) 24/192 Music Downloads...and why they make no sense

(http://ethanwiner.com/believe.html) Why We Believe: A common-sense explanation of audiophile beliefs.

The first one is kind of talking about the trend of what are essentially really high res MP3s as much as anything else, as though the number of bits in the file can create sound that wasn't there to begin with.. That's not the same thing as genuinely remastering for high res audio. But even that runs into the essential physics of formerly complex audio in a single electrical circuit and then some vibrational medium.

The second one is by a guy for whom acoustic spaces are his thing, so he knows his stuff, but equally- of course there's a spacial acoustic explanation. It does give a picture of why setting up even the most basic fair test of equipment is very hard. Even with a neatly set up switched test system, a slight volume change or a turn of the head could make some difference appear that isn't really there.

heywood on 10/9/2015 at 18:24

---

I am familiar with Monty's Ogg Vorbis work and his recent anti-Pono writings. I sort of agree with him and sort of don't. He made a video called "Digital Show and Tell" which is pretty good. It's basic undergrad EE stuff, but it's a good primer on digital audio. But some of his arguments against hi-res are pretty hand-wavy and not technically sound.

For instance, he says ultrasonic content in 24/192 recordings can produce harmful intermodulation distortion. This is FUD which is not substantiated by any measurements and contradicts some basic knowledge about audio systems. Pretty much all decent solid state amplifiers provide low distortion over their bandwidth, which is usually 50 KHz or more, well beyond the bandwidth of the loudspeaker. Tweeter response usually falls fast above 20 KHz, or above 30 KHz if it's a really wide bandwidth design. Frequencies above that are blocked by the rising impedance of the tweeter, so those ultrasonic frequencies he's worried about don't actually go anywhere. The toughest test of high frequency intermodulation is at frequencies within the system bandwidth, e.g. 19+20 KHz, which is a common and useful test. FWIW, I did try playing his ultrasonic intermod test files in my audio system, and naturally I got silence.

Another problem I have with Monty's arguments is that he ignores any affects of anti-aliasing and reconstruction filters. He basically declares modern filters to be perfect by fiat, and ignores the fact that the audio industry is still struggling with them.

Oversampling DACs became the norm because it was possible using long FIR filters to cram a super-steep brick wall filter in between 20 KHz and 22 KHz - thus achieving a flat response up to 20 KHz. That's looks fine if you're only looking in the frequency domain. But in the time domain, the impulse response of these filters has a lot of ringing. Some manufacturers opt to reduce the ringing by using a slower filter roll-off, which results in a small amount of attenuation above 10 KHz or so. Others prefer apodizing filters, or even minimum phase filters, to have only post-impulse ringing and no pre-impulse ringing. All kinds of things have been tried, even splines. Some are patented or branded (Pioneer Legato, Sony Super-Bit Mapping, Anagram, etc.). And it's fairly common for manufacturers to offer different filter options because they all sound slightly different. Some people prefer just a basic analog filter, or even no filter at all. There is no consensus in the industry or among listeners about which is best. And as long as the CD format is around I don't think there will ever be.

So I think the advantage of using a higher sample rate probably has nothing to do with being able to hear sounds above 20 KHz. The real advantage is that the transition band of these filters and all of their artifacts are entirely out of the audible frequency range.

Regarding dynamic range, I agree with him that 16 bits ought to be enough for replay, and I don't really understand why 24-bit playback would sound better. It is necessary to work with a 24-bit or larger word length during production though.

I think I also disagree with him philosophically. The only justification for throwing away data via downsampling or truncation, or for other format conversions, is saving space (e.g. trying to fit an hour of music on a CD, or hundreds of songs on an iPod shuffle). When space is not an issue, the logical approach is to keep things in the same format from recording through playback. When saving space is a priority, only then should we be relying on current models of human hearing to decide what information is important to keep and what should be thrown away. Monty is not just claiming that the CD format is good enough, he's actually claiming it's better than leaving the data in the form it was recorded and produced in. To me, that's just illogical. If extended frequency response is harmful to playback on an audiophile's system, why is it not harmful to playback on the engineer's systems?



I hadn't previously seen the Ethan Winer link you posted, but I am familiar with Ethan. I own a few of his Real Traps products; they were the first acoustic treatment products I ever purchased. They are good, but kind of pricey for basic cookbook designs out of Everest's Master Handbook. Since then I've mostly been building my own. Also, every time I Google something audio related and end up in a new audio forum, it seems like Ethan has a history there. He is a very prolific internet poster, and extremely sure of himself, although lacking an engineering background.

In the article you posted, Ethan has an interesting hypothesis, but the conventional wisdom is that all those comb filter cancellation nulls that look so ugly on an unfiltered in-room frequency response measurement are only audible at lower frequencies. 50 years ago, Manfred Schroeder published a theory that divided room response characteristics into a low frequency range and high frequency range, with the transition point called the Schroeder frequency which is determined by the room volume and its reverberation time. A typical Schroeder frequency for a domestic listening room is 200-250 Hz. For a recording studio control room that is acoustically dead, more like 150 Hz. Below the Schroeder frequency, the individual peaks and nulls (modes) of the response are deterministic, based on the shape and dimensions of the room and the locations of the loudspeakers and listener. They are also distinct, i.e. they have relatively little overlap in frequency, so you can hear the effect of an individual mode. Above the Schroeder frequency, the modes overlap to a degree such that the response is best modeled by statistics.

It is generally accepted in audio engineering that comb filtering is not audible above the Schroeder frequency, and instead the sound level we hear is best described by the values of the statistical distribution. The reason for that is that the frequency response of our ears is logarithmic but the comb filter pattern is not. We don't hear nulls that are very narrow in frequency, e.g. ones that don't even span one musical note or 1/12 octave. At lower frequencies, the comb filter nulls are wide enough on a log scale to be audible. At higher frequencies, they are not. That is where Ethan's hypothesis runs into trouble. A generally accepted view in audio engineering is that for mid and high frequencies (e.g. above the Schroeder frequency), the frequency response we hear is best described by a 1/3 octave filtered response. When you apply 1/3 octave filtering, all those comb filter nulls disappear.

On the other hand, when I look at his Figure #3, I see a difference of 6 dB over a full octave (roughly 3-6 KHz) between the "Listening Pos" and "Listening Pos + 1 inch" measurements. That difference would easily be audible: the former would sound too bright and the latter more natural. These are 1/3 octave responses, so it's not comb filtering at work. I have measured systems in 6 different rooms and don't recall ever seeing variation like that in the treble response with such a small change in listening position (I could upload an example if anyone is interested). So I wonder if it's unique to his system setup. If his result is typical, he might be on to something even if his comb filtering explanation doesn't work.

But I wouldn't count out the old placebo effect. The people who are on the tweak fringe of audio never fail to hear a difference with every tweak, but on top of that it's always an improvement. I'm not sure how Ethan's hypothesis would explain how their system keeps improving every time they sit down to listen after a tweak.

bjack on 11/9/2015 at 00:46

---

An old trick we used to do in the old days was to record our new vinyl records to tape right away, then play just the tapes. It saved the record from getting worn out. I have over a hundred records that have been played less than 10x in 30 to 40 years. Yes, the tapes did not sound as good as the records, but we could use them in our cars.

For the equipment I have now, MP3s are generally good enough. I can tell the difference, but it is not enough for me to care that much. Usually, if a WAV(CD) or MP3 sounds like crap, it is because of bad EQ and mix, not because the format is bad. When CDs first came out, they did sound a bit sterile as compared to vinyl. That difference is not so commonplace today. I do have to say that I prefer the sound of WAV over MP3 on some music. Still, it is not enough to really matter much to me. It is subtle, not glaring. Then again, it depends on the piece. I prefer older pre-digital music (pre mid-80s) to be on vinyl. Once into the 90s? Meh... mp3, CD, whatever.

ZylonBane on 20/9/2015 at 11:48

---

Jesus, this thread. Audiophiles really like the sound of their own voices, don't they.