Lots of us record at at 192k for the 4x sample rate improvement, but how does that translate to a more useful or flexible sound file? According to Tim Prebble, plug-ins are more effective at 192k (which I've found to be the case as well) but because they take longer to render I end up using them less due to time constraints.
Also, if most microphones are only capturing frequencies up to 20kHz, shouldn't we use "hi-def" mics (30 - 100kHz range) when recording 192k?
Your thoughts are apprciated!
I posed similar questions in this thread.
Afterward I did several tests (which are noted in my response later on in the same thread)
what I found was that even mics that aren't rated above 20k are perfectly capable of capturing great audio far above the audible range - and are capable of it using a pretty wide array of A/D converters and preamps.
everything test here is confirmed capable of recording good stuff well above 20k in my own tests:
I'll also reaffirm that higher sample rates really add to the flexibility of sounds. The doors that I recorded at 96k for Tim Prebble's doors project have already been pitched down and employed in all kinds of heavy impact situations, and because they retain high freq content they don't sound "pitched" at all - they just sound much bigger.
The way i think it works is that the samples are distributed equally across the20-20khz range to increase the resolution of the wave form for a more accurate "picture", as opposed to using the samples to register higher tan 20khz frequencies. That way the waveform has much more information and so suffers less degradation in processes like pitch shift, time stretch etc, because the signal processor has more to work with.
This doesn't come from any scientific source, it's just what makes sense in my head and would love to confirm or dis-confirm this; plus i just woke up, so i don't know if what i wrote makes any sense at all...
I discovered a LOT when dealing with the Fireworks Library at 192k eg I thought my HD2 PT rig was powerful (eg on films I often run out of voices, and it can playback 192 tracks at one) but at 192k it became underpowered! The track count, plugin use and even the timeline window became restricted! I would love to work even at 96kHz all the time for better plugin performance but even that would reduce my PT performance too much..
The moral of the story - there is definitely an advantage to using higher sample rates, but practically the workflow tends to be that of recording, editing and manipulating material at 96 or 192, but then do the syncing, layering, editing at 48k.... for now at least!
Here is an abstract you'll want to look at.
http://jn.physiology.org/content/83/6/3548.abstract
Talks about the impact of frequncies we can't heard in perception of sound. So yeah, it matters :)
The point of over sampling is not about hearing frequencies above 20k. The point of over sampling is to make it easier for the filters to filter everything out above the nyquist frequency. When you sample at lower sampling rates such as 44.1 or 48 the quality of the filter matters more. So if you record with an amazing ad converter with a really expensive filter, 44.1 sounds just fine and it becomes more difficult to hear the difference between 44.1 or 96 or 192. On the other side when you record with a less expensive converter, over sampling at a rate of 96k or 192 becomes extremely helpful in that you will have a cheap filter and it won't matter as much.
The issue with recording at really 96 or 192 with a lesser quality converter or interface is that the word clock tends to be poor as well which will bring about jitter issues. Therefore if you have a cheap converter and want to record at high sampling rates then it would be a good idea to buy a good external clock.
+1 on the increased resolution of the audible range.
There are definitely some interesting sounds above the 20kHz zone, but you don't necessarily need a mic rated up beyond that to capture them. Many mics will capture those frequencies, but they're response may not be as even as those rated to 50 or 100 kHz.
I wonder if a small diaphragm condenser mic would work better in those applications. Less membrane for the acoustic compressions/rarefactions to excite? Anyone have any thoughts on that idea?
There's no point in recording frequencies above 20 kHz unless you're going to do something with them, like put them through an effect that will modulate or stretch them back into the audible spectrum.
The main benefit of higher sampling rates is not to capture ultrasound, but to reduce audible things like aliasing, phase distortion, etc.
These can be a benefit for synthesis, too. If your square wave generator is poorly-written, it will produce an infinite number of harmonics, which are aliased and sound bad. If the sampling frequency is higher, this effect is reduced. Of course it would be better if the generator were written correctly, but you can't always control that.
For pitch shifting down you need more than a 0 - ∼22.05kHz bandlimited signal to try to maintain some highs in the downpitched version.
For digital recording you need more than 44.1kHz total sampling rate for 1. Making all gear-related conversion and input/output stage signal processing so that audible problems aren't induced to the audible range or the problems are minimal. 2. Cutting costs in engineering and manufacturing and coping with limits in technology. But improvements don't come above a certain point and 88.2kHz and 96kHz are optimal sampling rates (from the existing standards) according to engineer Dan Lavry: http://www.lavryengineering.com/pdfs/lavry-white-paper-the_optimal_sample_rate_for_quality_audio.pdf . It would be even better, if the audio industry wouldn't have the stupid double standard of 44.1kHz and 48kHz sample rates, which forces needless interpolated downsampling (e.g. 96kHz -> 44.1kHz).
For plug-in processing you need oversampling for the same reason as in digital recording, for keeping DSP induced problems (e.g. potential aliasing generated by DSP) out of the audible range. And of course for maintaining the recorded sample rate (88.2/96kHz) throughout the project until the audio is finally downsampled for distribution purposes.
But 192kHz. It seems unnecessary and has more downsides.
I realize this is an old thread, but perhaps I can add to the discussion.
To properly understand how greater sampling rates works, an understanding of how analog to digital conversion works is necessary. Fundamentally, when you record something (at least these days on "Digital" Recording Equipment) what you are doing is making a sound (Analog and Audible) and capturing a picture of it if you will, using a bunch of 1's and 0's. Analog sounds (like made by your instruments or voice) can be simplified for our understanding as sine waves at different frequencies, and they usually occur in combinations of frequencies and magnitudes (volumes).
Now, Sine waves are measured in cycles per second. A sine wave has peaks with respect to a certain reference point on the positive side and negative side of the reference (see picture at http://en.wikipedia.org/wiki/File:Waveforms.svg). So when the wave starts at the middle, goes up and then down, and then back to the middle, that is one cycle. The measurement of cycles per second is called Hertz or Hz. The lower the Hz, the deeper the sound (bass), the higher the Hz, the higher the sound (treble). That is a simplified way to think of it. The Human ear generally cannot pick-up frequencies above the 22,000 or 22KHz threshold, so for our intents and purposes, anything above that does not need to be captured in the recording.
Okay, so now I've just regurgitated a bunch of scientific mumbo-jumbo about frequencies and Hertz and the like, but how does that relate to the 192 KHz sampling frequency?
Here is why this is important. A sampling rate is quite simply, the rate at which a "sample" of the incoming signal (audio) is taken and "recorded" or represented with a series of 1's and 0's. So what does that mean? In the picture of the sine wave in the link above, imagine the wave form from the first middle starting point to the second middle crossing point (where it crosses middle on the upswing) occurs over the course of one second, if you sampled that waveform at 4Hz, or 4 times in one second, you would have a reasonable, but rather choppy representation of that wave form. If you increased the sampling rate, you would have more samples at shorter time intervals, which means a more accurate picture of the original waveform you are trying to capture. See the following link for an example http://artsites.ucsc.edu/ems/music/tech_background/TE-16/teces_164.gif as you can see in the picture, the waveform becomes smoother as the sampling rate increases (which means the nuances of how the audio sounds is much closer to the real deal then if the sampling rate were lower). Someone in a previous post said that you can "perfectly" represent a sine wave using just 2 samples. This is not true as stated, however you can reasonably represent a sine wave but it all depends on the frequency you are trying to capture and your sampling rate.
The commonly used sampling rate for audio has been for the most part 44.1 KHz, which is twice the 22.5 KHz that is at the top range of audible frequencies for humans. The reason for it being twice the amount is due to something developed by Nyquist-Shannon. Feel free to do your own research on that, but essentially, the 2x sampling frequency is to stop distortions to the signal waveform from occurring. Using a 44.1 KHz sampling rate will quite reasonably represent frequencies up to 22.5 KHz without Aliasing (a type of distortion) occurring. I won't go into Aliasing as thats another couple of pages I just don't have the desire to get into now. Just know that aliasing is bad!
In short the 192 KHz sampling rate will help in capturing frequencies above the 22.5 KHz range, but those are not particularly useful as we generally cannot hear them anyways :) The increased sample rate captures the audio with a way better resolution that represents the original audio much more accurately (or warmly for lack of better words). It's kind of like Digital Cameras, if you had the exact same cameras side by side, but one camera had the capability of 2.1 Mega-Pixels, and the other one had 5.1 Mega-Pixels, and they took the same picture, the 5.1 Mega pixel would capture more in-depth detail and if zoomed in on the screen when looking at it after, the 5.1 or "higher resolution" picture would prove to be better. This is much the same principle, as the increased sampling rate is increasing the resolution of the recording. Essentially higher sampling rate in recording = better quality of recording, but this means larger files on the back end requiring more hard drive space, and more Memory for processing the file prior to saving. This does not address however the audio playback of said file, which is another "long" post that again I just don't want to do right now. Ugh, I'm such a geek... :)
Hope this helps.
Fred
I came to this thread trying to figure out how people are using sample frequencies above 96KHz to positive benefit. The comments about dropping pitch and having content above the human hearing limit move into the audible range make sense. The arguments for easier/cheaper filtering above 44.1K makes sense but that seems to be solved by 48K. The arguments about content above 20KHz having a psychological impact on humans is interesting but not well documented and that approach would require everything in the recording and playback chain to support the higher frequencies right?
When I'm recording high quality audio I try to use 24 bit samples and as close to a 96KHz sampling frequency as possible. Some gear will only provide 48KHz which seems fine to me. Then, when I mix down I usually produce 24/96KHz, 24/48KHz, and 16/44.1KHz stereo files depending on the use. My reasoning on the use of 96KHz for sources is because it's twice the standard 48KHz end product tracks used in professional video and this could help avoid issues when mixing. My understanding of the benefits of sampling rates above 44.1KHz is that it's useful when combining, mixing, and applying effects to signals. Likewise bit depths above 16 bits are useful when capturing because the higher headroom and lower noise floor. You can't always maximize the use of the headroom in the capture so having a lower noise floor helps when you inevitably have to normalize the signal. I've stopped worrrying about trying to get my audio input signals to almost clip at capture. So, one of the benefits of 24 bit samples is avoiding the the risk of clipped digital signals. Just leave more headroom and then increase the volume later.
People often confuse PCM capture of audio with the analogy of capturing two dimensional photographs with digital cameras. I.e. higher resolution (higher sampling frequency) means you're capturing a more perfect version of the sine wave. Higher sampling frequency with PCM simply means that you are capturing higher frequency source signal without aliasing. This is what leads some people to claim that you never need anything more than 44.1KHz because of human hearing limitations. However, there are other benefits to capturing higher frequencies when the signals are going to be processed or mixed in some way, which they usually are.
This misconception of "audio capture resolution" is propagated by the standard text book examples of a sine wave with various slice samples showing approximations of each slice of the sine wave. Of course nobody wants ragged edge sound waves so the flawed logic is that more audio "resolution" requires higher audio frequency sampling. I thought I'd try to find an article that lays out the PCM vs this out clearly. This one is a start. https://www.soundonsound.com/techniques/digital-myth
Here's a relevant quote:
"Now, perhaps the greatest myth in digital audio relates to the misconception that digital signals are shaped like staircases, and that much of their 'brittleness' is a consequence of the steps. This is nonsense. Digital signals are not shaped like anything — they are sequences of numbers. Unfortunately, the type of representation in diagram 8 has led many people to confuse graphics with reality.
Let's be clear. When the samples in a digital signal are converted back into an analogue signal, they pass through a device called a reconstruction filter. This is the process that makes the Sampling Theorem work in the real world. If there are enough samples and they are of sufficient resolution, the signal that emerges is not only smooth but virtually identical to the analogue signal from which the samples were originally derived. Of course, it's possible to design a poor reconstruction filter that introduces unwanted changes and artifacts but, again, this is an engineering consideration, not a deficiency in the concept itself."
I dont agree that sampling frequencies higher than 40kHz increase the accuracy of the recorded frequencies below 20KHz...
Sounds are made up of sine waves and you only need 2 samples to be able to perfectly represent a sine wave since it is a mathematically perfect shape. If you store a million samples for one sine wave it will be no more accurately represented than with 2 samples.
But hey, I could be wrong
