You are here

96khz 192khz

All about the tools and techniques involved in capturing sound, in the studio or on location.

96khz 192khz

Postby iceman » Wed Feb 19, 2014 12:30 am

hi guys, im wondering are these worthwhile using? or do they just take up more disk space and mean that my interfaces have exciting looking numbers emblazoned on them?
User avatar
iceman
Regular
Posts: 279
Joined: Thu Oct 16, 2003 11:00 pm
Location: Liverpool

as metal as your nanna`s new hip


Re: 96khz 192khz

Postby hollowsun » Wed Feb 19, 2014 1:51 am

iceman wrote:or do they just take up more disk space and mean that my interfaces have exciting looking numbers emblazoned on them?
^^ That
User avatar
hollowsun
Frequent Poster
Posts: 2122
Joined: Thu Jan 20, 2005 12:00 am

Re: 96khz 192khz

Postby Jack Ruston » Wed Feb 19, 2014 7:03 am

It depends.

In my experience there's no argument for 192 and in fact smarter people have argued quite convincingly that it causes problems. I suppose if you were really reaching for a reason you could say that it imposes a track count and processing limit that some might find charming. But the idea that bigger numbers means better quality represents a misunderstanding. What it means is more, but not be necessarily better. It's like taking biscuits out of a small box and putting them into a big one in the hope that there will be more biscuits.

96k is a slightly different matter. On the minus side it gives you large files and certainly adds to the processing bill. But it moves one of the tricky parts of converter design, the low pass filtering, well away from what we can hear, which allows a less well designed budget unit to punch above its weight. It's not so much that we need to hear higher frequencies but if we are going to box things in tightly, we need to do it elegantly. 96k also shortens system latency which can be useful in tracking. Some plug ins work better at 96k than at standard rates.

But make no mistake, a very good converter working at standard rate sounds very good. Most records are still made at those rates and many electronic records start off with sources of much lower quality than that.

If you're working with lower track counts, a lot of acoustic instruments or classical recordings there's a good argument for 96. Most pop music with higher track counts, and more processing works very well at 44.1kHz or 48.

Some would argue that the final destination is important. I think that we could well see a future shift towards higher resolution delivery and perhaps there's an argument for using 88.2 or 96 for that reason.

In an ideal world sample rate would be 64kHz.

J
Jack Ruston
Frequent Poster
Posts: 3477
Joined: Wed Dec 21, 2005 12:00 am

Re: 96khz 192khz

Postby The Elf » Wed Feb 19, 2014 7:22 am

I do use 96kHz from time to time, since I like how it can affect the low end of some recordings, especially acoustic guitar, piano and small string ensembles. But in general 44.1 is just fine - especially if you're as metal as you suggest!
User avatar
The Elf
Jedi Poster
Posts: 9969
Joined: Mon Aug 13, 2001 11:00 pm
Location: Sheffield, UK

An Eagle for an Emperor, A Kestrel for a Knave.


Re: 96khz 192khz

Postby Ian. » Wed Feb 19, 2014 8:05 am

I don't understand how sampling rates above 44.1 could do anything to make the audio sound better, especially when converters oversample anyway. What is the good argument for 96k with classical recording? Please tell, i'm interested.
Ian.
Poster
Posts: 13
Joined: Sat May 11, 2013 11:00 pm

Re: 96khz 192khz

Postby John Willett » Wed Feb 19, 2014 8:15 am

I record classical at 24/96 or 24/88.2 all the time.

These higher rates avoid having the brick-wall filter at 20kHz and any artefacts that can ripple down into the audible frequencies.

I don't bother with 192kHz.
User avatar
John Willett
Jedi Poster
Posts: 4074
Joined: Tue Mar 07, 2000 12:00 am
Location: Oxfordshire UK

John
Sound-Link ProAudio
Circle Sound Services
President - Federation Internationale des Chasseurs de Sons


Re: 96khz 192khz

Postby Jack Ruston » Wed Feb 19, 2014 8:40 am

Yeah, if you use 44.1kHz you're having to filter off at 22.05kHz. But you can't have a 'vertical' filter slope, and steep filters are a tricky business. So the filtering starts to creep down into areas that you can hear. In most rock, pop and electronic we're not concerned with those frequencies anyway. Often we're getting rid of them. But in something like classical where you might have a single pair of mics with extended frequency response and a very 'pure' sonic picture, it makes a difference. And with a stereo recording like that, there's no real downside.

J
Jack Ruston
Frequent Poster
Posts: 3477
Joined: Wed Dec 21, 2005 12:00 am

Re: 96khz 192khz

Postby MarkOne » Wed Feb 19, 2014 9:45 am

The Elf wrote:I do use 96kHz from time to time, since I like how it can affect the low end of some recordings, especially acoustic guitar, piano and small string ensembles. But in general 44.1 is just fine - especially if you're as metal as you suggest!

If anyone can point me to a reason that a higher sample rate can effect in any way the low end, I would be happy to read it.
MarkOne
Frequent Poster
Posts: 971
Joined: Thu Feb 15, 2007 12:00 am
Location: Bristol, England, Earth, Peruses Gap, Milky Way

Re: 96khz 192khz

Postby James Perrett » Wed Feb 19, 2014 10:08 am

The only good reason for using 192kHz is for the times you want to sample signals with a bandwidth of up to 90 kHz or so. I'm not sure that it would ever be needed for normal audio work but there are industrial applications where it would be useful. I've certainly used 96kHz sample rates for recording sonar signals.

96kHz has its uses as Jack already mentioned. In addition, it can be useful when restoring old records as it makes click detection easier - any signal above about 20kHz must be a click (unless you are transcribing a CD-4 quadraphonic disc).
James Perrett
Moderator
Posts: 5816
Joined: Sun Sep 09, 2001 11:00 pm
Location: The wilds of Hampshire

JRP Music - Audio Mastering and Restoration.
http://www.jrpmusic.co.uk


Re: 96khz 192khz

Postby johnny h » Wed Feb 19, 2014 10:41 am

iceman wrote:hi guys, im wondering are these worthwhile using? or do they just take up more disk space and mean that my interfaces have exciting looking numbers emblazoned on them?
There are certain situations where higher sample rates make a difference but they are mainly due to poor converter design or poor plugin design.
johnny h
Frequent Poster
Posts: 3132
Joined: Sun Jul 23, 2006 11:00 pm

Re: 96khz 192khz

Postby Hugh Robjohns » Wed Feb 19, 2014 10:44 am

Jack Ruston wrote:I think that we could well see a future shift towards higher resolution delivery and perhaps there's an argument for using 88.2 or 96 for that reason.

In an ideal world sample rate would be 64kHz.

J


I'd go for 60kHz as the ideal -- it works better with all the video rates!

I'd also vote strongly against 'high resolution' delivery formats. The professional recording system should ideally be of higher quality than the delivery format, so that you have wriggle room in post production and can optimise the product for the end consumer. To me, that means 24/96 for recording, and 16/44.1 for delivery.

There is zero point in delivering higher wordlength for the typical home user since they won't have anything that can reproduce the potential dynamic range, and zero point in higher sampler rates since over-sampling delta-sigma converters avoid the potential filtering issues anyway.

From my perspective, 192 and higher rates are a complete waste of time. 96k is useful where you are concerned about high energy, high-frequency content and/or poor converter filtering. High sample rates are useful for some DAW processing, but most modern plug-ins now up-sample internally to deal with that anyway.

But other than that, it's 44.1 for general stuff and 48k if video related, and 96k for anything to be restored or used in sound design (where you might want to slow things down).

H
User avatar
Hugh Robjohns
Moderator
Posts: 16929
Joined: Thu Jul 24, 2003 11:00 pm
Location: Worcestershire, UK

Technical Editor, Sound On Sound


Re: 96khz 192khz

Postby Hugh Robjohns » Wed Feb 19, 2014 10:53 am

Jack Ruston wrote:So the filtering starts to creep down into areas that you can hear.

Actually, Jack, it's usually the reverse. The designers keep the pass band maximally flat up to 22kHz or so, and then roll off above that... but because the attenuation slope is not vertical the filter doesn't actually comply with Nyquists requirements, and so some material above half the sample rate is deliberately let through. Most off-the-shelf converter chips only manage 6dB attenuation at the Nyquist frequency!

Anything that is let through above the half-sample rate will alias back into the audio band, producing anharmonmic distortion, but the presumption is that, in general, there isn't much energy at extreme HF and so the stuff that does get through won't be audible, and therefore most people wont notice most of the time -- but it most definitely a bodge!

However, where you are close-miking sources with a lot of strong HF harmonics and noise -- cymbals, brass and string instruments, etc -- and recording theme at relatively high level, you can get some subtle but perceivable aliasing problems if working at 44.1k. If working in that kind of situation, recording the source at 96kHz can be helpful.

H
User avatar
Hugh Robjohns
Moderator
Posts: 16929
Joined: Thu Jul 24, 2003 11:00 pm
Location: Worcestershire, UK

Technical Editor, Sound On Sound


Re: 96khz 192khz

Postby Hugh Robjohns » Wed Feb 19, 2014 10:56 am

MarkOne wrote: If anyone can point me to a reason that a higher sample rate can effect in any way the low end, I would be happy to read it.

This is commonly reported but it most probably a perceptual illusion -- the cleaner top end seems to make the bottom end appear clearer too. Odd... but similar effects have been noted by loudspeaker designers where better tweeters appear to improve the low end!

Our sense of hearing is nothing short of weird!

H
User avatar
Hugh Robjohns
Moderator
Posts: 16929
Joined: Thu Jul 24, 2003 11:00 pm
Location: Worcestershire, UK

Technical Editor, Sound On Sound


Re: 96khz 192khz

Postby The Red Bladder » Wed Feb 19, 2014 11:41 am

Hugh Robjohns wrote:However, where you are close-miking sources with a lot of strong HF harmonics and noise -- cymbals, brass and string instruments, etc -- and recording theme at relatively high level, you can get some subtle but perceivable aliasing problems if working at 44.1k. If working in that kind of situation, recording the source at 96kHz can be helpful.

Exactly that!

In the good-old/bad-old days when men were men, almost all we worried about was noise and headroom. Then the first ADDA things appeared and we got the most terrible top end, esp. from early ProTools converters. According to Barry at iZ-Technology (Radar) they were poorly clocked and they had not been able to get the filtering (as Hugh describes above) quite right.

When mastering, I could always tell if something had been recorded in ProTools, because the spectrum analyser would stick to the side right at the top end at about 18-20kHz, rather like water to the side of a dirty glass. Somewhere up there, there was a nasty whistle that had to be filtered off.

With the introduction of the better clocked converters in the HD systems and of course others like Apogee and Radar, that nasty and distorted top end disappeared.

But for those of you who like to record at 96kHz, please remember that most studio microphones only go up to about 22kHz anyway. If you really push some of the newer designs, you can get up to nearly 30kHz, but then it is at -40dB or more, compared to the usual 20-20k that they all claim. A good converter at 44/48 is better than something mediocre at 96kHz.

Also, having equipment (desk, pre-amps, etc., etc.) that goes into RF brings its own problems, ranging from gain controls that can act as RF wobblers, to picking up Radio China. For that reason, many bits of kit (some of which even claim F ranges up to 200kHz) are capped all over their longer signal paths to keep the RF out. The customer naïvely likes to see a F range from zero to 200kHz, but for some reason, complains if the gain pot makes funny noises and longer unbalanced lines are able to receive a detailed report on shoe production in the Shaanxi province.

As for video/film compatibility - everybody and their mothers-in-law are batting on about 4K at 60 fps, so a 60kHz sample rate would indeed be ideal.

But then just as soon as we have all settled down at 4K/60fps, no doubt somebody will insist on 8K and 120 fps as being the 'new standard.'

And so we merrily go on!
The Red Bladder
Frequent Poster
Posts: 1878
Joined: Mon Jun 04, 2007 11:00 pm
Location: . . .

 


Re: 96khz 192khz

Postby Jack Ruston » Wed Feb 19, 2014 11:50 am

Hugh Robjohns wrote:
Jack Ruston wrote:So the filtering starts to creep down into areas that you can hear.

Actually, Jack, it's usually the reverse. The designers keep the pass band maximally flat up to 22kHz or so, and then roll off above that... but because the attenuation slope is not vertical the filter doesn't actually comply with Nyquists requirements, and so some material above half the sample rate is deliberately let through. Most off-the-shelf converter chips only manage 6dB attenuation at the Nyquist frequency!

Anything that is let through above the half-sample rate will alias back into the audio band, producing anharmonmic distortion, but the presumption is that, in general, there isn't much energy at extreme HF and so the stuff that does get through won't be audible, and therefore most people wont notice most of the time -- but it most definitely a bodge!

However, where you are close-miking sources with a lot of strong HF harmonics and noise -- cymbals, brass and string instruments, etc -- and recording theme at relatively high level, you can get some subtle but perceivable aliasing problems if working at 44.1k. If working in that kind of situation, recording the source at 96kHz can be helpful.

H

Ah interesting. Thanks Hugh!
Jack Ruston
Frequent Poster
Posts: 3477
Joined: Wed Dec 21, 2005 12:00 am

Re: 96khz 192khz

Postby Tim Gillett » Wed Feb 19, 2014 12:18 pm

James Perrett wrote:The only good reason for using 192kHz is for the times you want to sample signals with a bandwidth of up to 90 kHz or so. I'm not sure that it would ever be needed for normal audio work but there are industrial applications where it would be useful. I've certainly used 96kHz sample rates for recording sonar signals.

96kHz has its uses as Jack already mentioned. In addition, it can be useful when restoring old records as it makes click detection easier - any signal above about 20kHz must be a click (unless you are transcribing a CD-4 quadraphonic disc).

In your country the Graf company makes affordable high speed audio cassette "digitisers" built around 192khz PCI cards.
It means you can digitise cassettes at about 8 or 9 times normal speed at an effective sample rate of 22khz (10khz bandwidth). Another practical use for 192khz sample rate!

Tim
Tim Gillett
Frequent Poster
Posts: 1232
Joined: Wed Jan 30, 2013 12:00 am
Location: Perth, Western Australia

Re: 96khz 192khz

Postby Hugh Robjohns » Wed Feb 19, 2014 12:31 pm

Jack Ruston wrote: Ah interesting. Thanks Hugh!


No problem. Here's some more detail or you:

The plot below is of the 'transition band' of the Crystal Semiconductor CS5382 A-D converter chip -- often used in high-end pro-audio stuff -- taken from their own data sheet which you can find HERE.

Image

This is the usual kind of plot that you often see, and it looks impressive with stuff above 0.5Fs well below -95dBFS. However, if you look closely, the finite slope means that it presents only a few decibels of attenuation at 0.5Fs (half the sample rate). Crystal are not alone in this kind of design -- most of the high-end converter from all the usual chip manufacturers will look broadly the same.

Crystal also publishes in its data sheet a close up of the Nyquist region of the curve:

Image

Here you can see that at half the sample rate, the filter is only 3dB down. That nice Mr Nyquist said it should be way, way down -- minus 95dB or more for high quality audio! But I they did that roll-over would start further down into the wanted audio band, and people would get concerned if they read that the response was -0.5dB at 19kHz or similar -- stupidly and pointlessly concerned, but we all know that the marketing numbers are the only important thing! There is no marketing number for aliasing distortion!

However, what this means, in practice, is that any high level high frequency content just above the Nyquist frequency will be aliased back into the audio band, resulting in audible -- or at least perceptual anharmonic distortion. So stuff at 23kHz will appear at 21.1kHz. Okay, it's going to be attenuated a bit too -- probably by about 15dB in this specific case -- but our ears are quite capable of detecting unnatural distortion products and artefacts at much, much lower levels than that! Higher source frequencies will come back at lower in-band frequencies (albeit with slightly lower levels again), and the filter attenuation isn't really sufficient until you get above about 24kHz (as opposed to the theoretical requirement of 22.05kHz).

Cymbals are largely noise-like, and aliased noise is just more noise with a slightly different tonal colour. You may detect it... but the ultrasonic contributions from violins and trumpets are much more tonal and so sound more wrong when aliased!

But, as I said earlier, this only tends to be a problem if you close mic those kinds of sources -- because the air naturally filters a lot of the ultrasonic energy out with distance -- and record at high levels (so the artefacts are that much stronger) on the track. Record with more traditional mic techniques and headroom margins and there is little problem in practice... and that's what the chip designers are relying upon.

Nevertheless, moving up to a 96kHz sample rate moves the same filter response turnover up to 48kHz, well above the wanted audio, and the natural HF roll-off of the mics (and possibly the electronic chain) will attenuate the ultrasonic energy such that there won't be anything meaningful to alias around 48kHz anyway.

But as you can see, recording solo trumpet with an Earthworks 40kHz mic close up, into a budget interface running at 44.1kHz, and with the gain set to peak at -0.5dBFS really is a recipe for disaster!

Hope that helps.

H
User avatar
Hugh Robjohns
Moderator
Posts: 16929
Joined: Thu Jul 24, 2003 11:00 pm
Location: Worcestershire, UK

Technical Editor, Sound On Sound


Re: 96khz 192khz

Postby ken long » Wed Feb 19, 2014 1:07 pm

Tim Gillett wrote:
James Perrett wrote:The only good reason for using 192kHz is for the times you want to sample signals with a bandwidth of up to 90 kHz or so. I'm not sure that it would ever be needed for normal audio work but there are industrial applications where it would be useful. I've certainly used 96kHz sample rates for recording sonar signals.

96kHz has its uses as Jack already mentioned. In addition, it can be useful when restoring old records as it makes click detection easier - any signal above about 20kHz must be a click (unless you are transcribing a CD-4 quadraphonic disc).

In your country the Graf company makes affordable high speed audio cassette "digitisers" built around 192khz PCI cards.
It means you can digitise cassettes at about 8 or 9 times normal speed at an effective sample rate of 22khz (10khz bandwidth). Another practical use for 192khz sample rate!

Tim

not concerned with azimuth, apparently...
User avatar
ken long
Frequent Poster
Posts: 3161
Joined: Mon Jan 21, 2008 12:00 am
Location: The Orient, East London

I'm All Ears.


Re: 96khz 192khz

Postby Ian. » Wed Feb 19, 2014 1:08 pm

Hugh Robjohns wrote:
...designers keep the pass band maximally flat up to 22kHz or so, and then roll off above that... but because the attenuation slope is not vertical the filter doesn't actually comply with Nyquists requirements, and so some material above half the sample rate is deliberately let through. Most off-the-shelf converter chips only manage 6dB attenuation at the Nyquist frequency!

That's an analogue filter, yes?

As an aside, and not addressing the above quote, i don't get why any of this matters when digital filters oversample. Using a high sample rate to accommodate high frequency content in order to prevent aliasing distortion seems irrelevant since the converter is doing it for us anyway, no? It then goes through the anti-aliasing filter which, being digital, has no trouble fitting the transition band into the tight space required, and then the extra samples are just dropped leading to perfect replication of the signal with no distortion.

Also, because of the nonlinearity of our microphones, speakers and amplifiers, won't capturing higher frequencies with higher sample rates lead to even more distortion when we attempt to reproduce them? Especially considering that distortion tends to increase at higher frequencies.

Or am i confused and i've got this wrong?
Ian.
Poster
Posts: 13
Joined: Sat May 11, 2013 11:00 pm

Re: 96khz 192khz

Postby Hugh Robjohns » Wed Feb 19, 2014 1:44 pm

Ian. wrote:That's an analogue filter, yes?


Nope. Well, not quite!

The overwhelming majority of modern converters are delta-sigma types which initially sample at very high rates (eg, 128x wanted sample rate, so 5.6MHz for a 44.1 output), but a low wordlength. An analogue Nyquist anti-alias filter is required, of course, but it only has to stop input signals above a couple of MHz or so, which is easy enough to do and the slope can be quite gentle.

Once in the digital domain, the signal is 'decimated' to translate it into the required sample rate and wordlength (eg, 24/44.1 or whatever). Part of that decimation process involves a digital anti-alias filter to ensure compliance with the new sample rate, and the computations involved in that are related directly to the sample rate, so the filter shape tracks with all sample rates in the same way. The example shown above is typical and its response around the 0.5Fs rate are due in part to convenient simplifications in the digital filter design.

i don't get why any of this matters when digital filters oversample. Using a high sample rate to accommodate high frequency content in order to prevent aliasing distortion seems irrelevant since the converter is doing it for us anyway, no?


No. It is vital that the wanted audio always has a smaller bandwidth than the digital environment, other wise you will suffer aliasing. The high sample rate of the delta-sigma A-D converter is only an interim stage of the full converter process.

It then goes through the anti-aliasing filter which, being digital, has no trouble fitting the transition band into the tight space required


But even digital filters aren't perfect brick-wall designs -- at least, not in the practical implementations typically employed in chip converters. As the diagrams I posted above show. Those are digital decimation filters in action!

Also, because of the nonlinearity of our microphones, speakers and amplifiers, won't capturing higher frequencies with higher sample rates lead to even more distortion when we attempt to reproduce them? Especially considering that distortion tends to increase at higher frequencies.


The more linear each element of the chain, the better and (in the context of the points made earlier), recording at 96kHz can be a benefit in maximising HF linearity in some situations by avoiding potential aliasing distortion issues.

Or am i confused and i've got this wrong?


Yes But you're not alone in that!

H
User avatar
Hugh Robjohns
Moderator
Posts: 16929
Joined: Thu Jul 24, 2003 11:00 pm
Location: Worcestershire, UK

Technical Editor, Sound On Sound


Re: 96khz 192khz

Postby Tim Gillett » Wed Feb 19, 2014 2:43 pm

ken long wrote:
Tim Gillett wrote:
James Perrett wrote:The only good reason for using 192kHz is for the times you want to sample signals with a bandwidth of up to 90 kHz or so. I'm not sure that it would ever be needed for normal audio work but there are industrial applications where it would be useful. I've certainly used 96kHz sample rates for recording sonar signals.

96kHz has its uses as Jack already mentioned. In addition, it can be useful when restoring old records as it makes click detection easier - any signal above about 20kHz must be a click (unless you are transcribing a CD-4 quadraphonic disc).

In your country the Graf company makes affordable high speed audio cassette "digitisers" built around 192khz PCI cards.
It means you can digitise cassettes at about 8 or 9 times normal speed at an effective sample rate of 22khz (10khz bandwidth). Another practical use for 192khz sample rate!

Tim

not concerned with azimuth, apparently...

No, probably not the sorts of people who would normally buy these units for whom even the term "azimuth" would usually have no meaning. These units are all about speed and convenience.

Tim
Tim Gillett
Frequent Poster
Posts: 1232
Joined: Wed Jan 30, 2013 12:00 am
Location: Perth, Western Australia

Re: 96khz 192khz

Postby iceman » Wed Feb 19, 2014 5:54 pm

wow it seems i threw the cat among the pigeons here!! basically my setup is based around a motu 24 i/o and a motu 896HD i was just curious is all, ive recorded my own band indeed they are metal at 24/96 but couldn't really hear anything eye openingly amazing in comparison to a recording a few months earlier that was done at 44/24. the other reason i was wondering about this is because i seem to have a problem with any Daw ive ever used/trialed etc in the respect of when i bounce the mix that we are happy with to a stereo file basically what we get back as a stereo file doesn't replicate what im listening to in the mix, so i was thinking of as i have these two interfaces to record the output of one setup into another pc using the motu 896 so im recording exactly what im hearing we have tried several times ive even used demo's of the likes of motu digital performer, Reaper, sonar x3 etc etc and to a certain degree they all seemed to do it. i thought it might have been some kind of ear fatigue thing so even resorted to mixing one day leaving it a day then exporting and listening back. so basically i just wondered what the best sample rate would be to record onto my second pc would be. many thanks for all the insight and opinions!
User avatar
iceman
Regular
Posts: 279
Joined: Thu Oct 16, 2003 11:00 pm
Location: Liverpool

as metal as your nanna`s new hip


Re: 96khz 192khz

Postby Tomás Mulcahy » Thu Feb 20, 2014 1:10 pm

There are reasons for that, but first we need to eliminate some variables. Is this with an offline bounce like Reaper, Cubase etc., or a real time bounce like Pro Tools does? On these mixes, do you use plugins that have modulation?
User avatar
Tomás Mulcahy
Frequent Poster
Posts: 1528
Joined: Tue Apr 24, 2001 11:00 pm
Location: Cork, Ireland.

madtheory creations

Soundware Shop


Re: 96khz 192khz

Postby Jack Ruston » Thu Feb 20, 2014 1:26 pm

Hugh Robjohns wrote:
Jack Ruston wrote: Ah interesting. Thanks Hugh!

No problem. Here's some more detail or you:

The plot below is of the 'transition band' of the Crystal Semiconductor CS5382 A-D converter chip -- often used in high-end pro-audio stuff -- taken from their own data sheet which you can find HERE.

Image

This is the usual kind of plot that you often see, and it looks impressive with stuff above 0.5Fs well below -95dBFS. However, if you look closely, the finite slope means that it presents only a few decibels of attenuation at 0.5Fs (half the sample rate). Crystal are not alone in this kind of design -- most of the high-end converter from all the usual chip manufacturers will look broadly the same.

Crystal also publishes in its data sheet a close up of the Nyquist region of the curve:

Image

Here you can see that at half the sample rate, the filter is only 3dB down. That nice Mr Nyquist said it should be way, way down -- minus 95dB or more for high quality audio! But I they did that roll-over would start further down into the wanted audio band, and people would get concerned if they read that the response was -0.5dB at 19kHz or similar -- stupidly and pointlessly concerned, but we all know that the marketing numbers are the only important thing! There is no marketing number for aliasing distortion!

However, what this means, in practice, is that any high level high frequency content just above the Nyquist frequency will be aliased back into the audio band, resulting in audible -- or at least perceptual anharmonic distortion. So stuff at 23kHz will appear at 21.1kHz. Okay, it's going to be attenuated a bit too -- probably by about 15dB in this specific case -- but our ears are quite capable of detecting unnatural distortion products and artefacts at much, much lower levels than that! Higher source frequencies will come back at lower in-band frequencies (albeit with slightly lower levels again), and the filter attenuation isn't really sufficient until you get above about 24kHz (as opposed to the theoretical requirement of 22.05kHz).

Cymbals are largely noise-like, and aliased noise is just more noise with a slightly different tonal colour. You may detect it... but the ultrasonic contributions from violins and trumpets are much more tonal and so sound more wrong when aliased!

But, as I said earlier, this only tends to be a problem if you close mic those kinds of sources -- because the air naturally filters a lot of the ultrasonic energy out with distance -- and record at high levels (so the artefacts are that much stronger) on the track. Record with more traditional mic techniques and headroom margins and there is little problem in practice... and that's what the chip designers are relying upon.

Nevertheless, moving up to a 96kHz sample rate moves the same filter response turnover up to 48kHz, well above the wanted audio, and the natural HF roll-off of the mics (and possibly the electronic chain) will attenuate the ultrasonic energy such that there won't be anything meaningful to alias around 48kHz anyway.

But as you can see, recording solo trumpet with an Earthworks 40kHz mic close up, into a budget interface running at 44.1kHz, and with the gain set to peak at -0.5dBFS really is a recipe for disaster!

Hope that helps.

H

Wow. So if that filter were moved to achieve -90dB at the 22.05kHz it'd be starting to fall off at what...18kHz or something looking at that?

When Bricasti started making their stereo DA, I was lucky enough to have a studio visit and demo. Interestingly, they have adjustable filter settings on that box, and you can elect to have more extended HF at 44.1kHz with steeper filter, or allow some audible filtering in exchange for a more gentle filter. I personally preferred that sound.

J
Jack Ruston
Frequent Poster
Posts: 3477
Joined: Wed Dec 21, 2005 12:00 am

Re: 96khz 192khz

Postby The Elf » Thu Feb 20, 2014 2:06 pm

MarkOne wrote:
The Elf wrote:I do use 96kHz from time to time, since I like how it can affect the low end of some recordings, especially acoustic guitar, piano and small string ensembles. But in general 44.1 is just fine - especially if you're as metal as you suggest!
If anyone can point me to a reason that a higher sample rate can effect in any way the low end, I would be happy to read it.
I don't pretend to understand it either, but that's what I hear. A few year back we threw a few theories around, and Hugh has outlined a strong possibility.

It's subtle enough that I remain at 44.1 for the vast majority of my work.
User avatar
The Elf
Jedi Poster
Posts: 9969
Joined: Mon Aug 13, 2001 11:00 pm
Location: Sheffield, UK

An Eagle for an Emperor, A Kestrel for a Knave.


Re: 96khz 192khz

Postby Urthlupe » Thu Feb 20, 2014 3:38 pm

Very interesting thread for someone sitting in a hospital bed....

Thankyou everyone.

Loopy
Urthlupe
Regular
Posts: 424
Joined: Thu Sep 19, 2002 11:00 pm
Location: West Midlands, UK

Re: 96khz 192khz

Postby Hugh Robjohns » Thu Feb 20, 2014 3:57 pm

Jack Ruston wrote:Wow. So if that filter were moved to achieve -90dB at the 22.05kHz it'd be starting to fall off at what...18kHz or something looking at that?


Yep, given those filter parameters. And if they did that, most people probably wouldn't be able to hear the slightly reduced bandwidth, but might appreciate the cleaner high end under challenging HF content material. But the marketing people would really freak out! -3dB at 20kHz? Are you mad?

When Bricasti started making their stereo DA, I was lucky enough to have a studio visit and demo. Interestingly, they have adjustable filter settings on that box, and you can elect to have more extended HF at 44.1kHz with steeper filter, or allow some audible filtering in exchange for a more gentle filter. I personally preferred that sound.


Yes, a lot of high-end converters offer options like this. Different cut-off frequencies and slopes, linear phase or minimum phase, and so on. Sometimes you can hear the differences and sometimes not -- it largely depends on the audio material you're listening too. There is no perfect solution, the designers just have to choose the best compromise for a given set of parameters which include design complexity (and thus cost), manufacturability, delay, ringing, in-band ripple, bandwidth, aliasing, and so on.

H
User avatar
Hugh Robjohns
Moderator
Posts: 16929
Joined: Thu Jul 24, 2003 11:00 pm
Location: Worcestershire, UK

Technical Editor, Sound On Sound


Re: 96khz 192khz

Postby iceman » Fri Feb 21, 2014 12:31 am

Tomás Mulcahy wrote:There are reasons for that, but first we need to eliminate some variables. Is this with an offline bounce like Reaper, Cubase etc., or a real time bounce like Pro Tools does? On these mixes, do you use plugins that have modulation?

no just using eq's and compression the stock ones in this case that come with presonus studio one pro v2. i have actually done this before where we took one pc and recorded directly from it into another pc it did work well i just wondered if there were any do's and dont's for this approach or any opinions on the idea
User avatar
iceman
Regular
Posts: 279
Joined: Thu Oct 16, 2003 11:00 pm
Location: Liverpool

as metal as your nanna`s new hip


Re: 96khz 192khz

Postby Btyreman2013 » Sun Feb 23, 2014 10:15 pm

I record a lot of solo classical guitar and usually use 24/88 but also 24/96 sometimes. It shouldn't be a problem for space if you are using 1TB or bigger hard drives. I still use the higher sample rates though for other styles and think unless your computer is struggling to cope then you may as well use it, the main thing I have noticed is plug-ins especially saturation ones can sound cleaner and also seem to have more headroom.
User avatar
Btyreman2013
Regular
Posts: 64
Joined: Sat Apr 27, 2013 11:00 pm

Re: 96khz 192khz

Postby Goddard » Wed Apr 09, 2014 11:40 am

Reviving this thread to point out that it has spawned a "Sound Advice: Recording" Q&A in this month's issue of SOS:

http://www.soundonsound.com/sos/apr14/a ... 414-02.htm

Btw, an earlier "Sound Advice" Q&A may also be of interest on this topic:

http://www.soundonsound.com/sos/jul12/a ... 0712-1.htm

as may this even-earlier SOS article:

http://www.soundonsound.com/sos/sep07/a ... lmyths.htm

User avatar
Goddard
Frequent Poster
Posts: 995
Joined: Tue Apr 03, 2012 11:00 pm

 


Next

Who is online

Users browsing this forum: No registered users and 2 guests