Hey guys,

I'm aware that the higher the sampling rate, the less likely you are to have issues with aliasing, specially when dealing with plugins which introduce harmonics/saturation...It comes with a price though, which is a considerably increased CPU hit and bigger file size.

On the other hand, when using lower sampling rate, like 44.1 kHz, you make it a lot lighter on the CPU, but then you'll have to increase the oversampling of your plugins to avoid creating a cumulative aliasing mess, which will make your CPU work harder again.

So, let's say, between 88.2 and 44.1 kHz, both using the same bit depth, which one turns out to be more cost-effective overall?

The way I see it, you're compromising something either way, right?

So, what sampling rate and bit depth to you usually use for recording? Why?

Cheers,

Raf.

 

If you use 48k you probably have to convert less often, and my stuff is usually going to CD. And it is where a lot of things like interfaces and pre-amps with digital out are by default. I can't hear a difference in bit depth rate until I get to the mixing stage. When you start putting all those tracks together is when I think it may make a difference to my ears, but most digital audio theory says otherwise. My computer is designed for audio production, so it handles about anything I throw at it. I have to start piling up the plug ins to make it sweat.

 

Last edited: Dec 22, 2015

The optimal sampling freq is theoretically around 62kHz or so. I always use 24 bit. I;m using 48kHz these days so I don't have to SMUX my ADDAs. If you need to go higher for acoustic recordings, - for example - 88.2 works better for you than 96k, since you bypass a lot of artifacts by dividing 88.2 equally by 2, to get to 44.1. Look up Nyquist Theory.

I wouldn't worry about this. Professionals are using 44.1 as well. The only thing that matters is the "quality" of your converters.....and I do mean quality. Throw the Steinberg, Apollo, RME, and all of your prosumer interfaces out the window. Using Lavry, Benchmark, Mytek, *Weiss, and the like, will "utilize" 44.1 to it's fullest.

My advice, just use 24/48. Be mindful of what kind of plugins you use since some can introduce more aliasing than higher quality ones. Try a few sample rates with different projects and try and see if you can personally tell a difference on your system. If you're writing or recording a great tune, it really doesn't matter. It's the song itself that matters and makes your listener move.

 

Last edited: Dec 23, 2015

I've run 24bit/44.1 for years. last time i did some testing with bitrares, years ago, i could'nt hear any difference between 24bit and 32bit. could be me tho.

i would focus more on the room to record in and the mic positioning these days

 

Agreed. 24/48 is the industry standard for location sound recording dialogue for Film & TV production. You'd only really record higher rates if the audio is intended to be very heavily manipulated with special effects etc.

The human ear is far too ineffective to actually notice any difference recording higher rates (regardless of what audiophiles would have you believe).

 

Last edited: Dec 23, 2015

Fuck the sample rate that should be 2x to your 44100 final rate. A lot of src resamplers work much worse when you convert from 88200 to 44100 than resamplers which works fine converting from any sample rate to another, like from 96k to 44.1, for example finalcd, dbpoweramp 15.3, sox or izotope rx. No distortions below -170 dB for 32 bit int, linear phase "ideal" filter, full frequency range. So you can use any sample rate to work with. 21 century today. It's not 80s or 90s. To use 88 for 44 or 96 for 48 sounds like it is actual for 1990s. Algorithms now are much better. Don't worry. Just use what I said.

 

44.1kHz. You don't need anything more, as when you upload that to SoundCloud it's gonna end up being 44.1kHz anyway. Distortion, saturation plugins etc. (but also synths, like u-he Diva, even stock FL Studio plugins - think Maximus and Waveshaper) temporarily switch to higher sample rates (oversampling) - even up to 192kHz, which is ridiculously overkill - just to make sure no artifacts are being created. After the work is done, it bounces it back to something normal and usable, which is again 44.1kHz. If you're scoring for video/DVD movies, you might want to switch to 48kHz, as that is what DVD standard uses, but expect no audible difference.

For recording, pretty much any DAW is set out of the box to record in great quality. FL Studio, which is my go-to DAW, uses 32-bit float WAV. This means you can record way over 0dB and have no distortion. Some DAWs use 24-bit. It's what Pro Tools uses by default, and it is also a bit overkill, and gives plenty of headroom. You don't need to change that setting at all.

TL;DR: If your soundcard really supports 48kHz, go for it. Don't waste CPU and HDD by setting anything more. Record at 24-bit, but this is already set for you so no need to worry. Really, unless you're at 22.1kHz and 8-bit on a Commodore 64, you won't notice any lack of quality. Focus on making great music and mixing instead - this is where your "quality" comes from, and it changes way more than those settings.

 

Optimal in which way and for which reason?
If there's an theory that it is 62k there must be an research paper this theory is based on.

What are the technical shortcomings of RME converters in your opinion? Or is this more of an audiophile thing?

 

I record at anywhere between 44 and 96k always at 32bit float This should help

 

24/48 if you have any plans to submit for video, 24/44.1 otherwise (until high sample rate streaming media is the norm.)

If I had the processing power, I'd record at 192 all day long just to ensure the most accurate audio capture, but the CPU hit doesn't justify what nominal difference that might be "perceived" by going with 88.2 and up. Used to record at 88.2 until doing a few sessions at 24/44.1 and getting better mixes, because it had to do more with my recording technique, EQ, gain staging etc. not to mention workflow wasn't stifled by a large session choking while mixing it at 88.2.

 

If you're doing pitch shifting (auto tune, Melodyne, Synchro Arts plugs, etc), or if you're doing time stretching, a higher sample rate will produce less artifacts. This is a fact, and it is very testable. And when it bounces back down to 44.1khz, it would have reached this lower sample rate without enduring the artifacts because the pitch/time-shifting processing was done at the higher sample rate.

Aside from that, there aren't too many good arguments for using high sample rates. Essentially, though, there's no real need to record at ultra-high sample rates.

And lastly, I noticed someone said using 88.2 is better for the math than 96khz for stuff that will eventually go back down to 44.1. This has been proven to be false. We're dealing with computers that won't miscalculate -- not to mention that this particular truncation math is not difficult for a computer at all -- and thus won't introduce any real problems, considering you're truncating properly.

 

Last edited: Dec 23, 2015

48kHz/24bit

 

Last edited: Dec 24, 2015

32bits float is TOTALLY useless for recording.
Because all converters are 24bits only for now. An audio converter CAN'T be float, because quantification is...quantified
May be one day we will see 32bits linear converters. One day...

So basically, you record at 24bits linear...and convert to 32bits float.
You only loose ressources and space that way. Bigger is not always better. You need to know how it works. 32 float is a working format.

Apart this point, 24/44.1 for audio only. And 24/48 for general multimedia is enough for me.
I use 24/88.2 for audio mastering output.
Be careful : some plugins badly handle 48kHz

 

Had an 02R and use to record 24/48k for some 20+ years.
HDD and ADATS.

These days, I'm using a custom blown out Mac Pro, that
could handle higher rates but I mostly work 24/44.1Khz.
File sizes and rates don't tax the the CPU as much with all the ITB
work being done.

 

correct me if im worng but studio one 3 gives you options to even 64 float point? i use it 44.1 > 64 float

even if this can't help... is what i use...

but im no pro...

here is an article about studio one 3 link - http://www.studio-one.expert/studio-one-blog/2015/3/22/understanding-audio-setup-in-studio-one

Process Precision


This parameter refers to how accurately S1 generates audio signals.

We know that sound is a form of energy which travels as a wave. When we record audio into a computer, this audio has to be converted to and stored in a digital format. This process is called Analogue-to-Digital conversion (ADC).


Audio, being a natural phenomenon, exists as a continuous stream of energy. Digital systems cannot process this type of information, so it must convert it to a digital format. The computer does this by taking samples of the sound wave at regular intervals (typically 44,100 times a second). Thus, the information is stored as a large quantity of samples, which, when played one after the other, appear to re-create the original wave (more or less faithfully).

Digital systems, in their recreation of the original signal, must round data up or down to the nearest unit (a consequence of the computer requiring discrete values rather than continuous streams of information) This process of rounding and approximation is known as quantization, and it introduces quantization errors which distort the recreation of the original sound wave.

The maximum quantization error is constant, and its value expressed in dB relative to 0 dB FS is:

e = 20 * log (1 / 2^n)

32 bits -156dB

64 bits -331dB

Studio One supports 32 bit and 64 bit quantization.

As we have seen, the quantization noise (error) introduced for 32 bit float is -156 dB with respect to each sample value. This is well outside our hearing range, and since it is also relative to the signal level, we cannot hear it. Why then, would we want to use higher precision values?

With 64 bit floating point numbers the quantization error is -331 dB relative to sample value, which gives us really huge headroom. With 64 bit process precision, we can do a much larger number of computations (processes), while keeping the accumulated error well out of hearing range.

32 bit floating point allows a sufficiently large number of computations so that quantization noise is rarely audible, but in cases where the audio is being heavily processed, it is possible to introduce audible noise in 32 bit mode. So yes, in some cases 64 bit processing can make an audible difference, but only where a lot of processing is being done, or with audio where noise or distortion may be more noticeable. In most cases you should not expect it to be easily noticeable, if at all.

 

interesting I'm going to test that when I've the time.

I work at 44.1 at home with 24 bits just for the bigger difference between the noise floor and the clipping point.
In the studio that is 96, that grew from digital connected legacy equipment. where neither 44.1 or 48 were an option

my question is, Do people still have compatibility issues? I know for instance Cubase always failed to recognize or didn't even had a re-sample function. at that time importing wav's was a big joke.

 

Last edited: Dec 25, 2015

48kHz/24bit here.

 

Not really. You don't always record through your audio converter - recording can be done in software, too. Software synthesizers are a thing and using 32 bit float when recording them gives you some additional headroom. It makes barely any difference, but yeah.

 

Yes. Dan Lavry has explained this. I just went back over it, and it's actually 60kHz, not 62.
http://www.lavryengineering.com/pdfs/lavry-white-paper-the_optimal_sample_rate_for_quality_audio.pdf


You could look at it as an audiophile thing, if you're just recording a few tracks - cool. But, if you're mixing and summing a tune through outboard - in which you're stacking multiple tracks, then the qualities of both the AD & DA can build up. I just think that when you start out with good mics, pres, and DA, then you have a real solid recording to work with from the beginning. This means that your "track" will respond to EQ and processing much better, and will be more forgiving. Just my opinion.

 

This is why I record at 32bit Float