High Playback Sampling Frequencies, Why are they becoming popular? Options

[icstm]

OK, at the risk of sounding completely stupid?

Why do people listen to music with high sampling frequencies?
What is 88k+ providing them?
I understand why you might record at a high sampling rate, but why keep that for playback?

Looking through the FAQ, there are threads from 2003 that point out that the sampling frequency and bit-dept work in tandem. So the quantisation error of 16bit at 44.1k has the opportunity to be corrected sooner at a higher sampling rate, so in some ways is like a dithering pattern.

However given noise introduced in the analogue systems required to listen to music, a SNR within a 16bit signal of ~96dB seems pretty good.

So assuming that speakers struggle to produce the sounds that a 192k sampling frequency allow (eg 96kHz) and assuming that 16bits were sufficient when compared to the analogue equipment in the system, what have I missed in these high sampling playback formats?

[Prince Of All Sa...]

I'm not totally sure, but recalling my ever-so-tiny amount of knowledge pertaining to stereo CDs (44.1 khz, 22.05 khz per channel), I'd say the higher frequencies are for surround sound tracks or something of that nature.

Please correct me if I'm wrong, but I'm pretty sure humans cannot hear frequencies higher than around 20 khz.

[dhromed]

what have I missed in these high sampling playback formats?

The economic benefits of selling people larger numbers for more money.

[pdq]

People sell music with higher sampling rates for the same reason that they sell amplifiers whose volume controls go to 11, to make more money.

Edit: dhromed beat me.


This post has been edited by pdq: Feb 17 2012, 16:53

[dhromed]

I'm not totally sure, but recalling my ever-so-tiny amount of knowledge pertaining to stereo CDs (44.1 khz, 22.05 khz per channel), I'd say the higher frequencies are for surround sound tracks or something of that nature.

Surround sound is encoded in more than two channels, and all channels will be 16/44.1, or 24/48, etcetera.

The audio signal is the audio; it's not a carrier wave where extra content is somehow modulated into the higher frequencies.

This post has been edited by dhromed: Feb 17 2012, 16:56

[Prince Of All Sa...]

Exactly. That's why you see headphones with frequency ranges all the way up to 50+ khz, among other things.

When you don't know the answer to a question, assume it's money.


edit: Thanks for the info, dhromed. Didn't know that.

This post has been edited by Prince Of All Saiyans: Feb 17 2012, 16:59

[greynol]

stereo CDs (44.1 khz, 22.05 khz per channel)

The samplerate is 44.1kHz in order to properly preserve ~20kHz of bandwidth. It has nothing to do with it being in stereo.

[Prince Of All Sa...]

stereo CDs (44.1 khz, 22.05 khz per channel)

The samplerate is 44.1kHz in order to properly preserve ~20kHz of bandwidth. It has nothing to do with it being in stereo.

Got it. So back to the OP's question, since the frequency for 44.1khz or 48khz is per channel, even surround sound would be unnaffected by the higher frequencies of 96khz or 192khz. Right?

This post has been edited by Prince Of All Saiyans: Feb 17 2012, 17:55

[pawelq]

to stereo CDs (44.1 khz, 22.05 khz per channel)

It's actually 44.1 in each channel.



There is a slight and somewhat disputable evidence for minimal influence of frequencies above ~20 kHz on perception.


As others said, this mostly an advertising point, you can show that the numbers in your gear are higher than numbers is the other guy's gear.

Or in your brochures you can draw a smooth sine @192kHz and a (misleading and untrue) jagged sine @44.1k.

There are however people who are made very happy by this marketing phenomenon - I mean scientists who study audition using animal models. Most mammals' hearing range goes higher than human hearing range. 10 or 20 years ago getting equipment capable of producing sounds up to 40 or more kHz was very costly, now you just go and buy a $200 sound card, a $400 loudspeaker, and you're there.

[icstm]

10 or 20 years ago getting equipment capable of producing sounds up to 40 or more kHz was very costly, now you just go and buy a $200 sound card, a $400 loudspeaker, and you're there.

but is the 50-90khz range necessary or even achievable through speakers by the sampling afforded at 192k?
most roll off well before there (so the 24bit depth is even less useful...)

[Bartholomew MacG...]

I'm still confused as to why studios use really high sampling frequencies. I have a vague understanding of why higher bit depths might be needed for adujsting levels, but I don't get why they need higher sampling rates.

[saratoga]

I'm still confused as to why studios use really high sampling frequencies. I have a vague understanding of why higher bit depths might be needed for adujsting levels, but I don't get why they need higher sampling rates.

bigger number == better

[Walter_Odington]

For producing sound/music then higher bit depths and sample rates can have advantages (and disadvantages).

- time stretch and other granular based pitch/time changes allow smaller grain size, and relatively smoother grain size resolution. A clearly audible difference can be heard here.

- signal processing at higher depth/resolution can preserve signals above above audible frequencies that will affect the sound at audible frequencies i.e. control signals may have less audible artefacts at hight depth/resolution. Simular to above.

- resampling to change pitch is a common process. Resampling at higher resolution will remain more faithful to the original, and sounds outside of the audible range may be transposed down whereas if sampled at 44.1 you may end up with dull sounds when pitched down (although, who knows what is up there!).

- Presumably higher bit depth give you more range to play with when recording. Someone may correct me on that one.

- Higher depth/resolution hardware tends to be more inaccurate, jitter etc.

- Processing overhead and disk space usage could be quite insane.

[saratoga]

For producing sound/music then higher bit depths and sample rates can have advantages (and disadvantages).

- time stretch and other granular based pitch/time changes allow smaller grain size, and relatively smoother grain size resolution. A clearly audible difference can be heard here.

- signal processing at higher depth/resolution can preserve signals above above audible frequencies that will affect the sound at audible frequencies i.e. control signals may have less audible artefacts at hight depth/resolution. Simular to above.

- resampling to change pitch is a common process. Resampling at higher resolution will remain more faithful to the original, and sounds outside of the audible range may be transposed down whereas if sampled at 44.1 you may end up with dull sounds when pitched down (although, who knows what is up there!).

I don't think any of this is true with respect to sampling rate, since any effect processing could trivially upsample if it actually needed to prior to processing. Like I said before, I think the ral reason that higher sampling rates are used is that they exist and there is no real down side to using them so people use them. People really do like bigger numbers.

[Walter_Odington]

I don't think any of this is true with respect to sampling rate, since any effect processing could trivially upsample if it actually needed to prior to processing. Like I said before, I think the ral reason that higher sampling rates are used is that they exist and there is no real down side to using them so people use them. People really do like bigger numbers.

Depends on the software. For example, ableton live time stretch is based on the DAW setting, Reaktor has its own processing settings, and UAD has 192 sample rate as standard.

Im not saying higher is better for all situations, but these are some which do benefit from higher numbers. On the flip side, I could imagine a situation where a CD would have been better off coming from a 44.1 native project than a 96 kHz project due to quantisation error, so I don't think higher is better in ever situation. For me, it makes sense to work in as high a resolution as possible even if the delivery file is a 3gp, because the processing has a clearly audible difference. Plus I get drawn in to marketing speil :-p

[greynol]

the processing has a clearly audible difference.

Before posting again, please review the terms you accepted when registering, paying special attention to #8, the one you have violated twice in this discussion.

[saratoga]

I don't think any of this is true with respect to sampling rate, since any effect processing could trivially upsample if it actually needed to prior to processing. Like I said before, I think the ral reason that higher sampling rates are used is that they exist and there is no real down side to using them so people use them. People really do like bigger numbers.

Depends on the software. For example, ableton live time stretch is based on the DAW setting, Reaktor has its own processing settings, and UAD has 192 sample rate as standard.

You're just pointing out what I said: that people like higher numbers. But that says nothing about the importance of those numbers.

Im not saying higher is better for all situations, but these are some which do benefit from higher numbers.

Would it trouble you to mention them?

On the flip side, I could imagine a situation where a CD would have been better off coming from a 44.1 native project than a 96 kHz project due to quantisation error,

Quantization error has absolutely nothing to do with sampling rate. Perhaps you should look up what quantization error is prior to expressing your opinions on it . . .

[Walter_Odington]

I see I have walked in to the lions den!

I'll prepare an ABX test of ableton live at different sample rates.



I understand this is a forum for scientists, but I didn't realise scientists could be so tetchy! I'm not trying to get one over anyone, just puttin some perspective accross that might be relevant. I won't repeat or elaborate on my original points until I have some evidence, but I would like to respond that quantisation does have something to with sample rates when you change them.

I'm not a complete dunce Saratoga, sorry if ive offended the sensibilities of the forum, but you come across really petty in your post above. I will take the criticism on board and try and improve my conduct going forwards!

[greynol]

I won't repeat or elaborate on my original points until I have some evidence, but I would like to respond that quantisation does have something to with sample rates when you change them.

Would you like to provide some explanation rather than just say that it does in order to help those who don't have the same level of understanding as you?

[saratoga]

I understand this is a forum for scientists, but I didn't realise scientists could be so tetchy! I'm not trying to get one over anyone, just puttin some perspective accross that might be relevant.

I don't agree with that. When you come in assert something to be true that you do not know to be true, you are in fact trying to "get one over" on everyone. People will indeed get "tetchy" when you give them reason to suspect you.

I'll prepare an ABX test of ableton live at different sample rates.

You should start a new thread and discuss the setup for this experiment before committing a lot of time to it. It sounds like you may not be aware of some of the previous experience people have with ABX testing sample rates.

I would like to respond that quantisation does have something to with sample rates when you change them.

It does not. You may be thinking of rounding error, or possibly something else. Its unclear from your post exactly what you were trying to say.

[Walter_Odington]

I think I've misunderestimated the amount of effort that should go in to a post on here. I'll pipe down until Ive got something to say with discernable provanence.

Ps, reread my previous post, sorry if I come across as an arse saratoga - bit of a rash comment from me, I just felt like I was being hung out to dry when it wasn't warranted. Think I 'get' this forum that little bit more now, and the rigours required from participants. (although how did the guy who said each channel has half the samples get off Scott free?!? ;-)

[saratoga]

Ps, reread my previous post, sorry if I come across as an arse saratoga - bit of a rash comment from me, I just felt like I was being hung out to dry when it wasn't warranted. Think I 'get' this forum that little bit more now, and the rigours required from participants. (although how did the guy who said each channel has half the samples get off Scott free?!? ;-)

No problem, and I didn't mean to discourage you from posting in the future. Sorry if I came off like that.

[Walter_Odington]

Saratoga, just seen your latest. I appreciate I need to explain myself, but I don't believe I've actually said anything incorrect. Your arguments don't leave much room for any truth that may reside in my blabbering, yet there is something in what I'm saying.

What I've said previously re sample rates/depths has been too vague, which is evidently a cardinal sin round here, I will seek to improve that, but is it really right to rubbish my points because they are poorly expressed?


Edit - you've probably encouraged me to post better than not post, just felt like a bit of a sledge hammer!

This post has been edited by Walter_Odington: Feb 20 2012, 03:12

[saratoga]

Saratoga, just seen your latest. I appreciate I need to explain myself, but I don't believe I've actually said anything incorrect. Your arguments don't leave much room for any truth that may reside in my blabbering, yet there is something in what I'm saying.

Qunatization error is the error when quantizating a signal to a finite number of bits. For instance, 17 bit sample have half the quantization error of 16 bit samples. Sampling rate is the frequency at which samples are recorded. They are mathematically related in a very complex way, but for our purposes they have nothing to do with each other since the error per sample doesn't change if you record one sample per second and then upsample it to 2 Hz or 2khz or 2Mhz. Each sample would still have an error of approximately 2^-(N+1) for an N bit quantization.

With that in mind,

On the flip side, I could imagine a situation where a CD would have been better off coming from a 44.1 native project than a 96 kHz project due to quantisation error,

the problem with this statement is that it implies that changing sampling rate has something to do with quantization error, when (at least in this context) it doesn't.

What I've said previously re sample rates/depths has been too vague, which is evidently a cardinal sin round here, I will seek to improve that, but is it really right to rubbish my points because they are poorly expressed?

No I'm rubbishing them because I don't think you fully understand the question, and to illustrate that I'm pointing out that you're not using the terminology consistently with the depth of understanding required to give a useful opinion.

[Ron Jones]

Presumably higher bit depth give you more range to play with when recording. Someone may correct me on that one.

I'm not sure what you mean by "range". Are you referring to frequency response?

Higher depth/resolution hardware tends to be more inaccurate, jitter etc.

Not to my knowledge. It's possible, certainly, that this statement is true in some cases. I don't believe there's a tendency for this to be the case. In many cases the same components responsible for clocking and converting 'high-res' audio are the same as the hardware you'll find in more limited devices: the latter devices often just don't expose the capabilities of the underlying hardware, for one reason or another.

Processing overhead and disk space usage could be quite insane.

The increase in computational effort is generally negligible. As for disk usage, I suppose you could argue that, but you can also argue that the storage demands of 'high-res' audio is small potatoes compared to even compressed video or uncompressed imagery.

This post has been edited by Ron Jones: Feb 20 2012, 03:39