Take the 2-minute tour ×
Sound Design Stack Exchange is a question and answer site for sound engineers, producers, editors, and enthusiasts. It's 100% free, no registration required.

I use a firewire audio interface with a decent microphone pre-amp and the ability to sample up to 96KHz in 24-bit.

I get why I'd want to use 24-bit for the added dynamic range, especially when using a lot of DSP plugins. But I'm much less certain about the sampling frequency.

I understand why sampling at 44.1 KHz means frequencies approaching 22.05 KHz begin to suffer from sampling artifacts due to the Nyquist theorem. I can't tell the difference, to be honest. Once my signal leaves the digital domain, my analogue gear probably masks a lot of the details in this region and my hearing is definitely not sensitive to 20KHz anymore.

For the sake of obtaining slightly better fidelity of frequencies near 20KHz, I can switch to 48KHz. In my setup I can playback sound recorded and mixed in 48KHz "natively", but I know any potential listeners will be more likely to listen at 44.1KHz. If I burn to CD, it will definitely have to be resampled to 44.1KHz. My philosophy is then that I'd rather do the resampling than rely on whatever arbitrary resampler the listener has. I've been burned pretty badly by automatic resampling in Windows in the past.

Finally there's 96KHz. More than twice as many samples per second than a regular CD. I can see how this extra fidelity might benefit effects processing, especially if you do a lot of advanced processing. I can also see how you might benefit with extreme pitch or time changes, since there's more material you can throw away, yet still maintain good fidelity. I don't burn 96KHz audio DVDs -- do anybody actually use this today?

So the thing is, I can't really hear the difference when recording in 96KHz, but it takes up twice as much space and causes my DAW to work at least twice as much, meaning I can run much less concurrent DSP plugins at the same time.

I don't think I can blame my equipment much and I don't have any plans to upgrade any of it for the moment. I'm using the Mackie HR824 studio monitors in an acoustically decent location with the Mackie ONYX Satellite audio interface.

Am I missing something? Are my ears shot? Are there other properties of recording in 96KHz that makes it worthwhile?

share|improve this question

migrated from avp.stackexchange.com Jan 27 at 14:58

This question came from our site for engineers, producers, editors, and enthusiasts spanning the fields of video, and media creation.

    
There's also 88.2 kHz, which is simpler to downsample to 44.1. Related: audio.stackexchange.com/questions/212/… –  endolith Dec 15 '10 at 16:47
add comment

9 Answers 9

up vote 5 down vote accepted

I tend to prefer something higher than 44.1KHz because (like you said) it's pretty close to the minimum useful frequency (according to Nyquist), and I try not to assume that I'm going to be using the audio in the same way all the time.

The best example I can think of is timestretching algorithms. I'm a big fan of Ableton Live's warping, and at some point I want to invest in a DVS like Serato (or similar) and start playing my tracks at ridiculously high and low speeds, just to see what happens. When I do that, the extra resolution should become apparent. I don't KNOW this (never really experimented enough with it) but I'm pretty sure it's the case, and so I favor the higher rate.

For me, it's about keeping my options open.

share|improve this answer
    
It is the case, assuming your DAC is actually recording those high frequencies. –  endolith Dec 15 '10 at 16:48
add comment

I understand why sampling at 44.1 KHz means frequencies approaching 22.05 KHz begin to suffer from sampling artifacts due to the Nyquist theorem. I can't tell the difference, to be honest. Once my signal leaves the digital domain, my analogue gear probably masks a lot of the details in this region and my hearing is definitely not sensitive to 20KHz anymore.

You can't tell this when doing an A-B comparison between a 96KHz track and a 44,1KHz track, but you can tell the difference at the end of a 16h session in 1 day between "Why did id sound so good in the beginning?" and "Boy am I tired, let's continue tomorrow."

I am always getting a moment of the first case when recording on a low sampling rate. I can't tell exactly how this works psychologically but the dirty filtering stuff on low sample rates definitely has to do with it. It exhausts the part of your brain that has to process it and tells you there is no difference, I guess. I never got generally disappointed about quality while working on 96KHz during the day.

share|improve this answer
2  
While I can't hear a difference between sample rates on my setup I'll support this answer because I follow the rule that I should be using the highest resolution available to me "just in case" -- it's better to have to deal with too much data, than too little data. The only time I'd recommend scaling back the sample rate is if you're low on disk space or your hardware can't handle it as you add more tracks to a project. –  Ian C. Dec 9 '10 at 19:07
2  
@Pelle have you tried blind testing your finding? I've done most of my projects in 96KHz just for the sake of it, but I still get tired after 16 hours behind the mixer. I think maybe even if you don't feel tired, taking a break after 16 hours is probably a good idea for the overall creative process anyway :-) –  Kim Burgaard Dec 11 '10 at 3:27
1  
I should also add that upgrading my loudspeakers made a much bigger impact in combating listening fatigue. I started out with a pair of regular stereo speakers, but upgrading to my first pair of studio speakers felt like pulling a rug away between the sound in the DAW and my ears. I still felt listening fatigue after longer sessions, but my last upgrade seems to have eliminated that. –  Kim Burgaard Dec 11 '10 at 3:35
    
I am not sure wheter I pointed out completely what I meant. What I mean is not really feeling tired in general (that always happens on a 12 hours audio day), but feeling you really have the question "Why did this exact same mix sound so damn good in the morining?" –  Pelle ten Cate Dec 12 '10 at 9:36
    
I think I see (um, hear) what you mean, except I haven't experienced it myself... so let this be the exception where I disagree with you and put my vote where my ears are :-) –  Kim Burgaard Dec 14 '10 at 2:58
show 2 more comments

independently of hearing or not, there's the nyquist frequency, but I don't think that that should be to much of a factor to worry about. If you do hear the difference then do something about it, if you don't just let it be - it'll be less drain on your hard drive, RAM and processor, not to mention have to re-apply algorithms when bouncing etc. And why? because most ADC present you with something called oversampling. 4x and 8x oversampling isn't uncommon even for a consumer level cd recorder.

On the other hand, there's the filter. For those of you a bit more tech savvy, try and figure out how many pole low-pass filter you would have to have to cut any sound beyond 22.05kHz leaving anything up to 20kHz untouched (remember the cut-off point already has a 3dB reduction. If you did figure it out, you realize that is practically unfeasible to build something like that and the distortion that would result from such a steep curve would be really unpleasant!

Yes, true, despite all this there might still be a factor of how converters handle things. And I like to believe that I do hear a difference in most converters to justify a bump to 48kHz.

I was doing some maintenance work on a NEVE DFC with some chaps from AMS-NEVE. It turns out that the console doesn't like to swap sample rates a lot. Then we tried to design a more successful workflow for that console and this is what I heard coming from them. We were talking about the similar DFC's installed at Pinewood Studios and where my colleague works, Lypsinc. They have theirs set up at 48kHz and the only reason they justify going upwards (96kHz) would be for animation due to extreme editing demands that can arise due to the nature of the job.

I agree. I definitely struggle to do any proper elastic audio based editing with 44.1 and even Serato doesn't give me what I want.

On a budget recording studio? I wouldn't think of more than 44.1kHz. Unless you are doing heavy TCE editing, you will have costumers coming in. They shouldn't walk out of your studio empty handedly. I always use the monitor path of my console to give the artist a downmix even in the very first session. You can see the pain - we have to spare half hour or so just to downmix from 48kHz just because I'm being "posh" :P Also, it depends on how much you want to spend on memory.

Cheerios!

share|improve this answer
add comment

24bit of course. 88.2 kHz for CDs and 96kHz for Video. They divide by 2 perfectly.

share|improve this answer
1  
+1 for this answer. Moving between perfectly-divisible sampling frequencies doesn't introduce any aliasing. Moving from 48kHz to 44.1kHz creates lots of aliasing artifacts that must be compensated for by the resampling algorithm. Just avoid it. –  lukecyca Aug 23 '12 at 15:32
1  
@lukecyca: while it's certainly the best bet to go with an integer ratio, arbitrary resampling (with the proper sinc algorithms) is far less of a problem than your comment suggests. Most aliasing comes from the ADC and from FX plugins, so 48 kHz might still be a good idea for effects-heavy CD productions, when 88.2 is not an option for some reason. –  leftaroundabout Sep 24 '12 at 21:14
    
@leftaroundabout, the difference between 48kHz and 44.1kHz is so little, it's not worth introducing artifacts, however minimal they may be. –  d-_-b Sep 25 '12 at 1:59
    
@toor: exactly, it's not worth having the ADC and FX introduce any more artifacts than necessary. Which is why it's rather a good idea to use 48 kHz when there's the chance they they might do so, for even that small difference is probably more than the artifacts introduced through resampling. –  leftaroundabout Sep 25 '12 at 9:24
    
@leftaroundabout, I think you misunderstood me. I meant exactly the opposite. You can do a test yourself at lower frequencies where your ears can notice the 3.9kHz difference. –  d-_-b Sep 26 '12 at 1:17
add comment

I generate audio in the best quality possible, this means for my equipment 24bit/96 kHz and I am "only" doing a podcast.

Reason for this is to have all possibilities open for processing after recording. If this is too much data for you to process, the first step could be to copy the original data to a file with less quality.

Please check page 15 of this presentation (pdf). Even if it is in German you will get the point.

share|improve this answer
2  
That presentation is about pixelated visual data -- audio data isn't analogous. We don't hear pixelated audio data, we hear continuous functions, not discrete functions. Whether you record at 44.1, 48 of 96 kHz, the output is still a continuous function. –  Ian C. Dec 9 '10 at 18:33
    
The message behind the picture is, that you get better results with higher "resolutions" when you process the data afterwards. –  ddeimeke Dec 10 '10 at 6:23
1  
Ian C.: A sampled function is not continuous. It approximates a continuous function if the sampling frequency is high enough. –  lukecyca Aug 23 '12 at 15:35
add comment

If you can't tell the difference, I wouldn't sweat it. You may, someday, be able to tell the difference, but if you can't now, and your clients can't, I think this is a premature optimization of your setup.

There may be a technical reason to do so, such as interchange with other systems/people.

share|improve this answer
    
Even if you can't hear it, doesn't mean someone else can't. This is why good mastering engineers use tools along with their ears. Though, most people are fine staying amateurs. –  d-_-b Aug 24 '12 at 0:42
add comment

You get far more mileage out of going to 24 bit than you do by increasing the sampling frequency, especially when multitracking.

The only reason I would go above 44.1 kHz is if I was especially interested in preserving the phase characteristics of very high frequency sounds (cymbals, for example), since the anti-aliasing filters do not have to be so steep at those sampling frequencies.

share|improve this answer
add comment

I have asked this question to myself a few times in the past, too. If I can be a bit objective after I gained more experience:

The difference between 44.1 or 48.2 and 99.6 will never, ever, amount to 0.2% of the difference between an inspired or non-inspired performance that's captured, or whether it's played in a good or a bad instrument.

Bit rate is yes more important, keep it on the 24, but really, for sample rate to be a differentiator you gotta be on a very, very high level with the other things such as what you are actually recording.

share|improve this answer
    
By "bit rate", I think you actually mean "bit depth". Bit rate usually refers to the overall quantity of data per second (the bit depth multiplied by the sampling frequency, multiplied by the compression factor, if any) –  lukecyca Aug 23 '12 at 15:29
add comment

I just stick with 44.1khz. sometimes I've ported projects over to another location and the sound equipment can't handle anything about 44.1khz. Plus there isn't much difference in the sound.

The only advantage I can think of in using >44.1 is more data to process after mixing down/bouncing/recording..

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.