Hi, I am currently on a project which is on 44.1 Khz and 24 bit. Is it useful to go at 88.2 or even 176.4 ? And what about the bit, is it useful to go on a 32 bit ? When i add samples which are not from me, usually they are on a 44.1 khz and 16 bit, so it will do the convertion, but does it works ? or what.. I know that if i have a project in a better quality, it's better after for the mix and mastering. But when i have the half of my samples which are not at the same quality, what should i do ?
Well any samples that are not in 24bit will just be inflated to grow in size, they are still sonically 16bit, but thakes the space of 24bit. If the majority of your samples are 16bit, they will be of less quality, nothing can solve that. 32bit is just internal processing in the DAW for even mroe headroom, it is good to use, but not really a big benefit from 24. 24 is easier for compatibility.
Ok so if i understand, it's better to stay in 24 bit, but the samples which are in 16, i don't convert them in 24, i keep them in 16 bit. What about the Khz for the project, i know the norm is 44.1 but i heard that if i had the possibility to be in 88.2 it's better. Again same question, if some samples i would like to import into the project are only 44.1, do i have to convert them in 88.2 or keep them in 44.1. If i have some 44.1 samples into my 88.2 project does it give some problems for the next like the final mix, mastering ?
Read this and you be sorted for the rest of the time. if you're tracking stuff, meaning recording live material to disk it is best to go high (96k/32b) for dynamic purposes and headroom. But in the end everything comes down to 44.1/16. You can work in 24b or 32b for headroom. Thats up to you. No audible difference there. Cheers!
My experience: It's not the encoding, but rather what you make of it. And how you plan to release your stuff. One thing to add to what artwerksi already said: Some VST(i) have only mediocre internal filters and processing, so it can pay off to go both with higher sampling rates and/or higher bit depth.
recording is always in 24bit, your DAW always works in 32bit (or 64bit) internally, and it's OK to go to 88.2khz or 96khz if you know what intermodulation distortion is and how it can negatively effect your sound at high samplerates like 88 / 96. Going higher than 96khz is pointless. So generally it's a good idea to stick to 44.1 / 48, unless you know 'the risks' for 88.2 / 96. You record in 24bit because every analog-to-digital converter out there has a precision of around 20 to 22 bits. Not even the full 24bit, not even the $5000+ converters. So recording with more is just useless, recording with less means your throwing stuff away. So 24bit it is. 32bit is a bit tricky. You got to be careful and know if you're talking 'integer' or 'floating point'. Generally, when software or people talk about 24bit it is 24bit-integer, and when they talk about 32bit it is 32bit-floating-point. The good thing about floating point (dumbed it down a lot), is that it can't digitally clip. That means it's safe to go in the red _and over it_, as long as somewhere down in the path or in your session you bring it back down into sane regions again. 32bit floating point is also the way VST plugins 'talk', so trust me on this: Your DAW always works in 32bit floating point (and some in 64bit floating point). Your 'project settings' don't mean anything, besides the default you use to record in. At least, it's this way in Reaper, Studio One, Cubase and I'm betting a lot of others. If your DAW supports VST plugins, it works in 32 or 64bit floatingpoint, simple as that. A thing most people don't realize, is that '32bit floating point' has about the same precision as '24bit integer'. You're not gaining much (and like I said, your converter isn't precise enough anyway) but you do gain the 'can not digitally clip' thing which is nice. Going into 88.2 khz / 96khz can do good things, but it can also do bad things. The simple bad things are that it costs more CPU power. The simple good thing is that it can get better latency sometimes, specially on cheaper USB interfaces. The complicated good thing is that you recordings gain a bit of precision in a region you can't hear (but the data is there), and plugins _might_ get better precise results because of it, which can effect the eventual sound of your mix. It's about capturing and saving sound frequencies you can't hear, but might effect a plugin or effect down the line. It is a bit snake oil though (in theory there is an advantage, but I doubt you will find it in real life). The complicated bad thing is that your interface might work at 96khz, but it's producing worse quality at 96khz because of cheaper components. There can also be plugins (no, there _are_ a lot of plugins) that don't really expect or test with 96khz and don't have good algorithms to deal with the extra frequency data, creating a more harsh sounding mix in the end. Intermodulation distortion is a big factor in this. So it can be good, can be bad. In the end, a _lot_ of (pro) people out there work in 44.1 or 48 and it turns out just fine, so I don't really see a reason to go higher (unless you're recording / mixing for some Dolby audio track that needs to be in 96khz or something). So there we are at the beginning again: Set your project to 24bit / record at 24bit. Stay in 44.1 or 48 according to your preference. Make sure the _master_ fader / _master_ track is not clipping when you are playing or rendering your mix, and all is good. If you're targeting CD playback (or music on the internet) make sure you convert your final rendered file to 44.1 / 16bit (and make sure you have some sort of dithering to 16bit enabled in your DAW when you render or when you convert.. or that it is a function in your master limiter or something like that). Don't worry about anything else.
+1 for everthing dipje said. If you are primarily doing audio for video, use 48K instead of 44.1. 24/44.1 is fine for most people, 24/48 for video, 24/96 if you want to maximise quality of some instruments and/or can record in an environment/gear that can make use of it, but that's not your typical home recording environment with a couple of budget mics...
I agree with everything said above, but IMO it also depends of your Converters / Plugs / Mix processing / And of course your project. For Music and Cd project, you could stay at 44.1 but if your system can stand it, then work at 96K.
I had learned somewhere undocumentable by now that ideally, sample rate should be twice the bitrate - that is, 24 bits at 48kHz. Mixing down from there to "Red Book" CD audio is no problem. If you are recording vocals or instruments on which you'll be using Melodyne or time-shifting such as in Logic Pro (or its Melodynesque pitch-shifting), it's better to record them in 24 bits, for more information to work with.
without spitting out facts from a book or a website, you still have an insane amount of headroom in 16bit in the digital realm. tracking should definitely be done in at least 24, but i also used to work for a studio that recorded/mixed many famous artists, and until this year, they were only running 24bit 48k...food for thought. technically 96 is better, sure, but if you are worried about it really sounding THAT much better, i hope you have excellent monitors, cables, and converters to hear the difference. a lot of places also "upsample" stuff just to appease clients that think they need 96k. technically better? yes. necessary? not by a long shot. some of your favorite songs that sound excellent were done in 24bit/48k. also, when you get to 32 bit float, it starts to kill your processor so make sure you have an excellent setup to be doing so, or you will have more headaches than breakthroughs. otherwise, listen to everything dipje said!
oh 24 bit/ 48khz is good but tape is best hahaha No seriously, Ive read a nice article (in german though) about that and it claims that if the end product is 44.1 khz, the working sample rate of 88.2 (double of 44.1) is good as it makes the downsampling easier. https://www.adt-audio.de/ProAudio_WhitePapers/GrenzenDerDigitaltechnik_1.html (GERMAN!) I think its a good idea and use it from now on. i mean if you take the "stepped" audio from for example 96 khz, how in heaven can downsampling to an odd number like 44.1 be a good idea? someone said something about intersampledistortion? what is that?
THE MOST RIGHT/CORRECT ANSWER FOR YOU IS HERE! That's how I work: 96kHz production/sound-creation, arrangement, mixing/processing, mastering (everything, but not including limiting/clipping/maximizing). Synths settings are maximum quality etc. Then I find audiosamples that are good in my project (from my large database of sound libraries - 99,9% usually one-shots of drums, percussions, FX; no loops). All samples and other audiodata that are not 96kHz (99,5% sample libraries are usually 44.1k), I upsample and resave as an upsampled copy. I use the most highest quality SRCs (SoX, iZotope RX 64-bit SRC, SSRC, dBpoweramp's 16.0 converter etc.). They are all tested and do not harm the sound, or do harm at the most possible minimum damage (linear-phase's pre/post ringing which is innoticeable at 22kHz range; lowest level of intermodulation and aliasing shit (lower than -170dB) and the most steepest high-cut filter at around 22kHz (so, widest range saving up to 99,9% of frequency range; linear-phase cause no phase distortions; if samples are clipped, I do gain down i.e. quieter a bit (usually less than 1dB)). Why upsample? Because if you don't, your DAW automatically will do this, and it will be much worse (aliasing, intermodulation distortion, unknown filtering etc.) - watch SRC here and compare: src.infinitewave.ca/ Usually I do some editing/restoration for sample to remove some stuff, after that samples are extremely clean, without any artifacts, as it should be, but this waste of time, but my choice is quality, not the time. Then when I complete this stages, I do pre-final mastering. Render before limiter/clipper/maximizer at 96kHz/24bit(32bit) stereo pcm wave, and then downsample it with highest quality SRCs (iZotope RX, finalCD, dBpoweramp 16 src, SSRC, SoX; I found Voxengo r8brain Pro and Weiss Saracon are worse than lised above) and save as 44.1kHz/24 or 32 bit stereo wave pcm. Again, no distortions at all, your frequency is full up to 22kHz with steepest high-cut filter near the Nyquist end, and linear-phase causing no phase-distortion at all (only pre and post ringing at 22kHz area, which is extremely innoticeable). Some audio editing is possible if you need it (usually fade-in, fade-out, pauses, silence before track start and after it stops, etc.). Then clip/limit/maximize with high-quality clipper/limiter, with oversampling, intersample-limiting option (aka true-peak) and final ceiling on your final limiter is -0.3dB (it is headroom for your true-peaks, because they are higher than digital peaks if we reconstruct them into analog true-peaks). Then dithering/noiseshaping for 44.1kHz/16bit wave/flac/alac export and for CD. After that, no processing anymore! And remember, dithering after 16 bit conversion doesn't work, it will be just noise above spoiled audio; so you should apply dithering before the moment that your audio will be final 16 bit file. Don't dither for mp3, they are exported from 24 bit wavs with ceiling(output level) -1dB or a bit more (-1.2dB for example)! Because dithering for mp3 is the same as to polish broken glass (it will be lossy-compressed noise which is only the noise and this will no save your audio in mp3 or other lossy formats). Why -1dB output level or quieter? because after mp3 compressing you will get peaks rised up to about 0dB (if you leave -0.3dB, after that your mp3 file will be clipped, even wih level +1dB or maybe higher). I recommend to compress at 320kbps mp3. Oversampling at 96kHz production/mixing/processing/premastering : in some cases when you saturate, hardcompress, excite etc. - you can use oversampling x2/x4 factor, usually it is much more than enough. It is for sounds which have a lot of highs, harshness etc. 44100 Hz/48000 Hz production/mixing/processing/premastering usually need (more) oversampling in some cases. Don't work at 176.4 kHz or higher. Not always the quality is better, sometimes it may be worse at these settings and highest oversamplings, also not all plugins operate at 176.4kHz or higher. 96 kHz is enough. Higher sample-rates and oversampling offer less aliasing, so your highs and mid-highs might be cleaner, fresher and less harsh, less digital. Speaking about maths like resampling from 192/96/48 to 44.1 or so - there is the 21th century, powerful PC, good software. Posted above SRCs work with extremely top-grade quality with resampling from X->Y kHz, than your DAW that can't do right even 88.2 into 44.1 !!! Just use special SRC which I post here, they are specially made for resampling. 24 bit is for recording/digitizing from analog (mic, hardware synths etc.), because in ADC/DAC the number 24-bit is possible maximum, but practically they are 18-22 bits; rarely you can find high-end hi-fi hi-definition 32 bit ADC/DAC devices, but there are not so much of them (1-2 of 1000 models), and hardly expensive. I know only one ADC with 156dB dynamic range. Usually they are 108-128 dB dynamic range. So for this tasks 24 bit is finally enough! Don't clip while recording audio! Headroom is needed. You may use this for render your audio as well. As well as for final distribution as hi-res audio. 32 bit float = quality is the same 24 bit but with clipping-preventing above 0dB; but I don't recommend to clip everything, headroom is the key to clean quality and good dynamics, but recommended for rendering. 32 bit integer(fixed) = 24 bit * 1,5 = 192 dB of dynamic range. I use this. I don't clip (so, 32 bit float is useless for me because of headroom below 0 dB) and I don't need intermodulation/aliasing stuff, and as well all sounds including quiet ones must be clean. So, precisely it is better than 24 bit. 16 bit with dithering/noise-shaping - for finally mastered FLAC, CD, WAV, other lossless - for to be distributed throughout the Internet and CD-shops. Dithering at 16 bit offers quiet parts to sound cleaner, less harsh. Noise-shaping masks the noise of dithering for to be less noticeable. iZotope MBIT+ in Ozone and RX is an excellent one. I use Moderate or Medium or High modes, rarely Ultra or TPDF. Dithering is after clipping/limiting/maximizing! SRC (sample-rate conversion into 44.1) is much better before limiting/clipping/maximizing! Never use SRC after dithering/noiseshaping, because noise shape for 96 kHz and for 44.1 and for other sample rates are different, so conversion will kill the part of dithering range (both spectrum and dynamic) and this dithering won't work anymore, it will be just the noise. So dithering is the last one in the chain! And remember! SRC after limiting will cause peaks rising (with linear phase it will just rising, with minimum phase peaks will rise much higher, because of extreme phase shifts), so it can lead to clipping of digital peaks and true-peaks (analog reconstructed peaks)! If somebody disagrees - relearn some theory, and imagine what is it all for and imagine all the processes in your head! Last edited: Aug 11, 2016
music is music. quality is quality. I won't listen to Martin Garrix in Hi-Res I won't listen to Pink Floyd's DarkSideOfTheMoon in mp3 ;D
yeah man those are different worlds! but either has its .."charm" but someone told me "no one will care what daw you used or what sample rate you recorded if its a good song"
So, if it doesn't matter, we can just use mp3 compressed samples and VSTs from 90s and the earliest 00s, do not pay attention on mixing and mastering and render final 192 kbps audiotrack. Who cares) if the idea of track is excellent, fuckdaquality)
hahah noo but in the area of wav 44.1 to 192, or 16 to 32 no one cares if basic principles are being considered... ;) sure it sounds different. but great musicians could make music with a toaster and two toothbrushes and you'd still tap you foot ))))
