3

I've been brought on for editing audio on a project (the audio isn't going to a video at all it is just for the audience to listen). The recording engineer has recorded this audio at 48 kHz, even though it will never be used with video.

His point is the over sampling calculation; where audio going into the DAW needs to be twice frequency range of the incoming audio to avoid over tones/get a more accurate representation of the digital audio.

Now he used an SM57 whose frequency range is 40 Hz to 15 kHz so it wouldn't really make a difference because in the oversampling calculation the double of 15 khz is 30 kHz so the oversampling would fit perfectly in the 44.1 khz format.

Now remember his pro tools session is in 48 kHz format (therefore recording at 48 kHz) so basically my question is if it is ABSOLUTELY necessary to record at 48 kHz that's not for film where the final audio will be bounced to 44.1 format anyway?

  • 2
    Not at all - but now you're stuck at 48k, stay at that til the very final mix output, then downsample it. – Tetsujin Nov 20 '14 at 10:38
  • Ja i know hey, it is so dumb plus he is doing the mixdown so I have to give him a 48 session – JM V Nov 20 '14 at 12:36
4

As a possible side bar to the excellent answer provided by JCPedroza above,

If human ears are the only listeners (i.e., signals between 20 and 20,000 Hz) then it is never* absolutely necessary to record at 48kHz.

[H]e used an SM57 whose frequency range is 40 Hz to 15 kHz so it wouldn't really make a difference because in the oversampling calculation the double of 15 kHz is 30 kHz so the oversampling would fit perfectly in the 44.1 kHz format.

Exactly. However, just to be clear, doubling the max frequency gives the minimum rate not the "oversampling" rate....

Nyquist Sampling Theory: A sampled waveforms contains ALL the information without any distortions, when the sampling rate exceeds twice the highest frequency contained by the sampled waveform.
...
It is important to realize that the end result yields a waveform where the values are correct, not just at sample times but at all times.
- source

It's important to remember that each sample contains only single value that represents the amplitude of the signal being sampled - the value does not contain any frequency information. See Sampling fallacies and misconceptions


Sources and links

* There are always hypothetical exceptions to "never." However, hypotheticals are only real if documented by experiment (in which case, of course, they are no longer hypothetical).

  • So are you saying for this case, the SM57 he used and the fact that this in this case he was was only recording voice overs and the end product is only voice overs that would be bounced as 44.1, he was correct in recording the voice overs at 48 kHz? or should he have recorded at 44.1 Khz. Because all the wonderful comments are applicable but I really would like to know because there is a bone of contention, – JM V Nov 24 '14 at 10:09
  • 1
    @JMV If they are just voiceovers then recording at 22500 would have captured all of the data. But there is no harm in recording them at 48000. In fact that's what I would have recorded at, but only because that's the rate I typically use for everything. I do so purely for convenience and for file/stem/clip consistency. I suspect that your recording engineer, in the absence of specific instructions to the contrary, used 48000 for the same reasons. If, however, he is saying that a voiceover recorded at 48 is "better" or has more "resolution" than one recorded at 44.1 he is mistaken. – Rusty Nov 24 '14 at 11:17
5

is ABSOLUTELY necessary to record at 48 kHz that's not for film where the final audio will be bounced to 44.1 format anyway?

We know that at 44.1kHz we can accurately record and playback the frequencies that live in the human hearing threshold, so oversampling might seem an overkill. Most of the time this is the case, but some scenarios can benefit from higher sampling rates.

Ask yourself these questions:

Are you recording and/or working with ultrasonic frequencies? (above 20kHz)

You already knew this if you know about the Nyquist Frequency and Nyquist Rate.

A kind of exotic scenario, but it happens. If you are recording a sound that has frequencies (fundamental or overtones) above 20kHz, you need a higher sampling rate, which must be double the highest frequency you'll sample.

Even if you'll bounce at 44.1kHz, you might want to record ultrasonic sounds and then lower the pitch to a frequency range that 44.1kHz can handle.

Will you use a process that benefits from working with oversampled audio?

My DSP knowledge is very poor, so I need confirmation on this one. Is there a process that can benefit from working with oversampled audio? I don't know of any.

I know there are processes that benefit from working at higher sample rates (more on that later), but I'm not sure about the benefits of processing oversampled audio (which is not the same).

My first thought was about processes that use granular synthesis (like elastic audio). It at first makes sense that these would benefit from more resolution for the grains, making time and frequency manipulation smoother, but it might not be the case.

Anyway, if you find yourself using a process that can benefit from working with oversampled audio (if there's not such thing right now, there might be in the future, who knows), then that's a good reason to record at higher sample rates.

Will you use a process that benefits from working at higher sample rates?

Here's what I do know: some processes benefit from working at higher sampling rates. Some processes do the oversampling internally, though. To justify working at higher sample rates here, you would need to be working with a process that benefits from oversampling, but doesn't do the oversampling itself internally.

From article The Science of Sample Rates:

When you go beyond the mere recording and playback of sound and into the world of digital signal processing, it becomes clear that higher sampling rates actually can help. But the solution might be a different one than you’d expect.

When it comes to some non-linear audio processors like a super-fast compressor, a saturator, a super-high-frequency EQ, or a vintage synthesizer emulation, oversampling can be a major benefit. This in and of itself might seem like a great excuse to immediately jump up to 88.2 kHz or higher.

But not so fast: most plugin designers, knowing this full well, have written oversampling into their code. Even in a 44.1kHz session, plugins that benefit from oversampling automatically increase their internal sampling rate. To gain the full benefits of this, it’s important to note that the audio doesn’t have to be recorded at this higher sample rate, it’s just the processing that must happen at the higher rate.

So unless you are using plugins that have taken shortcuts and neglected to include oversampling in their code, then converting an entire audio session to a higher rate would make your mix take up more processing power without adding any sonic benefit.

Very few of the latter still exist. And thankfully so, because oversampling has led to huge improvements in the quality of digital processing. Finally, after decades of people trying, there are actually some software compressors that I like. A lot.

So, in short, it's only absolutely necessary to oversample if you plan to record and work with ultrasonic frequencies. It's not necessary but can improve your results if you find a process that can benefit from either oversampled audio or working at higher sampling rates (that doesn't already perform the oversampling itself internally).

These 3 are the only oversampling justifications I can think of in a recording, mixing, mastering, production, and processing context. Can anyone think of more?


Interesting articles on the subject:

  • Guys - take this argument to chat. It's not for comments. – Rory Alsop Dec 2 '14 at 7:52
-5

Which is the best choice between 44.1KHz and 48KHz is in my opinion not a question of video or audio - it's all a question if the material will be released in any way on CD or not.

44.1KHz samplerate was never a good idea, but originally there was no choice. The rate was chosen only because the modified video tapedecks used by the time for mastering the audio (for exact facts on this, here's a link: http://www.cs.columbia.edu/~hgs/audio/44.1.html), simply couldn't be set any other way the way they where constructed, and the samplerate had to be over 40KHz at any cost for it to be actual full-range.

Nowadays though, and even in the early 80's, that wasn't a problem anymore. I've never even seen those original tape-machines myself, not even on pictures, and haven't even ever met anyone who has as far as I know, so the old DAT format didn't have to obey that anymore. Unless it wanted to, of course. So they did the obvious thing - raising the sampelrate to full 6x8 spectral resolution - 48KHz.

CD is a dying format. Maybe not the disc in itself, I reckon it might be around for quite a while more (at least I hope it will), but the format itself. Everything else, period, supports at least 48KHz - from MP3's, videos, TV, even cellphones, so there is absolutely no reason to let that format live to see another day if you ask me.

You never said what the audio is intended for, but as you mentioned "audience", I interpret it as it might have something to do with public playback in some sort, and if that is the case, and the audio will be played from like a laptop, iPod, iRiver or something, and not a physical CD, then I say the technician did the right thing. Let 44.1 die in silence. If, however, his instructions indeed was to mix a CD, then he has made a really unnecessary error. The material should always be mixed and preferably recorded in the same samplerate, with possible full multiples (like for 48, mixing in 96 or 192KHz is a really good thing), as the intended medium it is intended to be played on, and though it's no big deal, depending on the software, to re-sample 48 to 44, it shouldn't be necessary. As I said, all complications not truly vital for some reason is bad complications.

[EDIT: I've updated the post as I realized I accidentally wrote dividers of 8, but in spectral resolution it's a matter of strict multiples.]

  • This comment thread wasn't going anywhere. Christian - I've removed the paragraph which made no sense (as 44.1k is very divisible by 2, and even 4) and deleted all the comments. – Rory Alsop Dec 2 '14 at 7:48
-7

Yes, It is necessary. The problem with the CD format (44.1kHz, 16bit) is twofold. On one hand, you have the problem with 16bit: any sound that has less than full dB is to all intents and purposes not 16bit, as any sound recording only has that full 16 bit fidelity at the peaks. The problem with 44.kHz isn't that you can't play the full frequencies, but that the sound waves can not be presented in the exact place in time where they actually occur or in the exact shape. This is most obvious to our ears when you are listening to a pure acoustic solo instrument or voice with the natural reverb of a room, but it is also a problem when instruments are fighting for the same frequency space. Some phantom frequencies might occur that wouldn't otherwise, and vice versa.

Those are the problems if you are working on an acoustic project where you are not planning on using any effects or eq. As soon as you go into reverb, eq, compression, saturation, reverb, etc., you have the much more serious problem of mathematical rounding errors: the lower resolution, coupled with a smattering of effects, sound sources mixed lower into the mix and mastering, sound bites moved right or left, equal many, many mathematical equations whith huge, compounded rounding errors. With 44.1kHz, your sound is only 44.1kHz as long as you don't do anything to it. 48kHz still isn't enough to avoid those rounding errors, but it gives you a bit of headroom before the overall sound starts to degenerate below 44.1kHz. It's something people who have grown up with CDs and MP3s tend to be unable to hear. And in modern electronic music production, the finished album is already way below that minimal sound quality standard. If you have listened to a lot of high definition vinyl records, live acoustic concerts or have mixed a lot of live concerts, or if you work in the classical, world or jazz world, you will have an ear more attuned to the loss of "warmth", "liveliness", room and all the other things people tend to call it.

And before you say: "vinyl records aren't as high definition as CDs"; Yes they are. In the lower dBs, they have a much, much, much, much higher definition, which is most often called "warmth" or "liveliness" or detail. Sound mixers wrongly try to emulate this by saturating tracks, but this only emulates the part of of the vinyl record that is inferior to the CD. CDs have more fidelity around the peaks and quick transients than LPs.

A useful way to think of it, is to imagine trying to take a photo of a tall building from above with a lens that has a relatively narrow focus. You can't get the whole building into focus, so you have to decide whether the top or the bottom of the building is sharp and in focus. With the CD, 44.1kHz 16bit, the top of the building is in focus while the bottom is fuzzy. with the vinyl record, the top of the building is fuzzy, but the bottom is very accurate. when you saturate a digital track, it's like the top and the bottom are both fuzzy, with just the middle of the building being in focus. If you want the whole building to be in focus, you need a much higher resolution, probably 192kHz 32bit and upwards. 320 bit MP3 in this scenario is just as clear at the top of the building as the CD, but the bottom is a complete blur.

  • 2
    I’m under the impression that the Sampling Theorem states that any frequency under the nyquist will be digitally represented with 100% accuracy. So it would not be possible to have an audible difference between 44k and 48k sampling. And that the noise floor created by quantization error at 16 Bit could only be a problem if you have around 80 db of dynamic range. Would you mind giving a more technical explanation of what is missing in warmth and liveliness? – user3841 Nov 20 '14 at 21:39
  • 1
    @Haffa - I'm not sure if your post is poorly stated or just flat out wrong. "any sound that has less than full dB is to all intents and purposes not 16bit, as any sound recording only has that full 16 bit fidelity at the peaks" makes no sense. 16 bit fidelity is based on the number of steps between the loudest and softest parts. It has nothing to do with only being 16 bit at the peaks. There is under-utilization of the bit depth due to the loudness common in most commercial recordings now, but that is because of the lack of using the floor and has nothing to do with the peaks. – AJ Henderson Nov 20 '14 at 21:59
  • 1
    The sampling bit seems radically off as well. It sounds like you have some basically correct statements (such as records had wider dynamic range in practice than most modern CDs) but the reasons are entirely wrong. There is no reason a CD can't have higher DR than the record, it's just a stylistic choice made for dealing with crappy stereo equipment resulted in mixes becoming incredibly flat things that push lots of loudness and no real DR. That's probably the difference you are hearing. It has absolutely nothing to do with the format though, just the way it is used in retail. – AJ Henderson Nov 20 '14 at 22:01
  • Perhaps I'm misunderstanding what you are trying to say, but it doesn't seem accurate from how it read to me. – AJ Henderson Nov 20 '14 at 22:03
  • I think someone is confusing frequency with bit-depth... best this one just dies quietly at the bottom of the pile... – Tetsujin Nov 21 '14 at 22:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.