Audio Frequencies

[SileEeles]

Something came up the other day that has been bugging me ever since. Doesn't refer to Minidiscs as such, more to audio hardware in general, but I figure you people are a knowledgeable bunch, you'll fill me in.

I was always under the impression that 96Khz is the is the highest that most audio hardware is capable of. Personally, I have never used it whatsoever, even for my own audio recordings. The default for me has always been 48Khz, no idea why, just has.

If what I've read is correct, humans can only hear as high as 20Khz (Obviously this varies between people and as you age this apparently decreases). 44.1/24 is what Minidiscs support, CD's are 44.1/16 if I'm not mistaken. I couldn't say what it is for Vinyl, I don't know, only that having more audio on one side of the record decreases the dynamic range that is possible.

I found out that 192Khz is now possible on some audio hardware. My friends motherboard supports it, advertising itself as "Absolute Pitch 192kHz/ 24-bit True BD Lossless Sound".

If BD is blu-ray, am I not correct in my assumption that blu-ray uses 48Khz for the most part? Even if it can go higher, 192Khz must take up quite alot of space for not alot of noticable improvment over what already exists, certainly at least that we can hear, audibly?

I gather at this point it's a thing to draw people in, so they'll buy it and "believe" that there's a difference when there isn't, but I'm stumped as to why it exists. Theres only so much our ears can pick out.

[trott3r]

I thought the khz was sanple rate? rather than frequency limits.

Most vinyl rips on the net are 96khz and 24bit.

The 20khz upper limit was apparently what thet thought at the end of the 70s when cd was being developed by sony/phillips.

It seems since then (80s) we do notice if >20khz is present or not.

It is true though that we do lose the ablity to distinguuisgh between 19khz and 20khz as we get older.

That how i understand it but hopefully better qualified ppl will chip in.

[PhilippeC]

This 96kHz means that the audio signal is cuted 96000 times per second. This is "oversampling" vs 44,1kHz.. Nothing to do with the usual "20Hz-20kHz" you can also find in any specifications and which is supposed to cover human earing capacity (with loss of high frequencies with age),

Stephen (sfbp), Azureal or MDietrich are more competent than me about this subject.

Before comparing any CD/DVD/Blu Ray/MD/etc... units together, just think that you just can correct your audio tracks before the MD recording, in particular the "brillance" : the goal is to enhance the loss of high frequencies with MP3 compression or old magnetic bands. The audio DSP plugin DFX audio enhancer (5 corrections) do that but it could be too agressive and even not acceptable with very good audio sound cards (or more generally with lossless tracks). With foobar2000, try also the "noise sharpening" plugin.A DAW (Digital Audio Workstation) make a more accurate work.

[SileEeles]

I use a DAW during my daily-ish routine because I record alot of my own music. I'm not concered overly with the result that ends up on a Minidisc, usually what my computer optica out gives off is decent to me. I just wonder what the point 192Khz at the moment, if it makes a significant improvement over what we already have, or if its just one of those things to suck people in when they buy stuff.

[Soundbox]

One thing 192 KHz allows is for greater sound modifications such as slowing down to extremes. I recorded a 192/24 file on my Sony flash recorder (birdsong) and I got some interesting effects when I slowed it right down.

As for music, 92 is plenty. Then there is DSD recording - that is 1 bit but at over 2 mHz sample rate.

[SonicFields]

You may be interested in one of my own 'Sonic Fields Site' blog posts back in November 2011....sorry about the text version attached here as the MHTML file is too large to upload.

Beginner's Guide Part III.txt

[MDietrich]

I was always under the impression that 96Khz is the is the highest that most audio hardware is capable of. Personally, I have never used it whatsoever, even for my own audio recordings. The default for me has always been 48Khz, no idea why, just has.

192 kHz sampling frequency has been the maximum 'sane' (if it´s sane or not depends on how you generally look at digital audio) value since 2005/2006. Nowadays some ADCs are capable of 384 kHz.

If what I've read is correct, humans can only hear as high as 20Khz (Obviously this varies between people and as you age this apparently decreases). 44.1/24 is what Minidiscs support, CD's are 44.1/16 if I'm not mistaken. I couldn't say what it is for Vinyl, I don't know, only that having more audio on one side of the record decreases the dynamic range that is possible.

Yes, our ears can only hear as high as 20 kHz - and only if we´re very young and weren´t exposed something like constantly loud noises (like visiting loud concerts, clubs, airports etc.). With age, our ability decreases rapidly until it reaches a horrible upper limit of roughly 7 kHz when we´re 70 years old. Now this is all based on averaged statistical data; take me: I can still hear up to 16,5 kHz even though I´m 38. But some people my age can only hear up to 12 kHz so the statistic is still valid.

Secondly, we don´t listen with our ear. Our brain is in fact the major part of our hearing system since it deletes what we don´t need for survival. In general it can be said that 80% of the things our ear hears won´t be used by the brain for further processing, it´ll be discarded. That´s BTW the only reason why lossy formats like MiniDisc, mp3 or AAC work, they erase what our brain would erase anyway.

Regarding your vinyl question: when translated to the world of digital, a common vinyl record would have a bit depth of 12 bit and a samplerate of 32 kHz. Should the record be brandnew, should the same apply to the pickup, the needle, should the azimuth be aligned perfectly and in case the necessary RIAA amp doesn´t produce the usually typically frequency errors you´d have a maximum of 14 bit and roughly 64 kHz. But play this vinyl once and you´re rapidly approaching radio quality. Vinyl is a playback system where every playback further destroys the medium it plays.

I found out that 192Khz is now possible on some audio hardware. My friends motherboard supports it, advertising itself as "Absolute Pitch 192kHz/ 24-bit True BD Lossless Sound".

If BD is blu-ray, am I not correct in my assumption that blu-ray uses 48Khz for the most part? Even if it can go higher, 192Khz must take up quite alot of space for not alot of noticable improvment over what already exists, certainly at least that we can hear, audibly?

192 kHz is not only possible on some but on most hardware. Even for portable devices. Look at this nifty little player I bought in October last year: http://www.amazon.co.uk/FiiO-X3-Portable-MP3-Player/dp/B00DQBWY04/ref=sr_1_1?ie=UTF8&qid=1392244662&sr=8-1&keywords=fiio+x3

Fill it up with 192 kHz FLAC and it´ll play them in full quality (I´ve measured it). Today space requirements aren´t a problem anymore when a 2 TB harddrive costs a mere 100 Euros and when you can have memory cards as small as your fingernail with 64 GB of space.

Audibility? Actual differences between 44.1 and 96 or 192 kHz are tiny. Really. I´ve tested it so many times over the past ten years... the actual differences are so small that 98% of all people won´t ever hear them. A shitty recording will still sound like shit in 192 kHz, it´s not the magical medicine that makes everything better. I believed so myself many years ago - but hitting the bottom was a mindblowing experience. Whenever I hear someone saying that this or that recording does sound sooooo much better in 192 kHz I think to myself that this person wants to believe that (excluding mastering differences). This 'want' subsequently creates his/her experience (this is called 'placebo effect').

EDIT: To clarify, I´m a firm believe in 96 kHz - for reasons explained below. I just wanted to put things in perspective, HiRes is just not as wonderful as people believe it to be.

[MDietrich]

This 96kHz means that the audio signal is cuted 96000 times per second. This is "oversampling" vs 44,1kHz.. Nothing to do with the usual "20Hz-20kHz" you can also find in any specifications and which is supposed to cover human earing capacity (with loss of high frequencies with age),

Stephen (sfbp), Azureal or MDietrich are more competent than me about this subject.

Before comparing any CD/DVD/Blu Ray/MD/etc... units together, just think that you just can correct your audio tracks before the MD recording, in particular the "brillance" : the goal is to enhance the loss of high frequencies with MP3 compression or old magnetic bands. The audio DSP plugin DFX audio enhancer (5 corrections) do that but it could be too agressive and even not acceptable with very good audio sound cards (or more generally with lossless tracks). With foobar2000, try also the "noise sharpening" plugin.A DAW (Digital Audio Workstation) make a more accurate work.

Ooooh... I´m sorry, but 96 kHz has nothing to do with oversampling. Oversampling is something completely different perfectly explained by Wikipedia: "In signal processing, oversampling is the process of sampling a signal with a sampling frequency significantly higher than the Nyquist frequency. Theoretically a bandwidth-limited signal can be perfectly reconstructed if sampled at or above the Nyquist frequency. Oversampling improves resolution, reduces noise and helps avoid aliasing and phase distortion by relaxing anti-aliasing filter performance requirements."

The DSPs you mentioned are trying to fix errors where none exist, I think I´ve even said that to you some months before. Mp3 sounds on occasion slightly muffled - but not because frequencies are removed, it´s because the timing resolution isn´t small enough (AAC fixes that).

[MDietrich]

I use a DAW during my daily-ish routine because I record alot of my own music. I'm not concered overly with the result that ends up on a Minidisc, usually what my computer optica out gives off is decent to me. I just wonder what the point 192Khz at the moment, if it makes a significant improvement over what we already have, or if its just one of those things to suck people in when they buy stuff.

If you routinely employ different DSPs (like EQ, stereo field tools, exciters, reverb, etc.) it makes sense to do all of them on 192 kHz files with a bit depth of 32 bit floating point or 32 bit integer. Many DSPs can´t handle 44.1 very well, they will introduce errors simply because they have been programmed shoddily. Reason: with 44.1 kHz the aliasing cutoff frequency is very close to the audible frequencies (the available window for aliasing rejection is just 2.050 Hz small). Bad DSPs miss an aliasing filter and will therefore introduce severe aliasing errors into the audible frequency band.

Good DSPs on the other hand oversample so that the problem will become moot. Yet, it´s even better to resample yourself because you effectively reduce the possibility of processing errors to 0. Whenever I do some processing on audio files I always do it at 192 kHz. It´ll take longer of course... but only because the HDD has problems keeping up in case it´s fragmented.

Now to HiRes in general. Have a look again at my 'available window for aliasing rejection' statement. Before the CD was introduced to the market in 1982, the engineers were arguing about the best samplerate. They wanted a samplerate of roughly 60 kHz so that they could construct a reliable and not harmful aliasing rejection filter that was 10.000 Hz wide. This 10 kHz wide window would have allowed for a harmless aliasing filter which in turn would have guaranteed perfect passband (20-20.000 Hz) quality. Well, it turned out differently, with CD (or MD) the window is only 2.050 Hz small. So they devised oversampling to get rid of aliasing in a more elegant way. It works sufficiently but the aliasing rejection mechanism embedded into the oversampling filter is still close to the passband which led to the adoption of higher samplerates. With these higher samplerate you can construct aliasing rejection windows even wider (in case of 192 kHz theoretically 76.000 Hz wide). It relaxes the frequency cut off and removes even the slightest possibility of reconstruction errors.

I´m a staunch believer in 96 kHz for exactly these reasons. 60 kHz would have been sufficient IMO but it doesn´t exist as a format. The next best thing is 88.2 kHz but that´s still not very compatible to the bunch of available hardware (so many DACs - even new ones - are incapable of playing them back). So it´s 96 kHz because of fantastic compatibility. I upsample every bit of music to 96 kHz and I do this with a special upsampling algorithm that avoids the small 2.050 Hz window I talked about above yet comes at the cost of introducing severe imaging errors ('aliasing' above the passband). A note: please consider that I´m mad and an audiophile; and audiophiles can´t be trusted.

You see, your generic audiophile is a person with too much money and not enough brains. He will buy everything as long as its marketed with technical pseudo-babble, looks expensive and was built from rare and expensive materials. Just mention that you have invented a magical de-jitter machine and he will eat right out of your hand, even if you hold a vial of poison. Cynical, I know... but I´m talking from experience as I´ve seen myself for decades in the mirror. My blog changed quite a bit of what I deem to be true... still, a true skeptic would have cried out loud if he read this post.

[SileEeles]

It's quite alright, mad people are the best

Thank you for your very informative replies, it certainly clears up much of what was on my mind, and what I was thinking in the first place ... It's nothing overly brilliant.

For me, I have no idea why I stick to 48Khz. Or why I don't go higher when I can. Ultimatley the difference between them isn't noticable to me, and 96Khz wasnt compatible with Minidisc, so I ignored it from then on really.

[MDietrich]

It's quite alright, mad people are the best

Thank you for your very informative replies, it certainly clears up much of what was on my mind, and what I was thinking in the first place ... It's nothing overly brilliant.

For me, I have no idea why I stick to 48Khz. Or why I don't go higher when I can. Ultimatley the difference between them isn't noticable to me, and 96Khz wasnt compatible with Minidisc, so I ignored it from then on really.

You probably use 48 kHz because it´s even more compatible than 96 kHz. For years, hardware for PCs wasn´t able to sample at 44.1 - the quasi-standard was 48 kHz. Should your hardware be of good quality I don´t think that you´d need to worry about it anymore as long as you´re happy with it.

[MDietrich]

You may be interested in one of my own 'Sonic Fields Site' blog posts back in November 2011....sorry about the text version attached here as the MHTML file is too large to upload.

BTW, your article is very well written, love it

[SileEeles]

You probably use 48 kHz because it´s even more compatible than 96 kHz. For years, hardware for PCs wasn´t able to sample at 44.1 - the quasi-standard was 48 kHz. Should your hardware be of good quality I don´t think that you´d need to worry about it anymore as long as you´re happy with it.

Seems (after looking) that it is capable of 192Khz. I have a fairly mid-range system (I built it myself), but I've yet to see it struggle with anything. Games, audio, HD footage, you name it ... So many people bash the CPU I have and yet it is so capable its ridiculous.

[MDietrich]

Seems (after looking) that it is capable of 192Khz. I have a fairly mid-range system (I built it myself), but I've yet to see it struggle with anything. Games, audio, HD footage, you name it ... So many people bash the CPU I have and yet it is so capable its ridiculous.

Off topic:

Out of interest: what CPU do you have? I ask because a year ago I replaced our old Core2Quad (oc'd to 3 Ghz) with an AMD APU (Trinity)... judging from all the rewievs I´ve read one could gain the impression that it's slow as a turtle and draws as much power as a 500 watt bulb.

As it turned out, the CPU part alone is 80% faster than the old CPU, draws at maximum clockspeed 30 watts less (including fully utilizing the GPU parts, everything at stock settings), and was 50% cheaper.

Back to topic:

Audio on PC is computationally so easy, even with high sample rates and bit depths. This hasn´t been a problem for PCs since 2006 / 2007.

[SileEeles]

Off topic:

Out of interest: what CPU do you have? I ask because a year ago I replaced our old Core2Quad (oc'd to 3 Ghz) with an AMD APU (Trinity)... judging from all the rewievs I´ve read one could gain the impression that it's slow as a turtle and draws as much power as a 500 watt bulb.

As it turned out, the CPU part alone is 80% faster than the old CPU, draws at maximum clockspeed 30 watts less (including fully utilizing the GPU parts, everything at stock settings), and was 50% cheaper.

Back to topic:

Audio on PC is computationally so easy, even with high sample rates and bit depths. This hasn´t been a problem for PCs since 2006 / 2007.

AMD FX4100 - Quad Core (although there is some debate about that). Has a default clock speed of 3.6Ghz but since I have an aftermarket cooler, I can manage 4.2Ghz without a problem and stay between 30 and 35 degrees C at minimal use - pushes about 50 degrees at the highest for certain games. That is paired up with a (somewhat dated some would say) HD5750 graphics card, which still manages to play many of the latest games at very decent settings, I feel no need to upgrade it yet.

And indeed. I remember being back on a laptop from such a time, which only had a single core Intel M celeron or something similar, at 1.8Ghz? When I was using Reason (Not a DAW as such ... more MIDI based), it would constantly tell me that I don't have enough CPU power, and even on a Pentium 4 machine at 2.8Ghz, when I pushed that it would give the same warning. Never had it once since moving to multiple cores.

[PhilippeC]

Ooooh... I´m sorry, but 96 kHz has nothing to do with oversampling. Oversampling is something completely different perfectly explained by Wikipedia:

Nevermind. This is not an understanding error, this is my poor english. First I wrote upscaling, then upsampling, then oversampling...

[MDietrich]

AMD FX4100 - Quad Core (although there is some debate about that). Has a default clock speed of 3.6Ghz but since I have an aftermarket cooler, I can manage 4.2Ghz without a problem and stay between 30 and 35 degrees C at minimal use - pushes about 50 degrees at the highest for certain games. That is paired up with a (somewhat dated some would say) HD5750 graphics card, which still manages to play many of the latest games at very decent settings, I feel no need to upgrade it yet.

And indeed. I remember being back on a laptop from such a time, which only had a single core Intel M celeron or something similar, at 1.8Ghz? When I was using Reason (Not a DAW as such ... more MIDI based), it would constantly tell me that I don't have enough CPU power, and even on a Pentium 4 machine at 2.8Ghz, when I pushed that it would give the same warning. Never had it once since moving to multiple cores.

So you have AMD's Bulldozer architecture too. I knew it . And the people uttering doubts regarding the number of cores are partly right as the architecture is - like Intel's Hyper Threading (Simultaneous Multithreading) - Clustered Multithreading. Yours and mine have four integer cores but only two floating point cores... so calling it a 'four threads' CPU would be more realistic. Have you thought about switching to Linux? Our AMD processors are much faster there.

In the near future I´ll buy a new mainboard (with more stable voltage regulation for undervolting), I´d like to have the new Kaveri architecture as I expect it´ll be roughly 20-30% faster than mine. Hell, the new A10-7850K (4 cores) is as fast as an FX-6120 (6 cores). And the GPU part is almost as fast as your GPU.

Here´s my system BTW: http://marlene-d.blogspot.de/2013/04/a-new-pc-and-lot-of-problems.html

Reason was problematic all those years ago. And that your Pentium 4 had problems with it is not surprising; it´s the dreaded Netburst architecture. I´ve had a Pentium 4 3.8 Ghz HT and it wasn´t able to overclock, a clock increase of 5% didn´t translate to a speed increase of 5%... at least not without voltages shooting through the roof. I´m glad that they´re gone.

[SileEeles]

Have you thought about switching to Linux? Our AMD processors are much faster there.

I have in the past dual-booted Linux, but I've only ever had problems getting everything to work properly. That, and I play a lot of games on PC and they just simply aren't available to Linux. I mean they might run via WINE but I haven't ever had the patience to be dealing with that. It just works in Windows, so I stick with it.

This is my system: http://sileeeles.tumblr.com/mypc

[MDietrich]

I have in the past dual-booted Linux, but I've only ever had problems getting everything to work properly. That, and I play a lot of games on PC and they just simply aren't available to Linux. I mean they might run via WINE but I haven't ever had the patience to be dealing with that. It just works in Windows, so I stick with it.

This is my system: http://sileeeles.tumblr.com/mypc

Yeah, I gave up with virtualization too.

Nice system

[Azureal]

MDietrich, I could your stuff all day long and not get bored! Great topic and great conversation ...