48000Hz vs 192000Hz aka What The Heck Is Going On?

[TFHT]

I've been hearing a lot of things about Sample Rate and Bit Depth, and the controversy that ensues when they are talked about.

So, I made a little test song to try out different Bit Depths and Sample Rates and see how they affect my music.

I made a little test track in FL Studio. ASIO4ALL was selected, I recorded an acoustic guitar part with my Microphone on channel 10. Turned off the mic, sent the audio file to that mixer channel, added reverb, chorus, an instance of Dblue Glitch, and an instance of mda Tube Screamer. I also sent the stock FL Studio snare and hat to that channel, kept the kick on channel one, added a compressor to the master channel. I also had 3XOSC on bass, not sent to the mixer, and a many octave spanning C chord in an instance of Zebra 2, also not assigned to a mixer track.

Basically the massive chord plays first, fading into the drums, bass, and guitar immediately.

I made 6 renders, .wavs, 3 in 32 Bit, and 3 in 24 Bit. Each bit depth file had a different sample rate - 22050, 48000, and 192,000 in that order.

For some reason, the 2 wav files at 22050 Hz didn't even play in Foobar or Winamp. The others played fine, and I found that while I could not discern an immediate difference between the two Bit Depths with my headphones, there was a HUGE difference between 48000 and 192000 Hz. Zebra 2 sounded amazing at 192000Hz - very full and sounded like it was resonating. The rest of the song sounded even in volume, but slightly muffled / less loud as a whole. Zebra 2 sounded louder at 48000Hz, but sounded brittle compared to 192000Hz. The rest of the song at 48000Hz sounded louder than at 192000, but it was very uneven mixing.

WHAT IS GOING ON HERE?

Are there parts that simply sound better with one sample rate than the other, and vice verse? Is this a mixing issue? Do some sounds and effects lack and ability to be properly rendered in certain formats?

Considering the test song, I want the loudness of the guitar and drums from the render at 48000Hz, but I want that awesome resonating effect from Zebra at 192000Hz? How can I get the best of both worlds?

I, like many of you, want the best possible sound quality for my finished products, so that those with the ability to hear them in the best light actually can. Considering this, if my music is eventually put on a CD, does it even matter if the song is 192000Hz, when the Sample Rate is reduced once it goes to CD?

Thank you all in advance for your help!

[James Perrett]

What SRC algorithm were you using? What sample rate did you record the guitar at?

A few points to think about...

The actual number of bits used is probably the same in your 32bit and 24 bit files - you're just wasting space with the 32 bit files.

Being loop based, I would guess that the stretching algorithms used in Fruity Loops could be obscuring any differences due to sample rate although I don't really know the software so don't know if these algorithms can be switched off.

Why are you using ASIO4ALL? That's normally used with low end consumer interfaces that could never handle 192kHz sample rates properly.

[Folderol]

In all seriousness, if you can hear a significant difference between 48k and 192k recoded under the same conditions and at the same bit depth, I'd suggest there is something very wrong with your setup.

As for 22.05, why such a peculiar frequency? It's half the lowest most stuff will handle.

[largon203]

My experience (17 years) tells me that mixing with 44.1 is a bit easier than anything higher. I like 48 for natural recordings like piano with very little compression or jazz though. Controlling those harmonics in 48 can be challenging and since I primarily record Pop, Rock, Hard Rock, and HipHop R&B the mixing techniques I employ become destructive at 192, much less 48. I Prefer 44.1 24Bit. Also unless you have some real serious monitors,amps, and very high end cables; your not going to hear what 192 has to offer. If there is one nearby, stop into an audiophile shop and have a listen to a pair of Stenheims or Wilson Alexias Powered by dual Mono Block BAT amps and a high end NAD DAC set to 192 and listen to their demo material then listen to the same thing at 44.1 it is not a subtle difference. But in the real world where people can't afford $250,000 USD (on the low side of things) for an audiophile playback setup.... again in the real world you cannot hear the difference.

-Dave K

[Tim Gillett]

I think the core question is "can humans hear ultrasonics?" ie: frequencies above 20khz, the normally accepted absolute upper limit.
As far as I am aware there is no solid evidence that we can hear much if anything above 20 khz. Maybe babies can.

Tim

[largon203]

Very true however on those high end systems I'm referring to your not hearing higher frequencies as much as your now hearing things more linearly. What's going on in those stratosphere frequencies is pretty important. To be a little poetic... It's the sonic glue that holds the whole thing (on a really interesting level) together. After you've heard 192 that was recorded at 192 on a high end audiophile system and then you hear the same track at 44.1 it becomes obvious what your missing. The 44.1 track sounds ailiased (is that a word? My spell check doesn't think so) well regardless I'm not arguing with you because your right except for a few random lucky people and dogs most of us can't hear higher than around 19-20k but again it's not that your hearing higher frequencies but that due to the ultra high sample rate those 15k-20k we can hear are WAY better represented with the overtone structure that is there, albeit that you can't hear even at a live performance, but that you can feel.

-Dave

[James Perrett]

As far as I am aware there is no solid evidence that we can hear much if anything above 20 khz. Maybe babies can.

I used to find the 19kHz pilot tone on my mono FM radio pretty annoying and, when I first started work, the HP computer terminals put out a line scan frequency higher than that so I reckon many young people can probably hear to slightly more than 20kHz but nowadays I have problems hearing the 15kHz line scan frequency from my CRT TV.

However, in my opinion, the issues with 44.1kHz are mainly down to the anti alias filtering. Many convertors opt to roll off the high frequency response at 20kHz slightly in order to reduce any filter artefacts. A higher sampling rate gets you well away from the filter's transition band. 88.1 and 96kHz sampling rate are probably closest to the ideal (although even 48kHz doubles the width of the transition band) but I've yet to hear any seriously convincing argument for 192kHz sampling rate.

[Tim Gillett]

Very true however on those high end systems I'm referring to your not hearing higher frequencies as much as your now hearing things more linearly. What's going on in those stratosphere frequencies is pretty important. To be a little poetic... It's the sonic glue that holds the whole thing (on a really interesting level) together. After you've heard 192 that was recorded at 192 on a high end audiophile system and then you hear the same track at 44.1 it becomes obvious what your missing. The 44.1 track sounds ailiased (is that a word? My spell check doesn't think so) well regardless I'm not arguing with you because your right except for a few random lucky people and dogs most of us can't hear higher than around 19-20k but again it's not that your hearing higher frequencies but that due to the ultra high sample rate those 15k-20k we can hear are WAY better represented with the overtone structure that is there, albeit that you can't hear even at a live performance, but that you can feel.

-Dave

Not so. The 15khz to 20khz are represented quite adequately even at 44.1 sample rate. Anything that is an "overtone" of 15 khz to 20khz is by definition a frequency higher than 20khz and therefore out of the passband.

Yes those ultrasonic overtones/frequencies are WAY better represented... but we cant hear them...It is not obvious what you are missing if you cant hear what you're missing. You're missing it either way!

Prove beyond reasonable doubt that a representative sample of people can hear ultrasonics and you will probably become famous. There's your incentive. Go for it!

Regards Tim

PS, if you're worried about aliasing you can always record at 88khz and resample back to 44khz.

[muzines]

I regularly miss motorway turn offs because all that ultra-violet light the eye cannot detect messes with my visual perception so much I can't read the signs.

That pesky ultra-violet bit of the electro-magnetic spectrum! Grr...
(This is presumably the reason they invented sat-nav...)

Note: soft synths can and do often sound better when running at higher sample rates, due to aliasing and other DSP-related reasons. There is an argument for something higher than 44/48 as it requires less steep roll-off filters. Other than that, there is absolutely no point in recording at 192KHz.

Double-blind, confirmation bias etc etc and all that...

[largon203]

Hey Tim

I'm afraid that what I heard and experienced wasn't really up for debate. When you get a chance to listen to a proper demo on an Audiophile system that is tuned properly costing upwards around 1/4+ million USD you let me know what you experienced.
Kudos mate!

-Dave

[TFHT]

WELL, it looks like I came to the right place for in depth responses! That said, I'm very new to what I am talking about here, so, and I don't say this often, please talk to me like a child.

I tested this out on a more complete song. In this song, Albino 2 is the only thing changing, especially on the instance that has distortion effects. Other than Albino, nothing else changes in both the 48000Hz and 192000Hz versions. The cleaner instances of Albino sound fuller at 192000HZ, but once the song gets to the distorted bass instance of Albino, there is some white noise and distortion, like it is making the bass too deep, causing thebass notes to "come undone" like they need to be raised an octave.

What SRC algorithm were you using? What sample rate did you record the guitar at?

Why are you using ASIO4ALL? That's normally used with low end consumer interfaces that could never handle 192kHz sample rates properly.

I don't even know what an SRC algorithm is. If its any consolation, the FL Studio ASIO panel says the microphone's IN and OUT are both 2x 6.4-48kHz, 16 bits, and "Always resample 44.1kHz 48kHz, what ever that means. I'm guessing it is only recording 16 bit files at 48kHz, and it is being resampled? And, I use ASIO4ALL because it is the only thing I know of right now that allows me to MIDI control with virtually no delay, play the song in FL with half the cracks and pops as the PSD, and it allows me to use my mic and other inputs very easily.

In all seriousness, if you can hear a significant difference between 48k and 192k recoded under the same conditions and at the same bit depth, I'd suggest there is something very wrong with your setup.

As for 22.05, why such a peculiar frequency? It's half the lowest most stuff will handle.

There very well could be. Got any ideas for diagnosis? I used 22.05 because, well, I'd never heard of that sample rate before and I thought that if I was testing my middle and highest sample rates, I should probably test the lowest for good measure.

----------

What is (anti)aliasing, how can I tell if this is happening, and what can I do about it?

And for good measure, what would happen if I did this test with something like Asynth, which can oversample 8x?

Basically, it sounds like I shouldn't be hearing any difference given my human hearing and the quality of my speakers, but since I am, something must be wrong.

[James Perrett]

It looks like your tests are pretty meaningless as your soundcard can't handle more than a 48kHz sampling rate.

[James Perrett]

When you get a chance to listen to a proper demo on an Audiophile system that is tuned properly costing upwards around 1/4+ million USD you let me know what you experienced.

Dave - I'm afraid that audiophile systems don't have a good reputation around here. I've heard expensive audiophile systems that sound worse than my 80 quid kitchen stereo. In my experience audiophiles seem to like "nice" sound rather than accurate sound and audiophile systems often include deliberately introduced distortion to create this "nice" sound.

And anyone advocating 192kHz sampling should read this white paper from Dan Lavry before going any further.

[muzines]

Yep. It take some skill and experience to set up unbiased technical tests that you can draw meaningful conclusions from.

It doesn't sound like you are there yet (no offence intended!) but I suspect your soundcard is resampling everything internally if you have set your DAW to 192KHz but the card can only support 48KHz.

As for instruments - as always, it depends how they are programmed. The better ones will be implemented properly and may sound clearer with less artifacts at higher sample rates because it's much easier to deal with aliasing (however, some will oversample anyway, and all that happens at higher samples rates is the oversampling need is removed or made less expensive). However, lesser instruments may not even be tested at those rates and may have DSP problems or other issues. It's very difficult to draw meaningful conclusions from testing the behaviour of one or two instruments, because they will doubtless be internally very different.

Basically - it's good to be curious and to test things out for yourself, but unless you can do good tests you may well take away the wrong information - it's wise to read up on the accepted standard and how things work and start from that basis, that start without that knowledge and try and work it all out from scratch, without the correct experience and equipment to do it properly.

There is lots of info and knowledge on this forum though, so you're in the right place to discover..!

[chris...]

dual Mono Block BAT amps and a high end NAD DAC set to 192

I can maybe see why they're called "bat" amps.

[TFHT]

It looks like your tests are pretty meaningless as your soundcard can't handle more than a 48kHz sampling rate.

Is a test really meaningless if it is educational? Could you please expand on this response, and perhaps answer the other questions I asked you? Judging from your signature, you seem like you are the guy to ask.

SOMETHING is happening hear, and I don't know what it is. If my soundcard cant read 192000Hz, shouldn't it just NOT play that file at all? Is it dropping the Sample Rate to 48000 automatically?

I'm having a hard time finding out exactly what Realtek soundcard model I have, but when I check the Realtek HD Audio Manager, select the "Speakers" tabs, and click on "Default Format," I can choose from 16 Bit, 44100Hz, to 24 Bit, 192000HZ, however, on "Stereo Mix" I only have two options - 16 Bit, 44100HZ and 16 Bit, 48000Hz. Since my Acer Aspire 5553g didn't come with any other cables or external hardware for audio, it SOUNDS like I SHOULD be able to properly render in 192000Hz. But, I'm the newb, so you guys tell me.

Can I turn off aliasing?

AND, since I used Dblue Glitch in the test song to purposely lower the sound quality to something like a video game, might that be a culprit as well? One of the options in the VST is the ability to reduce Bit Depth...

[hollowsun]

It's my opinion that you over-thinking this and taking it all a bit too seriously.

Your Realtek soundcard seems, from what I know, a consumer thing for average audio listening (i.e. MP3 listening). It is by no means a 'professional' audio I/O. To think superlative results are possible is, I think, a little naive ... but I appreciate that you're new to all this and yes, it CAN be damned confusing sometimes.

The Acer you have is a fine little thing for general computing, gaming, music listening, video/Netflix/YouTube viewing but I'd be loathe to use it for any serious recording, especially without some quality external audio I/O interface.

I hate to bring the news to you but having a laptop does not necessarily give you a recording studio. And sorry, 192kHz recording is total arsewash - 44.1/16-bit is perfectly adequate. At a push, 44.1/24-bit is (arguably) better but this will have to be dithered down to 16-bit for CD release and if your target market is an MP3 release, 192kHz really is overkill ... or, to use a technical term, horsesh!t with more than a hint of "The Emperor's New Clothes" about it.

Sorry to be blunt

[Richie Royale]

It should be noted that the main difference he appears to notice is with a plug in instrument, which desmond has already explained about; the discussion about the soundcard seems to be a bit of a red herring as I can't see any reference to recording the guitar at various rates or any improvement regarding it.

there was a HUGE difference between 48000 and 192000 Hz. Zebra 2 sounded amazing at 192000Hz - very full and sounded like it was resonating. The rest of the song sounded even in volume, but slightly muffled / less loud as a whole. Zebra 2 sounded louder at 48000Hz, but sounded brittle compared to 192000Hz. The rest of the song at 48000Hz sounded louder than at 192000, but it was very uneven mixing.

[James Perrett]

Is it dropping the Sample Rate to 48000 automatically?

That's the problem with many consumer audio chips - they do all kinds of conversions without telling the end user. It all started with Creative Labs Audigy range around 20 years ago which resampled everything to 48kHz and, unfortunately, resampling at that time wasn't particularly good. Other consumer audio manufacturers followed this lead so it would appear that resampling is pretty much the norm in these chips. A decent audio interface will either not resample at all or, if it does, it will use a much higher quality algorithm.

Did you follow the link to Dan Lavry's paper that I posted? He also has an introduction to digital audio on his site which would be worth reading - as would anything by Hugh Robjohns on this site. Unfortunately Hugh is on holiday this week, otherwise I'm sure he would have answered your questions much more fully than most of us can.

[Tony O'Shea]

Hi,
Just to second James' recommendation of Dan Lavry's paper. It is well worth reading particularly in the context of this debate.
Best,
Tony

edit - opps, apostrophe in the wrong place.

[Goddard]

It's my opinion that you over-thinking this and taking it all a bit too seriously.

Your Realtek soundcard seems, from what I know, a consumer thing for average audio listening (i.e. MP3 listening). It is by no means a 'professional' audio I/O. To think superlative results are possible is, I think, a little naive ... but I appreciate that you're new to all this and yes, it CAN be damned confusing sometimes.

The Acer you have is a fine little thing for general computing, gaming, music listening, video/Netflix/YouTube viewing but I'd be loathe to use it for any serious recording, especially without some quality external audio I/O interface.

I hate to bring the news to you but having a laptop does not necessarily give you a recording studio. And sorry, 192kHz recording is total arsewash - 44.1/16-bit is perfectly adequate. At a push, 44.1/24-bit is (arguably) better but this will have to be dithered down to 16-bit for CD release and if your target market is an MP3 release, 192kHz really is overkill ... or, to use a technical term, horsesh!t with more than a hint of "The Emperor's New Clothes" about it.

Sorry to be blunt

Just for the record, support for multi-channel 192kHz/24-bit audio is in fact required for playback of a number of defined Blu-Ray primary audio stream formats (and hence, for the recording and production of same). Which probably explains why it was also made part of Intel's High Def Audio spec (and thus, why onboard HDA codec chips like Realtek's include it) as well as why it was added in the USB audio class type 2 spec.

[largon203]

dual Mono Block BAT amps and a high end NAD DAC set to 192

I can maybe see why they're called "bat" amps.

BAT is an acronym for Balanced Audio Technology. Google it...

-Dave.

[largon203]

When you get a chance to listen to a proper demo on an Audiophile system that is tuned properly costing upwards around 1/4+ million USD you let me know what you experienced.

Dave - I'm afraid that audiophile systems don't have a good reputation around here. I've heard expensive audiophile systems that sound worse than my 80 quid kitchen stereo. In my experience audiophiles seem to like "nice" sound rather than accurate sound and audiophile systems often include deliberately introduced distortion to create this "nice" sound.

And anyone advocating 192kHz sampling should read this white paper from Dan Lavry before going any further.

Apparently the "Audiophile" systems you listened to were not properly setup or just a poorly matched set of equipment. My father specializes in the setup of such systems he has a process that includes several mods that remove/lower the noise floor by 60+ or should I say - dec. I'm sure your experience comes from listening to systems setup by wannabe audiophiles. The gentlemen I know and respect, recognize that All audio equipment has color and they generally go for the least intrusive. FYI even your studio monitors have color / distortion... It's unavoidable. We just look for equipment that we can work with. It took me about 6 years figure out my Hafler TRMs and an additional 5 to get good with them. I'm just not at about 14 years with them starting to make some really great mixes. They are very colored but I know how they translate. I LOVE my fathers Beauhorns that incorporate an ultra efficient single louther driver and a wooden knob mounted in it to disperse the high frequencies properly in a folding horn designed box. Lots of geek audiophile gear that most of us mortals simply can't justify spending the $$££ on. I assure you his system is WAY more accurate than your whatever quid kitchen stereo. LMAO.... No disrespect intended.

-Dave

[ernzo]

-What about the Nyquist theorem?

i'm not an expert by any means, but it seems that samplerate could be a bit like framerate..

If higher samplerates (192khz) mean more frames of captured audio per second, not only could it mean being able to record higher frequencyes, but maybe also record "normal" or all frequencies with more detail?

more precision in time and more samples of it..

as music made by real instruments is of very complex and dynamic harmonic nature, the faster and bigger number of samples can only be good.. and even better being able to record a wider spectrum of sound freqs..

anyone can elaborate on this?

-I never saw the need to go higher than 88.2khz tho (wich resamples better to 44.1khz than 192khz), and most of what I do is made at 44.1 from the start... only for acoustic stuff I might try 48khz or 88.2 if it was a super production

-Whats much much more noticeable to me is the change from 32 to 16 bit... it's like night and day, I woud never mix at any less than 32bit

[largon203]

And anyone advocating 192kHz sampling should read this white paper from Dan Lavry before going any further.

Not saying I advocate 192 but I read that whole article and he does a great job of maki g a mountain out of a molehill. I tell you I almost disregarded the whole article after busting out laughing when I got to the end of the very first paragraph. I did Finnish the article for entertainment value though. While this very articulate scientifically thought out observation seems sound it is filled with a load of bunk. To say a higher sampling rate is going to capture too much info is like saying vinyl record players provided too much info. C'mon man! If there is no sound up at 180k and recording up to 192 you are recording nothing. This guy implies you are recording BAD frequencies. He has clearly way over thought this. Ha ha ha. Makes for a good read though if you can stomach his blathering. Sorry if that offends mate but I couldn't disagree more. I'll end with I never record higher than 44.1 NOT because its better than 48 - 192 but because when you compress heavily as I sometimes do it becomes more difficult to control what happens to the sound IMO. That is not to imply that Bad sound gets recorded at higher frequencies. It's important sound IF your going for natural with very little if any compression. Such as a classical piano or a stereo mic'ed symphony.

-Dave