What Sample Rate and Bit Depth Should I Use?

 

What Sample Rate and Bit Depth Should I Use?I have been encountering questions about sample rates and bit depth a lot recently and been looking into the topic in the past week. I have been reading different articles around the web and browsing forums like Gearslutz to try and develop some sort of understanding of the general consensus in the field at the moment.

So let me share with you where I stand on this  right now. Now, I am not a very techy person so please feel free to correct me if my assumptions are wrong, and please share your thoughts. Let us continue the discussion in the comments.

 

Please note: Since writing the original article I have learned a lot – not least from you guys – so thank you very much for your comments and pointing me in the right direction! I have updated this article to reflect my current views on August the 10th 2014.

 

What Sample Rate to Use?

So what’s the fuss – why wouldn’t you just go as high as your setup allows? First of all it takes up a lot more resources from your system to go for the higher bit rates such as 88 kHz or 96 kHz. When sample rates double, so do the file sizes on your drive. And not only that, but the CPU gets hit a lot harder as well. The first obvious result is that you won’t be able to run as many plugins.

Second, even if you had the system resources to run high sample rates, it is very questionable if there is anything to achieve by doing that. In fact, it seems running high sample rates could in many cases be worse (as opposed to running a sample rate of 44.1 kHz or 48 kHz).

Instead of going into detail on why that is, I am referring you to this article by Monty at Xiph.org. He explains things much better than I ever could. The article approaches the question from the point of view of music downloads, but the theory behind digital audio is no different when talking about music production. Please make a priority to read this article if you want to get your head around this topic.

Keeping these things in mind, I have come to the conclusion that working in 48 kHz is the best choice for me. Going higher than that doesn’t seem to offer sonic benefits, and is very taxing on the system.

44.1 kHz is also perfectly fine and good, but 48 khz probably has a tiny advantage in the highest end of the spectrum (again, read Monty’s article to understand why).

44.1 kHz and 48 kHz are the sample rates most professional electronic music producers work in.

 

What Bit Depth to Use?

Now, the question about bit depth is more simple to answer. When recording and bouncing audio, you should always use a minimum resolution of 24 bits.

24 bit audio gives you a theoretical dynamic range of 144 dB, as opposed to 96 dB with 16 bit audio. More dynamic range means better signal-to-noise ratio, better precision when mixing and less worrying about headroom as you don’t have to run your levels so hot. 32 bit floating point is even better, but the benefits there over 24 bit audio seem to be pretty much indifferent (again, please correct me if you think I’m wrong).

 

What Are Your Thoughts?

Let me know where you stand on this guys. I know there are some very technologically savvy people out there reading this, so please drop a comment with your thoughts.


Accept responsibility of your success. We Spin shows you the rest. Join me there:
  • Bit

    “24 bit audio gives you a good dynamic range of 144 dB”

    Does this really matter in todays (overcompressed) EDM?

    • http://www.resoundsound.com/ Ilpo Karkkainen

      Good question! I would say yes – of course the dynamic range of the final mastered piece of music is nowhere near that. But it is important to have that dynamic range while working on the music, to have resolution in the mix and to be able to push things loud without bringing up the noise floor too much. This especially matters to those who record anything live – vocals etc.

  • http://drzhnn.com/ Denis Druzhinin

    IMO the “higher is always better” principle is only true for live recordings ;)

    I usually work inside the box in 44100 and only switch to 96000 when I need to avoid aliasing when working with curtain plugins or when very high frequency modulation is used, like super fast LFOs, FM, ring modulation. And for extreme pitch shifting as well.

    There’s a couple of things about higher sample rates to keep in mind. First is online and offline rendering. Some plugins can apply advanced techniques during offline mixdown, like higher quality and more CPU intense algorithms, multi-pass or oversampling. For example filters in U-HE Zebra are tuned “by hand” for every major sample rate and the same patch recorded at 44100 and 96000 may sound quite different. Second is while working at 44100 we can always choose higher sample rate for offline project mixdown, without even changing audio driver settings. This can be dangerous though if used carelessly.

    • http://www.resoundsound.com/ Ilpo Karkkainen

      Thanks for the comment Denis. Very interesting. Never thought of working at 44100 and bouncing higher. Makes sense – definitely going to try this out.

  • Jordan

    I made the decision to buy the 16 bit A/I over the 24 bit one for several reasons,

    Price was one of them. I’m still a small studio making NO money, working with NO clients yet, and couldnt justify the price for a box that does the same third, but with the addition of the number ‘8’.

    Another reason was seeing how many people, even with the 24 bit A/I, still had to put it in 16 bit to work properly.

    Another reason is because I know how to track proper levels, and have never ever ever run into any issues what so ever with 16 bit. Let me reiterate that, I HAVE NEVER EVER EVER EVER HAD ANY SINGLE CONCEIVABLE ISSUE WITH RECORDING 16 BIT. 96 DB of headroom isnt enough? What monstrous clipping, distortion application are you applying for NINETY-SIX DB to not be enough?

    Maybe if you need 24 bits to track, you need to work on your tracking.

    And of course, the age old tale of the Beatles recording everything in mono to 4 track machines, and making better music than any of us ever will.

    I think waiting for a 24 bit A/I is an excuse to not make music. No mix has ever been made or broken by the addition of 8 bits. IT’s simple inconsequential.

    • http://www.resoundsound.com/ Ilpo Karkkainen

      The question here is not wether you can make good music in 16 bits – surely you can – like you said music is about ideas and not equipment. But we don’t live in the 60’s – what’s ideal for electronic music production right now?

      I have never had a problem because of working with 24 bit audio, what kind of trouble are you referring to that your friends have been experiencing?

      Finally, I really don’t think the price is an issue for most people in electronic music production. Even many of the very cheapest end “semi-professional” audio interfaces come at 24/96 these days. I understand if you need an interface with tons of I/O, this could be different – however there’s no denying 24 bit format is the current industry standard in production and even at consumer level (while CD is 16 bit, DVD and Blu Ray do 24).

  • DrumEd

    I can bounce at 32bit whereas my recording input can only reach 24bit. I was told that bouncing at 32bit gives you more headroom? I work at 44.1 and then bounce my final pre mastered wav at 48hz

  • Rob

    32bit floating point allows you to hit lower frequencies. For me it comes down to what your ear-holes tell you.

    • Rob

      also, big-ups Resound

    • lambdoid

      The bit depth only affects amplitude, not frequency(the x axis on a graph). A higher bit depth increases the dynamic range of the audio and increases the signal to noise ratio/lowers the noise floor. However, 32 bit floating point does not add any more dynamic range than 24 bit integer(since no DAC exists that can use the full dynamic range of 24 bit audio) and is used internally in DAWs to avoid word length truncation and to improve accuracy at lower amplitudes. The sampling rate only affects the range of frequencies that digital audio can represent(the y axis on a graph) and determines at what frequency aliasing occurs which is 1/2 the sampling rate(Nyquist-Shannon sampling theorem). Low frequencies are represented well unless you’re using a ridiculously low sampling rate(extremely unlikely unless you’re going for a vintage sound). It’s the high frequencies that are more problematic at a lower sampling rate.

  • Lasse

    Hi,

    This is an interesting post and something I was looking into myself about a year ago. Some plugins may sound better at higher sample rates – others may do internal supersampling so you will get pretty much the same result with 44.1kHz. You can get some lower latencies with higher sample rates as well.

    What is important for sample rates is the quality of the conversion, if you are using higher sample rates. Take a look at this site and your DAW graph: http://src.infinitewave.ca/

    Back when Reason 6.5 was the latest version I was running it at 96kHz and I had some aliasing issues when exporting songs to a 44.1kHz audio file. If you take a look at Reason 6.5’s conversion graph from that site, it shows terrible aliasing that reaches into the audible band. So if you use high sample rates, make sure you are using quality converters when it’s time to bounce something back into 44.1kHz or 48kHz. Personally I only use 44.1kHz now.

    As for 24bit vs 16bit: 24bits gives more headroom for the working stage. 16bits dithered is well enough headroom for an end result (like a CD), where the gain staging, leveling and dynamics are already done. However, when editing it’s a whole lot easier to have more headroom (more bits) so not all of your recordings need to be recorded as hot as possible without clipping.

    • http://www.resoundsound.com/ Ilpo Karkkainen

      A very cool link, this is great – thank you!

      So basically, the less lines you see in the graph, the better, right?

      Reason 6.5 is definitely looking bright.

      Ableton 9 seems to have some bad stuff going on as well. Very interesting.

      • Lasse

        Yes, the sweep test should produce a single clear line. It is all explained in the help section, which also has some other interesting information on how SRC works.

        • http://www.resoundsound.com/ Ilpo Karkkainen

          Yeah, checked the help and FAQ sections, good info. Looks like I’ve made a solid choice switching to Pro Tools for mixing. But then I again my ears told me that already. ;) Thanks a lot Lasse.

  • Lasse

    So in short (my previous post), higher sample rates actually can be worse in some cases.

    By the way, in these discussions it is quite necessary to separate recording sample rate from the sample rate you are working with in a DAW with plugins and soft synths.

    For recording you could argue, that by the

    Nyquist–Shannon sampling theorem 44.1kHz is sufficient for all sources, since you can record frequencies up to 22.05kHz which goes well beyond the human hearing range. However, since A/D converters are not perfect, you might get some distortion in the 20 – 22k range, which is why I understand that the standard for digital film was chosen to be 48kHz to give some room to work around the cutoff point.

    Now if you record some live source at 96kHz, you are recording audio frequencies up to 48kHz, which are inaudible (as no one hears above 20kHz really), but may contain some audio information. If you convert this recording to 44.1kHz with a bad sample rate converter that has aliasing, you will bring all the inaudible supersonic information into the audible range. And in that case you would have gotten a better result recording straight at 44.1kHz, since it would not have picked up anything above 22.05kHz anyway.

    As for DAWs, mixing, plugins, soft synths I think you will find that many people perceive better quality with higher sample rates in online discussions on the topic. So try some higher sample rates and see if they are useful for your setup, but I wouldn’t worry at all if you’re stuck at 44.1kHz. Like said, it gives more CPU headroom and at least for my purposes the results have been exactly the same as with higher rates.

    • http://www.resoundsound.com/ Ilpo Karkkainen

      Thanks for the expert insights Lasse, much appreciated.

  • lambdoid

    z3ta has an oversampling option. I use the 48khz sampling rate in Reaper and when I switch the oversampling on in z3ta, you can clearly hear the difference in the high frequencies. This is great for pads and other sounds with a lot of high frequency content, but basses often sound a bit more grungy at the DAW sampling rate or utilizing z3ta’s undersampling option. You can do this for both online and offline, especially if you have an older CPU that struggles with z3ta’s high demands.

    • http://www.resoundsound.com/ Ilpo Karkkainen

      Interesting! So that means z3ta can work independently for example at 96k when your project is at 48k?

      • lambdoid

        Yes. I think so, although at higher sampling rates it would make less difference.

      • lambdoid

        There is an audible difference when using the oversampling option. It’s not always desirable to use it though. For dirty basslines, it’s usually better to use 1x the sampling rate(or sometimes even 0.5x), but for high-pitched sounds like pads etc the oversampling helps improve clarity.

  • Dj Pushups

    Yet again a very nice read!

    Personally I’ve come to conclusion that a 44.1khz and 24bit is enough in my bedroomstudio. I haven’t really given it even that much thought as to why that is; some producer friend of mine once just told me that it will sound better when I bounce my audiotracks in 24bit. And I suppose it did since I kept with that.

    I suppose that when this subject comes to a more professional level (say, mixing and mastering services for example) the samplerates and bitdepths matter more and more. My analogy here is that when you have better equipment to perceive and manipulate sound the more the quality of the signal matters.

    Thank you for your time!

    Regards, Dj Pushups

    • http://www.resoundsound.com/ Ilpo Karkkainen

      I think there is some truth in that – for instance since I upgraded my monitoring I have definitely started understanding what the fuss is about some plugins that I quite didn’t get before. Hearing more nuances. But yeah – working with 24 bit audio as opposed to 16 is still beneficial regardless of whether you can perceive a difference or not, as mixing/mastering guys can do a better job with it later like you said.

  • Andrew

    48khz for video

  • Michael_Mann

    96K is waste of CPU and disk space. 96K won’t make any difference whatsoever in terms of sound because:

    1) no one hears anything above 19K in an ordinary listening environment
    2) you still must be extremely careful with all frequencies above 12K
    3) you’re forced to downsample everything down to 44.1K sooner or later. Your one billion dollars worth converter won’t change the mathematical fact that downsampling creates always aliasing and that’s going to be the very last thing you want to hear in your finished mix
    4) decently coded plugins sound identical in 44.1K and 96K or 192K because of oversampling. There should not be difference in the sound!!!!
    5) automation won’t sound smoother in 96K either because everything you are going to hear in the end is aliasing caused by above-mentioned downsampling
    6) 44.1K won’t prevent you from getting 3-5ms latency if you have a good sound card and you know how to use it and you don’t forget to freeze tracks

    Keep your mix sparse enough and use send reverbs cunningly – that’s how you are going to get that sheen that 96K using people talk about.

    • http://www.resoundsound.com/ Ilpo Karkkainen

      Thanks for the comments Michael!

      I recently stumbled upon a great and very in-depth article on this topic:

      http://people.xiph.org/~xiphmont/demo/neil-young.html

      Highly recommended for anyone who is trying to get their head around this stuff.

      • Michael Mann

        Thanx Ilpo for the great link!

        It’s funny that the article mentions Neil Young because I wrote my first comment with him in my mind. This 192K/24-bit nonsense is pretty much coming from a man who’s notorious of excessive concert volumes and has reportedly damaged his hearing to the extent that he’s basically incapable of follow an ordinary discussion. At the same time, this mass producer of the noise pollution is praised of his awareness of environmental issues by the media. And last but not least, Young’s music has nothing to do with any kind of high fidelity or exceptional musicality. So, go figure….

  • Eugene Eugene

    Wow, this article you gave a link to (and vids from it!) is pure gold. Thanks!

    • http://www.resoundsound.com/ Ilpo Karkkainen

      Isn’t it! One of those ones you want to bookmark and come back to every once in a while.

  • Mav @ Scientific

    very interesting. i was still using 16bit, 44.1khz as i was used to this since forever, but i will move up to 24bit, 48khz from now on. cheers ilpo!

    • http://www.resoundsound.com/ Ilpo Karkkainen

      Thanks for the comment Mies good to hear :)

  • Jim Spratling

    Probably a bit off topic but… what bit rate should I use for ripping vinyl? I am using an Artcessories USB Phono plus. I have an iMac with Ableton running. I have encounters some strange glitches when running at 48 or 44.1 kHz. I have increased the sample rate and now have less glitches. My question is… Am i doing the this the correct way and am I loosing any quality by increasing the sample rate?

    • http://www.resoundsound.com/ Ilpo Karkkainen

      That is a great question!

      The glitches of course should not be happening in any sample rate, so that is something you may want to look into (unfortunately I am not very qualified to help there). You could possibly post about the glitch issue on a forum like Gearslutz (http://www.gearslutz.com/board/) and find help there.

      One thing you should try though, is increasing the buffer size inside Ableton. For ripping vinyl you can max it out. This might help getting rid of glitches.

      Live -> Preferences -> Audio -> Buffer Size

      To answer your actual question, I would personally use 48 kHz and 24 bit format for ripping vinyl. That offers plenty of resolution for capturing everything. Also make sure you are recording loud enough (watch for clipping though, but you probably knew that).

      You won’t lose any quality if you want to record (rip) in higher sample rate. But some slight deterioration might happen if you later convert that high sample rate file into lower sample rate (CD compatible 44.1 / 16 bit for example). The quality of the conversion depends entirely on the converter, some are better than others.

      So to recap, 48/24 is good enough, but if you want/need to record in higher bitrate, you won’t lose any quality – as long as you make sure you use a quality converter if you later on want to convert the sample rate.

      Does that make sense?

      • Jim Spratling

        That is very helpful, thank you Sir!

  • Noodlez

    Just came across this website about working ITB at higher sampling rates – http://varietyofsound.wordpress.com/2012/11/02/working-itb-at-higher-sampling-rates/

  • robertrobin10 .

    24 bit will sound better than 16 bit, remember its 16 bits only for a sound at 0db, if your peaks are at -6 you have already lost 1 bit of resolution and the average of your music maybe 20 below that!!And dont forget thats only for the bass and mids, treble has little energy because its only harmonics so they maybe another 20db below!!So your 16 bit playback is only sounding as a good as a 10 bit system!!You should be playing back at a 22bit rate to get the sonic benefits of a 16 bit sound.I believe this is the only reason why digital sounds fatiguing. As far sampling rate, the higher the better,recorders have sharp filters to reduce alaising distortion which can cause sonic problems,the higher the sampling rate the higher or gentler the filters can be.