Do Sample Rates Really Matter? How to Make the Choice (With Science!) [SSP
Вставка
- Опубліковано 25 сер 2024
- Sampling rates. 44.1k. 48k. 88.2k 96k. 192k??
How do they impact your sound, and how can you make the best choice for your projects?
Justin Colletti dives into the science behind sampling rates in a fun and informative talk that demystifies what they're really all about, and puts to bed much of the confusion surrounding this topic.
Thanks to Soundtoys, Sonarworks and Eventide for sponsoring this podcast.
Get more from Justin at masteringdemys... or mixingbreakthr...
For the articles mentioned in this episode try:
sonicscoop.com...
sonicscoop.com...
This video, and the one on Bit Depth, are an absolute treasure. I'd done several courses in person and online focused on recording audio for classical music, and I'd read a couple of books (Richard King, John Eargle), and I had never understood some very important stuff until I watched these two videos. Huge thanks for the time, effort and generosity you put into sharing your knowledge. You really are a great teacher.
yessssss. this guy is somethin else. i always come away knowing so much more than i need to know for whatever i'm doing.
After 19 years of producing and engineering this is the I say THE best explanation of sample rate I've ever heard. Not even when I was in engineering school was it broken down so simply. Thank you very much for this dope video.
Really helpful! Love your pauses at the end of certain explanations, that's something you don't see much in UA-cam nowadays...
Our man out here has such a smooth and articulated speaking voice. A pleasure to listen to.
Thanks Justin, this was really helpful. In the future I'd love to see a similar video about bit depth.
slowpogo 24 bit. Not 16 bit.
At 96k, every VST synth sounds the best I've ever tried (and there have been quite a few). Even the Omnisphere, in the manual of which it is written that there is no point in using 96k, because all the samples in it are recorded in 44.1k or 48k - sounds better in 96k. Obviously. This is immediately audible, perhaps because all the fx processing in the synthesizer starts to sound better when the sampling rate is higher. The same patches sound different, and not in favor of 48k. This is not fiction, but real long-term experience.
When I mix in 96k, I also make fewer mistakes (especially in the high-frequency ranges), the work of the saturators is more audible.
Also in 96k latency decreases, and recently in my studio, even guitarists monitored themselves with effects (FF Saturn 2) through the Logic Pro, and did not feel any discomfort! It would be impossible at 48k - fact.
Yes, it requires more processing power and disk space, but it's worth it. IMHO.
I tell my assistant at the studio " The more someone is worried about sample rates or what kind of converters/clocks we have - the WORSE their music is." It's true. These are the people obsessed with audio quality (even engineers) who don't realize the lyrics suck, the melody is boring, or the track is generic. They worry about 18kHz and dither. They have no clue and point it out by worrying about nuanced minimal things like tech stuff.
I own a ton of great records with terrible sound issues. And they are still GREAT.
@@OFR HA!!! i always say this! i mean... all of that money. equipment. time. mics. thousands and thousands spent everyday and the music sucks. i feel bad for you guys in the studios with those sorts, because you wonder where the real music is that actually _moves_ people. and it's all trap beats and me-toos out there. a million sheerans and swifts and futures. and yes, i obsessed over getting the interface with the quietest preamps. but believe me. the dream is there. in my music, there is always a dream. if it wasn't organically dreamed up- that melody, that concept,- those lyrics didn't become a thing. as much as i obsessed over mogami cables and signal-to-noise ratio, i obsess over developing concepts and melodies that _move me._ the lyrics are born from those like little lively children, developing and waiting to be set free. also, i am an 80s baby, so when i say dreams, i mean those kinds, colored with neon from the 80s, and also some bright shapes from the 90s.
Often that is true. However, many great performers I have known and/or worked with are obsessed with both. Sadly, there appear to be fewer and fewer studios and performers that seem to care about either, let alone how to achieve it.
*cough* dan worrall *cough*
I find running plugins that don't do oversampling and create harmonics like Soundtoys Decapitator, there's an obvious difference in sound at 96k vs 44.1k. And when I do a null test there's a big difference in the lows and highs. The differences are hard to hear otherwise.
I don’t understand why this video isn’t getting more likes and views! You got Gold sitting on your channel mate! I’ve been blessed just by watching it
Thanks for sharing!
I’ve heard this info 1000 times but never explained this thoroughly. Like…I understood it theoretically. But now it makes intuitive sense.
Important to remember that the waves created in air are more like circles / rings in 3 dimensional space, so the the wave on a timeline is a 2D representation of a kind of 3D spiral.
Great vid!
48kHz gang we out here
Nowu of the North I use 48000 & 96000
We out. I strongly prefer 48 over 44,1.
A great video! One thing that's worth a mention, though: if you're really into extreme sound design, that's the one justification for massive sample rates. If you're going to stretch sounds dramatically, it might be subtly advantageous to use giant numbers. I should add, I do that kind of extreme sound design but I still don't use huge sample rates (I use 48k because that's what the games/tv/film industry demands, and as you say in your video there's a possibility that 48k *might* be better than 44.1k so I'll go that far).
One of the most underrated podcasts when it comes to audio!
I am surprised no one mentions sample rate importance when audio gets stretched. Like with quantizing drums, or stretching loops, etc. The higher the rate is, the more you get to stretch audio without loss ;)
It finally came up in the comments. It's great point. Hadn't thought it through but make perfect sense.
do you know any article or video where i can read more about this?
@@javadis @Javier Martinez no need to read, just find a 96k drum loop,
convert it to 44k and see what happens when you stretch them both.
Foley recordings are made with high sample rates like 192Kh for this purpose in fiction productions
Bottom line. I don’t care if it’s easier for people to work with lower sample rates than 88 KHz-that should be the minimum-or how exacting you have to be above 44.1 KHz. I don’t care if computers are pushed harder. I don’t care about the hearing limitations of my next door neighbor or the engineer who mixed and mastered something. I don’t care about CD’s low standard. I don’t care about a studio’s economics. I value music and sound quality. No one would tell a great painter to save money by watering down his paint because the average person just doesn’t notice the difference. I don’t care about excuses. I know from personal experience that 88 KHz -192 KHz, when down correctly, sounds noticeably superior and reveals much more information. I know it from SACDs, from DVD-Audios, from Apple Hi-Res No Loss tracks (even when listened to through Bluetooth which dumbs down the signal). Anyone claiming it doesn’t make a difference is either hearing-impaired or a liar.
The problem with super high resolution audio isn’t that the “average person” can’t tell the difference. It’s that trained listeners can’t (past a certain bit depth and sample rate) in properly controlled blind tests.
Thought experiment: Why not advocate first a 1,000,000,000,000,000,000khz sample rate and bit depth?
At a certain point, there would cease to be an improvement for human beings, no?
Can you think of a better way to find out what that point might be than properly controlled blind listening tests of some kind?
Hope that helps,
Justin
@@SonicScoop SACD’s and DVD-Audio sound light years better than CD’s. Music originally recorded on SACD’s sound miraculous. Simple as that. No arguments change those facts.
@ReaktorLeak 3 possibilities. 1) you heard a subpar SACD-very possible. 2) you heard it in a bad listening environment-the A/B test would’ve negated that. 3) you honestly can’t tell that much of a difference even with a great SACD like Philip Bailey’s Soul on Jazz-I believe it was recorded in SACD-or Mary Chapin Carpenter’s masterful Time Sex Love. I don’t believe we’re medically advanced enough yet to deal with those kinds of hearing issues. There is a vast difference between a proper SACD/DVD-Audio and CDs.
@ReaktorLeak Both stereo and surround sound significantly better when properly done. The cable connections must be analogue and not digital which might not always be the case in a listening room if they don’t know about that requirement. Otherwise it’s not SACD or DVD-Audio but instead a lower resolution output. Also, The Rolling Stones SACD was just OK but far from an example of what SACD can and should be.
My mixer made the decision for me. It can only use 48k, so I just use 48k.
wow. listen. you should have been my teacher in high school. i just listened to 39 min of the most informative sonic education non-stop, and even revisited several parts.. and it didn't even feel like learning. it was just consistently interesting. i have watched _A LOT_ of videos, and you always offer so much more than what is commonly regurgitated, which is why i can't stop watching. i want all of those bits that no one else explains... and this video was the straw that broke the camel's back. this video did it for me.
i'm ready to put a ring on it. lol awkward. sorry. #alittlesorry
seriously, thank you for making all of these videos!!!!!
In summary: higher sample rates are useful for when you're using plugins that make a lot of harmonics (like limiters, clippers and saturators) and don't use Oversampling or any other anti-aliasing technique inside. Because of this, delivering a finished song at higher sample rates than 44.1/48 is useless.
THANK YOU , I dont have 40 minutes every video Id like to watch
Sample rate is not about what you can or cannot hear, but the way you may process the sound.
As someone below me has already said: sample rates matter a lot more with regards to processing than reproduction. I probably couldn't tell the difference between a sample rate of 32kHz and 96kHz just listening to a record. But I certainly can hear a difference when rendering a mix with a bunch of nonlinear processing at 44.1kHz vs 96kHz. It's a subtle difference, but it costs me virtually nothing so I do the latter.
Aliasing artefact might be quiet, but I strongly believe that they're responsible for giving digital audio a bad rep.
Michael Anderwald 96k costs you virtually nothing? It doesn't double the load on your system?
@@Tazmanian_Ninja - thats exactly what its doing.
thanks for saying the truth, thats the reason I decided to go with 88 khz, and my processor can take it and I think I can aford the extra electricity the processor will use. just sone thing audio as is 32kHz you might be abble to tell, but starting at 44.1 and 48, 88.2, 96 to me sound all the same without processing any eq plugins ect...
A clarification- When you talk about Nyquist, you imply that digital didn't start happening till around the 80's. AT&T began using Pulse Code Modulation on T1 carrier in 1959 as a method to get 24 voice channels on to one line, for long distance service.
Nyquist Sampling Theorem translated into lame man’s terms: 44.1 kHz of Sampling Rate creates a perfectly recreatable digital version of the full frequency spectrum a human is capable of hearing, actually, even below and above it. 1-22050 Hz. If you play it back, there is no measurable difference, let alone an audible one. If you can hear a difference on a higher sampling rate, your audio interface cannot capture at that sampling rate properly anymore. Or maybe your playback device cannot play it back correctly. Building a ADC or DAC of higher frequency ranges as perfect as lower ones is very hard to do, especially on the hunt for ever cheaper production cost or selling price tags. That’s as easy as explaining all of this is. But yea... I get it: It sounds too simple to be true with all of that esoteric offerings for better sound out there.
I'm half way in the video but I had to write this:
Thank you - I'm blind so I can't understand visuals. But in your presentation now I got how soundwave and digital representation of it really works behind the scene. Especially the part where you're having fun explaining the moveing forward and back of everything.
Good points, all true. Many audiophiles want to believe that that next piece of gear is sounding better. I noticed recently that my Focusrite Clarett 4 pre, though it samples at very Hi Res rates, on the spec sheet it says that its analog inputs are all limited to 20kHz! So you are sampling non existent frequencies anyway.
Great. I really appreciate your effort in making these podcasts.
Awesome to hear! Glad to be helpful.
The best explanation I’ve ever heard! Very well explained without trying to sell us on industry dogma. Loved it!!!
Awesome to hear Jeffrey! You might also like our video on but depth :-)
ua-cam.com/video/ShmvwTTHDJs/v-deo.html
Very best,
-Justin
Absolutely brilliant if you can explain it in a simple way it makes you smile then you know this man knows what he is talking about. Well done thank you
The 18 dislikes for this video were obviously clicked by disgruntled bats.
exactly. 31 dislikes now, from angry, bleeding bats who wanted more, so they chose to record jet engines at 24/192k.
Thanks Justin. I've wrestled with that question a lot. For me, it comes down to how much processing power I need to do what I want to do with a mix quickly. I'd rather stay at 44.1 or 48 and not have to offline process or freeze tracks. That slows down the mix and gives me way too much time to second guess everything.
one main thing people should realize, if your waveform is at 44.1, it's exactly the same sinusoidal waveform as 192.
You can not "add" quality by converting that to 192.
like he says in the vid. you only need a peak and a trough. dac's produce the perfect waveform from those datapoints.
Thanks for a great video, and wonderful explanation about this subject. Must say its the best explanantion and breakdown i have seen and heard! :) Thanks man! :)
Superb information. This video coupled with your bit depth video completely explains digital audio capture. Thank you.
Once you mentioned Nyquist I know you are on the right track, thanks for clearing up the air.
Though at 13:04, the samples taken don't have to be at the peak nor trough for the math function to reconstruct precisely the wave form of the sound to be recorded, as long as the other criteria are met like you described in the video so beautifully. In the cases where people claim that they could tell the difference of various sampling results through A/B test was largely because of inferior low pass filter being applied during sampling, hence noticeable artifacts from aliasing. Do take into consideration when doing blind A/B tests, majority of mics and speakers do not record nor play anything meaningful beyond 15k, rendered the AB tests irrelevant for comparing anything beyond 44.1
Though at 13:04, the samples taken don't have to be at the peak nor trough for the math function to reconstruct precisely the wave form of the sound to be recorded, as long as the other criteria are met like you described in the video so beautifully. In the cases where people claim that they could tell the difference of various sampling results through A/B test was largely because of inferior low pass filter being applied during sampling, hence noticeable artifacts from aliasing. Do take into consideration when doing blind A/B tests, majority of mics and speakers do not record nor play anything meaningful beyond 15k, rendered the AB tests irrelevant for comparing anything beyond 44.1
I have an inquiry. I was thinking, if you have a sample rate of 2 samples per second (2hz), and a bit depth of 10 bits, and let's say for a second you draw out a 1 hz sine wave at it's max amplitude on a graph that has samples mapped on the X-Axis and Bits mapped on the Y-Axis. And let's say the sample points are exactly on the peaks and troughs. You would have one sample at 10 bits and one sample at -10 bits And let's say what you draw is the signal getting sent through your DAW to the speakers. The DAW will take the original sine wave signal, and capture it as 2 points, a peak at 10 bits (at the first sample), followed immediately by a trough of -10 bits (at the second sample). Your Digital to audio converter will perfectly send those 2 signals to the speaker in sequence over a 1 second time period. 1 signal telling the speaker to move forward because of the signal point at 10 bits, the other telling the speaker to move backward because of the signal point at -10 bits. In my mind, this means that no matter what kind of wave you draw, be it a square wave, a sine wave, a triangle wave, or whatever wave the actual signal is. At a sample rate of twice the frequency, no matter what wave you draw, you will have the exact same signal being sent from the DAW to the speakers. Two points, one at 10 bits, one at -10 bits. The benefit to increasing the sample rate beyond twice the frequency of the original signal wave, is that you will be able to replicate the specific shape of the original signal wave drawn. If you had more samples in between, you could connect the dots in a fashion that looks like a sine/triangle/square wave or whatever, and have the speaker accurately represent the signal that you drew. So increasing the sample rate beyond twice the frequency would yield a more accurate wave because it would be able to more accurately differentiate which kind of wave the original signal is. Just learning this concept so I might be wrong but this is just what I was thinking. Please tell me if I am. Idk. Big fan of your stuff btw. (oh, I think I just figured out why you're right and I'm wrong, because our ears only capture audio at 20,000 hz. meaning we don't even have the ability to shape the wave with extra samples in between. If our ears are capturing a wave and turning it into data points that our brain can interpret, we wouldn't even be able to shape a wave at 20,000 hz because we wouldn't be physically able to map out more than 40,000 data points per second in our ear canal. We would take any wave at that frequency, and interpret it the same way, just as the DAW did in this example, as 2 data points to be sent to our brain to interpret as sound. I do wonder though, if even though our ears can only hear up to 20khz, if we can still differentiate between certain types of waves up until that point. You could test it out with a synthesizer in a super high sample rate session But I think 44.1khz is good enough, lol, I listen on crappy headphones anyway.)
Yeah man, atleast thats how i understood it. This video has it drawn out on the X and Y axis. Hopefully this helps. ua-cam.com/video/9z5HOZ4C1KY/v-deo.html
You are right. 44.1 is really fine.
Wrong
Eh, I like 48k better. Or better yet, something like 60k just to be sure you know :)
@@Beos_Valrah To be sure of what?
@@380stroker Really, why?
@@jordillach3222 Of the absence of anti-aliasing effects/artefacts caused by low-pass filters.
From my experience, stretching audio recorded at higher sample rate, the transients are more natural. Pitch tunning also may benefit from recording at higher than 44k sample rate; i'm talking about processing after recording.
ตí sհմ Your point about pitch processing is a great one. And it's one of the few types of processing that can’t be helped by oversampling. A niche case for sure, but if you often find yourself doing really significant amounts of time or pitch shifting in the downward direction or know you need to do it in a specific case, it’s a good reason. Thanks for weighing in.
ตí sհմ it certainly can. I discovered that in Ableton live - you get better pitch shifting but not time stretching. Weird huh?
@@infinaneek Well, "better" time stretching: I guess it makes sense because it is just a very controlled granular synthesis. So in theory you have a "smaller grains" resolution with higher sampling rates, but it doesn't mean implies that the "granular result" of it would sound better (expect maybe if you're DAW's sampling rate is the same as your audio recording). This is not a scientific answer, btw.
I listen to a lot of 192/24 albums on Qobuz and do a bit of Hi-Res mastering.
While there isn't much musical information outside of the 20k range per-se, it is my perception, there is less "PCM sound" at at 192kHz. Filter ringing (what you referred to as resonant peaks) in the audio band inevitably interacts with the musical signal. Even 96k doesn't quite get there to my ear.
It's not about resolution as much as transparency of the medium.
For real-world production level work, there are a thousand more important decisions than sample rate, but for the audiophile remastering world, I welcome the 192/24 releases.
Excellent video and brilliantly explained Justin.
This was very helpful thank you!
Looking fly as always Justin!
I have been recording in 24b/96kHz but I have watched this video twice (I "double sampled" it, lol) and am wondering if I should be using 24b/48kHz instead. I am thinking this is going to be the route I take, but one question: When dithering down 24b recordings to master to 16b/44.1kHz for CD... is there ANY sonic advantage in tracking the mix in 96k vs. 48k when you know its going to be dithered down to 44.1kHz? Thanks for what you do, Justin!
Nah man. Just eating up your disc space and processing power. Stick to 48 especially if you're gonna dither down. The 96 is just way too much overkill with no reward. Great question though!
I use 48kHz when working with video projects as it is the standard sample rate for most non-linear editing systems...
Yes it is theoricaly possible and practical to process your mix in 88.2 or 96 kHz even if you end up with a mp3, the distortion created by aliassing stay in the audible range and will be present no matter what you do after. The question is can you hear it ? you might think you wont hear it and some will say its placebo but it is real and the Nyquist-Shannon sampling theorem is still true and respected.
listening to you explaining sounds physics is the best ASMR I have ever experienced.
Very interesting!
Glad to hear you've been liking them Yuri! Thanks for tuning in :-)
-Justin
10:25 Beautiful analogy.
Thank you, I learned a lot watching this video.
I needed to hear this, thanks!
Great vid and I'm only 5 mins in. There is a limit to our hearing when it comes to certain freqs and I think that rule has ruined modern production.
People scoop out the bass etc automatically in guitars and bass to make the perceived volume louder (because radio is still a thing aparently and the volume wars rage on.. Sarcasm). They do this because the human ear can't hear those sub freqs so why have them there?
Producers don't even listen to the music, they scoop out the lows via a parametric eq AUTOMATICALLY on bass and guitars because of that perceived freq rule. Why do I think this is shortsighted? SPEAKERS can hear those sub freqs. Even earbuds react different to these more lopsided mixes.
You can see monitors pop and flex when you let that bass fly. What does that do? Creates thump and waves to your chest and creates analog speaker distortion which can be pleasing (similar to the harmonic distortion of pushing tape being more ear syrup vs digital limitations/cracks and pops) .
This modern mixing rule is my number one qualm and I think filling the gaps with white noise and perceived volume is why people switch off modern guitar music after the initial impact wears off. It's just not moorish.
Bass flub should not be a dirty word IMO. Every damn mix I see on UA-cam shows people sucking out low freqs regardless of a 4 string standard or a 6 string bass. Is anyone questioning that these producers have not had guitar based hits for decades? Since (sic) DAW replaced tape?
Yes I just had coffee.
When it comes to sample rates I do want to go to town with cymbals and see what's what. In my experience cymbals are the best medium to really hear that digital scree when something is lossy/glitchy.
A lot of good points in the video, but I still keep hearing people talk about so much that has absolutely nothing to actually do with the concept of why sample rates are actually considered important (or not), which really confuses a lot of people, which is apparent by some people's comments. Understandable because the idea is a bit counter-intuitive: what's happening in those non audible ranges is potentially affecting what is in the audible range...
It's to do with the processing... it has nothing really relevant to do with human/hearing ranges AT ALL, that's not why sample rates are important, it's really to do with the aliasing and artifacts being incorrectly produced in the audible spectrum that should not be being produced in AD PROCESSING in the FIRST place... then on top of that you have the potential for additional aliasing... depends what you are doing... and that's not even touching bit depth. Or dithering... yuk.
In practical terms think about this: You plug a condenser mic into the preamp of your standard entry level interface, it IS receiving freqs way above 22k, these are hitting the (likely) cheap converters that are not recognizing these freqs properly at all, (or using various digital FX etc creating ultra high freqs) and instantly you've got aliasing... subtle artifacts creating very cheap digital sounding mixes... Of course this is a gross over-simplification and not always the case, but it shouldn't really be that surprising.
Whilst gear is certainly not everything, Invest in a good preamp/interface ( - if you are serious, and making good music... if not... why bother reading this? ) , it really is one of the most important pieces of digital kit you have, use a grass roots-ground up philosophy, this is well known by producers young and OLD, it's not some stupid gear myth like pretty much everything else, and is worth your money. I mean, while they're not bad, do you really think a shitty 200 dollar multichannel interface can even remotely compare to a state-of-the-art preamp and AD converters? No... of course not... but that's a whole other can of ear worms! Do you need it? Of course not!... Does it still sound good? yeah sure... kinda... but does it sound like good ol' rock n roll? Fucking hell no!
My 2 cents is basically this: ultimately it's not what's important and the MUSIC is always #1, but it does matter... kind of... just because most people don't care doesnt mean we should just lower the standards of the Earth... (laughs) that's exactly what's wrong with the world! But I digress...
A good song is a good song! If the music's good.... it's good! Always focus on that....
HOWEVER... It is unfortunately an unavoidable aspect of the digital era and medium, and worth understanding properly as an engineer/producer or person who wants to take any sort of professional level recording seriously...
Personally i tend to use 88.2k, - because i'm exporting to 44.1k... the freq cutoff of 88.2k is not so close as to create excess audible artifacts, and seems to make a noticeable difference, it's also perfectly double to the out rate of 44.1 so you don't run into other problems (or it used to be, 48k is actually the common standard now because of the advent of dvds and digital media, also a lot of modern music hardware runs at 48k, if that is you're target out rate then choose 96k). I don't advocate higher than those, more is not better, but hitting the right mark based on your output sample rate IS. You won't harm things by down sampling to 44.1 or 48 from 88.2 or 96k respectively, the damage was either done in the first place or not by that point. Processing power shouldn't be an issue, you've got nothing to lose.
BETTER STILL... Do some proper tests for yourself at different rates at least with what you've got, it really depends on what you are doing, you can certainly notice a difference, without a doubt, but again it all depends on what you're doing and what you're dealing with.
That's a lot of rambling, but It's very late here and im also a bit drunk. Basically, IMO choose either 88.2 or 96k depending on whether you are exporting to 44.1 or 48k... (double) think about it...
Peace.
Age doesn't correlate to lack of hearing frequencies. It would if some guy was a rocker and constantly around loud noises. My mastering engineers are 55+ and they can hear amazingly well. They've taken really good care of themselves throughout their lives and protected their hearing. It's like saying some Oreo eating 22 year old being out of shape vs Tom Cruise, and that Tom Cruise must be fat because he's older. No. It's habits. Also genetics.
Higher frequencies do matter if we're talking about pitch shifting, sound design, frequency foldback etc. The filter being higher is better.
One thing many people don't mention is the Gibbs Phenomenon. When a signal is digitized it picks up a vibration or ringing when truncating. It can create a ringing at step transition in the time domain. Higher frequencies lessen this effect and put the filter far outside the hearing range of humans.
In Reaper when I upsample with the option "Use project sample rate for mixing.... " unchecked, my 48khz mix will be processed at 96khz and saved to a 96khz file. The processing is different , the aliasing is different, and even my reverbs sound a bit more 3D! So why people say there is no difference?
because people relly one science and visual/mathematical data, not actual sound.
One issue you did not cover has to do with the difference of sample rates as they pertain to virtual instruments. Some virtual instrument developers use the 44.1 kHz sampling rate in their delivered products, while other developers use the 48 kHz sampling rate. This poses a problem for those of us who are media composers. Doing projects with VI's that use the 44.1 kHz while having to deliver finished cues (or stems) at 48 kHz requires sample rate conversion. This, to my ears, causes a degradation in the sound quality of the resulting music. Also, some projects have a requirement to deliver the music at 96 kHz. It is simpler to just double a project which used VI's at 48 kHz rather than having to employ sample rate conversion because of a VI at 44.1 kHz. No client has asked me to deliver a project at 44.1 kHz or 82.2 kHz. Therefore, I have made the decision to use only virtual instruments which are delivered at 48 kHz. Any thoughts on this aspect of sampling rates?
VST developers started using 48khz because the math was easier during conversions than 44.1, especially because of the .1 after the 44. It was just easier math and thats why alot of people jumped on using it during development. Especially because there isnt a huge difference in hearing 44.1khz and the 48khz (unless of course you are converting afterwards). You should be able to choose your sample rate and bit depth in any DAW before starting a session.
Thats probably also why no client asks for a sample of 44.1khz because its pretty much useless these days and they arent asking for sample rates of 88.2 because no one is sampling at that rate. After 48khz the sample rate jumps to 96khz.
This might be helpful. ua-cam.com/video/9z5HOZ4C1KY/v-deo.html
Brilliant video as always. Thank you.
Hey, really nice that you speak about it, I spent some hours on that subject (and read the full Lavry paper also ^^).
Just to be clear, we do agree that you need slightly more than double the maximum sampled frequency to get sampled, right ? I am aware of the way converter work with sincs, but if you would try to record a perfect 20kHz sine wave with exactly 40kHz sample rate, you would end up with constant (DC) signal, right ? so you need that extra kHz or so to get the information needed to recreate the 20 kHz ?
Also, why would the anti aliasing filters be close to half the sampling theory, and not simply cut above the hearing limit ? why not for example sample at 88.2kHz, but have the filter at 21, 22 or 23 kHz to get rid of all the "sampled artifacts" ? You still get more samples for time-stretching / working with plugins that do not oversample, but without having potential intermodulation or so, I would guess ? Hope my questions were clear.
Have a good day !
Loved this thank you
Awesome, glad you liked it! There’s a lot more like it if you want to subscribe :-)
Best wishes,
Justin
Great video!! echoes what my friend has told me about using 44.1. CD quality is 44.1 16 bit so if its good enough for the final product its good enough to do the initial recording in.
Pretty much! With that said, there are fair reasons for recording at 24 bit instead of 16 bit. You can give yourself a larger window of dynamic range and lower noise floor when tracking very quiet or dynamic sources. It basically gives you more room for error. (About 144 dB instead of about 96 dB, which is already a lot.)
But assuming proper gain staging, yes, you could probably get away with tracking at 16 bit without much issue at all-if any. The chances of it making or breaking a record's sound quality are slim indeed. But there, I think the arguments may be slightly better. At least for tracking purposes. Will have to do an episode about that sometime! It is also an area that is very much misunderstood.
Thanks for another great video! Any chance you can dedicate some time discussing whether you’d mix and/or master 100% in the box or go with a hybrid 2-Bus setup (including Comp/EQ/Color) that can also be used for single track/bus processing or maybe even a hybrid setup that starts off with analog summing then to Comp/EQ/Color? I’m less interested in the controversial summing piece, but thought I’d include the option for completeness. All this assumes great hardware, great converters, a treated studio, and quality monitoring. The debate would be whether there is enough of a tangible benefit for a hybrid 2-Bus setup (Comp/EQ/Color) to justify a higher noise floor. Your thoughts??
This is a great topic. The short answer is "test things out and do what works best for you". I'm inclined to say these days that any sound I want to get I can get in the box, because that has been my experience of it in recent years. But it's not always qas fast at first. (Though it's much faster to recall!) And there are very satisfying workflow, speed and creative advantages to having some tactile analog gear in the studio.
Ariel Borujow was on the podcast last week talking about this. He used to work with an analog hybrid system, and abandoned it for digital with equally great results. Then, he tried using his analog hybrid system for a recent mix and loved the results. So he tried it again on another mix. But it didn't come out right and he had to do it again all digital! :)
Suffice it to say, I think there are much more powerful factors at play in mixing than what gear you use, or whether it's analog or digital. That said, using tools you love using can only help. And there are some analog tools I have really LOVED using and have gotten me where I want to go fast. Whether they are worth the expense and recall complication is something you have to figure out for yourself.
Yes: I'm inclined to agree that the "summing" aspect of analog is oversold and probably not what's actually helping for people who like it. That's not to say that these boxes don't have a sound. They can. But I'm not convinced the sound is coming from the summing part itself. There are other more plausible explanations, I think.
But yes, analog gear can definitely have a "sound", and some gear can certainly be identifiable, even blind, and even on very subtle or zeroed out settings. So can many plugin emulations though :) They might not be as unique in that regard as hardware, but some software companies are trying to address that with some success.
This would be great to do a whole episode about sometime. Thanks for the thoughts! Hope that helps in the interim.
Justin Colletti that definitely helps in the interim! I’ll also have another listen to last week’s podcast. Thanks again for taking time to explain the details of these important topics!!
Great explanation. Thanks.
This is all very profound. If we can concentrate on bringing the 16 to 20K (high bound) the oversampling that we do engage in will be more valuable in the spectrums (and fucking harmonics) that we actually interpret. Fucking science!
Beautifully explained.
So useful and beautifully articulated!
I trust my ears, work, apply plugins and specially use external peripherals at 96k or better will make a drastic difference on the final mix. Try it! Even if you have to deliver a final 44.1k mix. Logic X has the tools for that. I'm not alone, research more and will see the importance of higher sample rate. More important, when you pass your audio through an external hardware at 96 maintain the track intact, without any plugin. Summing with AD tracks will result in a better mix than DD tracks.
some people have no "ears" and are obsess with what they think they understand as science. the theory is real and can be proven with math but in reality its sound better so why care about science.
Best explanation around the internet about this misunderstood problem :)
But I wonder, because you mentioned the blind test and Placebo, can this be the case of some audio interfaces?
I didn't hear someone actually doing a blind test with few interfaces to hear if the very expensive ones
are in fact 5 times better than the average, when in fact can be the Placebo effect also, if someone knows which interface
they are listening :)
Wonderful video. I'm at 96/24, much because I stretch and pitch down a lot for ambient music. Also some people won't buy my vinyls unless the resolution is higher than on the CDs. I did learn a lot here. Everything you're saying sounds right. Big thanks.
What do you mean by "resolution"?
By the way, have you ever heard that vinyl records cannot deliver more than 60 or 70 dB of dynamic range. They can't deliver the 96 dB of dynamic range that the 16 bit of CD standard can deliver. What do you need 24 bits for, then?
@@jordillach3222 Here I'm referring to both bit depth and sample rate. I have some limited insight on the range of vinyl. I've been asked if the vinyls are made from higher resolution than the CDs, to which I can happily respond yes, and that leads to more sales.
@@mcpeko As a general statement, "vinyls are made from higher resolution than the CDs" makes no sense. I still
I have an apollo x16 and I mix a 88.2 and I can't say it sounds better...but to me, I can hear what I'm doing much better as far as eqing & compression etc etc.
I hear that my studio monithors sound much better in low frequencies and highs are better to listen with higher sample rate on soundcard
I was more interested in if 48000 is better than 44.1k because it's more divisible by 2 or because of some other conventional reason that keeps you from having to split a waveform in half or something. We already knew it didn't make a difference in audio quality before this 39 minute video
I swear that I can her a different tamber of the whole track at different bit rate
great video thanks so much! But sample rates help with time stretching, warping, and pitch shifting, no? retains the quality more.
But lots study shows us that ultrasound also can impact on visualising the sound in different way...In nature, we don't hear sound as pure sine wave, we take it as a harmonics ...
lots of musical instruments can produce ultrasonic sound that can give us different kind of musical test(may we can't hear ultrasonic as a sine wave , but in harmonics , it might give different colorization to the preceiving sound)..
There is only one study that I'm aware of, from Japan, decades ago that tried to suggest this, but there were some real flaws, and its results have never been replicated.
In this study, they played all sorts of music to subjects under brain scans. They found no difference in brain scans when comparing with and without ultra sound... except in just ONE of the pieces of music, which was a piece for gamelan chimes.
In that one piece of music, and only that one piece of music, the experimenters perceived a (subjective) difference in their reading of the brain scans of a small number of the small sample size of participants, and they believed these results could have been statistically significant.
The results from that study have never successful been duplicated.
So.... maybe? But there's no substantial or persuasive evidence of it, and a lot of existing evidence to the contrary.
Hope that helps!
-Justin
I’m not sure where I heard it but I have in my head that because video usually has its audio at 48k, it’s better have your music at 48k to avoid any additional conversion if it gets used in a music video or commersial, etc, and that multiples (96, 192) will convert cleaner than non-multiples (44.1, 88.2, 176.4). Any truth to that?
Thankyou God bless!
24/48, FTW (now and into the foreseeable tech future; for just music AND film purposes) ………. ?!?!?! 😬
I remember some people saying they used 88.2 (over 48 or 96), because it was exactly double 44.1, so when they exported at 16/44.1, ‘rounding errors’ that might/would otherwise occur when exporting from other sample rates were avoided (they may well have recorded in 32-bit, as well, rather than 24 … ‘because it’s easier for a computer to just halve data, rather than multiply by 2/3’ … 🙄😏 … and/or ‘because it somehow avoids rounding errors’).
… No doubt, they were concerned about quantizing … but, given how good computers are at math, that sounded like nonsense then, and it still does now … ?!
Hello Justin, thank you for all infos, really liked it! I have been testing this and I found that the 96k works better for me, simply because if you use any plugin that creates harmonics close to the barrier of 20k, in 44, the aliasing effect will reflect that harmonic back to the listenable frequency range. Even when you're using sounds that goes over 20k (like virtual synths, without the antialiasing) starts to reflect back, sometimes to 600hz. In 96k aliasing exists as well, I tested with some plugins and I could find some frequencies coming back to the audible range, but it was an extreme case. What do you think about that?
For instance, CLA compressor creates harmonics to try to emulate the hardware, if you compress cymbals that goes to that airy highs you could get some aliasing right?
Is there anything, besides the oversampling in some plugins, that we could prevent this to happen in 44?
thank you
best
Yes, higher sample rates equal lower latency, and less ailiasing in the audible range. One must manage imd however above 44.1k sample rates.
Lower sample rates shouldn't cause more aliasing and higher sample rates shouldn't cause less aliasing. A properly designed anti-aliasing filter should reject all frequencies above the sampling rate, no matter what that sampling rate is.
So with a lower sample rate, you wouldn't hear any aliasing. What you theoretically COULD hear is the sound of the anti-aliasing filter, which would amount to a slight EQ difference in the extreme high frequencies.
But with modern digital antialiasing filters that employ oversampling, this isn't so much a thing anymore. When it was more of a thing the difference was pretty subtle, generally speaking, and often occurred in the frequency ranges that are usually inaudible to people over the age of 30 or so.
Latency is a bit of a double edged sword. A higher sampling rate will give you less latency at the same buffer size. However, the extra processing load of the larger sample rate will usually require you to use a larger buffer size, all else being equal.
If you have half the sample rate and half the buffer size, or twice the sample rate and twice the buffer size, the resulting latency is the same.
Hope that helps!
-Justin
@@SonicScoop the plugins on the session are less likely to alias in the audible range with higher sample rates. If all makers properly implemented oversampling for any non linear processes in the plugin, a high session sample rate would not be required.
If tracking you can get away with a lower buffer and higher sample rate, to reduce latency, since even a modest computer can handle quite a few inputs live. So the trade off benefits higher sample rates, as buffer can be adjusted later when latency isn’t an issue.
Hi Justin. Thanks so much for your channel. It’s really helpful and really accurate.
I have a question. So what about these really expensive DAC’s able to reproduce 286 kHz at 32 bits, and these platforms like Tidal where you can stream Hi Resolution files?
Do I understand by this videos that I’m not gonna hear any difference at all? Does it all comes down to 24 bits better then 16?
I have a question about a very specific use case. Guitar Amp VSTs. Some can be quite noisy. I know improved bit depths have a reduced noise floor. Is there any benefit in going from say 48 to 96 in that regard? I assume no but am curious if I'm wrong.
Tchad Blake is 44.1k and I know Serban Ghenea WILL downsample to 48k/44k WHATEVER sample rate your recorded in. perspective :)
I've never heard anyone make a simple mathematical calculation so damned complex and black-magic sounding. The 44.1 standard was created assuming an insanely steep high-cut filter to eliminate the harmonics. Like 18+ db per octave. You can't have one without the other. They assumed the current-day economics of the 1970's and the least expensive way to meet the minimum spec of reproduction, so 44,100.
We now have 192 khz all day long, and the hardware filters to match. Is it better? Well, I record at 96k, but only when high freq is present, and I'm pretty picky with my mix. Could I record at 44.1 or 48k? Yup, but at 48k, I *might* detect some issues with harmonics, and the integer conversion from 48k to 44.1 is problematic in that case. If I know there are no cymbals or high-freq percussion, then I just track at 48K to save some space.
I prefer to just side-step all the edge-conditions and use 96k when transients are expected (yeah, my SSD hates me). If we assume 30 kHz to include all the 2nd order harmonics from a cymbal emanating 15K, then 60 k samples/S would be enough, using Nyquist's math. I use 96K, to avoid any "rounding errors" found in real life, because 48K is a wee-bit shy of my personal comfort zone for harmonic overhead. Once mixed, then the output sampled at 44.1 kHz is fine, but inside the mix, I prefer to track at 96k for high freqs.
Love it🎼💯
Which sample rate should I choose to streaming with obs ? 44.1k or 48 k ?
Finally the full truth about sample rate. And given to us so clearly. Thank you a lot!
Can you make a video for "BitDepth" as well?
Yes! I plan to. Hope to do one on that and one on mp3s, AACs, streaming formats and the like. Look for those coming soon.
@@JustinColletti No need to look - there is a Bell Icon for this purpose (((:
Cheers!
What if you take a 96khz file and and cut out the 0 - 20~ khz, then play it on a system with speakers that claim to produce up to 50khz? That way you'll only play 20 - 48~ khz.
A little background, 48k comes from the time with DAT recorders, and it is to this day relevant in studio's. I worked in a studio, recording masters to DAT, very annoying work. I did it one year and threw the towel. Enough was enough, 75% were rubbish songs. OMG.
#1. To begin a true relationship with God Almighty and His only Begotten Son Jesus Christ, first read John 3, John 14 and John 17:1-3. #2. To receive remission/forgiveness of sins; read and Obey Acts 2:38. #3.To receive the Holy Spirit, and to see how some believers received the Holy Spirit in the Holy Bible; read Luke 11:13, Acts 1:1-8-14, Acts 2:1-4,16-21,38, Acts 5:32, Acts 8:1-17, Acts 9, Acts 10, Acts 19:1-7. #4. To learn how to live a Holy Life unto God Almighty; Read John 14:15-26, John 15:1-10, Galatians 5:16-26, Romans 6:1-23, Ephesians 4, Ephesians 5, Colossians 2, Colossians 3 etc. #5. To be ready for the Rapture and the day of the Lord; read Matthew 24:44-51, Matthew 25:1-13, Revelation 16:15, 2Peter 3 etc.
Steps to recieve Jesus Christ as your Lord and Saviour and to live a Holy Life:
#1. John 3 and John 14.
#.2 Acts 2:38 and Mark 15:15,16.
#3. Luke 11:13 and Acts 5:32.
#4.Romans 6:1-23 and Galatians 5:16-26.
#5. Matthew 25:1-13 and Matthew 24:44-51.
Excellent, the best explanation I've ever heard! THANK YOU!
You've been speaking about reproduction of the sound and mastering of the sound. What about recording?
So to end the subject, is there a benefit to record sounds at highest sample rates? Like 24/192Mgz? Seams not, as this would be to capture frequencies for bats? Then again, mangling the sound to the extreme, like in sound design, where sound passes through lots of processors, is sampling higher at recording stage helps anything? Sorry if I ask something that should be obvious after listening to you, but I want to close the subject once for the good :)
Thank you
Thanks! Yes, the one good use case I can think of for music production is if you want to slow down sounds substantially in a sound design setting. If you cut the speed of a track in half, some frequencies that WERE supersonic before will now be in the domain of human hearing.
So, if you wanted to slow down a track to 1/2 speed without losing everything above what would now be 10kHz, then recording at a high sample rate would allow you to do that. Or if you wanted to slow down to 25% speed without losing everything above what would now be 5kHz and so on.
This is a pretty niche application and is not done often outside of sound design for film and the like, but if you ever wanted to do this and not have the slowed down sound be quite AS dark, this would be a potential way to do it.
That said, most instruments put out so much less content above 20k than below that for most instruments you might not hear a super substantial difference. Things like cymbals and such could seem somewhat brighter once slowed down if you used a higher sample rate though.
Give it a try and let us know if it's a significant enough difference for you in that context! That's the one good argument I can think of, but it doesn't apply to everyone.
Hope that helps!
-Justin
Does recording (or just rendering) at higher rates make a difference with fx like delay and reverb?
Does bit depth make a huge difference?
Basically as a beginner in this, I want to know the best to use for general all genres and styles of music in a home studio.
Bit depth just means you have a lower floor for any unwanted noise. 16 bit is fine. 24 bit is perfect, anymore is just a waste of time imo
This is great, i've always thought all this. Science b!7£h! The snake oil that we are fed as producers is surely the highest in any industry other than moisturising creams and make up!
By constantly returning to the issue of playback fidelity you've glossed over the benefits of working with higher sample rates. There's little to no controversy that material with a 44.1kHz sample rate can transparently reproduce the entire audible frequency range but this in no way negates the benefits of working with higher rates during processing. Not sure how in the same video you can sing the praises of modern anti-aliasing algorithms while arguing that higher sample rates offer virtually no benefit... Higher sample rates often greatly reduce or eliminate artifacts created during processing. Lower rates despite being able to faithfully reproduce the entire audible range are often forced to either employ filtering to suppress artifacts created during processing or resample on the fly to prevent them in the first place. As good as anti-aliasing algorithms may become they will always be judged against the benchmark of quality offered by performing the same operations at sample rates where aliasing is no longer an issue.
Will most modern software automatically resample and filter as needed negating much of the benefit that using higher sample rates from beginning to end during processing has to offer? Sure... but is using a higher sample rate for a project during processing and down sampling once for final output always just a waste of disk space and processing power? No. There will always be software that handles issues with low sample rates poorly and certain types of processing will always favour high sample rates.
I believe I mention the benefits of over sampling in both the article and the video.
Yes, there’s a good case for oversampling in converters to improve the anti-aliasing filters and inside certain plugins for processing.
But I don’t believe there’s a good case for super high session and audio file sample rates.
Hope that helps!
-Justin
too long didn't watch; I'm a C programmer and music producer.
CD quality is 44100 sampling rate. During recording this means it samples the voltage on the input 44,100 times per second. Nyquist frequency is half the sampling rate, so you can only record up to 22050Hz audio before you start to lose the information (literally over this, the latter half of the wave starts getting chopped off, so information lost). This also happens to correlate with most of the human hearing range.
DVD quality is 48000 sampling rate. Little bit more highs; 24,000Hz max; little bit more detail captured.
Spotify, UA-cam, iTunes.... every streaming platform... uses lossy compression that throws out most of these details anyway. CD quality is good enough.
Higher than this is just asking the system to process a bunch of data you'll never use. ... It's like an open book test, but you decided to bring 4 books, but you still have to get the test done in the same time.
It’s funny but I just turned 40 and can still hear up to 20k in my left ear and as high as 23k in my right.
True story. Had my hearing tested a couple of years ago and the tester was very shocked!!!
It may be down ever so slightly - but consider my a rarity!
Why does the latency drop when u raise the sample rate?
@MF Nickster effective explanation
Sonicscoop . What sample rate do you use?
Hi Justin, are you still using these Focal monitors?
Sadly, they asked for the back after I demoed them along with others in their category for a while. Right now I have some very nice Ex Machnia speakers up. Very nice.
-Justin
@@SonicScoop Graphene tweeters... Very nice indeed 👍🏻
@30:10 I actually make music for bats -- Batronica Dance Music
Is this man saying that a squarewave and a sinewave is the same thing? kinda getting that vibe 13:00
what if you do sound design, you might get extra fidelity when you pitch down the audio a few octaves?
YES. Absolutely. This is one of the good reasons to record at a higher sampling rate. You can slow down audio significantly without losing high frequency content.
-Justin
i simple use 48khz and 24bit. just slightly Overkill it and you are gold lmao.
Make a video on dsd and pcm