C
chacho313rd
Future music icon
i've been recording in 96 khz, just cuz i can lol. But i really dont hear that much of ah difference after mixing and mastering the track. Any input people's.
deRaNged 4 Phuk'dup
New member
It's arguable, not definitive. But if you're going to CD/MP3 in the end, you'd be better off using 88khz for conversion(in the end the audio will be converted down to 44.1khz for media).
Most engineers i know use 44.1khz 24bit. So if you want a "professional sound" exceeding that wouldn't be the "secret".
Most engineers i know use 44.1khz 24bit. So if you want a "professional sound" exceeding that wouldn't be the "secret".
A
anomalouscode
New member
yeah its not the secret, and to be honest, with stuff like hip hop i only do 88khz cos i can too,... although apparently some specific vst's like nebula will take advantage of higher bit rates in general, if its going out as mp3 or cd,... not sure how much comes through in the end.
on the other hand, i know that the location recording remote guys seem to hear a difference between 96 and anything lower, more shimmer up top,... so when recording acoustic/traditional/classical i do 96khz.
apart from "because i can" which i do too,.... i think its nice to have something that can be transfered to higher fidelity stuff in the future when a new standard comes through,... like the hidef flac stuff going around the net at the moment.
on the other hand, i know that the location recording remote guys seem to hear a difference between 96 and anything lower, more shimmer up top,... so when recording acoustic/traditional/classical i do 96khz.
apart from "because i can" which i do too,.... i think its nice to have something that can be transfered to higher fidelity stuff in the future when a new standard comes through,... like the hidef flac stuff going around the net at the moment.
C
chris carter
New member
There are two benefits to high sample rates:
1) The bigger one is that the filter on your A/D and D/A converters is much higher and shallower so it moves further beyond what you can easily hear. Much of the coloration from a converter comes from this filter. So with a higher sample rate, you essentially hear less of the nyquist filter. Interestingly enough, at the D/A it doesn't really matter if the source was 88.2 to begin with, or 44.1 upsampled to 88.2 (or 96kHz), because it's NOT the amount of data, it IS where the filter is and how steep it is. For this reason, if you mix entirely ITB then you will only get the benefit at the A/D as there is no D/A. Mixing with outboard you'll get both benefits.
2) The mathematical distortion (rounding error) is the same amount regardless of the sample rate, it's just a matter of how much data it's spread across. So at 88.2 the same distortion is spread over twice as much data as 44.1. The result is half the distortion for the same amount of data and equals half the audible distortion. So every single time a math calculation is made (ie. any DSP of any kind) you get half the audible distortion. Bear in mind that we are talking about insanely low levels of distortion so this is not nearly as big of a deal as (1) above.
For anything with high track counts, or that will use a lot of plugins, it's not worth it because the system just can't handle it. For projects with lower track counts that require less plugins, where the system can easily handle the extra hit, then it can obviously work. That said, for most electronic based music it's very questionable whether there's much to be gained. Whereas with an all acoustic project like classical or folk, there is more to be gained.
I generally work at 44.1 unless there's some specific reason to go higher. Every record I've mixed has been at 44.1 (a few at 48, and only one that came in at 88.2). Most pros I know generally work at 44.1. R&B is pretty much always 44.1 (or 48) because there's almost no way for any computer to handle those kinds of track counts at 88.2 or 96.
1) The bigger one is that the filter on your A/D and D/A converters is much higher and shallower so it moves further beyond what you can easily hear. Much of the coloration from a converter comes from this filter. So with a higher sample rate, you essentially hear less of the nyquist filter. Interestingly enough, at the D/A it doesn't really matter if the source was 88.2 to begin with, or 44.1 upsampled to 88.2 (or 96kHz), because it's NOT the amount of data, it IS where the filter is and how steep it is. For this reason, if you mix entirely ITB then you will only get the benefit at the A/D as there is no D/A. Mixing with outboard you'll get both benefits.
2) The mathematical distortion (rounding error) is the same amount regardless of the sample rate, it's just a matter of how much data it's spread across. So at 88.2 the same distortion is spread over twice as much data as 44.1. The result is half the distortion for the same amount of data and equals half the audible distortion. So every single time a math calculation is made (ie. any DSP of any kind) you get half the audible distortion. Bear in mind that we are talking about insanely low levels of distortion so this is not nearly as big of a deal as (1) above.
For anything with high track counts, or that will use a lot of plugins, it's not worth it because the system just can't handle it. For projects with lower track counts that require less plugins, where the system can easily handle the extra hit, then it can obviously work. That said, for most electronic based music it's very questionable whether there's much to be gained. Whereas with an all acoustic project like classical or folk, there is more to be gained.
I generally work at 44.1 unless there's some specific reason to go higher. Every record I've mixed has been at 44.1 (a few at 48, and only one that came in at 88.2). Most pros I know generally work at 44.1. R&B is pretty much always 44.1 (or 48) because there's almost no way for any computer to handle those kinds of track counts at 88.2 or 96.
H
HakimCallier
Guest
If you have to ask that question then it is not worth it for you. You do not have the resources. And your setup was not not designed with recording at high sample rates in mind.
When the time comes you'll know because you won't have to ask. The compromise won't be significant.
EDIT: I failed to read/understand OPs statement "i've been recording in 96 khz, just cuz i can" so disregard the previously written post above.
When the time comes you'll know because you won't have to ask. The compromise won't be significant.
EDIT: I failed to read/understand OPs statement "i've been recording in 96 khz, just cuz i can" so disregard the previously written post above.
Last edited by a moderator:
moses
hardliner
As mentioned above, there's absolutely no gain if you're not into the technical details.
In fact, there is a huge benefit (more below). But you need a tight workflow and especially a high-end sample-rate converter (which is not that easy to get). One simple mistake and you'll do more harm than good.
There are at least four good reasons for using higher sample-rates under certain circumstances:
1. The ADA conversion (chris mentioned it already). Especially in the case of 44.1kHz, the nyquist filter only has a quarter octave to filter 96dB (for 16bit quality) to 144dB (24bit) away! This is an extremely steep and most probably bad sounding processing. No matter if it's about recording or playback, higher resolutions tend to deliver better results because they substantially relax the nyquist filter.
I made a few graphs to visualize what happens in a AD/DA converter. Every frequency above Nyquist must be attenuated by at least ~96dB to deliver a dynamic range comparable to 16 bit. Here's how this looks like at different rates:
Pic 1: As you can see, a sample-rate of 44.1kHz asks for a crazy steep filter (a 351 taps FIR filter was required to create this response).
Pic 2: At 48kHz, the filter has more room to work, which is likely to introduce less negative side-effects (only 91 taps for this filter).
Pic 3: At 96kHz, the filter has an very easy job (only 19 taps for this filter).
2. Processing quality. Now this is where the "wow" is. I've already written countless posts about the topic, so a search might help. It's not that obvious, so I need a few lines...
The bandwidth of a digital system is clearly limited and defined by the sample-rate. The nyquist frequency is exactly about half the sample-rate (google for nyquist if this is a new concept for you, it's digital audio basics).
So, a samplerate of 40kHz results in an available bandwidth of 20kHz. But, there's an important detail here. While signals exceeding the bandwidth in an analogue system are simply attenuated, something radically different happens in its digital counter-part. The exceeding frequencies are instead mirrored back from the nyquist point into the audible spectrum (inverted of course).
Now, all non-linear processes are know to create new frequencies. Non-Linear processors are all kind of dynamics processing effects like compressors, saturators and clippers. As well as certain special effects like ring modulation, frequency modulation, gain modulation, exciters, transient designers, all kind of pitch/time stretcher and more. Most modern EQs are also non-linear because they all work with saturators to create pseudo analogue "color". Most of these new frequencies are new harmonics based on (and placed above) the signal (in fact, these are even and odd multiples of the signal's frequency).
This simply means that a say 8Khz tone being slightly saturated will create harmonics at 16kHz (the 2nd harmonic), 24kHz (the 3rd), 32kHz (the 4th) and so on, decaying to infinity.
But the bandwidth of the 40Khz samplerate system won't be able to carry the 3rd and 4th harmonic! These two harmonics mirror at 20Khz. The 3rd 4Khz above the limit, so its alias will land at 16kHz. The 4rth is 12Khz above the limit, so its alias will land at 8Khz. Of course, these aliases are now harmonically completely unrelated and make the saturation sound ugly. This is just an example, real life is much worse. BTW this is the main reason why high end analogue EQs and compressors are still unmatched. They don't have the unnatural bandwidth behaviour typical digital systems have.
A simple sine-sweep, a cheap saturator and a FFT visualizer will demonstrate it clearly (and also make the aliasing VERY audible).
Or, just watch a helicopter flight on TV, the rotation is faster than 24 fps (which is 12Hz), so it creates aliases that seem to move backward.
The benefits are huge in these cases. Not only do non-linear effects sound much much better at higher rates, even the cheapest synths deliver way better results at higher rates. But again, no proper sample rate convertion for the final jump to 44.1kHz and you fvck up everything.
3. Media requirements. Some DVDs and all Blue-Ray discs ask for a 24bit/96kHz resolution. You can use a lower sample-rate, but keep in mind how cheap DVD players are - their DA converters are of extremely low quality and built to work best with high resolution material (i.e. they use cost-efficient nyquist filters). See point 1.
4. Future developments. It may be useful to have a high resolution mix in your archive, even if the current technical consumer standard is lower. A future standard my ask for it.
Some AES papers mention a ~100kHz bandwidth requirement for impeccable high end digital audio processing(!). This is very interesting since the analogue world also has that rough "you need 100Khz of bandwidth for high end processing". This would be a sample-rate around 192kHz. Some radio processors like the Orban tools even oversample 16x for their clipping algorithm.
Ok, enough tech talk. Just work in 44.1kHz like you did before, but at least try it out and render your project at a higher rate. Reason for example allows to change the sample-rate on the fly, the before/after is shocking, really. Convert it down to 44.1kHz with a decent SRC later. You will hear a difference, it will sound much rounder, deeper, crispier and smoother at the same time.
Don't just think about your ears, also give the math some room to work. Even the cheapest EQs and synths start to excel when being oversampled (i.e. "using more samples the ear needs").
High end quality SRC recommendations:
Isotope 64bit SRC (delivered as bonus bundle with soundforge). Very nice, really. The Weiss Saracon is technically slightly better, but more expensive (Still very fair ~850$, I mean it's a Weiss product
).
If both above are too expressive or hard to find, check out Voxengo's R8Brain free, which is also of very high quality.
Finally, never ever trust the sample rate converters included in your DAW/Audio Editors. No matter if it's logic, PT, Cubase, Sonar, Wavelab, Soundforge, all have really bad SRCs built in.
Overview here: http://src.infinitewave.ca/
In fact, there is a huge benefit (more below). But you need a tight workflow and especially a high-end sample-rate converter (which is not that easy to get). One simple mistake and you'll do more harm than good.
There are at least four good reasons for using higher sample-rates under certain circumstances:
1. The ADA conversion (chris mentioned it already). Especially in the case of 44.1kHz, the nyquist filter only has a quarter octave to filter 96dB (for 16bit quality) to 144dB (24bit) away! This is an extremely steep and most probably bad sounding processing. No matter if it's about recording or playback, higher resolutions tend to deliver better results because they substantially relax the nyquist filter.
I made a few graphs to visualize what happens in a AD/DA converter. Every frequency above Nyquist must be attenuated by at least ~96dB to deliver a dynamic range comparable to 16 bit. Here's how this looks like at different rates:
Pic 1: As you can see, a sample-rate of 44.1kHz asks for a crazy steep filter (a 351 taps FIR filter was required to create this response).
Pic 2: At 48kHz, the filter has more room to work, which is likely to introduce less negative side-effects (only 91 taps for this filter).
Pic 3: At 96kHz, the filter has an very easy job (only 19 taps for this filter).
2. Processing quality. Now this is where the "wow" is. I've already written countless posts about the topic, so a search might help. It's not that obvious, so I need a few lines...
The bandwidth of a digital system is clearly limited and defined by the sample-rate. The nyquist frequency is exactly about half the sample-rate (google for nyquist if this is a new concept for you, it's digital audio basics).
So, a samplerate of 40kHz results in an available bandwidth of 20kHz. But, there's an important detail here. While signals exceeding the bandwidth in an analogue system are simply attenuated, something radically different happens in its digital counter-part. The exceeding frequencies are instead mirrored back from the nyquist point into the audible spectrum (inverted of course).
Now, all non-linear processes are know to create new frequencies. Non-Linear processors are all kind of dynamics processing effects like compressors, saturators and clippers. As well as certain special effects like ring modulation, frequency modulation, gain modulation, exciters, transient designers, all kind of pitch/time stretcher and more. Most modern EQs are also non-linear because they all work with saturators to create pseudo analogue "color". Most of these new frequencies are new harmonics based on (and placed above) the signal (in fact, these are even and odd multiples of the signal's frequency).
This simply means that a say 8Khz tone being slightly saturated will create harmonics at 16kHz (the 2nd harmonic), 24kHz (the 3rd), 32kHz (the 4th) and so on, decaying to infinity.
But the bandwidth of the 40Khz samplerate system won't be able to carry the 3rd and 4th harmonic! These two harmonics mirror at 20Khz. The 3rd 4Khz above the limit, so its alias will land at 16kHz. The 4rth is 12Khz above the limit, so its alias will land at 8Khz. Of course, these aliases are now harmonically completely unrelated and make the saturation sound ugly. This is just an example, real life is much worse. BTW this is the main reason why high end analogue EQs and compressors are still unmatched. They don't have the unnatural bandwidth behaviour typical digital systems have.
A simple sine-sweep, a cheap saturator and a FFT visualizer will demonstrate it clearly (and also make the aliasing VERY audible).
Or, just watch a helicopter flight on TV, the rotation is faster than 24 fps (which is 12Hz), so it creates aliases that seem to move backward.
The benefits are huge in these cases. Not only do non-linear effects sound much much better at higher rates, even the cheapest synths deliver way better results at higher rates. But again, no proper sample rate convertion for the final jump to 44.1kHz and you fvck up everything.
3. Media requirements. Some DVDs and all Blue-Ray discs ask for a 24bit/96kHz resolution. You can use a lower sample-rate, but keep in mind how cheap DVD players are - their DA converters are of extremely low quality and built to work best with high resolution material (i.e. they use cost-efficient nyquist filters). See point 1.
4. Future developments. It may be useful to have a high resolution mix in your archive, even if the current technical consumer standard is lower. A future standard my ask for it.
Some AES papers mention a ~100kHz bandwidth requirement for impeccable high end digital audio processing(!). This is very interesting since the analogue world also has that rough "you need 100Khz of bandwidth for high end processing". This would be a sample-rate around 192kHz. Some radio processors like the Orban tools even oversample 16x for their clipping algorithm.
Ok, enough tech talk. Just work in 44.1kHz like you did before, but at least try it out and render your project at a higher rate. Reason for example allows to change the sample-rate on the fly, the before/after is shocking, really. Convert it down to 44.1kHz with a decent SRC later. You will hear a difference, it will sound much rounder, deeper, crispier and smoother at the same time.
Don't just think about your ears, also give the math some room to work. Even the cheapest EQs and synths start to excel when being oversampled (i.e. "using more samples the ear needs").
High end quality SRC recommendations:
Isotope 64bit SRC (delivered as bonus bundle with soundforge). Very nice, really. The Weiss Saracon is technically slightly better, but more expensive (Still very fair ~850$, I mean it's a Weiss product
).If both above are too expressive or hard to find, check out Voxengo's R8Brain free, which is also of very high quality.
Finally, never ever trust the sample rate converters included in your DAW/Audio Editors. No matter if it's logic, PT, Cubase, Sonar, Wavelab, Soundforge, all have really bad SRCs built in.
Overview here: http://src.infinitewave.ca/
Last edited:
H
HakimCallier
Guest
This is a really good thread. I'm giving it stars.
moses
hardliner
Sticky for a while
Bryst
New member
This thread rocks. I find all this stuff incredibly interesting even though I don't understand 95% of it yet...
I do pretty much 100% electronic instruments with my recordings. (Occasionally an acoustic guitar, but almost never) The vocals are the only thing recorded acoustically. I record everything at 48khz, and I usually keep it there even when I go to CD. I feel like there's no reason to go down to 44.1 for CD's if you dont have to... don't know if that's right or not...? (Obviously I compress to make MP3s for internet etc.)
I do this because 48k is the highest my system allows, and I never have CPU issues at this rate.
Like I said, I know almost nothing about this.
Think I'm doing the right thing for my system?
I do pretty much 100% electronic instruments with my recordings. (Occasionally an acoustic guitar, but almost never) The vocals are the only thing recorded acoustically. I record everything at 48khz, and I usually keep it there even when I go to CD. I feel like there's no reason to go down to 44.1 for CD's if you dont have to... don't know if that's right or not...? (Obviously I compress to make MP3s for internet etc.)
I do this because 48k is the highest my system allows, and I never have CPU issues at this rate.
Like I said, I know almost nothing about this.
Think I'm doing the right thing for my system?
deRaNged 4 Phuk'dup
New member
Wow.Great frikkin posts in this thread. Brought a tear to my eye, lol. 
Dre Major
Mix Engineer
Moses couldn't have chosen a better screen name.
C
chacho313rd
Future music icon
thank you for the input guys it was very informative, thanks for taking the time to comment.
---------- Post added at 04:53 AM ---------- Previous post was at 04:27 AM ----------
thank to all very informative i appreciate the feedback. an i really was just curious.
---------- Post added at 04:53 AM ---------- Previous post was at 04:27 AM ----------
thank to all very informative i appreciate the feedback. an i really was just curious.
Last edited:
krushing
Active member
I'm nominating moses' post as the best post EVER on FP 
moses
hardliner
Lol, thanks man
Dre Major
Mix Engineer
I'm nominating moses' post as the best post EVER on FP
Seconded.
Moses I implore you to never tire of FP and leave.
FP needs people like you.
Last edited:
EddyCutz
New member
Wow, that was a great lesson for me. I just ordered my new 96khz audio card and was wondering myself. I love the nerdy technical details (being a software developer in the day) so it made perfect sense to me, thanks.
C
Cool Sounds
New member
Maybe not...
I once attended an experiment where producers and sound engineers listened to recording made in 96 , in 88 and in 44 KH. They had to say which is which...
The answers they gave were not correct. It is very dodgy, I doubt if people hear it. I would leave the 96 alone, it's not worth the time and effort. I would invest my time and cpu in something else.
But if you can hear, go for it.
I once attended an experiment where producers and sound engineers listened to recording made in 96 , in 88 and in 44 KH. They had to say which is which...
The answers they gave were not correct. It is very dodgy, I doubt if people hear it. I would leave the 96 alone, it's not worth the time and effort. I would invest my time and cpu in something else.
But if you can hear, go for it.
Synapsis
The Mepsyah's Dipsyples
So if I am recording my Rolands at 96 khz the sound is better? I have the option to put the output from my v synth to 96 khz and my X8 will only have 44.1 khz ,if I am recording at 96 khz on my pc and the spdif output from my Fantom is 44.1 will it automatically upsample the rate?? I am about to start uni soon to sort this shit right out
moses
hardliner
Please read carefully.
The general consensus is: Do not play around with the sample-rate if you're unsure why you should do it (i.e. do not understand the theory behind). Better stick to your target rate (44.1kHz in most case).
Also, we're not really talking about the playback quality of the carrier. It is clear that in most cases a 44.1kHz playback will sound exactly like a 96kHz playback. The big improvement is oversampled processing not oversampled playback.
The general consensus is: Do not play around with the sample-rate if you're unsure why you should do it (i.e. do not understand the theory behind). Better stick to your target rate (44.1kHz in most case).
Also, we're not really talking about the playback quality of the carrier. It is clear that in most cases a 44.1kHz playback will sound exactly like a 96kHz playback. The big improvement is oversampled processing not oversampled playback.
Last edited:
Synapsis
The Mepsyah's Dipsyples
So....recording 96khz, then processing your mixdown through some waves processing tools, L3 etc, or whatever compressor/Limiter processing you use, the end result from bouncing the final down to 44.1/16 bit for cd will be sharper and a bit more"analogue" for the records what is your opinions on T racks mastering suite?