Slightly disapointed in this video compared to the rest of your videos. Whilst I agree with your conclusions, you don't touch on a lot of the reasons why CD was at 44.1kHk and DVD was at 48kHz. Or why 24bit is better than 16bit. Sampling rates always have to be double the maximum frequency you want to capture (Shannon-Nyquist). So 44.1kHz sampling rate produces an audible frequency of up to 22.05kHz, the reason this is above the max human hearing range was to allow for filters. Instead of having a hard cut of filter to stop any frequencies entering the ADC at exactly 20kHz, a filter with a sloped cut of is used as the filter would be easier to implement and therefore cause less artifacts in the sound. Ultimatley, It was, as you rightly say, to be efficent in saving space/data. However, 48kHz is popular on DVD/films/TV because it is easier to sync to a variety of frame rates. Not because they want 24kHz of audible sounds vs 22.05kHz. Human ears can't hear that high, and even if we could the difference between the notes they would produce is small (e.g. the difference between 200hz tone vs a 500hz tone is much greater than 14.2kHz vs 14.5kHz) our ears have better resolution at lower frequencies. 48kHz is purley for ease of maths vs 44.1kHz in implementation and filters. 16 bit vs 24 bit determines the noise floor/dynamic range available. The human ear in ideal conditions has a noise range of 120-130ish dB. However, we all have to deal with background noise and the equal loudness curves show that its only at certain frequencies our ear can hear down to near 0dB. When CD was introduced it was taking over from tape. 96dB (16bit audio) was seen as more than enough of a Sigal to noise ratio compared to tape (as in Cassette tape even with Dolby-NR, not studio/reel to reel etc). However, 24bit (which provieds 144dB of noise floor/dynamic range) more closley resembles the human ear. So again, it was a compromise to, as you said, save space. Also in a recording environment, having that headroom is incredibly usefull - especially when recording classical music. So - I completly agree with your conclusion, just not the way you came to it!
I agree completely. Part of the problem is the manner in which he represents "headroom", i.e., drawing it in such a way that it makes it look like the amplitude of a waveform will exceed the range of 16-bit audio. This is, of course, completely wrong. 0db is 0db whether in 24-bit or 16-bit audio, and a waveform that exceeds 0db will clip in 24-bit just as much as it clips in 16-bit. Nevertheless there is indeed more headroom with 24-bit. It gives you the luxury of recording at a lower volume and applying more aggressive compression or other dynamics without losing dynamic range as compared to 16-bit. But this is all about mixing and mastering. Headroom ceases to be an issue with distribution, and 96db of dynamic range is more than enough for the most dynamic classical recording. As for 48Khz being more "edgy" than 96Khz? That makes me cringe. That's most likely a DAC issue with poor filtering and/or the result of driving speakers with high harmonics what are inaudible, but introduce audible distortion.
I think your explanation lacks one point. It is correct that 44,1 is basically implementation and filters BUT we are also getting only 4 measure points a 10khz. That makes high frequencies which are still very noticeable pretty low resolution. That'd be a reason to upscale recording or listening. Of course it has its pro and cons but I think it is very worth adding to your great note! :)
@@adrianzadi No, that's not the case. That's the beauty of the Nyquist theroem. Mathematically, at 44.1Khz, you have all the points you need to perfectly reproduce every waveform up to 20Khz, with a 2.05Khz buffer for the filter. There is no noticeable "pretty low resolution" at those frequencies. None whatsoever. The resulting waveform is perfect in every case, with all source material. Again, this is for distribution. For production it's an entirely different story, especially if you are doing time or pitch shifting. Then higher resolution is absolutely beneficial. It is also beneficial if you rely on saturation effects and are getting audible and distracting aliasing. Then 96Khz is beneficial. I generally produce and mix at 96Khz/24 and distribute at 48Khz/16.
Please. Sample rate is about frequency content. The higher the sample rate, the higher the frequencies the converter is going to registrate. Read about the Nyquist theorem to really inform yourselves about this and why you should care (the effect is called aliasing, which it is something we do need to avoid). I like Colt videos. I think he discusses many interesting topics about the art of mixing and how to run a professional studio. It is a magnific work he does. But he missed the point in this video.
This is true if we can say that digital anti-alias filters have no “sound”. The more of a brick wall the brick wall filter has to be the more likely it will impose some kind of sonic footprint because it is more extreme. For 44.1k it has to go from passband to stopband between 20k and 22.05k. Increasing the sampling rate means the filter has more space to roll off, is less extreme, there is less ringing on square waves and transients and is less likely to affect the sound quality.
My computer can barely keep up with 44.1. It's a 2019 Macbook Pro. So it's not like it's old. I'm constantly adjusting my buffer size, turning off plugins while recording, freezing tracks, turning on low latency monitoring, etc. I still have issues no matter what. "Computers are more than powerful enough." Yes, if you can just drop thousands of dollars at the drop of a hat. Many of us can't. For many of us, buying a new computer would require months and months of saving. If I am going to be spending thousands, it's going to be on acoustic treatment, new microphones, etc. Not gear so I can upgrade my sample rate.
There's something wrong on your side, not with many of us being unable. My desktop computer is from 2019 and it still creates bangers of a song and i can go to 24 bit if i wanted to and it's fine. Besides, Macbooks are built for audio and sound design which is why I'm curious to know what is going wrong. If my windows computer can mix audio better than your Macbook, then I think you need to make some changes or do something about it....
Same goes for anything that’s not so “dense” that you can’t discern the aliasing. Like he says in the clip more intimate music and what I consider more natural music like Bluegrass he samples higher, for good reasons.
The human perception of Sample rates above 44k is probably mostly imagined or placebo. What these higher sample rates really do is move nyquist very far out so some EQs and plugins that would normally cramp or unpleasantly distort (especially non oversampling plugins that have a steep high frequency boost or cut) will no longer cramp or distort any where close to the frequency ranges we can perceive. Unless we all start to develop mutant hearing or you are doing some really crazy $hit there is no reason to use 96k sample rates. The size of the files and burden of working with them doesn’t make them worth the struggle. At 24 bits you get 16,777,216 steps to represent something like 144db of dynamic range. You are lucky to get 119db of dynamic range into the A/D of most interfaces. 96 db of range is pretty typical. Also, most modern music has the shizzle compressed out of it and it winds up with almost no dynamic range. Opening most mastered songs in a wave editor will display the post loudness war modern music WAV log shape. You can probably fit that level of dynamic range in 12 bit.
I always figured any benefits of any sample rates greater than 44.1k has to do with plugin processing. I always go 48kHz. Now, as far as I know, higher sample rates on playback should be subtly audible. Completely unimportant, but subtle because, correct me if I'm wrong, the DACs in our gear have some of their work cut out for them in smoothing out those samples into continuous analogue signals. All DACs also oversample (4x-8x) for a smoother sound, but higher sample rates reduces their oversampling rates, and less oversampling will reduce how much the DACs impart their own sonic character on the sound, too. Splitting hairs here, though, you need high-end monitors to hear the subtle difference, it does come out in a recording when you've got Burr Brown DACs versus the ubiquitous AKM DACs for playback, but very subtle and requires high-end headphones or speakers to even hear clearly. I think I remember reading a long time ago that DAC performance improves with higher playbck sample rates, but this is for nerdy electrical measurement stuff that no one cares about, lol.
Can You imagine that everyone knows 12bit is more than enough and 60db of SNR is all they need in listening? Whole industry will struggle due to lack of selling new "high-res" stuff. 32-bit 1Ghz sample rate on the way.
After some experimenting I found that 48/24 is the rate that gives me the best results for what I do, on my system with my converters. There's an argument for recording at 96 and downsampling to 48 when mixing, but then again I'm a little sloppy and wouldn't want to constantly change sample rate... 🤷🏻♂. Also, let's not forget that the results will vary depending on the converters and clocking you are using! Peace and peaches y'all! 🍑
88.2kHz/32bit has been my default for probably 20 years now. I settled on it at the time because it divides evenly to 44.1kHz when downsampling, so in theory it would be less prone to artifacts. The converters still only had 24bit resolution, but saving as a 32bit file gives it much more headroom when editing and matched the internal resolution of the DAW processing.
88.2 kHz is from Bob Katz, right? But I wonder, did you ever try to downsample from 88.2 to 48/24 or to 44.1/24? I'm quite new to this and try to understand but there are a lot of bids out there.
You in theory shouldn't need to divide evenly when downsampling if your DAW does things correctly. It is entirely possible to reconstruct the continuous waveform for any frequencies in a lower resolution by the higher resolution samples.
I've been on 88.2 for the past 5 years.. 24/64 bit thou studio one default. No reason to switch from 44.1 to 48 since that's a 8.8% change in the sample rate only. At the most the filter on the interface changes slightly.
In my humble opinion… switch to 48 from 44.1 if your client base are working with this sample rate. But don’t use it if you think it sounds better… there is no audible difference between 44.1 and 48. 88… hell yeah… 96…absolutely and 192… mainly for cinema scoring… insane difference. I was asked by AES to attend a listening session at McGill university where George Massenburg use to teach…. The test was a blind shootout to listen to short snippets of a song in all sample rates. 44.1 to 48 was impossible to hear a difference… but as soon as it went up to 88 and above…wow… huge noticeable difference in how much more open not to mention the depth was. FYI… Just in case… I’m a Juno nominated songwriter/producer who has recorded for Dream Theater, Corey Hart, Harvey Mason Junior, Debbie Gibson, Jim Vallance. My 2 cents 🙌🏽
Must be a personal taste thing among those that can hear a difference. I believe Jack Joseph Puig prefers 44.1 to 48k saying 44.1 sounds more musical to him.
That bugs me too....big JJ Puig fan....i'm currently sticking to 48....I tried a 96 and ran into some conversion hassles? something was not as easy...maybe I will give 96 another try at some point.
44.1kHz will sound the most musical as that is the frequency at which DACs apply the greatest oversampling rates that they do to generate the smoothest analogue signal. Higher sample rates causes them to dial back the oversampling. This stuff is freaking subtle, mind you, and requires high-end playback systems to notice. For most people, it is likely just an imaginary difference.
It's really a 'headroom thing' as he says. 16/44.1 was chosen for CD, because it satisfies what the average human being can actually hear, so it makes sense to utilise the increased headroom afforded by 24/48 and 'chopping off' the redundant unused upper bits above 16 to produce a final product that uses as much of the information available without clipping. Sony did this for years with their 'Super Bit Mapping' of 20 bit recordings. Above 48K sampling frequency, there's quite a lot of measurable audio 'junk' which can impinge on audible frequencies, so is best filtered off, or simply not recorded, in the first place, which produces a cleaner perceived sound within the humanly audible frequency spectrum.
Something to think about: If you switch from 44.1k to 96k your plugins will use 4x as much CPU. Also, the higher the sample, the higher the Nyquist Limit. Meaning there is more resolution before signals bounce back into themselves and create Aliasing.
Why 4x if it's only ~2x the rate? Also, most plugins oversample anyway. When I first switched over to 96k from 48k, my CPU hit was ~50% more with the same plugins.
That's not entirely true. Most plugins today use upsampling to avoid aliasing, therefore working in 96K many times make the plugins less CPU intensive as the upsampling becomes redundant.
This Dan Worrall video "Samplerates: the higher the better, right?" is worth a watch for sure, but 2x the SR = 2x CPU. Sure, you can get audible aliasing at 96k if you have strong harmonics above Nyquist, but very few acoustic sounds have significant info up there, and any good saturation plugin (like Fab Filter Saturn) has internal OS and filtering. Not saying it's necessarily better to record at 96k, but there is much less likely to be any audible aliasing and the smoother LP filtering won't produce intermodulation in the audible range. The reason 48k sounds slightly more "edgy" is likely due to the intermodulation distortion. By all means, save the CPU if you don't mind that effect. It is very subtle. However, larger bit words don't use significantly more CPU, and won't generally task disks more due to the packet size streams. Yes, 16bit final recordings may sound nearly identical to the human ear, but that's because the entire 16bits is used. You'll want to leave at least 2 bits headroom when recording in most cases. Digital recording systems went to 20bits in the early 90s to eliminate low level hash in quiet sections. Low to upper midrange waves do exhibit some intermodulation distortion below -65dBfs in 16bit recordings. That may be audible in low level passages, and may be partially why hipsters find CDs "harsh". It doesn't happen until well below -100dBfs in 24bit recordings. I'll stick to recording at 32bitFL for that reason, and to ensure no digital overs during rendering.
The only difference between 44.1k and 48k is that that the later has content above 22.05khz (up to 24k), below 22.05khz the content will be and sound absolutely identical! Any difference is not rate related, you either have aliasing going on or analog converting issues.
Or "The number 48 is greater than the number 44.1" issues...😀. I didn't here one single reason for why 48 would better than 44.1 besides "CD:s have been used as a format for so many years". And people just seems to buy this here left and right! Because more is always better...
@@Magnus_Loov My thoughts exactly! Other than the fact that clients started requesting it, I see no reason to go above 44.1 if you're not working video material. Bit depth, on the other hand. makes sense to increase. I have mine set at 32 bit float for the unlimited headroom, but export the master at the usual 16. What I'm not sure about is if I should rather be exporting at 24 bit...🤔
Also gonna chime in with the rest who feel this was a bit disingenuous. Even among engineers, double blind tests have shown us that we can't consistently & reliably tell a difference between 44.1 & 96-192. Unless your final intended destination is DVD/Blu Ray, there's no reason to go against the established standard of mastering to 44.1 16 bit. Switching between sample rates that aren't perfectly divisable introduces aliasing from rounding errors for one thing, and what about music that gets a digital only release? Plenty of us still have CDs in our cars. Have you forgotten what happens if you don't dither to 16 bit? I've bought stuff from bandcamp that's been at 24 bit that I want to put on my phone, I really don't want to run it through my own software & dither it myself first. Also have to pick on his testing method. ANY time you record & compare between different performances, you invalidate your test. No matter how good & consistent your musician may be, there are ALWAYS going to be differences that influence the sound. The only accurate way of doing a sample rate comparison would be to split the signal & send it to 2 different machines. If he's as experienced as he claims, this whole video is really disappointing. Only going to unnecessarily confuse & complicate things for newcomers. I'm sorry dude, I've enjoyed other videos of yours but you should know better. Oh and "we have more than enough CPU power & hard drive space these days" is not a reasonable argument. Just because you can, doesn't mean you should.
its not that simple it entirely depends on the filtering methods in the converters / software. But yes your correct be it issues at low samples rates with or conversion back from higher rates there is always a compromise.
The idea of 22.05khz audio content is theoretical. Think about it for a moment. That scenario would have only one bit for the peak and one for the trough of each wave. Essentially reducing the input to a square wave. But even worse is that filtration and dithering are a necessary part of the A/D conversion and these cause foldback artifacts well below the input frequency. These artifacts are now living in the audible spectrum (most people can't actually hear 22khz). The best argument for 48k is pushing these foldback artifacts up out of the audible spectrum.
I feel the only reason to go higher than 48 is because of aliasing introduced by non linear processing. But a lot of plugins (most that I use) have an internal upsampling or the "HQ" button so even that is taken care off. I'm almost 100% sure nobody will be able to hear the difference between 48 and 96 if there is no aliasing going on from processing.
it gets a lil more complicated too, there are situations where lower sample rates with upsampling can actually produce less aliasing in multi-non-linear chains than equivalent higher sample rates. there's a fascinating paper on the subject by either Vladislav or Fabian who code for Tokyo Dawn.
A second reason would be if you intend to offer yourself greater flexibility in time-stretching or pitch-shifting on your sample, for using your sample in future sound design purposes.
Well, the benefit of high resolution audio is the same as high framrate video. If you wanna alter the speed of the audio you have more headroom before it starts sounding distorted. 48khz leaves you with around a 20% wiggle room in time stretching the signal. 96khz gives you over 50%. So if you're planing on doing some heavy retiming or speed alterations, record in a higher resolution. It's as simple as that.
@@bitman6043 It probably depends on the internal processing resolution of the plugins. Some process sample at 48khz while others sample at much higher resolutions. Some plugins even allow you do oversample internally.
This is a real world benefit right here. Sound manipulation , just like graphical image manipulation is best using a higher quality source as poissible to prevent artifacts.
Correction: 24bit does not have more volume than 16bit. It is just able to give more values, ie. 2^24 steps instead of 2^16. between the "zero" and the clip point (which is determined by you preamp and not your bit depth). Would add that lower sample rates only change the reproduction on frequencies higher then 20kHz, interpreting them as low frequencies, which *theoretically* may cause artifacts in the audible range. This may explain the "cleaner" sound you got at 96kHz vs the 48kHz. (which would have been nice to hear, even thou YT compresses). Keep in mind most instruments don't have an above 20kHz content in the first place. Also most synths have an anti aliasing option to mitigate these artifacts. Now would it be audible in a contemporary pop/rock mix, where songs are extremely loud, distorted, and compressed? IDK. probably not. Would be interesting to test.
This correction makes intuitive sense, but it's not actually correct. In PCM audio, each bit in the sample gives you 6db of dynamic range. It doesn't get scaled by your preamp or converter, you don't get finer gradations of volume level within a fixed dynamic range as you add more bits. 16-bit has a maximum theoretical dynamic range of 96db. In practice, dither and noise eats into that dynamic range. Similarly, 24-bit has a maximum theoretical dynamic range of 144db, but in practice you'll usually see good 24-bit gear with a dynamic range spec around 126 db (actual 21-bit resolution). On paper, 32-bit float has a dynamic range of 1528db, but that's also theoretical, because that's a bigger dynamic range than what is hypothesized to be the loudest possible sound pressure level on earth. As far as the stateent about sample rates, it's important to know that digital sampling introduces a ringing noise around the Nyquist frequency (half the sample rate), and that needs to be filtered out. For 44.1 kHz audio, that means you have sample noise at 22.05 kHz, and this becomes a problem, because no perfect brick wall filter exists. A low-pass filter that allows audio through at 20 kHz and blow, but effectively kills the signal before it hits 22.05 kHz, is extremely difficult to create, and those filters usually introduce undesirable effects in the audible frequency range. Using higher sample rates allows the use of shallower filters, with fewer detrimental effects on the audible range. Most acoustic instruments *do* have harmonics above 20 kHz, but most people can't hear them, and most microphones are going to have trouble capturing them accurately.
The Problem is, that the 48khz sells and streams on those platforms makes absolutly no sense for the end consumer. Because you can't hear the difference. Sell something for more money, that doesn't have any advantage for the listeners is a cash grap. 48khz makes sense from a producer standpoint and nothing else. But sure, if the costumer want it, he'll get it.
Sampling frequencies over 44.1/48KHz is mostly snakeoil/marketing hype to sell more gear, these frequencies can perfectly describe the soundwave, there is no need to waste computer resources and storage on higher rates. 24 bit matters quite a lot at the recording stage though to be sure to get the most out of the dynamic resolution, for reproduction of a mastered track 16 bits is just fine, the people that invented the CD really knew what they were doing. I record at 44.1/24.
I repectfully disagree. I'm not in the industry, but I do have a decent HiFi. I own two copies of Aftermath by the Rolling Stones, one at 24/88.2 PCM, the other is DSD64. And, even at those resolutions, I can hear the difference. The DSD64 version sounds better. I'm not talking about by a tiny amount either.
Sampling up to 192 allows for easy use of common digital reconstruction filters of the slow type for better impulse response without HF roll off without needlessly taxing hardware or software upsampling resources. Upsampling to 352/768 is a joke. DSD native is often much better recorded content with much more dynamic range and care taken in mastering but is huge in file size. Upsampling to DSD is an absolute waste of resources. Most DSD native content is super niche or esoteric or classical. Tape to DSD is the best way to distribute master tapes and you can purchase FLAC versions if file space is an issue.
I've been mixing and mastering music for 15 years and I've come to find my favorite sample rate and bit that is 192000/16 -bit. I used to recorded master 192000/24-bit and I lost a lot of the low-end frequencies and the bass out of my tracks, and when I put it back to 16-bit I get the bass back.
@@The.Original.Mr.Xyou can disagree all you like. The math that goes into reconstruction of sound waves requires no more than 48k to reproduce all tones heard by the vast vast majority of the population. Anything more than that is solely to record inaudible frequencies and slow them down.
If you have good ears and a good system you can easily tell the difference. Listen to the clarity and high frequencies especially. The higher the sample rate, the higher the frequency range because of the increased nyquist frequency
Excellent presentation, thank you. I gotta smile when I realize that all of this technology ends up going into 'earbuds'. I've been doing this since 1961, my studio monitors back in the day were 12 1/2 cu. ft. apiece with 15" woofers, etc., custom built. And today kids think 'subwoofers' deliver bass. Thanks again, Bill P.
It seems everyone has forgotten two things about the sample rate of music: if you record or synthesize frequencies higher than the cutoff, aliasing is highly likely to occur, which basically means they start bouncing back in (seemingly random) areas in the audible frequency spectrum. Even if there was little to no aliasing occuring, you don't have to be able to physically hear past 20KHz to hear the effects of the frequencies past that range. Certain frequencies can cancel out or amplify other frequencies, even modulate other frequencies, many of which could be potentially audible. This is most applicable when using speakers and not headphones, because the frequencies travel through an open medium where frequencies can physically clash or cancel out in space. It's important to note that I don't think these various effects happen digitally, so these things will be experienced to a minimum through headphones with little to no medium to travel through except your ear canal and head/skull. When you play the same track out loud in the open, the issues could potentially reveal themselves. This is why I personally believe it would be *best* to export using the highest sample rate to avoid aliasing and other artifacts from "lost" frequencies, however, I too use 48kHz because other sample rates can increase CPU usage or even the chance of incompatibility or having to convert to 44.1kHz/48kHz (takes a while). The differences usually won't be too noticeable anyway. Correct me if I'm wrong, as I'm not extremely experienced with audio, but I take in a lot of info and run a lot of tests to prove my theories, and I've seen, with a spectrum analyzer with my own two eyes, serious aliasing from producing frequencies higher than 20KHz on 48KHz sample rate. I can't confirm the second theory but I am pretty sure anyone can agree based on the reasoning. This was a reply to another comment, so I'll add context: The cutoff of frequencies audible is dependent upon the sample rate. 44.1kHz - reproduces up to 22.05kHz frequencies in the exported audio file 48kHz - reproduces up to 24kHz frequencies in the exported audio file Filtering is used to cutoff the remaining frequencies nearing this limit. The higher the sample rate, the smoother this filtering can be (so less artifacts occur)
The good news is that the converters actually sample at a very high resolution both in bit depth and SR. This means that the info is captured and decimated to the target output rate. So difference between rates is not super significant.
I love your other content but please stop talking about technical stuff you don't fully understand because what you're saying in this video is bs, sorry... Please do some research on how A/D-converters work and what's the Nyquist theorem because then you'll realize that the sample rate only determines the highest frequency that is digitized which is roughly half the sample rate, so 22kHz when using 44.1kHz as a sample rate. I don't know the conditions of your drum test comparing 48k to 96k but I guess you somehow tricked yourself because technically the only difference is that one file contains information up to 24kHz and the other file information up to 48kHz. Since our hearing ends at under 20k you shouldn't hear any difference at least in the raw files. Because yes: when using dynamic processors or saturation higher sample rates have their right to exist because it reduces aliasing. But recording at 44.1k and then using oversampling in these kind of plugins is the way to go.
Completely disagree. On a purely technical level, Tidal will stream up to 9216 kb per second, that is equal to a 32 bit, 96K file. So I think you are better off going with the highest resolution you can handle. Not once did I mention the nyquist curve because it’s irrelevant to my point here, which I think was well explained and I stand by it. Thanks for watching.
Regarding the bit depth I totally agree with you: bith depth DOES matter! But nobody in the world should hear a difference between 44.1k/32bit and 96k/32bit. That Tidal offers 96kHz streaming is technically bs but I guess they offer what customers pay for and if people are willing to pay more for getting 96kHz sample rate instead of 44.1 they're happy 🤷♂️ and as already mentioned in another comment here: a correctly upsampled 44.1k file should sound exactly the same as a 96k file.
this is the only good and correct comment. Another reason why you would want to record with a high samplerate is because you want to manipulate the sound by for example pitching it down but as you said there will be precisely no audible difference between listening to a 16bit 44.1kHz wav file compared to a 32bit 96kHz. Anything that does sound like there is a difference is either from the samplerate conversion which should also be pretty much inaudible or it's plain placebo.
48k gives you just a smoother filter in the ad conversion. This filter and its settings (differ per sample rate) is what gives the sound difference in your ad converter when switching between the sample rates.
Some home audio equipment can not produce a flat frequency response to 20kHz, as it should, with 48kHz sample rate, the highest frequencies droop. When the digital file is converted to analogue music, the hardware has to filter out the alias image. This filtering (different DACs use various mathematical formulas) can create distortions (whether you can hear them or not) in the time domain, it’s often called ‘ringing’. Some Hi-Fi companies choose filtering that only produces a correct frequency response with 96kHz to avoid this type of distortion. It’s pretty common, especially on high end equipment. So if you want all listeners to hear what you are recording or mixing as you intended, you need 96kHz sampling rate.
How do you go about using samples that are 44.1khz (which most samples are) in your projects. If youre running your project at say 48khz the computer will need to upscale all 44.1 samples to 48 and bring loads of artefacts.
When you are talking about 96k sounding clinical vs. 48k you are talking about double blind testing where you weren’t told what you were listening to and just had to make a call of what sounded better based on how it sounded and you wouldn’t be biased by your preconceived notions right?…Right????
If you understand the math about sampling and sample rate you know that 44.1k is literally more than enough. Bit rate you are right, but you should read more and not just base your opinion on experience. Read about the nyquist theorem, and stop saying things like 96kHz sounds clinical because it is bs, no one could ever actually hear a difference. Normally i love your videos but this is kind of misinformation.
To clarify: i do work on 48kHz for video, and do agree that people should switch, but some of what you said is wrong, and this being an educational channel, you should be more careful
@@ColtCapperrune i think that you should tell him to look for example on iztope RX and have look on the amplitude sound and headroom if he dont hear or feel defrience than maybe he came see 😂😂😂
@@RollRollunio 44.1 and 48 both have the same headroom, bit rate defines headroom, sample rate only affects the frequency limit of what you can theoretically sample...
Been 48/24 since I started mixing. But then, started digitally with ADAT with 48k 16bit, mixing on a mackie 32x8 - back when gain staging every step was critical to get a good digital mix.
for what it's worth, i saw an interview with pete townshend (the who) quite a few years ago where he said he recorded everything at 96kHz, but was just waiting until computers were powerful enough to run 192 kHz
Yikes, 16 bit does not have "less overall headroom in the volume or amplitude department" than 24-bit. And obviously the signal chart you drew would clip if recorded at 16-bit, and we know that doesn't happen. 24-bit has the same maximum, it just has smaller increments between possible values (like drawing with finer graph paper). Yes, I agree it's a dumb idea to record and work in 16-bit when 24-bit and 32-bit float is so cheaply available. 48k allows slightly more flexibility in the antialiasing filter requirements, if the goal is 20k bandwidth, or the same filter spec but just a tiny bit more bandwidth. I don't fault anyone for preferring 48k over 44.1, but it's just allows a slightly more relaxed filter or a tiny more bandwidth-it's not the world of difference that you imply. It's less than 10% difference, and we hear in log frequency, which makes it a scant difference.
A lot to think about, and weird that this video came across for me right after I finally switched from 44.1kHz to 88.2kHz lol. When I first started learning all this, I read a lot of "use 44.1kHz because CDs, and dithering down to 44.1 from 48 or 96 is wonky/noisy)" and just kinda stuck with it since. But I finally told myself "this isn't a thing anymore!" I did take in what all you said, and so for most of my stuff being in a metal arena I'll go for 48kHz and save 88.2/96kHz for when I pretend I'm a composer (lol...). I have always used 24-bit, at least. thanks for sharing!
Yep that’s why “high res” even though it’s a marketing term 24 bit SHOULD be the standard. The truncation of bits from 24 to 16 with adding dithering white noise is pointless in todays worlds. The difference between 16 to 24 bit is only about 200 mb. There’s no reason why the consumer can’t have the “master copy” not to mention the 24 bit is better from them to process third party dsps in playback much like processings plugins when mixing and mastering I have re-recorded 24 bit to 16 bit mixes on ADC and it’s fine with no added noise and it’s sounds better than the truncated 16 bit version as well but there’s no point for it in this day in age. I don’t like the use dithering and truncating the bits as well.
I'm curious, the listening test with 48 vs 96 you did was a double blind test? Also, the debate on 48 vs 44.1 is actually about where the artist thinks his music would end up (Tidal .. DVD etc) not a difference in audio quality. Because, again, a double blind test will tell that is no difference. Fun fact: Metallica The Black Album was mastered from DAT tape 16 Bit/ 44.1 KHz. Sounds awesome even today compared to any production.
(Apologies if this has already been addressed) * If you want to mix in Dolby Atmos, it’s supposed to be 24/48. * The higher sample rates are more useful when dealing with non-linear processes like harmonic distortions/saturation. However, it is better to use oversampling within the plugin rather than natively work in that higher sample rate. * If your source for a vinyl record is digital, they actually recommend 24/44.1. This has to do with the cutter heads and not overloading it with supersonic frequencies that are beyond what the heads can cut, let alone human hearing. (Feel free to correct anything that is wrong)
Yep, Atmos is at 48 kHz. But the bit about supersonic frequencies disrupting cutting heads doesn't make sense to me. It was the super low frequencies that disrupted cutting heads.
@@Technoriety I did some more reading on this. Frequencies above 20kHz can overheat the cutting heads and one has to be careful in the high end (sibilance especially) as well as the low end. That said, it seems many of the pressing plants accept digital files with a range of sample rates. They *may* downsample it to 44.1 on their end to make it easier or if it proves to be difficult. This comment about the sample rates also came from a friend of mine who has a lathe cutter and has volunteered to make some lathes for me. Again, feel free to correct me if I am wrong. =]
I only record at native CD rates when only directly producing to a CD. Everything else is higher bit rates. Edit, In regards to bit depth, remember when recording there is headroom. For every 6DB, there is one bit of resolution. So if you record with 18DB of headroom, by recording at -18dbFS, the upper 3 bits are never used, just there if something peaks you don't want clipped. This makes a 16 bit recording use only 13 bits. However, if you master at 24 bits, now you are working with an effective 21 bit recording, When you normalize the recording for a 16 bit CD, you have 16 bit depth on the CD, not 13. It is in the math of binary. This is a purely technical discussion, and does not address the "Sound or Feel" This is about accuracy and signal to noise. To stuff your pipe on 16 bit the renoun Neuman microphones of many thousands of dollars have much less dynamics than a 16 bit recording. I know the microphone is not digital so sue me. They do have a Signal to Noise ratio. The signal to noise ratio is 80 dB for the U 87 Ai. The signal to noise ratio of 16 bit audio is 96 db, which means the microphone noise floor can be accurately recorded in 16 bit audio.
as a working studio musician. 95% of the studios I've worked at including sunset sound, capitol, universal, resonate.. use 48k 24 bit w few exceptions 88.1 or 192. I can't tell you why. However, when it comes to art. The STROKES 1st album was tracked on a digi 001 at 44.1 24bit
Aside from the fact that he was talking about sampling frequency - you`re correct - it is true. But all that one actually needs to know about 24 bit is, that 24 bit gives you 144(!)dbfs of headroom. Which says pretty much everything. That is _way_ more than _any_ analog noise floor . So essentially the noise floor of even the highest quality outboard gear - be it whatever it may be you're recording - will mask any possible quantization errors by a huge margin, where the quantization errors of 24 bit is effectively at such a low level, that _no one_ will ever be able to hear it - especially considering that there's actual musical signal at a way higher level on top of it. Bottom line: Everything higher than 24 bit is in no way ever of interest for human hearing - even for the best ears that ever existed/will exist. At a typical maximum dynamic range of the program material of say from -2 dbfs (to avoid intersample peaks on the output) down to -50 dbfs - even considering for a moment that there' s absolute no noise floor of the recorded gear - you'll still have way over 60 dbfs headroom, before these quantization errors actually may occur.
@@broklanders4730 Well put. However the mention of 144db of headroom, is actually dynamic range. Headroom is the difference between signal and 0dbFS. To ship a 24 bit recording for mastering, the headroom is removed, so a -18db recording when normalized, is no longer 24 bit as the recorded depth is less.6 DB per bit, so a 24 bit recording at -18DBFS, only has 21 bits of information. This is the primary reason masters for CD are recorded at higher bit depth, so when the headroom is removed, and gain boosted with compression in mastering, the final product still has some valid data in the lower bits.
@@isettech - it's a common misbehavior (even falsely requested by mastering engineers) that there's any benefit by delivering the to be mastered track at -18dbfs. you'll be all good if you're just not clipping the individual tracks when bouncing stems or the summary master - and even that won't hurt nowadays, because we're in 32 bit/64 bit systems. so the mastering engineer would only have to normalize the track(s) prior mastering and he'd be good to go without any damage... and if the track(s) are too loud (even if not clipping) - he'd normalize all tracks o his working level anyways, which would be done in just a second.
I did wanna say colt didn’t really explain what the actual advantage is to using a higher sample rate. Humans can hear from 20hz to 20khz obv. So when your computer has to sample the analog signal 44.1 kHz sample rate will capture up to abt 21 kHz in analog (because a single wave has + and - values). The only real reason to work at high sample rates is if you actually need information that is above human hearing. For example changing the pitch of a sound or over sampling with a LPF outside of what we can hear so that data isn’t reflected back and causes aliasing. Essentially 48 kHz sample rate is typically seen as the best balance compared to other sample rates, because it is efficient with CPU (being marginally harder to run than 44.1 and 50% of 96) and it gives a bit more overhead for higher frequency sounds. 44.1 is allowing almost no overhead. Anyways hope that helps! Great video Colt
My question is this, if the highest frequency a human can hear is 20 kHz (which it isn't and closer to 18 or 16 kHz), so why then do you need to have a higher sample rate than 44.1 kHz, which still snapshots more than twice a second for the highest frequencies we can hear?
When you use processing that generates harmonics (distortion, saturation and to a lesser extent compression), those harmonics technically go past 20khz, although they have a smaller amplitude. If the plugin is not properly written, those harmonics can be "reflected" back down below 20khz (imagine a mirror at 20khz ish) and you can get some weird stuff in the high end. This phenomenon is call aliasing, and oversampling usually attenuates its impact. There is a great video on the Fab Filter youtube channel about this topic if you want to go deeper.
Also, some people (not me) claim to perceive a difference in transient response at higher sample rate. The idea here is that the very first part of a transient is so quick that if you analyze its frequency spectrum it goes above 20khz, and while we can't perceive a sustained tone at 22khz for example, we might be able to perceive it as a transient. This assumes of course that the whole chain ( from microphone to speakers) is able to reproduce those frequencies.
In playback you don't. The exported audio file can be 44.1. It will make no difference. This is a video about mixing. You want to record at a higher frequency to avoiding ailising.
Audio information present in 96k but not in 48k is inaudible. Actually audio information present in 48k but not in 44.1k is inaudible. When processing, oversampling can make sense because of aliasing issues. But for a final product, more than 48k is unnecessary. A 32bit float has a 24bit mantissa. The advantage of 32bit float is for signal levels significantly below 0dbFS. (If you are working in 24bit, and you're signal values, consider as being in the range [-1,1], stays below 0.5, then you've lost one bit of resolution. But with a 32bit float, you have the same 24bits of resolution, and this keeps on being the case. If you turn a 24bit sample down 36dB, an then up 36dB, you lose information, with float you lose a lot less.)
Nearly every recording we do in Hollywood is done at 96kHz minimum. Then we do any and all time stretching or alignment. Then it’s usually downsampled to 48kHz when being handed off to mixing. Mastering is then done either at 48kHz or 96kHz (upsampled). I’ve personally heard the same 48kHz mix going thru the exact same master chain and converters at 48k and 96kHz, and the higher resolution was then downsampled back to 48kHz. I was very surprised that there was a very audible difference between them. The 96kHz was the clear winner. I wish they had also shown me a 192kHz downsampled master. My curiosity is killing me. 😂
This really doesn't make any sense... There shouldn't be any audible differences in the audio. The only way there would be is if harmonics generated by something in the FX chain led to aliasing, but you'd need to actively avoid oversampling options in your plugins and push them hard. I'd be curious to know what mix you were listening to where the difference was that obvious and just how high a frequency you can hear because you'd need superhuman hearing to hear a difference. (Human hearing tops out at around 20kHz and anything recorded in 96kHz has frequency content up to 48kHz, well above the audible range.)
Hy ..are there Problems (maybe Ableton) wen u record first in 44.1k and some clips in 48 k ?wen u mix this what u think are the results..?sorry my englisch..big Respect and best wishes from Vienna ✌️🇦🇹
I use to work for a guy who help develop the CD's for sony and 48k is what they liked the best but couldn't fit all on a cd... So I have always used 48k 24 bit. thanks for the conformation.
It's not really about anything other than aliasing. I work with Bluegrass and Folk more than anything and work at 96K and here is why. Aliasing causes transient smearing and inter-sample modulation distortions that basically cause frequency clashing and masking. I've worked a 96K since 2012 not because of any other reason except lower latency and more importantly less aliasing in the audible range. Modern computers have no issues recording at 96K and sessions aren't really that big when you compare them to say 15 minutes of 4K 60 H265 footage or any of the video codecs. 192 is problematic since some of the most popular plugins don't work correctly at sampling rates above 96K like Waves and some instruments like Arturia. On the point about 16 bit vs 24 you are not giving up headroom you are lowering the noise floor at 24 bits from the 96DB noise floor of 16 bits. Noise floor and headroom are two different things, the only time you gain headroom in comparison to bits is at 32 float which gives you 768db above 0 DBFS. Yes there are 32 bit float R2R ladder AD/DA converters like the sound devices mixpre and others but most audio interfaces on the market today are 24 or 32 bit fixed point. But in the end the song has to be good and like my mentor told me back in 1995 no one can see your gear on the radio. Cheers Colt!!!
Good post! Although it’s the filter that removes the aliasing that creates the smearing rather than the alias image itself, maybe that is why some people like non-oversamplig DACs.
This is such a bollox argument. The simple truth is regardless of how crystal clear/super hi fi a song may have been recorded, if a song is strong enough, unless the recording is full of hiss, a good song is a good song. What's the saying, 'you can't polish a turd'.
why not 88.2 and then downsample to 44.1? perfectly half. depending on the resampler, going from 48 to 44.1 can introduce artifacts that wouldn't otherwise be present if you just recorded in 44.1 from the get go. maybe what you're describing around 8:00 is that distortion (euphonic) that's being added in when you resample
@@vinylcabasse it seems you can get exact math with other SR too, with some other math: "to resample exactly from 44.1KHz to 48KHz you need to use the ratio 160/147 (and the inverse for the other direction)" it seems sample rate choice is more important for other stuff: ruclips.net/video/-jCwIsT0X8M/видео.html
That was very helpful and enlightening I've switched to 24 bit. Like you said there's more room on computers these days but stuck with the 44k. And didn't really know why I should go to 48k. But now I'm convinced. Thanks
Wait if you change that rate when you try to sync a music video most cameras are using the 44.1 but soon as you try to sync it becomes a drift nightmare
The human ear can not hear beyond 44.1. Certainly not 48khz. Anyone who says they can hear the difference is lying. 16 bit is old now. Ok agreed. But 48khz is an illusion. A bluff.
Not really true. I myself have a hard time differentiating it on a full mix, but it brings obviously a different sound at the recording stage. It is still subtle but you CAN hear it easily on a good system. There is also this one video on RUclips where they did 44.1 | 48 | 96kHz recording elecric guitar and it is easy to spot the differences.
Is the grit you hear in 48k possibly from foldback distortion/aliasing from not oversampling? Kind of like keeping more saturation which our brains like to hear.
Totally agree on 24/48 as the default. RE 96 some of what you are hearing may be jitter related and demands higher quality clocks. i.e. 48KHz is less demanding on having high quality clock generators.
I'm curious how you determined that it was sample rate in general rather than something about your specific converters that caused the differences you heard.
Interesting points BUUUUT… I prefer the sound of 44.1 and so do a lot of people. Why?.. who knows! It maybe that crystal clocks have been made in 44.1 for longer? Tighter clocking? I don’t know, but you should ask Jack Joseph Puig why he also prefers 44.1 over 48 🤷🏻
I have to disagree, but just on the delivery format for mastering. As 44.1kHz sets our nyquist frequecy just above the hearing swell, it is more than enough for anything "final". And as upsampling (for example for a music video) is often way more transparent (except if you have a lot of processing power and the correct software/filters) than downsampling, it only makes sense to upsample instead of downsampling if another samplerate is needed. Correct me if i'm wrong of course!
Yeah, I have to completely disagree with that. You always want the ability to have the highest resolution you will ever need. One day we will be listening to 24-bit 96K files over streaming. It certainly would be a shame if you only ever had 16 bit, 44.1 files, and then up sampled to 24/96. because you will absolutely hear the difference between 16/44.1, and 24/96. hope that helps!
having nyquist freq. right above our hearing range aint enuf from preventing aliasing 100%. it recochets back from the limit back to our hearing range (depedning on material). so it is better to have a higher sample rate just to b safe.
@@ColtCapperrune Fair point, but as i understand it, resolution is in the Bit-Depth. So you are right, it would be easy to tell 44.1/16 and 96/24 apart. But you aren't able to hear the difference between 44.1/24 and 96/24. That can proven by taking a recording at 96/24 and downsampling it to 44.1/24. If the downsampling has been done correctly (high filter steepness and using a linear phase anitaliasing-filter), the only difference in the files lies upwards of 20kHz, which is inaudible. If somebody downsamples incorrectly, there can be some loss in the upper frequencies. If somebody upsamples incorrectly, the only change would be imaging above the original nyquist. TLDR; In my opinion you are correct about the Bit-Depth, but not about the samplerate.
So you’re saying you can hear a difference between 48k and 96k ? You must have been blessed with golden ears. Most people can’t tell the difference between a 320 Kbps MP3 and a 44.1k wave file.
Thank you for a really interesting view on this Colt. I read that Bob Katz preferred to record in a samplerate of 88.2 kHz. Have you ever done that? And when you say that you record in 48 kHz 24 bit, does it mean that you go for the same settings from recording to mixing to mastering to release? I'm quite new to this and try to understand because I heard someone say that you alway should record, mix and master at a higer samplerate and then sample down to the release value, whether it's 44.1 or 48.
Thanks for your opinion, lately I indeed was wondering if the 44.1k would still make sense, since there's no more CDs to be bound with. But at the same time I was wondering if audio streaming services are still sticking to the 44.1k, which is not so clear to me. But nowadays I'd gladly do everything in 48k at least to avoid samplerate conflicts in Windows that now seems to use 48k as a system standard.
Yeah, I was thinking the same. But my conclusion was, record at the highest sample rate that you reasonably can. Then you will be covered. But then, you have conversion to worry about. I think aliasing is worse for us than sample rate conversion.
When the CD standard was developed it was based upon what humans can hear and also what was practical at the time from an electronics perspective. 700MB CDs were huge at the time. 44.1khz allows sounds up to 22khz to be recorded and reconstructed faithfully. 16 bits is plenty of dynamic range for most audio purposes. Most computer hardware was 16 bit capable but 32 bits was not common in early 80s. So for playback 24 bit 48khz is more than ample especially considering the bulk of the audience cannot hear it and listen on Bluetooth or highly compressed audio formats. Higher bit depth will increase quantisation noise and so will require dithering to be removed. Higher sample rates will allow you to sample supersonic (higher than human hearing) signals. The only justification I've heard for higher sample rates is when processing effects to make it easier to filter as aliased signals will be more supersonic requiring filters that are less complicated. That said if you over sample you won't hurt anyone and worst case you will be using more CPU and disk space for diminishing improvements in audio quality.
How its played is irrelevant. Ailising can only be introduced in the mixing stage. Once the final song is exported. It can be converted back down to 44.1khz without a flaw. 44.1khz allows perfect reconstruction of the entire audible range. There are no advantages to playback at 48khz response or even higher.
Sorry but this is misinformation, the only way you can listen the 96k is if for some reason your hardware performs differently than 48k (LPF aliasing etc), human ears are unable to listen the extra frequencies that are included even at the 48k. 48k has a highest freq of 24kHz and our ears at the best case can listen up to 20k. There are some use cases for high sample rates such as pitching down an audio file (higher freq that 20k are going down to the audible range) or aliasing issues in some old plugins but none of the stuff you are claiming are scientifically true.
Another great tutorial from The Man!! I recorded my song at 96, sounded great! Saved to it usb, took it to my car, and the car stereo wouldn't play it! ha! Only played my previous version I recorded at 44.1..
Not mentioned here or in the comment: non-linear processing (compression, saturation) without upsampling have serious aliasing artifacts at lower sample rates. Many waves plugins have this problem. Worth it to at least mix at 96k unless all your plugins have internal upsampling.
If you're using Lightpipe and need a lot of channels, 48k/44.1k give you eight. 88k/96k fold those channels and only give you four. On the flip-side the higher the sample rate, the lower the possible buffer size and subsequent latency. That's the determining factor for me: if I need extremely short latency, I use 96k. If I need more channels, I use 48k.
I track everything in 96khz 32 bit. I have the Apple silicon and latency has not really been an issue for me since I’ve made the switch. I can honestly say if I had to switch back to 24bit 48 kHz tomorrow it wouldn’t bother me to much. Great video man. 👍🏻
Hey there, 96k inherently has less latency than 48k. Juliane Krause' videos do a great job of demonstrating measurements for different sample rates and buffer sizes. Switching from 48k to 96k requires more bandwidth, so the threshold where an audio signal will start clicking or dropping out will decrease. Latency improves because audio is being sampled faster.
If you are entirely in the box I would suggest staying at 44.1. If you're not capturing analog signals then there are scenarios where up sampling is actually worse for your audio depending on the processing used.
@@Robangledorf if you want the full explanation my knowledge is referring to a white paper by either Fabian or Vladislav of Tokyo Dawn fame. Dan Worral has also done a good video on high sample rates. But to try to sum it up if you are using non-linear processes in series it can be better to let them upsample and sample internally as they won't 'pass on 'as many alias-able frequencies to the next process.
I think it's interesting to note, for historical perspective, that the only real reason why CD's were limited to 44.1/16 was because that was the rate and depth at which a reasonable amount of music (about the same amount as an LP) could fit onto a CD in PCM format. It was a technical and marketing choice, not one of quality. Now, sure, the marketing at the time claimed that this was the highest quality one could ever hope to hear. Heck, I STILL encounter people TODAY who claim that the human ear can't detect any differences above that standard -- even though all valid research proves otherwise, and the fact of the matter is that the human ear and brain ARE capable of detecting bits as opposed to a full waveform, at least on the subconscious level, at lower audio resolutions. In fact, your observation that 96k sounds "clinical" goes toward proving that: it's closer to the full waveform, so it's going to sound cleaner than anything below it. But that's a whole other tangent for another time. My point is simply 44.1kHz at 16 bits was a standard established for pragmatic reasons in the late 1970's and should not be viewed as any sort of gold standard today. Technology has advanced, and in this case, newer truly is better.
Yes, the CEO of Phillips wanted a certain piece of classical music (can't remember which) to fit on a single CD and the typical playing time of that symphony was 74 minutes.
As a former senior sales engineer at Sweetwater I was surprised to discover there was so much difference between the quality of AD/DA conversion paths from different manufacturers. Apogee for instance for quite a while produced a MUCH more accurate sound in 16/44.1 than several of the newer 24bit/48kHz and even 24/96 converters from other manufacturers. This is when the conversation began about the importance of the analog portion of the path in this process as well as the quality of the anti-aliasing filters. Many manufacturers would include higher sample rates as a way of getting around their poor anti-aliasing filters. Over time converter paths have become better over all, but there are still very noticeable differences in the quality of AD/DA conversion from manufacturer to manufacturer and within brands themselves. I mention this because while bit depth and sample rate are important (and I fully agree 16/44.1 needs to be dead because of current streaming capabilities), it's super important that the user is aware of the weakest links in their chain so they don't unknowingly sabotage their path. For instance. There are a lot of people out there (who I've dealt with personally) who are using 24bit/96kHz or higher who are using microphones that aren't capable of recording frequencies higher than what 44.1kHz sampling allows capture of. The highest frequency that can be recorded is half the sample rate (Nyquist theorem), so if someone is recording in 96kHz that means they are able to - in theory (depending on the quality of the anti-aliasing filtering) capture frequencies as high as 48k. But if their microphones are only sensitive to 20kHz, why are they recording in 96kHz? 48kHz also handles beyond 20kHz, but again, if your mics aren't sensitive beyond 20kHz then what's the point? I use 24/48 because I mix for film and TV and all audio projects coming in are 24/48 or 24/96 so I work with what I'm given. Because computers and drives can handle that amount of data processing and storage ok now I'm ok using those higher rates, but 99.9% of the time it's just wasted extra disc space to use any sample rate beyond 48kHz. The dynamic range and signal to noise ratio improvements we see in 24bit are DEFINITELY worth it but sample rates are negotiable for most folks. So my point is, everyone needs to do their homework and determine what they are actually capturing - which is determined by the weakest link in their path whether it be mics, converters, cables or even just the noise in the room - and then use the settings that make the most sense. If vocals are being recorded, or bass or guitar how high of a frequency are you actually WANTING to capture? Most of the time NOT up to 20kHz. Other than classical and jazz, most projects being recorded today still only really need a great 16/44.1 converter path because they aren't needing beyond 15kHz usually and dynamic range is so minimal in pop, R&B, rock music etc that 16bit is MORE than ample. But, to give people a little assist in getting away from poor anti-aliasing filters in less expensive home equipment, sure, go ahead and record in 24bit/48kHz. It probably won't help your sound depending on what you have plugged into the front of the path, but it won't hurt either. Plus Colt makes a great point about being able to deliver to mastering in the bit depth and sample rate that mastering will use and will ultimately be used for delivery. This is especially true now that RUclips and other streaming services now accept and stream in 24bit/48kHz. As a sound designer one of the advantages of using 96kHz-192kHz is the ability to manipulate and slow down sound for some pretty crazy cool sound effect creation. Other than that, upper sample rates are pretty pointless except for those using it as a mandated standard procedure for archive purposes.
The audio interface seem to allow higher buffersizes at 48k and in 44.1 aliasing is much more audible. Just open a synth without oversampling and make a sine sweep in the high frequencies. Sounds like an AM radio until switching to 48.
48 khz/24bit is more than enough for most genres. I encourage every mix engineer to learn about aliasing, and use oversampling when appropriate (Reaper can do that for you pr. plugin ;-) ), but remember that oversampling does not improve all plugins! Also there are a few places where 96 or higher might be cool...like when doing samples and stuff like that. But for 99% of us 48/24 is the better choice. Oh and sometimes aliasing can sound BETTER than when removing it! og and did I mention that Reaper can oversample pr plugin instance 2,4,8 or 16x ?.. and often that oversampling even sounds better than the build in oversampling that some plugins have .. lol.. I know.. I'm a Reaper fanboy. NB: oversampling on EQ's does rarely give you anything .. so save it for compression, saturation and the like -> stuff that is non-linear
Thx for the upload! After watching I quickly starting comparing sample rates from my last few tracks. You are definitely right, sounds better in 48 > 44. I also compared 64,000 SR to 48 and found that 64 sounds a bit cleaner and still gives that slight ruggedness ❤️. I’ll be bouncing to 64:) PS. If your sub rate is acting a little funky it’s because I unsubbed to re-sub;)
@@ColtCapperrune I mean, if your clients are requesting 48k files for higher quality streaming platforms it makes sense to switch even just for that reason.
@@ColtCapperrune yea there are some sensitive subjects when you step into the digital technical realm, samplerate is definitely one of them. You have the best mastering engineers and tech experts in the world disagreeing, and if you take a look on forums it's a deep rabbit hole. You could be right or wrong, provide a professional take on what you hear, but if it doesn't match the information they've accepted as correct, then they devalue your opinion !
But Dear Colt, since first thing that Spotify does to your audio, is to convert it to 44.1kHz, if your source file is 48 kHz, doesn't that introduce aliasing and distort the actual sound a bit? Wouldn't it then be making more sense to run it at 88.2 kHz as it's a full octave up?
yes. the conversion back is always ugly, even from double. but certainly less ugly from an integer multiple. I doubt somehow any of the streaming services are using dedicated high quality offline conversion, that's a whole science in itself. and definitely not if it's realtime lol.
@@JiihaaS sorry that's 100% wrong. aliasing is always introduced in SRC downwards. You're trying to remove information above the new nyquist it's a guaranteed consequence, much in the same way quantisation clipping distortion is a consequence of reducing bit rate. As an extreme have you never heard a 'bitcrusher' do sample reduction?
@@Bthelick that's why anti-alias filters are for. The information above the new Nyquist is low passed away before the SR conversion. Are you saying aliasing still happens somewhere in the process? Where and how?
@@JiihaaS yes because no filter is perfect. At all. Otherwise we'd just record at 44.1??? And the whole debate would be moot. And the quality of those SRC filters is always worse in non dedicated programs like daws. Put it this way, the hardware converters in your interface will likely be working at 128x the sample rate to avoid aliasing on capture. And unless you're using a very expensive interface with burr/brown etc converters even at 128x that process is not invisible. Daws have come a long way in the last 10 years but it's truly never truly clean, even if you use a dedicated separate program SRC like one of the best r8brain by Voxengo.
Excellent video. I also am a 24/48 guy. I think that your description of more aggressive sound is a good one. Many folks don't think about how the filters applied to avoid artifacts of conversion affect the sound in our "20-20k" range. I was always feeling that those had some harmonic color there that altered the sound somehow. Even notice it with my burl bomber.
I record in 96 and you’re actually not the first person to say that. Once is a one off, twice will make you rethink your decision. It works great when I’m recording jazz stuff but I will say there is an edge that’s missing when I record rock songs. I thought it was my mixing. Edit: dang, you definitely got a point. I just recorded in 48 and you do hear difference. It’s slightly more biting and even sounds like…kinda tighter.
If you are willing to use electronic music as reference, its actually easy to check how different sample rates sound. Just make a simple project with only soft synths/instruments and export in 44.1, 48, 96 etc.
there's no any noticable difference between rates if we talk about sampling rates as storage media. The difference you might hear when changing rate tells you have a poor clock in your audio device, The higher sample rates are needed only for production to avoid aliasing. But all modern plugins use oversampling, so we mustn't worry about it too much.
I used to record at 48/24. I switched to 44.1/24 and couldn't be happier. I have tons of analog gear and mostly mix outside the box. JJP seems to agree, too.
44.1 & 88.2 are the sweet spot on converters I built. 96khz sounds too bright & washy. 44.1/88.2 vs 48/96 will sound different on various converters. ADC more-so than DAC.
This is a nice video but there are some key errors in your explanation of more = more. Nyquist and Dan Worrall have proven that as long as you have double the rate of the frequency you're trying to recreate, you can, with perfect accuracy. See here: ruclips.net/video/-jCwIsT0X8M/видео.html For eg to recreate 400Hz, you'd need a sample rate of 800Hz to recreate it. The human hearing range of 20Hz-20KHz is completely covered by 44.1KHz. 20KHz is recreated by 40KHz just fine, and higher sampling does not increase the accuracy of the recreation at all. The problem with higher sample rates is that the closer you get to Nyquist the more aliasing you create in the ranges above what you can hear, and these issues literally fold backwards into the audible range you can hear. I think, though I'm not sure exactly why, that most people in video work with 48KHz because the framerate is usually 24fps, so it's easier to time align frame by frame when you're matching 2 cycles to 1 frame perhaps? The higher you go above 44.1KHz the more crucial it is to cut the sound above 20KHz on your tracks to stop aliasing bouncing back into the audible range. More cycles per second DOES NOT provide better accuracy in recreating 20Hz to 20KHz, there is little benefit at all to using higher sample rates. Watch Dan Worrall's video on this subject and you'll see.
![Seungmin "그렇게, 천천히, 우리(As we are)" | [Stray Kids : SKZ-PLAYER]](http://i.ytimg.com/vi/kAzmhLHePqU/mqdefault.jpg)
![Blox Fruits Dragon Rework Update [Full Stream]](http://i.ytimg.com/vi/EqDAp8udhm0/mqdefault.jpg)
Slightly disapointed in this video compared to the rest of your videos. Whilst I agree with your conclusions, you don't touch on a lot of the reasons why CD was at 44.1kHk and DVD was at 48kHz. Or why 24bit is better than 16bit.
Sampling rates always have to be double the maximum frequency you want to capture (Shannon-Nyquist). So 44.1kHz sampling rate produces an audible frequency of up to 22.05kHz, the reason this is above the max human hearing range was to allow for filters. Instead of having a hard cut of filter to stop any frequencies entering the ADC at exactly 20kHz, a filter with a sloped cut of is used as the filter would be easier to implement and therefore cause less artifacts in the sound. Ultimatley, It was, as you rightly say, to be efficent in saving space/data. However, 48kHz is popular on DVD/films/TV because it is easier to sync to a variety of frame rates. Not because they want 24kHz of audible sounds vs 22.05kHz. Human ears can't hear that high, and even if we could the difference between the notes they would produce is small (e.g. the difference between 200hz tone vs a 500hz tone is much greater than 14.2kHz vs 14.5kHz) our ears have better resolution at lower frequencies. 48kHz is purley for ease of maths vs 44.1kHz in implementation and filters.
16 bit vs 24 bit determines the noise floor/dynamic range available. The human ear in ideal conditions has a noise range of 120-130ish dB. However, we all have to deal with background noise and the equal loudness curves show that its only at certain frequencies our ear can hear down to near 0dB. When CD was introduced it was taking over from tape. 96dB (16bit audio) was seen as more than enough of a Sigal to noise ratio compared to tape (as in Cassette tape even with Dolby-NR, not studio/reel to reel etc). However, 24bit (which provieds 144dB of noise floor/dynamic range) more closley resembles the human ear. So again, it was a compromise to, as you said, save space. Also in a recording environment, having that headroom is incredibly usefull - especially when recording classical music.
So - I completly agree with your conclusion, just not the way you came to it!
🔥
I agree completely. Part of the problem is the manner in which he represents "headroom", i.e., drawing it in such a way that it makes it look like the amplitude of a waveform will exceed the range of 16-bit audio. This is, of course, completely wrong. 0db is 0db whether in 24-bit or 16-bit audio, and a waveform that exceeds 0db will clip in 24-bit just as much as it clips in 16-bit.
Nevertheless there is indeed more headroom with 24-bit. It gives you the luxury of recording at a lower volume and applying more aggressive compression or other dynamics without losing dynamic range as compared to 16-bit. But this is all about mixing and mastering. Headroom ceases to be an issue with distribution, and 96db of dynamic range is more than enough for the most dynamic classical recording.
As for 48Khz being more "edgy" than 96Khz? That makes me cringe. That's most likely a DAC issue with poor filtering and/or the result of driving speakers with high harmonics what are inaudible, but introduce audible distortion.
Thank you for the insight!
I think your explanation lacks one point. It is correct that 44,1 is basically implementation and filters BUT we are also getting only 4 measure points a 10khz. That makes high frequencies which are still very noticeable pretty low resolution. That'd be a reason to upscale recording or listening. Of course it has its pro and cons but I think it is very worth adding to your great note! :)
@@adrianzadi No, that's not the case. That's the beauty of the Nyquist theroem. Mathematically, at 44.1Khz, you have all the points you need to perfectly reproduce every waveform up to 20Khz, with a 2.05Khz buffer for the filter. There is no noticeable "pretty low resolution" at those frequencies. None whatsoever. The resulting waveform is perfect in every case, with all source material.
Again, this is for distribution. For production it's an entirely different story, especially if you are doing time or pitch shifting. Then higher resolution is absolutely beneficial. It is also beneficial if you rely on saturation effects and are getting audible and distracting aliasing. Then 96Khz is beneficial. I generally produce and mix at 96Khz/24 and distribute at 48Khz/16.
arguments - 0
idle talk - 100
Dude doesn't even know the difference between bit rate and bit depth.
Plot twist: The audio to this video is 44.1
Yes and no .... youtube compression algoritm
Opus. 48KHz Sample Ratw with a low pass at 20KHz.
@@tyjuarezAnd HPF at 20hz right?
@@bharatmadho3742 not that i know of. according to the documentation, the only filter on by default for Opus codec is an LPF at 20K
@@bharatmadho3742 , 10 or even 3 Hz (i not remember)
Please. Sample rate is about frequency content. The higher the sample rate, the higher the frequencies the converter is going to registrate. Read about the Nyquist theorem to really inform yourselves about this and why you should care (the effect is called aliasing, which it is something we do need to avoid).
I like Colt videos. I think he discusses many interesting topics about the art of mixing and how to run a professional studio. It is a magnific work he does. But he missed the point in this video.
You are 100 percent right
This is true if we can say that digital anti-alias filters have no “sound”. The more of a brick wall the brick wall filter has to be the more likely it will impose some kind of sonic footprint because it is more extreme. For 44.1k it has to go from passband to stopband between 20k and 22.05k. Increasing the sampling rate means the filter has more space to roll off, is less extreme, there is less ringing on square waves and transients and is less likely to affect the sound quality.
"The higher the sample rate, the higher the frequencies.." What's the highest frequency you can hear?
My computer can barely keep up with 44.1. It's a 2019 Macbook Pro. So it's not like it's old. I'm constantly adjusting my buffer size, turning off plugins while recording, freezing tracks, turning on low latency monitoring, etc. I still have issues no matter what.
"Computers are more than powerful enough." Yes, if you can just drop thousands of dollars at the drop of a hat. Many of us can't.
For many of us, buying a new computer would require months and months of saving. If I am going to be spending thousands, it's going to be on acoustic treatment, new microphones, etc. Not gear so I can upgrade my sample rate.
There's something wrong on your side, not with many of us being unable. My desktop computer is from 2019 and it still creates bangers of a song and i can go to 24 bit if i wanted to and it's fine. Besides, Macbooks are built for audio and sound design which is why I'm curious to know what is going wrong. If my windows computer can mix audio better than your Macbook, then I think you need to make some changes or do something about it....
Dan Worrall from FabFilter has a fantastic video on this as well.
“If the song is good and the mix is good, no one will care about a bit of aliasing.”
That quote is gold.
he's the best!
If the orchestra is great and the soundstage is great and the composer is great they are really gonna care about even a little bit of aliasing,
Same goes for anything that’s not so “dense” that you can’t discern the aliasing. Like he says in the clip more intimate music and what I consider more natural music like Bluegrass he samples higher, for good reasons.
The human perception of Sample rates above 44k is probably mostly imagined or placebo. What these higher sample rates really do is move nyquist very far out so some EQs and plugins that would normally cramp or unpleasantly distort (especially non oversampling plugins that have a steep high frequency boost or cut) will no longer cramp or distort any where close to the frequency ranges we can perceive. Unless we all start to develop mutant hearing or you are doing some really crazy $hit there is no reason to use 96k sample rates. The size of the files and burden of working with them doesn’t make them worth the struggle. At 24 bits you get 16,777,216 steps to represent something like 144db of dynamic range. You are lucky to get 119db of dynamic range into the A/D of most interfaces. 96 db of range is pretty typical. Also, most modern music has the shizzle compressed out of it and it winds up with almost no dynamic range. Opening most mastered songs in a wave editor will display the post loudness war modern music WAV log shape. You can probably fit that level of dynamic range in 12 bit.
Interesting
As an audiominded EE, thank you for the boner
No, i like slowing down things a lot, and recording with higher sample rates makes it sounds better.
I always figured any benefits of any sample rates greater than 44.1k has to do with plugin processing. I always go 48kHz.
Now, as far as I know, higher sample rates on playback should be subtly audible. Completely unimportant, but subtle because, correct me if I'm wrong, the DACs in our gear have some of their work cut out for them in smoothing out those samples into continuous analogue signals. All DACs also oversample (4x-8x) for a smoother sound, but higher sample rates reduces their oversampling rates, and less oversampling will reduce how much the DACs impart their own sonic character on the sound, too. Splitting hairs here, though, you need high-end monitors to hear the subtle difference, it does come out in a recording when you've got Burr Brown DACs versus the ubiquitous AKM DACs for playback, but very subtle and requires high-end headphones or speakers to even hear clearly. I think I remember reading a long time ago that DAC performance improves with higher playbck sample rates, but this is for nerdy electrical measurement stuff that no one cares about, lol.
Can You imagine that everyone knows 12bit is more than enough and 60db of SNR is all they need in listening?
Whole industry will struggle due to lack of selling new "high-res" stuff.
32-bit 1Ghz sample rate on the way.
After some experimenting I found that 48/24 is the rate that gives me the best results for what I do, on my system with my converters. There's an argument for recording at 96 and downsampling to 48 when mixing, but then again I'm a little sloppy and wouldn't want to constantly change sample rate... 🤷🏻♂. Also, let's not forget that the results will vary depending on the converters and clocking you are using! Peace and peaches y'all! 🍑
"There's an argument for recording at 96 and downsampling to 48 when mixing..." but not a mathematically solid one!
I am a simple man, Colt uploads, I watch, I learn.
I appreciate you!
Me too, but I also read other opinions, compare them, think about it, get deeper in the stuff and then I descide what I´ll do with this information.
I am a complex man, I never use copy paste comments like this.
88.2kHz/32bit has been my default for probably 20 years now. I settled on it at the time because it divides evenly to 44.1kHz when downsampling, so in theory it would be less prone to artifacts. The converters still only had 24bit resolution, but saving as a 32bit file gives it much more headroom when editing and matched the internal resolution of the DAW processing.
88.2 kHz is from Bob Katz, right? But I wonder, did you ever try to downsample from 88.2 to 48/24 or to 44.1/24? I'm quite new to this and try to understand but there are a lot of bids out there.
You in theory shouldn't need to divide evenly when downsampling if your DAW does things correctly. It is entirely possible to reconstruct the continuous waveform for any frequencies in a lower resolution by the higher resolution samples.
That’s crazyyy me toooo! 💯💯
I've been on 88.2 for the past 5 years.. 24/64 bit thou studio one default. No reason to switch from 44.1 to 48 since that's a 8.8% change in the sample rate only. At the most the filter on the interface changes slightly.
My computer starts smoking when I try to record stuff in 48k. And my Pappi says if 44.1 was good enough for him, it's good enough for me.
In my humble opinion… switch to 48 from 44.1 if your client base are working with this sample rate. But don’t use it if you think it sounds better… there is no audible difference between 44.1 and 48.
88… hell yeah… 96…absolutely and 192… mainly for cinema scoring… insane difference.
I was asked by AES to attend a listening session at McGill university where George Massenburg use to teach…. The test was a blind shootout to listen to short snippets of a song in all sample rates. 44.1 to 48 was impossible to hear a difference… but as soon as it went up to 88 and above…wow… huge noticeable difference in how much more open not to mention the depth was.
FYI… Just in case… I’m a Juno nominated songwriter/producer who has recorded for Dream Theater, Corey Hart, Harvey Mason Junior, Debbie Gibson, Jim Vallance.
My 2 cents 🙌🏽
Must be a personal taste thing among those that can hear a difference.
I believe Jack Joseph Puig prefers 44.1 to 48k saying 44.1 sounds more musical to him.
That bugs me too....big JJ Puig fan....i'm currently sticking to 48....I tried a 96 and ran into some conversion hassles? something was not as easy...maybe I will give 96 another try at some point.
44.1kHz will sound the most musical as that is the frequency at which DACs apply the greatest oversampling rates that they do to generate the smoothest analogue signal. Higher sample rates causes them to dial back the oversampling. This stuff is freaking subtle, mind you, and requires high-end playback systems to notice. For most people, it is likely just an imaginary difference.
@@Jason75913 Right. Another consideration is the audible effect of decreasing of clock accuracy with sample rate, in individual systems.
It's really a 'headroom thing' as he says. 16/44.1 was chosen for CD, because it satisfies what the average human being can actually hear, so it makes sense to utilise the increased headroom afforded by 24/48 and 'chopping off' the redundant unused upper bits above 16 to produce a final product that uses as much of the information available without clipping. Sony did this for years with their 'Super Bit Mapping' of 20 bit recordings. Above 48K sampling frequency, there's quite a lot of measurable audio 'junk' which can impinge on audible frequencies, so is best filtered off, or simply not recorded, in the first place, which produces a cleaner perceived sound within the humanly audible frequency spectrum.
Something to think about: If you switch from 44.1k to 96k your plugins will use 4x as much CPU. Also, the higher the sample, the higher the Nyquist Limit. Meaning there is more resolution before signals bounce back into themselves and create Aliasing.
Someone’s up on their Dan Worrall 😉
Why 4x if it's only ~2x the rate? Also, most plugins oversample anyway. When I first switched over to 96k from 48k, my CPU hit was ~50% more with the same plugins.
You got the the Nyqvist issue before I even got here. But yeh ...
That's not entirely true. Most plugins today use upsampling to avoid aliasing, therefore working in 96K many times make the plugins less CPU intensive as the upsampling becomes redundant.
This Dan Worrall video "Samplerates: the higher the better, right?" is worth a watch for sure, but 2x the SR = 2x CPU. Sure, you can get audible aliasing at 96k if you have strong harmonics above Nyquist, but very few acoustic sounds have significant info up there, and any good saturation plugin (like Fab Filter Saturn) has internal OS and filtering. Not saying it's necessarily better to record at 96k, but there is much less likely to be any audible aliasing and the smoother LP filtering won't produce intermodulation in the audible range. The reason 48k sounds slightly more "edgy" is likely due to the intermodulation distortion. By all means, save the CPU if you don't mind that effect. It is very subtle.
However, larger bit words don't use significantly more CPU, and won't generally task disks more due to the packet size streams. Yes, 16bit final recordings may sound nearly identical to the human ear, but that's because the entire 16bits is used. You'll want to leave at least 2 bits headroom when recording in most cases. Digital recording systems went to 20bits in the early 90s to eliminate low level hash in quiet sections. Low to upper midrange waves do exhibit some intermodulation distortion below -65dBfs in 16bit recordings. That may be audible in low level passages, and may be partially why hipsters find CDs "harsh". It doesn't happen until well below -100dBfs in 24bit recordings. I'll stick to recording at 32bitFL for that reason, and to ensure no digital overs during rendering.
The only difference between 44.1k and 48k is that that the later has content above 22.05khz (up to 24k), below 22.05khz the content will be and sound absolutely identical! Any difference is not rate related, you either have aliasing going on or analog converting issues.
Or "The number 48 is greater than the number 44.1" issues...😀. I didn't here one single reason for why 48 would better than 44.1 besides "CD:s have been used as a format for so many years". And people just seems to buy this here left and right!
Because more is always better...
@@Magnus_Loov My thoughts exactly! Other than the fact that clients started requesting it, I see no reason to go above 44.1 if you're not working video material. Bit depth, on the other hand. makes sense to increase. I have mine set at 32 bit float for the unlimited headroom, but export the master at the usual 16. What I'm not sure about is if I should rather be exporting at 24 bit...🤔
Also gonna chime in with the rest who feel this was a bit disingenuous. Even among engineers, double blind tests have shown us that we can't consistently & reliably tell a difference between 44.1 & 96-192. Unless your final intended destination is DVD/Blu Ray, there's no reason to go against the established standard of mastering to 44.1 16 bit. Switching between sample rates that aren't perfectly divisable introduces aliasing from rounding errors for one thing, and what about music that gets a digital only release? Plenty of us still have CDs in our cars. Have you forgotten what happens if you don't dither to 16 bit? I've bought stuff from bandcamp that's been at 24 bit that I want to put on my phone, I really don't want to run it through my own software & dither it myself first.
Also have to pick on his testing method. ANY time you record & compare between different performances, you invalidate your test. No matter how good & consistent your musician may be, there are ALWAYS going to be differences that influence the sound. The only accurate way of doing a sample rate comparison would be to split the signal & send it to 2 different machines. If he's as experienced as he claims, this whole video is really disappointing. Only going to unnecessarily confuse & complicate things for newcomers. I'm sorry dude, I've enjoyed other videos of yours but you should know better. Oh and "we have more than enough CPU power & hard drive space these days" is not a reasonable argument. Just because you can, doesn't mean you should.
its not that simple it entirely depends on the filtering methods in the converters / software.
But yes your correct be it issues at low samples rates with or conversion back from higher rates there is always a compromise.
The idea of 22.05khz audio content is theoretical. Think about it for a moment. That scenario would have only one bit for the peak and one for the trough of each wave. Essentially reducing the input to a square wave. But even worse is that filtration and dithering are a necessary part of the A/D conversion and these cause foldback artifacts well below the input frequency. These artifacts are now living in the audible spectrum (most people can't actually hear 22khz). The best argument for 48k is pushing these foldback artifacts up out of the audible spectrum.
I feel the only reason to go higher than 48 is because of aliasing introduced by non linear processing. But a lot of plugins (most that I use) have an internal upsampling or the "HQ" button so even that is taken care off. I'm almost 100% sure nobody will be able to hear the difference between 48 and 96 if there is no aliasing going on from processing.
Thank god being the only smart person here there is so much bs around floating around and this guy is lowkey spreading it as well
it gets a lil more complicated too, there are situations where lower sample rates with upsampling can actually produce less aliasing in multi-non-linear chains than equivalent higher sample rates.
there's a fascinating paper on the subject by either Vladislav or Fabian who code for Tokyo Dawn.
@@4lanimoyo553 are you saying the op is wrong? How exactly?
A second reason would be if you intend to offer yourself greater flexibility in time-stretching or pitch-shifting on your sample, for using your sample in future sound design purposes.
Totally agree.
Well, the benefit of high resolution audio is the same as high framrate video. If you wanna alter the speed of the audio you have more headroom before it starts sounding distorted. 48khz leaves you with around a 20% wiggle room in time stretching the signal. 96khz gives you over 50%. So if you're planing on doing some heavy retiming or speed alterations, record in a higher resolution. It's as simple as that.
what if I record in 192khz but run signal through plugins? does it deteriorate quality?
@@bitman6043 It probably depends on the internal processing resolution of the plugins. Some process sample at 48khz while others sample at much higher resolutions. Some plugins even allow you do oversample internally.
@@toastbrot97 so basically what I thought - you have pick and research every plugin if you want non 44.1/48khz
Exactly. A lot of people thinks that more is better but it only defines how high is the frequency you can record.
This is a real world benefit right here. Sound manipulation , just like graphical image manipulation is best using a higher quality source as poissible to prevent artifacts.
The best channel on youtube when it comes to Mixing.
Correction: 24bit does not have more volume than 16bit. It is just able to give more values, ie. 2^24 steps instead of 2^16. between the "zero" and the clip point (which is determined by you preamp and not your bit depth).
Would add that lower sample rates only change the reproduction on frequencies higher then 20kHz, interpreting them as low frequencies, which *theoretically* may cause artifacts in the audible range. This may explain the "cleaner" sound you got at 96kHz vs the 48kHz. (which would have been nice to hear, even thou YT compresses).
Keep in mind most instruments don't have an above 20kHz content in the first place. Also most synths have an anti aliasing option to mitigate these artifacts.
Now would it be audible in a contemporary pop/rock mix, where songs are extremely loud, distorted, and compressed? IDK. probably not. Would be interesting to test.
This correction makes intuitive sense, but it's not actually correct. In PCM audio, each bit in the sample gives you 6db of dynamic range. It doesn't get scaled by your preamp or converter, you don't get finer gradations of volume level within a fixed dynamic range as you add more bits. 16-bit has a maximum theoretical dynamic range of 96db. In practice, dither and noise eats into that dynamic range. Similarly, 24-bit has a maximum theoretical dynamic range of 144db, but in practice you'll usually see good 24-bit gear with a dynamic range spec around 126 db (actual 21-bit resolution). On paper, 32-bit float has a dynamic range of 1528db, but that's also theoretical, because that's a bigger dynamic range than what is hypothesized to be the loudest possible sound pressure level on earth.
As far as the stateent about sample rates, it's important to know that digital sampling introduces a ringing noise around the Nyquist frequency (half the sample rate), and that needs to be filtered out. For 44.1 kHz audio, that means you have sample noise at 22.05 kHz, and this becomes a problem, because no perfect brick wall filter exists. A low-pass filter that allows audio through at 20 kHz and blow, but effectively kills the signal before it hits 22.05 kHz, is extremely difficult to create, and those filters usually introduce undesirable effects in the audible frequency range. Using higher sample rates allows the use of shallower filters, with fewer detrimental effects on the audible range. Most acoustic instruments *do* have harmonics above 20 kHz, but most people can't hear them, and most microphones are going to have trouble capturing them accurately.
You’re about to start a war with this one bro! 😂 The age old sample rate debate.
don't mix up bitrate and bit depth Colt! Bitrate is measured in kbps or mbps
Pretty typical for a youtuber to think audio and video are the same thing. Ugh.
The Problem is, that the 48khz sells and streams on those platforms makes absolutly no sense for the end consumer. Because you can't hear the difference. Sell something for more money, that doesn't have any advantage for the listeners is a cash grap. 48khz makes sense from a producer standpoint and nothing else.
But sure, if the costumer want it, he'll get it.
Sampling frequencies over 44.1/48KHz is mostly snakeoil/marketing hype to sell more gear, these frequencies can perfectly describe the soundwave, there is no need to waste computer resources and storage on higher rates. 24 bit matters quite a lot at the recording stage though to be sure to get the most out of the dynamic resolution, for reproduction of a mastered track 16 bits is just fine, the people that invented the CD really knew what they were doing. I record at 44.1/24.
I repectfully disagree. I'm not in the industry, but I do have a decent HiFi. I own two copies of Aftermath by the Rolling Stones, one at 24/88.2 PCM, the other is DSD64. And, even at those resolutions, I can hear the difference. The DSD64 version sounds better. I'm not talking about by a tiny amount either.
@@The.Original.Mr.X you’re hearing a different master rather than the different container for the digital information
Sampling up to 192 allows for easy use of common digital reconstruction filters of the slow type for better impulse response without HF roll off without needlessly taxing hardware or software upsampling resources. Upsampling to 352/768 is a joke.
DSD native is often much better recorded content with much more dynamic range and care taken in mastering but is huge in file size. Upsampling to DSD is an absolute waste of resources. Most DSD native content is super niche or esoteric or classical. Tape to DSD is the best way to distribute master tapes and you can purchase FLAC versions if file space is an issue.
I've been mixing and mastering music for 15 years and I've come to find my favorite sample rate and bit that is 192000/16 -bit. I used to recorded master 192000/24-bit and I lost a lot of the low-end frequencies and the bass out of my tracks, and when I put it back to 16-bit I get the bass back.
@@The.Original.Mr.Xyou can disagree all you like. The math that goes into reconstruction of sound waves requires no more than 48k to reproduce all tones heard by the vast vast majority of the population. Anything more than that is solely to record inaudible frequencies and slow them down.
Your 10 min videos have become part of my daily reading. Thx for your efforts Colt :)
In real world… in a blind test, the difference is…
zero
If you have good ears and a good system you can easily tell the difference. Listen to the clarity and high frequencies especially. The higher the sample rate, the higher the frequency range because of the increased nyquist frequency
Noticeable if using melodyne
@@metaldemonseanknels no... most dacs will cut off at 20khz and most studio monitors dont go above 20khz.
it's all about the digital aliasing that happens especially in saturation plugins that are ubiquitous now. it's very noticeable.
Excellent presentation, thank you.
I gotta smile when I realize that all of this technology ends up going into 'earbuds'.
I've been doing this since 1961, my studio monitors back in the day were 12 1/2 cu. ft. apiece with 15" woofers, etc., custom built.
And today kids think 'subwoofers' deliver bass.
Thanks again, Bill P.
It seems everyone has forgotten two things about the sample rate of music: if you record or synthesize frequencies higher than the cutoff, aliasing is highly likely to occur, which basically means they start bouncing back in (seemingly random) areas in the audible frequency spectrum.
Even if there was little to no aliasing occuring, you don't have to be able to physically hear past 20KHz to hear the effects of the frequencies past that range. Certain frequencies can cancel out or amplify other frequencies, even modulate other frequencies, many of which could be potentially audible. This is most applicable when using speakers and not headphones, because the frequencies travel through an open medium where frequencies can physically clash or cancel out in space. It's important to note that I don't think these various effects happen digitally, so these things will be experienced to a minimum through headphones with little to no medium to travel through except your ear canal and head/skull. When you play the same track out loud in the open, the issues could potentially reveal themselves. This is why I personally believe it would be *best* to export using the highest sample rate to avoid aliasing and other artifacts from "lost" frequencies, however, I too use 48kHz because other sample rates can increase CPU usage or even the chance of incompatibility or having to convert to 44.1kHz/48kHz (takes a while). The differences usually won't be too noticeable anyway.
Correct me if I'm wrong, as I'm not extremely experienced with audio, but I take in a lot of info and run a lot of tests to prove my theories, and I've seen, with a spectrum analyzer with my own two eyes, serious aliasing from producing frequencies higher than 20KHz on 48KHz sample rate. I can't confirm the second theory but I am pretty sure anyone can agree based on the reasoning.
This was a reply to another comment, so I'll add context:
The cutoff of frequencies audible is dependent upon the sample rate.
44.1kHz - reproduces up to 22.05kHz frequencies in the exported audio file
48kHz - reproduces up to 24kHz frequencies in the exported audio file
Filtering is used to cutoff the remaining frequencies nearing this limit. The higher the sample rate, the smoother this filtering can be (so less artifacts occur)
The good news is that the converters actually sample at a very high resolution both in bit depth and SR. This means that the info is captured and decimated to the target output rate. So difference between rates is not super significant.
I love your other content but please stop talking about technical stuff you don't fully understand because what you're saying in this video is bs, sorry... Please do some research on how A/D-converters work and what's the Nyquist theorem because then you'll realize that the sample rate only determines the highest frequency that is digitized which is roughly half the sample rate, so 22kHz when using 44.1kHz as a sample rate.
I don't know the conditions of your drum test comparing 48k to 96k but I guess you somehow tricked yourself because technically the only difference is that one file contains information up to 24kHz and the other file information up to 48kHz. Since our hearing ends at under 20k you shouldn't hear any difference at least in the raw files. Because yes: when using dynamic processors or saturation higher sample rates have their right to exist because it reduces aliasing. But recording at 44.1k and then using oversampling in these kind of plugins is the way to go.
Completely disagree. On a purely technical level, Tidal will stream up to 9216 kb per second, that is equal to a 32 bit, 96K file. So I think you are better off going with the highest resolution you can handle. Not once did I mention the nyquist curve because it’s irrelevant to my point here, which I think was well explained and I stand by it. Thanks for watching.
Regarding the bit depth I totally agree with you: bith depth DOES matter! But nobody in the world should hear a difference between 44.1k/32bit and 96k/32bit. That Tidal offers 96kHz streaming is technically bs but I guess they offer what customers pay for and if people are willing to pay more for getting 96kHz sample rate instead of 44.1 they're happy 🤷♂️ and as already mentioned in another comment here: a correctly upsampled 44.1k file should sound exactly the same as a 96k file.
this is the only good and correct comment. Another reason why you would want to record with a high samplerate is because you want to manipulate the sound by for example pitching it down but as you said there will be precisely no audible difference between listening to a 16bit 44.1kHz wav file compared to a 32bit 96kHz. Anything that does sound like there is a difference is either from the samplerate conversion which should also be pretty much inaudible or it's plain placebo.
48k gives you just a smoother filter in the ad conversion. This filter and its settings (differ per sample rate) is what gives the sound difference in your ad converter when switching between the sample rates.
Exactly. And that will vary substantially from DAC to DAC. Particularly cheaper DACs running at 96Khz which fail to filter high harmonics.
@@mwdiers I think he's talking about recording, not playback
@@Jason75913 So was I. Filtering is only done in the AD phase (i.e., recording), to exclude frequencies above Nyquist.
Some home audio equipment can not produce a flat frequency response to 20kHz, as it should, with 48kHz sample rate, the highest frequencies droop.
When the digital file is converted to analogue music, the hardware has to filter out the alias image. This filtering (different DACs use various mathematical formulas) can create distortions (whether you can hear them or not) in the time domain, it’s often called ‘ringing’. Some Hi-Fi companies choose filtering that only produces a correct frequency response with 96kHz to avoid this type of distortion. It’s pretty common, especially on high end equipment.
So if you want all listeners to hear what you are recording or mixing as you intended, you need 96kHz sampling rate.
How do you go about using samples that are 44.1khz (which most samples are) in your projects. If youre running your project at say 48khz the computer will need to upscale all 44.1 samples to 48 and bring loads of artefacts.
When you are talking about 96k sounding clinical vs. 48k you are talking about double blind testing where you weren’t told what you were listening to and just had to make a call of what sounded better based on how it sounded and you wouldn’t be biased by your preconceived notions right?…Right????
If you understand the math about sampling and sample rate you know that 44.1k is literally more than enough. Bit rate you are right, but you should read more and not just base your opinion on experience. Read about the nyquist theorem, and stop saying things like 96kHz sounds clinical because it is bs, no one could ever actually hear a difference. Normally i love your videos but this is kind of misinformation.
To clarify: i do work on 48kHz for video, and do agree that people should switch, but some of what you said is wrong, and this being an educational channel, you should be more careful
“You should read more and not just base your opinion on experience”…. That sounds like great advice, thanks 😂
@@ColtCapperrune i think that you should tell him to look for example on iztope RX and have look on the amplitude sound and headroom if he dont hear or feel defrience than maybe he came see 😂😂😂
@@RollRollunio 44.1 and 48 both have the same headroom, bit rate defines headroom, sample rate only affects the frequency limit of what you can theoretically sample...
@@sebastianvallejoperez9754 i agree with you,i knows that already 😁😁😁 Regards
Not quite right on DVD here Colt. DVD-A was 24bit at 96k But Audio synced to video AKA DVD-Video soundtrack is 16-bit, with a sampling rate of 48 kHz
Been 48/24 since I started mixing.
But then, started digitally with ADAT with 48k 16bit, mixing on a mackie 32x8 - back when gain staging every step was critical to get a good digital mix.
for what it's worth, i saw an interview with pete townshend (the who) quite a few years ago where he said he recorded everything at 96kHz, but was just waiting until computers were powerful enough to run 192 kHz
Article on a top guy in Tape OP recent mag records at 192.
Neil Young at 192 also, I think.
Pete Townsend and Neil Young are hardly the benchmark for golden ears!!
@@brin57 🤣
Yikes, 16 bit does not have "less overall headroom in the volume or amplitude department" than 24-bit. And obviously the signal chart you drew would clip if recorded at 16-bit, and we know that doesn't happen. 24-bit has the same maximum, it just has smaller increments between possible values (like drawing with finer graph paper). Yes, I agree it's a dumb idea to record and work in 16-bit when 24-bit and 32-bit float is so cheaply available. 48k allows slightly more flexibility in the antialiasing filter requirements, if the goal is 20k bandwidth, or the same filter spec but just a tiny bit more bandwidth. I don't fault anyone for preferring 48k over 44.1, but it's just allows a slightly more relaxed filter or a tiny more bandwidth-it's not the world of difference that you imply. It's less than 10% difference, and we hear in log frequency, which makes it a scant difference.
A lot to think about, and weird that this video came across for me right after I finally switched from 44.1kHz to 88.2kHz lol. When I first started learning all this, I read a lot of "use 44.1kHz because CDs, and dithering down to 44.1 from 48 or 96 is wonky/noisy)" and just kinda stuck with it since. But I finally told myself "this isn't a thing anymore!" I did take in what all you said, and so for most of my stuff being in a metal arena I'll go for 48kHz and save 88.2/96kHz for when I pretend I'm a composer (lol...). I have always used 24-bit, at least.
thanks for sharing!
Yep that’s why “high res” even though it’s a marketing term 24 bit SHOULD be the standard. The truncation of bits from 24 to 16 with adding dithering white noise is pointless in todays worlds. The difference between 16 to 24 bit is only about 200 mb. There’s no reason why the consumer can’t have the “master copy” not to mention the 24 bit is better from them to process third party dsps in playback much like processings plugins when mixing and mastering
I have re-recorded 24 bit to 16 bit mixes on ADC and it’s fine with no added noise and it’s sounds better than the truncated 16 bit version as well but there’s no point for it in this day in age. I don’t like the use dithering and truncating the bits as well.
Nice. I did not think of that you just upgraded my sound. Thanks Bro!
I'm curious, the listening test with 48 vs 96 you did was a double blind test?
Also, the debate on 48 vs 44.1 is actually about where the artist thinks his music would end up (Tidal .. DVD etc) not a difference in audio quality. Because, again, a double blind test will tell that is no difference.
Fun fact: Metallica The Black Album was mastered from DAT tape 16 Bit/ 44.1 KHz. Sounds awesome even today compared to any production.
What about the St. Anger CD? That sounds shit, but the St. Anger DVD sounds so much better
@@bear4759 If the same mastered version used for the CD would have been put on the DVD, it would sound equally shitty.
(Apologies if this has already been addressed)
* If you want to mix in Dolby Atmos, it’s supposed to be 24/48.
* The higher sample rates are more useful when dealing with non-linear processes like harmonic distortions/saturation. However, it is better to use oversampling within the plugin rather than natively work in that higher sample rate.
* If your source for a vinyl record is digital, they actually recommend 24/44.1. This has to do with the cutter heads and not overloading it with supersonic frequencies that are beyond what the heads can cut, let alone human hearing. (Feel free to correct anything that is wrong)
Yep, Atmos is at 48 kHz.
But the bit about supersonic frequencies disrupting cutting heads doesn't make sense to me. It was the super low frequencies that disrupted cutting heads.
@@Technoriety I did some more reading on this. Frequencies above 20kHz can overheat the cutting heads and one has to be careful in the high end (sibilance especially) as well as the low end. That said, it seems many of the pressing plants accept digital files with a range of sample rates. They *may* downsample it to 44.1 on their end to make it easier or if it proves to be difficult.
This comment about the sample rates also came from a friend of mine who has a lathe cutter and has volunteered to make some lathes for me.
Again, feel free to correct me if I am wrong. =]
I only record at native CD rates when only directly producing to a CD. Everything else is higher bit rates.
Edit, In regards to bit depth, remember when recording there is headroom. For every 6DB, there is one bit of resolution. So if you record with 18DB of headroom, by recording at -18dbFS, the upper 3 bits are never used, just there if something peaks you don't want clipped. This makes a 16 bit recording use only 13 bits. However, if you master at 24 bits, now you are working with an effective 21 bit recording, When you normalize the recording for a 16 bit CD, you have 16 bit depth on the CD, not 13. It is in the math of binary. This is a purely technical discussion, and does not address the "Sound or Feel" This is about accuracy and signal to noise.
To stuff your pipe on 16 bit the renoun Neuman microphones of many thousands of dollars have much less dynamics than a 16 bit recording. I know the microphone is not digital so sue me. They do have a Signal to Noise ratio. The signal to noise ratio is 80 dB for the U 87 Ai. The signal to noise ratio of 16 bit audio is 96 db, which means the microphone noise floor can be accurately recorded in 16 bit audio.
as a working studio musician. 95% of the studios I've worked at including sunset sound, capitol, universal, resonate.. use 48k 24 bit w few exceptions 88.1 or 192. I can't tell you why. However, when it comes to art. The STROKES 1st album was tracked on a digi 001 at 44.1 24bit
Aside from the fact that he was talking about sampling frequency - you`re correct - it is true.
But all that one actually needs to know about 24 bit is, that 24 bit gives you 144(!)dbfs of headroom. Which says pretty much everything. That is _way_ more than _any_ analog noise floor . So essentially the noise floor of even the highest quality outboard gear - be it whatever it may be you're recording - will mask any possible quantization errors by a huge margin, where the quantization errors of 24 bit is effectively at such a low level, that _no one_ will ever be able to hear it - especially considering that there's actual musical signal at a way higher level on top of it.
Bottom line: Everything higher than 24 bit is in no way ever of interest for human hearing - even for the best ears that ever existed/will exist. At a typical maximum dynamic range of the program material of say from -2 dbfs (to avoid intersample peaks on the output) down to -50 dbfs - even considering for a moment that there' s absolute no noise floor of the recorded gear - you'll still have way over 60 dbfs headroom, before these quantization errors actually may occur.
@@broklanders4730 Well put. However the mention of 144db of headroom, is actually dynamic range. Headroom is the difference between signal and 0dbFS. To ship a 24 bit recording for mastering, the headroom is removed, so a -18db recording when normalized, is no longer 24 bit as the recorded depth is less.6 DB per bit, so a 24 bit recording at -18DBFS, only has 21 bits of information. This is the primary reason masters for CD are recorded at higher bit depth, so when the headroom is removed, and gain boosted with compression in mastering, the final product still has some valid data in the lower bits.
@@isettech - it's a common misbehavior (even falsely requested by mastering engineers) that there's any benefit by delivering the to be mastered track at -18dbfs. you'll be all good if you're just not clipping the individual tracks when bouncing stems or the summary master - and even that won't hurt nowadays, because we're in 32 bit/64 bit systems. so the mastering engineer would only have to normalize the track(s) prior mastering and he'd be good to go without any damage...
and if the track(s) are too loud (even if not clipping) - he'd normalize all tracks o his working level anyways, which would be done in just a second.
wow, the legendary U87 only has 80dB of SNR?
I did wanna say colt didn’t really explain what the actual advantage is to using a higher sample rate.
Humans can hear from 20hz to 20khz obv. So when your computer has to sample the analog signal 44.1 kHz sample rate will capture up to abt 21 kHz in analog (because a single wave has + and - values).
The only real reason to work at high sample rates is if you actually need information that is above human hearing. For example changing the pitch of a sound or over sampling with a LPF outside of what we can hear so that data isn’t reflected back and causes aliasing.
Essentially 48 kHz sample rate is typically seen as the best balance compared to other sample rates, because it is efficient with CPU (being marginally harder to run than 44.1 and 50% of 96) and it gives a bit more overhead for higher frequency sounds. 44.1 is allowing almost no overhead.
Anyways hope that helps! Great video Colt
My question is this, if the highest frequency a human can hear is 20 kHz (which it isn't and closer to 18 or 16 kHz), so why then do you need to have a higher sample rate than 44.1 kHz, which still snapshots more than twice a second for the highest frequencies we can hear?
I can hear just over 20k prob close to 21,500 or 22k
When you use processing that generates harmonics (distortion, saturation and to a lesser extent compression), those harmonics technically go past 20khz, although they have a smaller amplitude. If the plugin is not properly written, those harmonics can be "reflected" back down below 20khz (imagine a mirror at 20khz ish) and you can get some weird stuff in the high end. This phenomenon is call aliasing, and oversampling usually attenuates its impact.
There is a great video on the Fab Filter youtube channel about this topic if you want to go deeper.
Also, some people (not me) claim to perceive a difference in transient response at higher sample rate. The idea here is that the very first part of a transient is so quick that if you analyze its frequency spectrum it goes above 20khz, and while we can't perceive a sustained tone at 22khz for example, we might be able to perceive it as a transient. This assumes of course that the whole chain ( from microphone to speakers) is able to reproduce those frequencies.
In playback you don't. The exported audio file can be 44.1. It will make no difference. This is a video about mixing. You want to record at a higher frequency to avoiding ailising.
@@lorenzocelata4107 Thanks Lorenzo! that was the missing piece for me but now it makes more sense
Audio information present in 96k but not in 48k is inaudible. Actually audio information present in 48k but not in 44.1k is inaudible. When processing, oversampling can make sense because of aliasing issues. But for a final product, more than 48k is unnecessary. A 32bit float has a 24bit mantissa. The advantage of 32bit float is for signal levels significantly below 0dbFS. (If you are working in 24bit, and you're signal values, consider as being in the range [-1,1], stays below 0.5, then you've lost one bit of resolution. But with a 32bit float, you have the same 24bits of resolution, and this keeps on being the case. If you turn a 24bit sample down 36dB, an then up 36dB, you lose information, with float you lose a lot less.)
Nearly every recording we do in Hollywood is done at 96kHz minimum. Then we do any and all time stretching or alignment. Then it’s usually downsampled to 48kHz when being handed off to mixing. Mastering is then done either at 48kHz or 96kHz (upsampled).
I’ve personally heard the same 48kHz mix going thru the exact same master chain and converters at 48k and 96kHz, and the higher resolution was then downsampled back to 48kHz. I was very surprised that there was a very audible difference between them. The 96kHz was the clear winner. I wish they had also shown me a 192kHz downsampled master. My curiosity is killing me. 😂
This really doesn't make any sense...
There shouldn't be any audible differences in the audio. The only way there would be is if harmonics generated by something in the FX chain led to aliasing, but you'd need to actively avoid oversampling options in your plugins and push them hard. I'd be curious to know what mix you were listening to where the difference was that obvious and just how high a frequency you can hear because you'd need superhuman hearing to hear a difference. (Human hearing tops out at around 20kHz and anything recorded in 96kHz has frequency content up to 48kHz, well above the audible range.)
Hy ..are there Problems (maybe Ableton) wen u record first in 44.1k and some clips in 48 k ?wen u mix this what u think are the results..?sorry my englisch..big Respect and best wishes from Vienna ✌️🇦🇹
I use to work for a guy who help develop the CD's for sony and 48k is what they liked the best but couldn't fit all on a cd... So I have always used 48k 24 bit. thanks for the conformation.
What do you do with project that’s in 44100 then, can you convert it. Or are you stuck with 16bit 44100 ?
It's not really about anything other than aliasing. I work with Bluegrass and Folk more than anything and work at 96K and here is why. Aliasing causes transient smearing and inter-sample modulation distortions that basically cause frequency clashing and masking. I've worked a 96K since 2012 not because of any other reason except lower latency and more importantly less aliasing in the audible range. Modern computers have no issues recording at 96K and sessions aren't really that big when you compare them to say 15 minutes of 4K 60 H265 footage or any of the video codecs. 192 is problematic since some of the most popular plugins don't work correctly at sampling rates above 96K like Waves and some instruments like Arturia. On the point about 16 bit vs 24 you are not giving up headroom you are lowering the noise floor at 24 bits from the 96DB noise floor of 16 bits. Noise floor and headroom are two different things, the only time you gain headroom in comparison to bits is at 32 float which gives you 768db above 0 DBFS. Yes there are 32 bit float R2R ladder AD/DA converters like the sound devices mixpre and others but most audio interfaces on the market today are 24 or 32 bit fixed point. But in the end the song has to be good and like my mentor told me back in 1995 no one can see your gear on the radio. Cheers Colt!!!
Good post! Although it’s the filter that removes the aliasing that creates the smearing rather than the alias image itself, maybe that is why some people like non-oversamplig DACs.
This is such a bollox argument. The simple truth is regardless of how crystal clear/super hi fi a song may have been recorded, if a song is strong enough, unless the recording is full of hiss, a good song is a good song.
What's the saying, 'you can't polish a turd'.
why not 88.2 and then downsample to 44.1? perfectly half. depending on the resampler, going from 48 to 44.1 can introduce artifacts that wouldn't otherwise be present if you just recorded in 44.1 from the get go. maybe what you're describing around 8:00 is that distortion (euphonic) that's being added in when you resample
Can you explain why different mathematical ratios would provide different conversion artifacts?
IIRC perfect multiples in sample rate conversion is a myth.
@@gius_taakstudio 44100 is *exactly* half of 88200, as 48k is exactly half of 96k
@@vinylcabasse it seems you can get exact math with other SR too, with some other math: "to resample exactly from 44.1KHz to 48KHz you need to use the ratio 160/147 (and the inverse for the other direction)"
it seems sample rate choice is more important for other stuff: ruclips.net/video/-jCwIsT0X8M/видео.html
Great Video and 48k is at the Atmos standard at the min also
That was very helpful and enlightening I've switched to 24 bit. Like you said there's more room on computers these days but stuck with the 44k. And didn't really know why I should go to 48k. But now I'm convinced. Thanks
Wait if you change that rate when you try to sync a music video most cameras are using the 44.1 but soon as you try to sync it becomes a drift nightmare
The human ear can not hear beyond 44.1. Certainly not 48khz. Anyone who says they can hear the difference is lying. 16 bit is old now. Ok agreed. But 48khz is an illusion. A bluff.
Not really true. I myself have a hard time differentiating it on a full mix, but it brings obviously a different sound at the recording stage. It is still subtle but you CAN hear it easily on a good system. There is also this one video on RUclips where they did 44.1 | 48 | 96kHz recording elecric guitar and it is easy to spot the differences.
true not heard. but definitely felt.
Is the grit you hear in 48k possibly from foldback distortion/aliasing from not oversampling? Kind of like keeping more saturation which our brains like to hear.
🍉
I actually recently switched to 48k/24bit! Glad I made the right choice. A lot of clients have been requesting that so I just stay in that now.
Totally agree on 24/48 as the default. RE 96 some of what you are hearing may be jitter related and demands higher quality clocks. i.e. 48KHz is less demanding on having high quality clock generators.
I'm curious how you determined that it was sample rate in general rather than something about your specific converters that caused the differences you heard.
Interesting points BUUUUT… I prefer the sound of 44.1 and so do a lot of people. Why?.. who knows! It maybe that crystal clocks have been made in 44.1 for longer? Tighter clocking? I don’t know, but you should ask Jack Joseph Puig why he also prefers 44.1 over 48 🤷🏻
I have to disagree, but just on the delivery format for mastering. As 44.1kHz sets our nyquist frequecy just above the hearing swell, it is more than enough for anything "final". And as upsampling (for example for a music video) is often way more transparent (except if you have a lot of processing power and the correct software/filters) than downsampling, it only makes sense to upsample instead of downsampling if another samplerate is needed. Correct me if i'm wrong of course!
Yeah, I have to completely disagree with that. You always want the ability to have the highest resolution you will ever need. One day we will be listening to 24-bit 96K files over streaming. It certainly would be a shame if you only ever had 16 bit, 44.1 files, and then up sampled to 24/96. because you will absolutely hear the difference between 16/44.1, and 24/96. hope that helps!
having nyquist freq. right above our hearing range aint enuf from preventing aliasing 100%. it recochets back from the limit back to our hearing range (depedning on material). so it is better to have a higher sample rate just to b safe.
@@ColtCapperrune Fair point, but as i understand it, resolution is in the Bit-Depth. So you are right, it would be easy to tell 44.1/16 and 96/24 apart. But you aren't able to hear the difference between 44.1/24 and 96/24. That can proven by taking a recording at 96/24 and downsampling it to 44.1/24. If the downsampling has been done correctly (high filter steepness and using a linear phase anitaliasing-filter), the only difference in the files lies upwards of 20kHz, which is inaudible. If somebody downsamples incorrectly, there can be some loss in the upper frequencies. If somebody upsamples incorrectly, the only change would be imaging above the original nyquist. TLDR; In my opinion you are correct about the Bit-Depth, but not about the samplerate.
I absolutely hear the different between 44 and 96. I also feel the dynamic difference recording an instrument into it.
@@claymillsmusic 🧢
Good material thanks bro
So you’re saying you can hear a difference between 48k and 96k ? You must have been blessed with golden ears. Most people can’t tell the difference between a 320 Kbps MP3 and a 44.1k wave file.
The difference is not his ears. It's in his head. There is objectively no difference in the analog signal being reproduced.
@@StevenAakre I know. I’m being sarcastic.
96k is a huge waste of CPU usage.
The other reason for higher sample rates is that you get fewer aliasing artifacts when using plugins with nonlinear distortion effects.
basically that's the only reason, everything else is voodoo placebo
That's interesting! Like a tube sim with warmth emulation for example? Now you say it, that's logical !
Thank you for a really interesting view on this Colt. I read that Bob Katz preferred to record in a samplerate of 88.2 kHz. Have you ever done that? And when you say that you record in 48 kHz 24 bit, does it mean that you go for the same settings from recording to mixing to mastering to release? I'm quite new to this and try to understand because I heard someone say that you alway should record, mix and master at a higer samplerate and then sample down to the release value, whether it's 44.1 or 48.
Thanks for your opinion, lately I indeed was wondering if the 44.1k would still make sense, since there's no more CDs to be bound with. But at the same time I was wondering if audio streaming services are still sticking to the 44.1k, which is not so clear to me. But nowadays I'd gladly do everything in 48k at least to avoid samplerate conflicts in Windows that now seems to use 48k as a system standard.
Yeah, I was thinking the same. But my conclusion was, record at the highest sample rate that you reasonably can. Then you will be covered. But then, you have conversion to worry about. I think aliasing is worse for us than sample rate conversion.
When the CD standard was developed it was based upon what humans can hear and also what was practical at the time from an electronics perspective. 700MB CDs were huge at the time. 44.1khz allows sounds up to 22khz to be recorded and reconstructed faithfully. 16 bits is plenty of dynamic range for most audio purposes. Most computer hardware was 16 bit capable but 32 bits was not common in early 80s. So for playback 24 bit 48khz is more than ample especially considering the bulk of the audience cannot hear it and listen on Bluetooth or highly compressed audio formats. Higher bit depth will increase quantisation noise and so will require dithering to be removed. Higher sample rates will allow you to sample supersonic (higher than human hearing) signals. The only justification I've heard for higher sample rates is when processing effects to make it easier to filter as aliased signals will be more supersonic requiring filters that are less complicated. That said if you over sample you won't hurt anyone and worst case you will be using more CPU and disk space for diminishing improvements in audio quality.
How its played is irrelevant. Ailising can only be introduced in the mixing stage. Once the final song is exported. It can be converted back down to 44.1khz without a flaw. 44.1khz allows perfect reconstruction of the entire audible range. There are no advantages to playback at 48khz response or even higher.
Sorry but this is misinformation, the only way you can listen the 96k is if for some reason your hardware performs differently than 48k (LPF aliasing etc), human ears are unable to listen the extra frequencies that are included even at the 48k. 48k has a highest freq of 24kHz and our ears at the best case can listen up to 20k.
There are some use cases for high sample rates such as pitching down an audio file (higher freq that 20k are going down to the audible range) or aliasing issues in some old plugins but none of the stuff you are claiming are scientifically true.
Don't be shy, do an AB test
Another great tutorial from The Man!! I recorded my song at 96, sounded great! Saved to it usb, took it to my car, and the car stereo wouldn't play it! ha! Only played my previous version I recorded at 44.1..
Not mentioned here or in the comment: non-linear processing (compression, saturation) without upsampling have serious aliasing artifacts at lower sample rates. Many waves plugins have this problem. Worth it to at least mix at 96k unless all your plugins have internal upsampling.
"More clinical". Seems bro forgot his aids
cool info Colt..Cheers
If you're using Lightpipe and need a lot of channels, 48k/44.1k give you eight. 88k/96k fold those channels and only give you four.
On the flip-side the higher the sample rate, the lower the possible buffer size and subsequent latency.
That's the determining factor for me: if I need extremely short latency, I use 96k. If I need more channels, I use 48k.
Fyi tidal don't care. I suggest you analyse their output. It is not very high.
I track everything in 96khz 32 bit. I have the Apple silicon and latency has not really been an issue for me since I’ve made the switch. I can honestly say if I had to switch back to 24bit 48 kHz tomorrow it wouldn’t bother me to much. Great video man. 👍🏻
Hey there, 96k inherently has less latency than 48k. Juliane Krause' videos do a great job of demonstrating measurements for different sample rates and buffer sizes.
Switching from 48k to 96k requires more bandwidth, so the threshold where an audio signal will start clicking or dropping out will decrease. Latency improves because audio is being sampled faster.
Yeah man, I say record the highest possible and worry about conversion later. You never know when you need extra bit rate and sampling.
I'm interested to know which interface allows you to track in 32 bits. I currently know 2 of them.
If you are entirely in the box I would suggest staying at 44.1. If you're not capturing analog signals then there are scenarios where up sampling is actually worse for your audio depending on the processing used.
Can you give me some examples
@@Robangledorf if you want the full explanation my knowledge is referring to a white paper by either Fabian or Vladislav of Tokyo Dawn fame. Dan Worral has also done a good video on high sample rates. But to try to sum it up if you are using non-linear processes in series it can be better to let them upsample and sample internally as they won't 'pass on 'as many alias-able frequencies to the next process.
I think it's interesting to note, for historical perspective, that the only real reason why CD's were limited to 44.1/16 was because that was the rate and depth at which a reasonable amount of music (about the same amount as an LP) could fit onto a CD in PCM format. It was a technical and marketing choice, not one of quality. Now, sure, the marketing at the time claimed that this was the highest quality one could ever hope to hear. Heck, I STILL encounter people TODAY who claim that the human ear can't detect any differences above that standard -- even though all valid research proves otherwise, and the fact of the matter is that the human ear and brain ARE capable of detecting bits as opposed to a full waveform, at least on the subconscious level, at lower audio resolutions. In fact, your observation that 96k sounds "clinical" goes toward proving that: it's closer to the full waveform, so it's going to sound cleaner than anything below it. But that's a whole other tangent for another time. My point is simply 44.1kHz at 16 bits was a standard established for pragmatic reasons in the late 1970's and should not be viewed as any sort of gold standard today. Technology has advanced, and in this case, newer truly is better.
Yes, the CEO of Phillips wanted a certain piece of classical music (can't remember which) to fit on a single CD and the typical playing time of that symphony was 74 minutes.
As a former senior sales engineer at Sweetwater I was surprised to discover there was so much difference between the quality of AD/DA conversion paths from different manufacturers. Apogee for instance for quite a while produced a MUCH more accurate sound in 16/44.1 than several of the newer 24bit/48kHz and even 24/96 converters from other manufacturers. This is when the conversation began about the importance of the analog portion of the path in this process as well as the quality of the anti-aliasing filters. Many manufacturers would include higher sample rates as a way of getting around their poor anti-aliasing filters. Over time converter paths have become better over all, but there are still very noticeable differences in the quality of AD/DA conversion from manufacturer to manufacturer and within brands themselves. I mention this because while bit depth and sample rate are important (and I fully agree 16/44.1 needs to be dead because of current streaming capabilities), it's super important that the user is aware of the weakest links in their chain so they don't unknowingly sabotage their path. For instance. There are a lot of people out there (who I've dealt with personally) who are using 24bit/96kHz or higher who are using microphones that aren't capable of recording frequencies higher than what 44.1kHz sampling allows capture of. The highest frequency that can be recorded is half the sample rate (Nyquist theorem), so if someone is recording in 96kHz that means they are able to - in theory (depending on the quality of the anti-aliasing filtering) capture frequencies as high as 48k. But if their microphones are only sensitive to 20kHz, why are they recording in 96kHz? 48kHz also handles beyond 20kHz, but again, if your mics aren't sensitive beyond 20kHz then what's the point? I use 24/48 because I mix for film and TV and all audio projects coming in are 24/48 or 24/96 so I work with what I'm given. Because computers and drives can handle that amount of data processing and storage ok now I'm ok using those higher rates, but 99.9% of the time it's just wasted extra disc space to use any sample rate beyond 48kHz. The dynamic range and signal to noise ratio improvements we see in 24bit are DEFINITELY worth it but sample rates are negotiable for most folks. So my point is, everyone needs to do their homework and determine what they are actually capturing - which is determined by the weakest link in their path whether it be mics, converters, cables or even just the noise in the room - and then use the settings that make the most sense. If vocals are being recorded, or bass or guitar how high of a frequency are you actually WANTING to capture? Most of the time NOT up to 20kHz. Other than classical and jazz, most projects being recorded today still only really need a great 16/44.1 converter path because they aren't needing beyond 15kHz usually and dynamic range is so minimal in pop, R&B, rock music etc that 16bit is MORE than ample. But, to give people a little assist in getting away from poor anti-aliasing filters in less expensive home equipment, sure, go ahead and record in 24bit/48kHz. It probably won't help your sound depending on what you have plugged into the front of the path, but it won't hurt either. Plus Colt makes a great point about being able to deliver to mastering in the bit depth and sample rate that mastering will use and will ultimately be used for delivery. This is especially true now that RUclips and other streaming services now accept and stream in 24bit/48kHz. As a sound designer one of the advantages of using 96kHz-192kHz is the ability to manipulate and slow down sound for some pretty crazy cool sound effect creation. Other than that, upper sample rates are pretty pointless except for those using it as a mandated standard procedure for archive purposes.
Hello Colt! Nice topic ! What about virtual synths and plugins working in the box and higher sample rates ?
The audio interface seem to allow higher buffersizes at 48k and in 44.1 aliasing is much more audible. Just open a synth without oversampling and make a sine sweep in the high frequencies. Sounds like an AM radio until switching to 48.
Great video! For the orchestral stuff - you mentioned you would record in 96k but what would you do for bit rate on that - 32 float?
Depending on what I'm recording, I use 88.2 pretty often. 48 is my demoing go to.
Thanks 👍, I've been trying to help others understand all the sound quality that they're missing.
48 khz/24bit is more than enough for most genres.
I encourage every mix engineer to learn about aliasing, and use oversampling when appropriate (Reaper can do that for you pr. plugin ;-) ), but remember that oversampling does not improve all plugins! Also there are a few places where 96 or higher might be cool...like when doing samples and stuff like that. But for 99% of us 48/24 is the better choice.
Oh and sometimes aliasing can sound BETTER than when removing it!
og and did I mention that Reaper can oversample pr plugin instance 2,4,8 or 16x ?.. and often that oversampling even sounds better than the build in oversampling that some plugins have .. lol.. I know.. I'm a Reaper fanboy.
NB: oversampling on EQ's does rarely give you anything .. so save it for compression, saturation and the like -> stuff that is non-linear
When you said on the end when récord vocal and guitar used 96k. You mean 24bit and 96k ?
Thx for the upload! After watching I quickly starting comparing sample rates from my last few tracks. You are definitely right, sounds better in 48 > 44. I also compared 64,000 SR to 48 and found that 64 sounds a bit cleaner and still gives that slight ruggedness ❤️.
I’ll be bouncing to 64:)
PS. If your sub rate is acting a little funky it’s because I unsubbed to re-sub;)
Haha thanks!! This one was significantly more controversial than I expected and lost more than a couple subscribers over it. Lol.
@@ColtCapperrune I mean, if your clients are requesting 48k files for higher quality streaming platforms it makes sense to switch even just for that reason.
@@ColtCapperrune yea there are some sensitive subjects when you step into the digital technical realm, samplerate is definitely one of them. You have the best mastering engineers and tech experts in the world disagreeing, and if you take a look on forums it's a deep rabbit hole. You could be right or wrong, provide a professional take on what you hear, but if it doesn't match the information they've accepted as correct, then they devalue your opinion !
But Dear Colt, since first thing that Spotify does to your audio, is to convert it to 44.1kHz, if your source file is 48 kHz, doesn't that introduce aliasing and distort the actual sound a bit? Wouldn't it then be making more sense to run it at 88.2 kHz as it's a full octave up?
yes. the conversion back is always ugly, even from double. but certainly less ugly from an integer multiple.
I doubt somehow any of the streaming services are using dedicated high quality offline conversion, that's a whole science in itself. and definitely not if it's realtime lol.
@@Bthelick how exactly is it ugly? Pretty sure it doesn't introduce aliasing as the op suggested.
@@JiihaaS sorry that's 100% wrong. aliasing is always introduced in SRC downwards. You're trying to remove information above the new nyquist it's a guaranteed consequence, much in the same way quantisation clipping distortion is a consequence of reducing bit rate. As an extreme have you never heard a 'bitcrusher' do sample reduction?
@@Bthelick that's why anti-alias filters are for. The information above the new Nyquist is low passed away before the SR conversion. Are you saying aliasing still happens somewhere in the process? Where and how?
@@JiihaaS yes because no filter is perfect. At all. Otherwise we'd just record at 44.1??? And the whole debate would be moot.
And the quality of those SRC filters is always worse in non dedicated programs like daws.
Put it this way, the hardware converters in your interface will likely be working at 128x the sample rate to avoid aliasing on capture. And unless you're using a very expensive interface with burr/brown etc converters even at 128x that process is not invisible.
Daws have come a long way in the last 10 years but it's truly never truly clean, even if you use a dedicated separate program SRC like one of the best r8brain by Voxengo.
Excellent video. I also am a 24/48 guy. I think that your description of more aggressive sound is a good one. Many folks don't think about how the filters applied to avoid artifacts of conversion affect the sound in our "20-20k" range. I was always feeling that those had some harmonic color there that altered the sound somehow. Even notice it with my burl bomber.
I had the same thought I think maybe Colt is hearing aliasing and getting into that vibe (it's not always a bad sound!)
I record in 96 and you’re actually not the first person to say that. Once is a one off, twice will make you rethink your decision. It works great when I’m recording jazz stuff but I will say there is an edge that’s missing when I record rock songs. I thought it was my mixing.
Edit: dang, you definitely got a point. I just recorded in 48 and you do hear difference. It’s slightly more biting and even sounds like…kinda tighter.
No talk about anti aliasing filters??
If you are willing to use electronic music as reference, its actually easy to check how different sample rates sound. Just make a simple project with only soft synths/instruments and export in 44.1, 48, 96 etc.
there's no any noticable difference between rates if we talk about sampling rates as storage media. The difference you might hear when changing rate tells you have a poor clock in your audio device, The higher sample rates are needed only for production to avoid aliasing. But all modern plugins use oversampling, so we mustn't worry about it too much.
I used to record at 48/24. I switched to 44.1/24 and couldn't be happier. I have tons of analog gear and mostly mix outside the box. JJP seems to agree, too.
Why the switch
44.1 & 88.2 are the sweet spot on converters I built. 96khz sounds too bright & washy.
44.1/88.2 vs 48/96 will sound different on various converters. ADC more-so than DAC.
This is a nice video but there are some key errors in your explanation of more = more. Nyquist and Dan Worrall have proven that as long as you have double the rate of the frequency you're trying to recreate, you can, with perfect accuracy.
See here: ruclips.net/video/-jCwIsT0X8M/видео.html
For eg to recreate 400Hz, you'd need a sample rate of 800Hz to recreate it. The human hearing range of 20Hz-20KHz is completely covered by 44.1KHz. 20KHz is recreated by 40KHz just fine, and higher sampling does not increase the accuracy of the recreation at all.
The problem with higher sample rates is that the closer you get to Nyquist the more aliasing you create in the ranges above what you can hear, and these issues literally fold backwards into the audible range you can hear.
I think, though I'm not sure exactly why, that most people in video work with 48KHz because the framerate is usually 24fps, so it's easier to time align frame by frame when you're matching 2 cycles to 1 frame perhaps?
The higher you go above 44.1KHz the more crucial it is to cut the sound above 20KHz on your tracks to stop aliasing bouncing back into the audible range. More cycles per second DOES NOT provide better accuracy in recreating 20Hz to 20KHz, there is little benefit at all to using higher sample rates. Watch Dan Worrall's video on this subject and you'll see.
Thank you :) my words :)
Hey, audio post guy here. Most TV is in 25fps, cinema movies are in 24fps usually.
Outstanding professional tip. Sold me.
Those highest sample rates also fit great to recordings for sound design purpose because of the possibilities that it gives to audio stretching.
yep! and also goodbye latency!
@@DAMIENrap how?
Do u can even ear the difference?
Would love to hear/test 44.1 vs 48 at 24bit. Maybe it’s also pretty subtle hair splitting, or even more aggressive sounding to do 44.1…?
Did you do the test? I'm curious.
Uploading to Spotify or iTunes ??