playing wav files from a ramdisk gave best sound
then moved on to memory play, initially SQ was worse.
found that a function called memcpy was the culprit, most memory players use memcpy and this is one of the reasons why memory play sounds worse ie digital sounding. Fortunately there is an optimised version of memcpy, using this version removes the hard edge produced by memcpy. the other thing I did was to close the file after reading into the buffer.
also most players use malloc to get memory while new is the c++ method and sounds better.
I only use locally sourced, gold-plated memcpy, for warmth.
A revolution in audio rendering:
Tags: computers, mad science
39 Responses:
What kind of dumb-ass mailing list archive display website doesn't have "next in thread" buttons?
Ha ha, great parody. I mean, it is a parody right? I checked the messages up and down the thread just to make sure but they're all in on the joke. Everybody keeps a straight face. It's almost as if... no no it can't be. Can it? Oh god...
I suppose there are sane audiophiles on the internet, but they seem to be scarce. Caring deeply about music bitrates and being completely insane seem to go hand in hand.
Audio is a solved problem. A friend of mine really, really cares about sound and video, he studied acoustics (or something like that) as an undergraduate, and then when it comes to setting up his own system on the audio side he just bought whatever looked nice and fit into his setup, because it doesn't matter at all. As a civilisation when it came to reproducing music we went way past "that'll do now" and kept going, back in the 1990s, ie twenty years ago. Video still isn't easily perfected, he spends time examining different projectors and has opinions about HDR and 4K and so on, but audio is done.
So that's why the audiophiles you see are the crazy ones. PCM works, 16-bits is plenty, 48kHz is plenty, human hearing is so poor anyway that any halfway competent engineer can make speakers that sound fine. There isn't anything much for non-crazy audiophiles to talk about. Ten years ago there was maybe a reason to argue about lossy compression. Is 128kbps MP3 acceptable? What about this thing from Apple? What about this Vorbis idea? And a tiny bit of that continues to this day, but who cares now? If I buy an MP3 online it's not going to be compressed to 128kbps, and it's not going to take an appreciable amount of time to download either, because typical home bandwidth shot up meanwhile.
But for crazy people it never ends. Once you reject reason entirely you aren't constrained to mere Tortoise-like denial of the consequent, you can start making up whole chains of reasoning in which arbitrary nonsense follows naturally from other arbitrary nonsense. Of course drawing on the CD with a felt-tip pen will make it sound different, and therefore cables must be ritually washed before listening in order to preserve the high-end, and the listener needs to be laying in a bath of warm tapioca in order to hear the music as it was originally intended, when nothing makes sense, everything makes sense.
Actually audio isn't a solved problem: in particular transducers almost always have significant levels of nonlinearity (aka distortion, but nonlinearity is a less loaded term) which are such that differences between designs are really easy to hear. Transducers mean microphones, pickups & tonearms for record players (the turntable bit makes no real difference so long as it's adequate), loudspeakers and headphones.
For a non-vinyl-based playback system (so no microphones & record players), that means speakers and headphones, which both tend to have large nonlinearities which vary significantly depending on design. This is easy to test, and you do not need magic ears to do the tests, because the differences are often that frequency response, for instance, has enormous bumps and dips in it: 'enormous' meaning 10dB or something, not to mention just missing the bottom octave of useful frequencies for anything you can lift. The differences between designs are immediately apparent.
And these problems are really physically hard to fix: so hard that fixes probably are not ever going to happen. This is not to say that good speakers & headphones don't sound fine: they do, they just sound different than each other in really obvious ways.
Everything else, as you say, is a wash: compressed digital audio is more than good enough, even with streaming applications, DACs and amplifiers are all essentially perfect, and the magic cable stuff is just a joke. Everything before the bit that turns electricity into noise is a solved problem, but that bit still matters.
Just in case: I am not a golden-eared hifi person, if this reads like I am then I have failed to get my point across.
So what I get from that is that audio electronics is a solved problem. It's turning the electricity back into sound (or vice versa) that's hard.
TLDR: If you care about music, shop for speakers in person and test them with your ears and your preferred kind of music. Everything else involved in playing audio, it doesn't really matter all that much, any old thing will do fine.
Maybe also invest in an EQ to correct for standing waves and other imperfections? I bought a cheapo behringer multiband eq to cut down on the boomiest parts of the bass and it made a world of difference.
First sentence was going in the right direction, then you immediately said "record players" and identified yourself as a die hard hifi dork identical to the sort in the initial post.
I think, in fact, I identified myself as English and old enough that that was what we called them. If you think I am a diehard hifi dork you didn't read what I wrote very carefully.
Record player is a modern term. Back in the day we called them "gramophones": https://m.youtube.com/watch?v=dSINO6MKtco
And yes, I did have a collection of 78's. Beethoven's 9th symphony took up 9 discs, hence the term album. And if you tell that to the young people today, they won't believe you.
Also, vinyl is a Johnny-come-lately, and a sad disimprovement over shellac, although admittedly less brittle. Just try making a Wimshurst machine using these modern vinyl things!
Weren't they 'radiograms' (I suppose only if they had radios (which would have been called 'wireless sets'))?
I'm not au fait with radiograms, "radio" being a modern American affectation for "wireless". But when I was a lad we did have a combined record player (78 rpm only) and wireless in an oak cabinet in the dining room. Of course the wireless would only receive the BBC Home Service, Light Service, and Third Programme, although occasionally the signal from a local police radio would get picked up and obscure "Desert Island Discs" with Roy Plomley.
Lots of people have collections of old vinyl that they don’t particularly care to rebuy or digitize, or is obscure enough to not be available on CD or for download. Some people just like the physical experience of putting on a record, without insisting that it sounds “warmer” or better.
That's me: I have lots of records, and I buy records and not CDs if I want something physical. But because they are nicer objects, not because I think they sound better.
I don't think you and Nick are actually disagreeing about anything? Audio is "solved" in that the shape of the problem space is well known, and all of the low-hanging trade-offs and workarounds are well-documented and cheap to implement.
The digital-domain part really is "done": 16b/48khz playback covers the range of human hearing, and 24b/192khz mastering gives you enough headroom to do whatever production tricks you'd want. 256kbps MP3 (or AAC or Vorbis) with a modern encoder is close enough to indistinguishable from redbook cd audio as to make no nevermind.
The analog-domain choices are, as you say, all about picking whichever pretty reasonably priced equipment sounds best on average for your room/ears and taste in music. (Or for your studio, on the recording end.) And there hasn't, as far as I'm aware, been any actual substantial advance in the technology available there for decades, because there are intractable physical limitations in the way.
No, I don't think we are. I just wanted to say that you can make audio systems sound better, or at least significantly different, you just don't do it by buying special magic versions of the bits that are effectively perfect already: to me 'solved problem' made it sound like you can't do that.
There is no actual benefit to sampling at 192KHz in music production, as explained by Monty here:
https://people.xiph.org/~xiphmont/demo/neil-young.html
There is some advantage to 24 bit samples, because the difference in dynamic range matters sometimes and you don't want your quantization noise to add up if you're dealing with many tracks.
You are of course correct; typing "24/192" is a bit of a reflex action from too many online arguments over this nonsense. :-) (Most of which involve me eventually posting that very same article of Monty's!)
I get referred to that article all the damn time, but Monty's own AB testing software has a huge bug in it that he's ignored fixes for years, so I take his advice with a pinch of salt
That reasoning doesn't follow. He could write truly horrible software and never want to fix it but the article could remain correct. The two are not logically connected.
I'm reminded of Maciej Ceglowski's talk about the Concorde, space travel and web design tools (here) -- we tend to look at a small window of explosive progress in a domain and assume that the same rate will continue indefinitely. If you came of age in the 80s and 90s, you saw the transition from vinyl LPs, cassette tapes, tube amplifiers and analog equalizers to CDs, DSPs and MP3s. It was a revolution! And it all sounded MUCH BETTER and was way more convenient to carry around to boot. So much better that we all replaced our record and cassette collections, possibly multiple times. So surely in another 20 years, we'd have some heretofore unimaginable ULTRA AUDIO, right? It's just around the corner, maybe if we double the bit depth again...
Just keep in mind mp3@320 is really not enough.
http://www.npr.org/sections/therecord/2015/06/02/411473508/how-well-can-you-hear-audio-quality
Try it yourself. I can identify lossless out of most of these, and all I've got is some cheap Sennheiser HD600 headphones plugged to a Schiit Fulla 2.
I got 320kbps almost all the time. In the a capella part I got the 128kbps one which is the worst. Only got two correct answers.
Was that post from about 1993? It sure brings back the memories. I remember the digital music scene from around then, and things like a bum memcpy or dumb malloc could make all the difference even in audio processing. The SUN 4 SPARC station I was using could only copy 7MB/sec. That sounds like plenty until you realize that systems back then copied everything a bazillion times and lets not talk about same day process switching.
Even more amazing was what people would put up with. A friend of mine was in the Boston music appreciation group and they would burn CDs where every track started with a glitch of static. It turned out that these were some kind of file header. My friend wrote a little program to blast the first 256 bytes or so off of each music file and got rid of them. He was a local hero.
In 2013, I don't know. It sure feels like a time warp. Shuffling music bytes around is a solved problem these days. As for turning music bytes into music, I doubt that will ever be a solved problem as long listening to music involves vibrations traveling through air.
Vibrations traveling through air is so two-thousand-and-been-there-done-that. The air medium is fit only for cantankerous cacophonies of chaotic calliope clutter for the pass-time amusement of mouth-breathing cavemen wiping the drool out of their beards and the wax out of their ears. The truest bits bypass that bump in the wire and go straight to the bone, that is, until there's USB to spinal column (which we already know will have first-gen bit-rate issues, Apple will have non-compatible hardware, and all the other adapters will need several attempts to plug in until you get which side is supposed to be up correct). And fuck facebook with their skin listening.
Audio is a solved problem in most playback environments, but car audio is still notoriously full of "peak RMS" and tinny little overpriced drivers and standing waves and lousy insulation. The same tat infested the headphone market when they started putting Bluetooth in them. $2 mylar drivers with a chunk of steel next to them, then sold for $400.
I know people who have some really interesting things to say about audio but are not motivated to blog etc because of the saturation of nonsense out there, and anything they did write would be mercilessly attacked by the 'philes ("of course your double blind test was inconclusive, you need to spend at least another $1000 on headphones to appreciate the tonal blah blah...")
Sure, but does his code align the magnetic polarity of the individual bits???
On a slightly more serious note, this is the only link on this blog where I've actually wondered "WTF was JWZ looking for when he encountered this?".
Previously
Only? Only.
I Have Extensive Files.
You can say what you want about audiophiles, but they're the only ones who know how to get the cleanest and purest bits from a malloc. We're talking about high end audio here, don't forget. And no oscilloscope can detect the edges in a memcpy'd buffer as well as a pair of human ears
Does Monster Cable do SATA? Because if you're losing fidelity reading those bits into the ramdisk in the first place...
If they had at least mentioned ECC RAM they'd have a better chance of roping in a few suckers with a semi-plausible theory...
This reminds me of cryptographic programming. For example power analysis makes it necessary to use very specific algorithms for memory movement and arithmetic.
There is actually a common problem: There are existing side-channel crypto attacks that, much like timing attacks, measure the "effort" of the cpu, by means of CPU voltage regulator: cpu current draw determines the mobo supply's chop waveform, and this spills through the connections onto adjacent electronics. Which in lab conditions can even be recorded by microphone. So, if the mobo is particularly crappy and this guy's audio card etc. is improperly shielded then yes, the software does affect the sound playback!
In short words:
https://en.wikipedia.org/wiki/Acoustic_cryptanalysis
https://www.tau.ac.il/~tromer/acoustic/
This guy is a bozo. Anyone with half a brain and at least two ears know that the quality of the bits you load into your system are way more important than whatever memcpy in use. Buy good quality zeroes. You get what you pay for. Ones are not as critical. I buy the basic GD house brand ones from Frys and notice no difference.
Replace your zeroes at least once a month. More frequently if you listen to raw vintage jazz or blues. I've found it is possible to recycle the ones for up to a year without any dropouts.
It took me a while to realise that pathological audiophilia is the same fundamental disease as "environmental sensitivity". It's the notion that discomfort is a symptom of physical superiority, and the bizarre models for this are a sign of the desperate need not to lose that sense of superiority.
I once found a site full of tourmaline crystals and the usual audiophile nonsense (did you know that crystals "de-sample" digital audio streams? Never mind that square waves are a myth anyway...) that actually sold a regimen of nutritional supplements to allow the skeptic to transform into a delicate sensitive Audiophile.
If there was ever proof that this is about a belief in physical superiority, that was it. Normal hearing diagnosed as nutritional deficiency! Alas, they re-did the whole site and forbade the internet archive crawl them from the start. Muppets.
Audiophilia is the only mental disorder I can think of offhand where it doesn't bother me to see hucksters ripping people off.