Does 96KHz really sound better???

[LeMauce]

[Showcase]

[Bender412]

I'm surprised no one has ever asked this before.

[S21]

Probably asked once a month since... Well since long before electricity was invented.

Everything is more important than sample rate.

But fast silicon is cheap these days.

[Andy_bt]

96kHz can sound quite different if you're using pitch shifting or samplers
A.

[chk23]

It might make a difference for ITB processing, but not for tracking, imho.

[BBG]

Originally Posted by Sean Evans

I would like to know whether anyone has noticed a difference in A-D conversion at 96K, is it really neccesary? Surely there are more important factors e.g. converter build quality, preamps, monitoring, mic placement, processing in/out of the box. I personally find 44.1 or 48 is fine. Is it my inexperienced ears?

without a 100% doubt it might be

[chk23]

Originally Posted by BBG

without a 100% doubt it might be

Yes, especially your dog will hear a big difference (at least if the rest of the recording chain is capable of performing adequately in the 24-48 kHz range...)

[PhilDW]

The problem is that this gets asked all the dang time! Just search for -sample rate- and read the threads that refer to "optimal" or "better".

[BBG]

Originally Posted by chk23

Yes, especially your dog will hear a big difference (at least if the rest of the recording chain is capable of performing adequately in the 24-48 kHz range...)

I think you meant yer moms innit?

[Lotus 7]

Originally Posted by Sean Evans

I would like to know whether anyone has noticed a difference in A-D conversion at 96K, is it really neccesary? Surely there are more important factors e.g. converter build quality, preamps, monitoring, mic placement, processing in/out of the box. I personally find 44.1 or 48 is fine. Is it my inexperienced ears?

Maybe, sometimes, possibly, yes and no, but not necessarily, and maybe not. Thank goodness you didn't ask about 192kHz sampling, whew!

[BradLyons]

Originally Posted by Sean Evans

I would like to know whether anyone has noticed a difference in A-D conversion at 96K, is it really neccesary? Surely there are more important factors e.g. converter build quality, preamps, monitoring, mic placement, processing in/out of the box. I personally find 44.1 or 48 is fine. Is it my inexperienced ears?

Yes and no..... Technical information is sometimes just that, technical....information. For example let's talk converters that are supposedly 24-bit. For every bit of audio it's roughly 6dB in signal, which means that a 24-bit converter has a theoretical dynamic range of 144dB. In reality, I'm not aware of any converters off the top of my head that have that much dynamic range. There are some 24-bit converters that are in the mid to upper 90's and others that are in the 120dB and up----all of which have a completely different sound. There was a company many years ago that made a converter called the A16, it was an 18-bit converter but it blew the pants off of every 24-bit converter that was out at the time. When it comes to the sampling rate, there are other things going on.

Understand that recording at 96kHz really is NOT recording at 96kHz. In fact you're recording at rates SIGNIFICANTLY higher than 96kHz. ON cheaper quality converters higher sampling rates sound much better because there is less down-sampling going on. They have cheaper converter chips, analog sections, etc so the less they are going from 2.6MHz the better off it sounds. But when you get into your higher grade converters , you may not notice as large of a difference between the frequencies. Not to say there isn't one, but it's not as night and day.

This can fool those to think that 96kHz really is superior in every way over 44.1kHz. I'm not saying there isn't a difference, but the difference you hear may not be exactly what you think you're hearing. My point is this.... a great converter @ 48kHz could sound better than another converter @ 96kHz. It happens ALL the time.

[Doc Mixwell]

Originally Posted by BradLyons

Understand that recording at 96kHz really is NOT recording at 96kHz. In fact you're recording at rates SIGNIFICANTLY higher than 96kHz.

Brad, please help me understand what you mean here

Nyquist says we're encoding up to 48kHz when the Sampling rate is set to 96kHz

Do you disagree with this finite math?

[BradLyons]

Originally Posted by Doc Mixwell

Brad, please help me understand what you mean here

Nyquist says we're encoding up to 48kHz when the Sampling rate is set to 96kHz

Do you disagree with this finite math?

Neither agreeing or disagreeing with Nyquist---what I'm referring to is the initial sampling rate of raw recordings, which is somewhere around 2.6Ghz to 2.8Ghz. Someday we'll be playing back at those sample rates as a standard. Look at VSTi's and how higher sampling rates have less latency than lower sampling rates, again it has to do with less down-sampling. Believe me I'm not the end-all expert on this, but it's what I've learned and studied for many years and has proven itself to be true, at least in my experience.

[-tc-]

Originally Posted by Doc Mixwell

Brad, please help me understand what you mean here

Nyquist says we're encoding up to 48kHz when the Sampling rate is set to 96kHz

Do you disagree with this finite math?

He's talking about 1-bit or multibit sigma-delta convertors that encode the analog signal digitally in the megahertz range, and then decimate to the chosen target pcm sample rate. Afaik, it's how virtually all modern converter chips work.

Regardless of that, what it amounts to is that you actually are recording at the decimated rate, unless you digitally store the pdm stream (e.g. as DSD). So imo Brad's point was misleadingly made. Recording at 96k is recording at 96k ... it's a 96k pcm stream that is stored.

[Doc Mixwell]

Originally Posted by BradLyons

Neither agreeing or disagreeing with Nyquist---what I'm referring to is the initial sampling rate of raw recordings, which is somewhere around 2.6Ghz to 2.8Ghz. Someday we'll be playing back at those sample rates as a standard. Look at VSTi's and how higher sampling rates have less latency than lower sampling rates, again it has to do with less down-sampling. Believe me I'm not the end-all expert on this, but it's what I've learned and studied for many years and has proven itself to be true, at least in my experience.

I had to re-read your post like 10 times to make sure I am comprehending it. By Raw Recording, do you mean Analog Signal? Perhaps though, what's good for software instruments might not be the same for analog signals/analog to digital/digital to analog converters.

I was really interested in your theory above that there was more than 96Khz bandwidth when using a 96khz sampling rate, which would of course be disagreeing with the math. Was that not what you meant? If you meant something else, please accept my apologies for mis reading it.

In my understanding, [please correct me if I am wrong] we need to sample at twice the highest frequency we want to accurately represent. Otherwise we will end up with Aliasing Distortion within that spectrum. This is what the "Anti-Aliasing" filter does. It prevents mis-representations from spinning back into reality. When the Converters are switched into higher sampling rates, this steep brick wall filter is pushed out of the audible_spectrum.

I see you are talking about the "theoretical discrete sampling rate of continuos time", which is not really something I care to discuss. But what does interest me, is your statement above. Having spent some time studying the basics of Nyquist, and learning how this sampling theorem works with ADA converter implementation, I am quite interested in what you meant by that statement.

[BBG]

Originally Posted by Doc Mixwell

Brad, please help me understand what you mean here

Nyquist says we're encoding up to 48kHz when the Sampling rate is set to 96kHz

Do you disagree with this finite math?

what is up with your signature? seriously, is all of that really necessary(doc)?

[ReckNC00]

NO, there is not an audible difference between 44.1KHz and 96KHz.

YES, there is an audible difference between a real life AD/DA converter running at 44.1KHz and 96KHz.

The mathematical defense only holds up if you are comparing physically flawless converters! Don't exist.

I get so frustrated when the question "Should I record at 96k?" is immediately met with MATHMATHMATHMATH. It's kinda the same thing but not as annoying as "What's a good mic for acoustic?" being answered with "Well, if the musician and the roomandthesongandtheplayerandthestringsandthewoodandtheweatherandthemoodandtheproducerandthecarpet aren't right, the mic doesn't matter.". Yeah, FINE. But, it just doesn't HELP anybody.

We all know Nyquist and I acknowledge it as truth. But, I've also compared recordings 44/96 on 6 different converters (from $40 to $500 a channel), and there is a HUGE difference. Not mathematically or theoretically, but ACTUALLY...

So, please... Lets be pragmatic musicians. Let's keep those early on in their home studio ventures from settling and then struggling.

Loveyoumeanit.

[S21]

Half the sample frequency gives you the Nyquist rate, but you only get about 80% of that in the real world because of the rolloff of antialiasing filters.

[Jazz Noise]

Just to reconfirm what's been said here.

We initaly began oversampling to give anti-alias filters (which were analogue at the time) breathing room. Current digital tech can go alot faster - so we can give these filters more breathing room initialy and then digitaly brickwall above our sample rate afterwards.

So to your A/D stage it's almost arbitrary - it's doing a buttload of work either way and sampling millions of times a second so that the ugly stuff between your "pass band" and "stop band" doesn't reflect down into the audio range.

It doesn't sound different in isolation - but when you time stretch/pitch shift/formant shift or modulate the audio at an extreme rate (Super fast compressor attack times, phaser plugin going at 400hz) you're going to hear artifacts being introduced. If the plugin upsamples itself, then you're fine, if it doesn't then you'll probably hear it. Variety Of Sound discusses this on his blog and books such as the Dodge Ad Jerse guide on Computer Music discuss aliasing within audio manipulation in better detail than we've space for.

I get the "We're artists" mentality - but oversimplifying doesn't help either. You have to understand the question properly before you can reach a meaningful answer.

[psyOPs]

Originally Posted by chk23

Yes, especially your dog will hear a big difference (at least if the rest of the recording chain is capable of performing adequately in the 24-48 kHz range...)

Speakers reliably produce frequencies that high?

[chk23]

Originally Posted by psyOPs

Speakers reliably produce frequencies that high?

Afaik, there are monitors that do so - but finding a decent mic with such a frequency range will be hard

[Lotus 7]

Originally Posted by chk23

Afaik, there are monitors that do so - but finding a decent mic with such a frequency range will be hard

Actually not so hard. Sennheiser MKH 8020s are almost flat from 10 Hz to 40 kHz and usable to 60 kHz. Using Grace M-101s, a sampling frequency of 192 k, and playing back through our Magneplanar ribbons (flat to beyond 45 kHz), our dog reports that the top octave is very "crisp".

[chk23]

Originally Posted by Lotus 7

Actually not so hard. Sennheiser MKH 8020s are almost flat from 10 Hz to 40 kHz and usable to 60 kHz. Using Grace M-101s, a sampling frequency of 192 k, and playing back through our Magneplanar ribbons (flat to beyond 45 kHz), our dog reports that the top octave is very "crisp".

Yes, I know, there are some mics with a range like that. But most mics used in the industry (especially the expensive vintage ones) aren't specified for frequencies above 20kHz.

[stinkyfingers]

sm 57 goes above 20kHz
most things do

[Jazz Noise]

Originally Posted by stinkyfingers

sm 57 goes above 20kHz
most things do

If you wan't to be able to reproduce 20Khz, chances are you'll still only be rolling off by 40. The rate of rolloff is another question!

But I thought the question of the thread was "Is 96Khz better" rather than "Can I hear above 20Khz?".

[chk23]

Originally Posted by stinkyfingers

sm 57 goes above 20kHz

Most mics somehow "go above 20 kHz". But they're far from being flat at this frequencies. As for the sm57: All specs I found on the web state a frequency range up to 15 kHz and according to the curves I saw the rolloff starts there, only -10dB remain at 20kHz...

[TheWiz]

[S21]

Strong signals beyond the nyquist folding frequency are a pain. Say you sample at 48kHz so the folding freq is 24kHz. A 30kHz signal input into the system will show up in the digitised output at 18kHz. Hopefully the antialiasing filters have rolled this off somewhat so it isn't too loud.

It is just simpler if the mics don't produce any signal beyond the expected range.