48K v 96K 24bit ? Huh ?

PaulBrewer

In my opinion 96k does sound better than 48k.

My question this time is this -

Which sounds better - 48k when tracked and mixed at 48k entirely through the project.

OR

Tracking at 96k and converting those tracks to 48k for the continuation/completion of the project.

Anyone every done suchlike ?

MilanPan!c

PaulBrewer wrote: »

In my opinion 96k does sound better than 48k.

My question this time is this -

Which sounds better - 48k when tracked and mixed at 48k entirely through the project.

OR

Tracking at 96k and converting those tracks to 48k for the continuation/completion of the project.

Anyone every done suchlike ?

I would def say that done correctly starting at 96k will always sound better..

PaulBrewer

MilanPan!c wrote: »

I would def say that done correctly starting at 96k will always sound better..

Is that a guess ?

madtheory

It depends on the converter I think. IME some sound better at 44 than at 48, and the reverse also. Depends on the filters and how the clock is divided down for each rate from the master oscillator.

If you're doing any sort of processing, a strong case can be made for 96k being better. For example, the Massenburg eq operates internally at double the sampling rate for that reason. Some others do also, but not all. So it depends on what plugins you're using.

Short answer to your question Paul- IMO it depends!

PaulBrewer

madtheory wrote: »

It depends on the converter I think. IME some sound better at 44 than at 48, and the reverse also. Depends on the filters and how the clock is divided down for each rate from the master oscillator.

If you're doing any sort of processing, a strong case can be made for 96k being better. For example, the Massenburg eq operates internally at double the sampling rate for that reason. Some others do also, but not all. So it depends on what plugins you're using.

Short answer to your question Paul- IMO it depends!

A long answer .... but not to the question I asked !

MilanPan!c

PaulBrewer wrote: »

Is that a guess ?

no... Though, it def DOES depend on quite a few things...

I know someone with a HUGE amount of experience recording at 24/96 and downsampling to 16/44.1.

I have heard his work at 48 and 96 and 96->48->44.1.

It's just my opinion abd if course you can easily test it yourself, but I'd def vote for 96 given the option.

kfoltman

PaulBrewer wrote: »

Tracking at 96k and converting those tracks to 48k for the continuation/completion of the project.

Theoretically, this should be better. With higher sample rate, there is less risk of introducing aliasing at A/D conversion stage (which is caused by imperfect lowpass filtering in the A/D converter). This is especially important for source signals contains harmonics outside of the audible range, due to harmonic distortion or anything else.

In practice, it all depends on quality of the A/D converter and the resampling algorithm used for downsampling to 48k.

SeanHurley

Are we talking about converting each individual track here Paul? I am no expert but I would have thought introducing an extra layer of conversion would not be desirable? I personally track and mix with 48/24bit but that has more to do with the limitations of PTLE and the 002.

PaulBrewer

SeanHurley wrote: »

Are we talking about converting each individual track here Paul? I am no expert but I would have thought introducing an extra layer of conversion would not be desirable? I personally track and mix with 48/24bit but that has more to do with the limitations of PTLE and the 002.

Yes ....

I'm wondering how much of the extra juice that's captured with 96 would be lost by converting to 48.

That compared to what is already lost just by going straight to 48 ....

MilanPan!c

PaulBrewer wrote: »

Yes ....

I'm wondering how much of the extra juice that's captured with 96 would be lost by converting to 48.

That compared to what is already lost just by going straight to 48 ....

Welll...

I think if you converted pre-master you'd get a LOT of that extra juice... however, if you did it per-track, pe-mix, I would wonder if you might not actually be hurting yourself...

Paul, why not just try it?

You're ears are prolly good enough to tell the difference... right?

I mean, if you can't punters certainly can't... if it's better than do it... if not, don't bother...

PaulBrewer

Niall - Dahlia wrote: »

I've never tracked and mixed a full session at 96k. I've always assumed the extra processing required by the plugins at 96k made it unsuitable. Any sessions we get here at 96k are converted down to 48k, mainly due to the AD8000s.

Is the processing stress/track count the reason behind converting down after tracking?

I have a HD2 rig that can more or less run a whole session with plugs at 96.

However it takes a big of rejigging with plugs to keep it going ....

Sonically there's a clear improvement by using 96 - more of the legendary 'AIR' that's so valuable.

However, is the trade off worth it for more plugins + speed by going 48 ?

Is the plan to track at 96 then convert to 48 the 'Best' compromise ?

PaulBrewer

MilanPan!c wrote: »

Welll...

I think if you converted pre-master you'd get a LOT of that extra juice... however, if you did it per-track, pe-mix, I would wonder if you might not actually be hurting yourself...

Paul, why not just try it?

You're ears are prolly good enough to tell the difference... right?

I mean, if you can't punters certainly can't... if it's better than do it... if not, don't bother...

I will ....

I was wondering had anyone else any insights

trackmixstudio

I record at 24/44.1
I tracked drums at 96k one time to hear it but there was no difference I could hear.
Big let downs are obviously double processing and hard drive space.
Audio drivers and disks have to work twice as hard when tracking and you loose half or your inputs when using an adat card.
Also if you want to record at high sample rates, 88.2 to 44.1 is an easier conversion for obvious reasons. 48 is not a good idea unless you are doing sound for film which is 48 default. The conversion from 48 to 44.1 for music is pointless and bad for your sound.

I remember reading somewhere (maybe gearslutz) that most commercially released records are done at 44.1.

madtheory

Ah right, gotcha Paul! Is the sample rate conversion in PT 8 the same as in PT 7? Because SRC in PT up to and including 7 is not very good. Best avoided.

Barbabatch, R8 Brain and Izotope are vastly superior.

Bear in mind that pretty much all converters are oversampling so there is on chip SRC already, to make the filter work properly. So recording at 96k and downsampling is a bit like getting your tech to calibrate the tape machine to xnW at +4, and then running it into the red all the time.

tweeky

Yeah, but how many people here with all the rates and bits have been in the middle of a session that everyone say's sounds really great... suddenly notice on the front of the A/D that it's set to 44.1/16 by mistake? ....and all that time.... go on hands up!

woodsdenis

madtheory wrote: »

Ah right, gotcha Paul! Is the sample rate conversion in PT 8 the same as in PT 7? Because SRC in PT up to and including 7 is not very good. Best avoided.

Barbabatch, R8 Brain and Izotope are vastly superior.

http://www.audioease.com/Pages/BarbaBatch4/Barba4SRCTest.html

On the Barbabatch website tests it shows that Protools Tweekhead is certainly not the worst of the major DAWs

madtheory

Agreed, but it is not the best either. I didn't mean to imply that it was the worst of all. Although I am of the opinion that it is a failing of PT. I think it's fair to expect better from a pro app.

splitrmx

In my (admittedly amateur) opinion, I would do all the work using one sampling rate and only let proper mastering engineers sample it down to whatever format it was going to end up on.

Quiggers

if the end format is mp3 then 44.1kHz 24 bit is more than good enough,
if its sound track work for film then you need to be at 48kHz or 96kHz,
88.2kHz is available but there is no point as everything in music gets compressed to maximize loudness. What you want to do is avoid non integer related sample rate conversion, dropping from 96k to 48k for video work is fine as its a simple average of the previous and next sample, but conversion from 96k to 44.1k will introduce artifacts, and should be avoided.

PaulBrewer

trackmixstudio wrote: »

I record at 24/44.1
I tracked drums at 96k one time to hear it but there was no difference I could hear.
Big let downs are obviously double processing and hard drive space.
Audio drivers and disks have to work twice as hard when tracking and you loose half or your inputs when using an adat card.
Also if you want to record at high sample rates, 88.2 to 44.1 is an easier conversion for obvious reasons. 48 is not a good idea unless you are doing sound for film which is 48 default. The conversion from 48 to 44.1 for music is pointless and bad for your sound.

I remember reading somewhere (maybe gearslutz) that most commercially released records are done at 44.1.

No difference at all ? I find that hard to believe from what I've heard.

I agree with the processing /space point , though I guess space is hardly an issue these days.

Your points regarding 88.2 etc are valid. I could have picked the 44.1 /88.2 combo. I take your point re 48 to 44.1 too excepting that a lot of mastering houses use Analogue gear with a play rig and a record rig so the higher the quality of the master the better the 16bit 44.1 kHz finished product.

MilanPan!c

PaulBrewer wrote: »

No difference at all ? I find that hard to believe from what I've heard.

I agree with the processing /space point , though I guess space is hardly an issue these days.

Your points regarding 88.2 etc are valid. I could have picked the 44.1 /88.2 combo. I take your point re 48 to 44.1 too excepting that a lot of mastering houses use Analogue gear with a play rig and a record rig so the higher the quality of the master the better the 16bit 44.1 kHz finished product.

Bingo

Quiggers

my point of view is from a keen hobbyist, post the track on sound cloud, most of us don't use mastering houses. and people earlier in the thread mentioned isotope and other software based sample rate converters, analogue obviously negates sample rate problems
provided you've got good DA and AD with low noise levels.
obviously if its a paying client for a comercial release either record at the highest bit and sample rates or bite the bullet and pay for tape

studiorat

We did an orchestral recording a few years ago on multi-track at 88.2 and converted the session down to 44.1 afterwards.

There was a big difference between the 88.2 version and the 44.1 version, in terms of the stereo image and how solid the whole mix sounded. For drums and vocals I guess the higher sample rate is a good thing. I find 88.2 much kinder to vocals when it comes to siblance. Though electric guitars etc don't go up that high; the amp limits their range, so I don't imagine thats much of an issue.

I also did some piano recordings at 192kHz more as an experiment than anything, the main thing I noticed was those recordings ate harddrives at an alarming rate!

I guess the difference was in the A2D conversion because the filters were much higher up and in the summing which I'd imagine would be more accurate at the higher sample rate. We used the digidesign conversion algorithm as usual.

Slightly related I wonder does it make much difference from 88.2 to 44.1 or 96kHz to 44.1? Does the conversion process interpolate or just snip out the extra samples?

madtheory

The "it's destined for mp3 so xx is good enough" argument is a fallacy. Why would one ever compromise on sound quality? Plus mp3 will not be here forever, but your music probably will. Best to keep the quality as high as possible (within reason) if only for the archive. A good master can end up lucrative for re releases, anthologies etc.

I think the sound quality depends more on the converter than on the sampling rate. Some are better than others. But I really need to do more tests with the higher rates...

SeanHurley

madtheory wrote: »

I think the sound quality depends more on the converter than on the sampling rate. Some are better than others. But I really need to do more tests with the higher rates...

I would agree with this, the difference between 48k on my Digi 002 and 48k on the Digi 002 with my Apogee Rosetta 800 is light and shade. I would goes as far as to say that using the apogee at 48k sounds as good as using the Digi at higher rates. I will qualify all this with the fact that I havent done exhaustive testing in this area just a bit of A/Bing when I got the Rosetta.

I personally would like to track and mix at 96k or 88.2k but I am unable to do this with the Digi 002/Rosetta 800 as the Rosetta 800 doesn't support SPDIF (only format Digi 002 accepts 96k over the optical inputs) and the 002 doesn't support SMUX.

woodsdenis

studiorat wrote: »

We did an orchestral recording a few years ago on multi-track at 88.2 and converted the session down to 44.1 afterwards.

There was a big difference between the 88.2 version and the 44.1 version, in terms of the stereo image and how solid the whole mix sounded. For drums and vocals I guess the higher sample rate is a good thing. I find 88.2 much kinder to vocals when it comes to siblance. Though electric guitars etc don't go up that high; the amp limits their range, so I don't imagine thats much of an issue.

I also did some piano recordings at 192kHz more as an experiment than anything, the main thing I noticed was those recordings ate harddrives at an alarming rate!

I guess the difference was in the A2D conversion because the filters were much higher up and in the summing which I'd imagine would be more accurate at the higher sample rate. We used the digidesign conversion algorithm as usual.

Slightly related I wonder does it make much difference from 88.2 to 44.1 or 96kHz to 44.1? Does the conversion process interpolate or just snip out the extra samples?

I agree with the Rat on this one. Any multi mike acoustic stuff defo sounds better at higher sample rates, would I bother for anything else, I dont think so.
Hard drive space might not be an issue but you half the amount of tracks coming off each drive too. On a 64 track session you would have 4 drives going if you follow Digi's guidelines. Either way its totally dependent on the source material whether its applicable or not. My gut would tell me if your going to go 96 you should do it for the whole session. You could of course have 2 Protools rigs and use the Digi Satellite having different sample rates on each rig.:eek:

Anima

Funny I was reading a manual today by one of the head guys of the CSound language. He said this about audio quality and sampling rates.

There is no audible difference in quality between 88,200 and 96,000 samples per second, but 88,200 can translated to CD quality by direct down-sampling, whereas 96,000 requires fancy filtering and lots of time.

Makes sense.

madtheory

...but Barbabatch can convert any rate to any other, and quickly... basically you just upsample to a rate that is an even multiple of the desired rate, and then downsample. Big numbers, but that's trivial these days.

Anima

Well its taken out of context really. He says "lots of time" in relative terms as well.

What you described is upsampling/downsampling by a rational fraction, which is more costly than by an integer. So why bother go to 96khz when you can go to 88.2khz for the same quality but less work, thats basically his argument.

The wiki article explains it well, http://en.wikipedia.org/wiki/Upsampling

Neurojazz

Why not record your source at 96 and dither down to 48k... sort of compromise.

Welease

Isn't the final mix, and what your audience cares about all that matters?

Sometimes these discussions seem like a turner prize discussion (art for the sake of art) It's largely irrelevant to 99.9999999999999% of the people listening to the track with the low level consumer level equipment they possess.

It's an interesting technical discussion, but my gut feeling is the outcome will go unnoticed by everyone bar the person making that particular decision.. can anyone here pick a known track and state (on listening) alone which rate is was recorded at?

Neurojazz

Welease wrote: »

Isn't the final mix, and what your audience cares about all that matters?

Sometimes these discussions seem like a turner prize discussion (art for the sake of art) It's largely irrelevant to 99.9999999999999% of the people listening to the track with the low level consumer level equipment they possess.

It's an interesting technical discussion, but my gut feeling is the outcome will go unnoticed by everyone bar the person making that particular decision.. can anyone here pick a known track and state (on listening) alone which rate is was recorded at?

It's a huge difference to the top end in reverbs - especially when coming down in quality - not an 'nth'

As someone said earlier in the thread - the air in the mix... it's a big obvious thing when you can hear it or a/b to see the difference.

MilanPan!c

Neurojazz wrote: »

It's a huge difference to the top end in reverbs - especially when coming down in quality - not an 'nth'

As someone said earlier in the thread - the air in the mix... it's a big obvious thing when you can hear it or a/b to see the difference.

Agreed.

I have heard dozens on 24/96 recordings... There's a big difference in many circumstances, but not always... If you have the resources and the right client, it's def worth experimenting.

Welease

Neurojazz wrote: »

It's a huge difference to the top end in reverbs - especially when coming down in quality - not an 'nth'

As someone said earlier in the thread - the air in the mix... it's a big obvious thing when you can hear it or a/b to see the difference.

But if it was that obvious then everyone would do it... And A/B'ing on pro equipment can show a difference (if one exists).. but when it's mixed down and the bit rate is reduced down, then replayed on consumer equipment.. is the obvious difference obvious anymore?

(don't get me wrong, I am just pushing the pro vs. consumer application question.... )

MilanPan!c

Welease wrote: »

But if it was that obvious then everyone would do it... And A/B'ing on pro equipment can show a difference (if one exists).. but when it's mixed down and the bit rate is reduced down, then replayed on consumer equipment.. is the obvious difference obvious anymore?

(don't get me wrong, I am just pushing the pro vs. consumer application question.... )

In short, yes it def can be!

Welease

MilanPan!c wrote: »

In short, yes it def can be!

but my point is to whom... you might.. can the intended consumer hear it?
If they can't (or don't care) then its art for arts sake..

If they can.. they why is this discussion even happening.. surely everyone would record at 96?

Neurojazz

Find out about dithering, experiment then understand - let your own ears hear the effect and reason

Welease

Neurojazz wrote: »

Find out about dithering, experiment then understand - let your own ears hear the effect and reason

I shall hehe

I'm just posing the question, that if it is that clear and that obvious (and i don't belive it is for the average iTunes downloader), then why would Paul with his wealth of experience even be asking the question..

I am (as a novice) not in a position to argue technically with you (or most of the folks on this board)... but I can push the logic that if it was that clear cut the question would not have been asked and there would have been no varying opinions

Edit - Maybe this makes it clearer where I am coming from.. I was on a course last week.. the tutor was explaining room sounds and waves.. he clapped and moved to various parts of the room showing the difference.. the difference was obvious.. as I looked at the few other students there, most seemed perplexed.. they didnt seem to actually listen to the detail of the sound.. hence my (and i am completely open to being wrong ) opinion that the khz/bit level details matter to those who "listen" (a minority).. but the vast majority of people just "hear"

Double edit - and the "consumer" to who I was referring, would not honestly be able to tell the difference between an 8/16/24 bit signal.. They will hear an old Motown/Prince/Leona Lewis track the same.

Neurojazz

Maybe someone knows of a simple analogy, but basically the better the source is (and the finer the sound) - the better the end result will reflect that original when dithered. It's like a process of averaging/shading (if you think of painting) - and if your source is crap then there will be harsh differences in that averaging.

I was shocked at the results.

MilanPan!c

Welease wrote: »

I shall hehe

I'm just posing the question, that if it is that clear and that obvious (and i don't belive it is for the average iTunes downloader), then why would Paul with his wealth of experience even be asking the question..

I am (as a novice) not in a position to argue technically with you (or most of the folks on this board)... but I can push the logic that if it was that clear cut the question would not have been asked and there would have been no varying opinions

Edit - Maybe this makes it clearer where I am coming from.. I was on a course last week.. the tutor was explaining room sounds and waves.. he clapped and moved to various parts of the room showing the difference.. the difference was obvious.. as I looked at the few other students there, most seemed perplexed.. they didnt seem to actually listen to the detail of the sound.. hence my (and i am completely open to being wrong ) opinion that the khz/bit level details matter to those who "listen" (a minority).. but the vast majority of people just "hear"

Double edit - and the "consumer" to who I was referring, would not honestly be able to tell the difference between an 8/16/24 bit signal.. They will hear an old Motown/Prince/Leona Lewis track the same.

There's never just one audience, if you can please various groups, do it.

Welease

Neurojazz wrote: »

Maybe someone knows of a simple analogy, but basically the better the source is (and the finer the sound) - the better the end result will reflect that original when dithered. It's like a process of averaging/shading (if you think of painting) - and if your source is crap then there will be harsh differences in that averaging.

I was shocked at the results.

Sorry I edited when you posted this..

I am not argueing on a pure sound academic level.. I am argueing on a consumer level..

Was 96Khz 24 bit available for Motown, Elvis, Sex Pistols, U2... yadda yadda 50's, 60's, 70's, 80's, 90's... Some great records that an engineer may find shocking, but the intended audience could not tell the difference

My wife might at a push spot the difference between a 50/60's tune and a 80's/90's tune.. she definately wouldn't tell me the reverb sounded harsh .. not through an iPod..

Welease

MilanPan!c wrote: »

There's never just one audience, if you can please various groups, do it.

Absolutely.. You guys are right.. I'm just offering some perspective from the average consumer.. the bulk of the work goes unnoticed.. and at the end of the day if it took at 4 disk systems and tons of extra pain.. be aware only sound engineers may notice the difference..

Let me give you a bad analogy... - I run a team of developers.. One was late developing a large amount of code.. seriously late.. When we sat down and tried to understand the problem, he was unhappy because his code was not elegant.. It did was it was supposed to do weeks ago, but from an "art"/"ego" perspective he knew there was a better smarter way of doing it.. he just hadn't cracked it yet.. My problem and the customers problem.. bluntly.. noone gives a sh1t.. it did what it was supposed to do.. (and was perfectly maintainable).. it didnt need an extra month of messing about for exactly the same output.. Sometimes if the end result is the same.. then the quicker/cheaper process is fine. (i just read that back.. in fairness he works on multiple projects.. so he was part assigned )

MilanPan!c

Welease wrote: »

Sorry I edited when you posted this..

I am not argueing on a pure sound academic level.. I am argueing on a consumer level..

Was 96Khz 24 bit available for Motown, Elvis, Sex Pistols, U2... yadda yadda 50's, 60's, 70's, 80's, 90's... Some great records that an engineer may find shocking, but the intended audience could not tell the difference

My wife might at a push spot the difference between a 50/60's tune and a 80's/90's tune.. she definately wouldn't tell me the reverb sounded harsh .. not through an iPod..

I understand this argument, but trust me thousands, more, clamor to hear high quality recordings... And not just audiophiles w oodles of cash, but kids with a desire to experinece, not just listen to, music.

Every effort should be made to make music immersive.

Many, most even, won't know why, but it wont stop them from feeling a bit more connected to songs... It's worth that.

madtheory

Actually it's sample rate conversion not dither... dither is for removing quantizing distortion when the bit depth is reduced...

Reverb does not need to go to 20kHz, far from it; it is reflected sound so it has reduced energy compared to the source, that's one of the big reasons why it sounds like reverb! So I'm wondering why the same reverb algorithm would sound better at 96 than at 44? Something else to try out... we have an SPX2000 here that goes to 96...

Anima wrote: »

for the same quality but less work

But it's not work at all, because it's so fast on a modern computer? Thanks for the wiki, I must look at that when I'm awake...

PaulBrewer

studiorat wrote: »

We did an orchestral recording a few years ago on multi-track at 88.2 and converted the session down to 44.1 afterwards.

There was a big difference between the 88.2 version and the 44.1 version, in terms of the stereo image and how solid the whole mix sounded. For drums and vocals I guess the higher sample rate is a good thing. I find 88.2 much kinder to vocals when it comes to siblance. Though electric guitars etc don't go up that high; the amp limits their range, so I don't imagine thats much of an issue.

I also did some piano recordings at 192kHz more as an experiment than anything, the main thing I noticed was those recordings ate harddrives at an alarming rate!

I guess the difference was in the A2D conversion because the filters were much higher up and in the summing which I'd imagine would be more accurate at the higher sample rate. We used the digidesign conversion algorithm as usual.

Slightly related I wonder does it make much difference from 88.2 to 44.1 or 96kHz to 44.1? Does the conversion process interpolate or just snip out the extra samples?

Thanks Gramps ....

PaulBrewer

woodsdenis wrote: »

You could of course have 2 Protools rigs and use the Digi Satellite having different sample rates on each rig.:eek:

:eek::eek:

PaulBrewer

Neurojazz wrote: »

Why not record your source at 96 and dither down to 48k... sort of compromise.

Sure isn't that what I'm askin ? :mad:

PaulBrewer

Welease wrote: »

but my point is to whom... you might.. can the intended consumer hear it?
If they can't (or don't care) then its art for arts sake..

If they can.. they why is this discussion even happening.. surely everyone would record at 96?

Snooze .......:mad:

PaulBrewer

Welease wrote: »

why would Paul with his wealth of experience .

Wealth of Experience me Whole ! :eek:

There's always more to learn and better ways of doing things ...

madtheory

studiorat wrote: »

Slightly related I wonder does it make much difference from 88.2 to 44.1 or 96kHz to 44.1? Does the conversion process interpolate or just snip out the extra samples?

It's a filter, rather than interpolation, isn't it? I think that's correct, but I'm not 100% certain.

Interesting interview in Resolution mag this month, where a guy is downsampling by dropping all the odd samples, using one of those boxes that split 88.2 into two ADAT cables, and only using one. My initial reaction is EEEK! but he appears to have many happy clients...

MilanPan!c

madtheory wrote: »

It's a filter, rather than interpolation, isn't it? I think that's correct, but I'm not 100% certain.

Interesting interview in Resolution mag this month, where a guy is downsampling by dropping all the odd samples, using one of those boxes that split 88.2 into two ADAT cables, and only using one. My initial reaction is EEEK! but he appears to have many happy clients...

That is CRAZY!