Check out all our courses here: ►Mixing Breakthroughs: mixingbreakthroughs.com ►Compression Breakthroughs: compressionbreakthroughs.com ►EQ Breakthroughs: EQBreakthroughs.com ►Mastering Demystified: MasteringDemystified.com Become a channel member here: ►th-cam.com/channels/qEKv3KIZoZrjv2LymD3iMg.htmljoin
Justin, I hope this makes its way to you and finds you well. My name is Lorenz. I’ve been following your channel(s) and have purchased several of your classes for the last so many years. I refer to them often. Thank you for your time and talent. Some time ago, in one of your videos, you mentioned (paraphrasing of course) “people take (or go to school) lessons to learn an instrument but, when it comes to recording/mixing/mastering, it’s not looked at the same way. I’ve always wanted to enroll, not only to learn the proper skills but, to be immersed in the culture. I’m surrounded by creative musicians and artists but, no audio engineers. As of late, several things have fallen into place and here I go. After 20 years of running my own home studio, today will be my first day in a classroom environment. I would like to believe that you had a small part in making this decision. I’m 56 years old (57 next month), I’ve spent the last 32+ years being a full-time working musician. I’m a local guy, doing local things. Thank you for everything. Wishing you a safe and wonderful New Year. Musically yours, Lorenz
Justin, thanks so much for this video. As an audio engineering instructor, I used to spend a lot of time covering minute details of A/D conversion. The longer I have taught this material, the more I have come to realize that much of this won't impact my students at all! As you say, at the end of the day, it's pretty difficult for the vast majority of people will hear no difference in sample rate differences above a certain point. And, it's far more important to focus on the music, mixing, and session in general. Sometimes it's easy for us to get caught up in concepts that don't matter enough to the end result of our projects.
An example to describe sampling graphically but with words could be this: It is a series of sticks of different heights and evenly spaced, with a hole in the upper part (similar to the eyelet of a sewing needle), through which a “thread” with certain characteristics must be threaded; A very good candidate for the “thread” is the fiber optic cable, which everyone knows cannot be twisted at an angle, no matter how small it may be (because it could break), so it can only be bent forming curves. more or less wide depending on the rigidity that would determine how more or less closed that curve would be. Precisely that rigidity (the tightest curve that we could form) would be the “analogy” of the maximum frequency that we want to record and reproduce (Nyquist frequency). In conclusion, the sampling is perfect (the bits are another story); With higher sampling rates we only get more bandwidth and it remains to be seen whether reaching ultrasonic harmonics will audibly influence the band that is audible. Greetings.
After all this wonderful information I still use 32 bit floating/48k. Not for me, …for my cat. Also I used to worry a lot about all of this until the first time I saw some kid playing music through a phone with earbuds. It’s like painting a Rembrandt only to see it in a museum lit by birthday candles.
I was in sound for many years and way back in the day I sat through many an AES presentation on the subject. Even blind listening sessions. I probably still have a box of studies in the attic. Not once did anyone get it right. This video explains the topic perfectly. Now don’t get me wrong, I can hear the difference some of bit depth/sample rates offer, but above 24/48 it gets to be a bit moot at my age.
Also I was a student years ago and one of my teachers was Fred Catero. He was legally blind and I swear he could hear out to 25k. He also memorized his eq etc settings by hearing. He could reset up a session by sound. I know because I was a good session note taker. After he re set up an eq I would confirm it in the notes and was always perfect.
especially considering most people arent listening to hi res audio with speakers or headphones and playback systems that make it easier to tell the difference. great video, thanks
Thanks for this, I too in the early days of digital was confused by this. I use 96k as an additional confidence in quality, rightly or wrongly. I also believe that its programme dependent how much difference we may perceive, as you mentioned the more complex the waveform the more you have to rely on the quality of the anti-aliasing filter at lower rates, particularly with the sounds that are saturated. Thanks again.
Justin, you just explain life frequency's in terms Ive never though, Im so happy this means so much to me as everything is vibration and Im studying Life.
When you said "boomers" talking about angry comments, I suddenly remembered the cassette struggle with regards to treble. Man I don't miss Dolby Noise Reduction LOL
Yo Justin, great video overall. The first bit about different waveforms and their harmonic content was a bit misleading, that's more like a "fundamentals of sound" concept rather than a sample rate concept. But I saw where you took it. You ended up nailing the part about aliasing and LPF Q factor so you're still the Superman of sonics / the Clark Kent of content.
16/44.1 is probably more than sufficient in most cases. However, to keep a larger dynamic "window" for recording at low levels without worrying about the noise floor or clipping, 24 bit is generally recommended. That said, if you do record at 16 bit, the chance of you "ruining" anything is extremely extremely low. Theoretically, if you record at absurdly quiet levels at 16 bit you could introduce some additional noise. But even then, it's much more forgiving in that regard than any analog medium, especially once dither is applied. 24 bit is probably overkill for any competent engineer. But most systems can handle it and it gives you extra room for error, so may as well! That's the long answer. Short answer is that 24 bit / 44.1kHz is still the "standard" for audio production in my book. Going higher than that is fine but not necessary. As a playback medium, no one on record has ever reliably heard any difference between 16 and 24 bit using properly mastered music in a blind test as far as I am aware of. Hope that helps! -Justin
Very interesting video. Thank you. May i ask you a question? Say you have a mastering chain Comp-> Clip->Limit so three devices that can be affected by aliasing . Where in this chain would you apply a Lowpass filter , say at 20khz in order to avoind aliasing? "After the Comp"? "After the Clipper" ? "After Comp+After Clipper“? Or am i seeing it wrong? As to Mp3 i always find it astonishing how many people claim they hear HUGE differences between 320 and High Res which is still to this day find hard to believe.
You don't need to apply a low-pass filter. The anti-aliasing filter he's talking about is part of your digital audio setup. It's automatically applied because data that tries to record frequencies above half of your chosen sample rate can be ambiguous, and the playback would have no way to discern which possible interpretation was the right one. To ensure those ambiguities never happen, the filter throws out the top half of the sample rate (so 44.1kHz sample rates only record up to 22050Hz).
@@joechapman8208 hi there. Thank you for the explanation. If i am honest i still dont get it. "is part of your digital audio setup.“ Lets say my setup is like the mentioned Comp->Clp->Limit. How do i know there even is a antialiasing filter at all in my devices ? "the filter throws out the top half of the sample rate“ Which filter and where exactly? The comp? the clipper? the limiter? How will i know which one has this filter if at all? It sounds so general to me. So i thought it might be wise not to hit them with information above 20khz? Dunno . This is one of the subjects that confuses me as i have never completely understood it in a way in which i could explain it to someone else,, which means i have not understood it till now.
@@saardean4481 When you set a sample rate and start working in it, every sound is affected by the device's anti-aliasing filter. There are some things like Dither which have to be toggled on or off but this filter is not one of them. Different soundcards/interfaces use slightly different curves, which Justin mentions at one point as a possible difference in sound, but these differences are very slight. One device's filter might curve down to 10000Hz before tapering to no effect while another curves to 15000Hz, but the differences between the two devices at any pitch above 15000Hz would only be fractions of a dB. This is one reason why people might favour a 48000Hz sample rate instead of 41000Hz; this raises the anti-aliasing filter curve a bit higher, hopefully keeping the curve just above the range of human hearing without significantly increasing the system workload. Every time you have listened to digital audio, you have heard audio with an anti-aliasing filter. The constant presence of these filters in your musical experience means that these filters are barely perceptible, and in fact completely imperceptible by most people. If the device didn't engage the filter automatically, digital audio production simply wouldn't work: it would be a complete mess, mathematically. The only time a plugin might need to apply an anti-aliasing filter is if it is using an Oversampling mode, in which case the plugin utilises a higher sample rate to improve its processing and then down-samples the results to your project's sample rate again. Oversampling plugins apply their own anti-aliasing filters as part of that process, so you never have to think about those either.
Thanks. It makes me think of the last Led Zeppelin remasters of 2015.They were available as 96kHz digital downloads.Overseen by Jimmy Page. He was 71 then, so what was he hearing ?!. He's 80 now ! I loved those remasters. Haha
This was very illuminating technically as well as in regard to the actual perceivable nature of audio given external influence (physical and cognitive). Thank you, sir 💎
Just wow! Thank you so much for this! I'm a novice / learner and your awesome video has sent me down a Bard-fueled brain-cram. So to be clear, are you claiming that (practical considerations aside like anti-aliasing and dithering) theoretically just satisfying the Nyquist-Shannon sampling theorem guarantees fully accurate transient reproduction (in addition to frequency / waveform shape)? If so, could you explain that (or provide a link). If not, then, how can you make the (implied) claim that from a theoretical standpoint we should only need a 40 kHz sampling rate to fully reproduce music for humans?
I've always used 44.1. When it comes to super high frequencies I rely on Foobar's spectrogram which I've programmed with a couple dozen colors depicting increments of amplitude. The closer to red, the lower the volume of a frequency, and the closer to blue the louder. It's the best plugin to watch casually OR for mastering, because it's a perfect representation yet colorful and beautiful as it scrolls across the screen. I used to have a short video explaining what it is and how and why to do it, but I hate how I speak on camera so it got deleted LOL *edit: and in the process of typing, I forgot to articulate my point. And so, in the spectrogram, I can see all kinds of things we can't hear. For example I can easily see if a song is a purely digital copy or ripped from vinyl. It's shockingly different looking, actually. When people describe a CD as sterile and the same vinyl as full, and I play both on Foobar and literally see the root frequencies of respective instruments being 4x THICKER on the vinyl version, it becomes difficult to deny. I would say if I DO have a secret weapon, it's that foobar spectrogram and how I have it set up and what I use it for. Sure I have good ears, but the spectrogram doesn't lie. The only unfortunate thing about it is that it only goes down to about 45Hz.
Sure, vinyl can definitely sound different than a digital capture! However, this is because it sounds further from the source (in a way that is cool!), rather than closer to it. Also, certain albums have been mastered differently for vinyl vs other formats, so that can be an additional factor on top of it. -Justin
I have a pretty dumb question, but I need to have this explained to me. What if I wanted to reproduce a 10 KHz square wave? Wouldn't I also have to be able to reproduce all the overtones that helps shaping the square wave to make it distinguishable from a sine wave? Like the following frequencies : 10.000 Hz (fundamental) 30.000 Hz 50.000 Hz 70.000 Hz 90.000 Hz 110.000 Hz What if I record and play back at 44.100 Hz, will it still sound as much like a saw wave than if I recorded and played it back at 192 KHz? Of course, if it's to be played back at 44.100 Hz, then recording at a higher sample rate wouldn't help in this situation, but would it make a difference if both recording and playback occured at a higher sample rate?
I can definitely hear differences in between 48khz sampling frequency and 192khz. But like you said, the differences we hear are affected by lots and lots of things, and might not even cause by limitation of sampling frequency. 1st I have to mentioning the fact that you had oversimplified the audio chain for something you can or cannot hear. To determine what you hear, not only the whole electronic audio processing chain should be way more complicate (which where I will focusing), but the receiving end (aka biological side) also tends to have unmeasurable effects on the results. For a simple example on the latter, your hearing can be affected by simple internal pressure or external pressure of your body. That's also the reason why high blood pressure people will sometimes hear ringing in ear when his heart pumping fast. And who knows, if earth turns 0.00000005% faster tomorrow then maybe I can hear frequency above 15.5khz.🙃Thus, I don't believe in an experiment that had too many uncalculatable constant. I mean, maybe I will do and enjoying it a little bit, but will refrain making conclusion out of it any time soon. From electronic audio chain alone, sampling frequency could mean for ADC or DAC, namely recording sampling frequency and playback sampling frequency. For recording frequency (first), theoritically it should be only had minor filter's effects that causing slightly EQ imbalance in top end (like you said), plus some minor losses in tone. Because although sine wave is a sound, but sound is not a simple sine wave. And we certainly don't have infinity sine wave capability in current electronic capability. I think the lesser dot in sampling frequency are meant to tell you that, not that directly connecting those dot and make a 20khz become 2khz, not while it's still fulfiling the criteria of nyquist theoram. So the "radio" sound effect might be partially right, but nothing in tone cannot be fixing by good mix, mastering and post processing, right? But the one i wan't to focus is how electronic's physics work, which lead to the topic of playback sampling frequency. The playback sampling frequency is the thing that I can definitely say I hear the most differences. For one, many subtle tonal quality in high frequency can now be clearly heard and seperate. And also the insane accuracy of the timing of those high frequency notes. IMO the 1st one is due to the filtering effect you are mentioning (EQ shift in high frequency), plus reemulate some of the tonal characteristic losses cause by using lesser sine wave to emulate the analog signal originally, and plus the extreme low noise floor and higher sudden transient details of the 24 bit depth. The reason of the latter (more accurate timing, thus seperationand attack) IMO are due to higher frequency quartz clock being used in the DAC(usually higher frequency = higher precision = lesser margin of error), higher power supplying for the processing chipset (lower latency), less throtle/congestion and etc. Basically when you put the DAC in 48khz sampling frequency, it's like putting your computer in power saving mode. So, the conclusion is although I do not sure if the differences were caused by sampling frequency limitation, but the reemulate of the DAC signal can definitely sound better by using a higher resolution up sampling's sampling rate. The caveat is it's not actually the definition of high fidelity. Cause the truth is every time when you use any emulation/processing, a portion of the sound characteristic will be changed.
I've gotten clicks and extremely short beeps when converting from 88200 or 44100 to 48000 and adding some effects. Could we touch on that? Haven't bothered to reproduce it with the most recent version of my DAW and plugins, or with another interface, and it's been ages.
My dac lets me pick between 2 different anti-aliassing filters, and I can easily tell them apart in a double blind test at 48khz, but at 96khz or higher, I can't. Better placement of sounds in a stereo image, but more distortion, versus a bit more fuzzy placement of sounds, but audible less distortion.... but you'll need a pretty decent setup to hear that clearly, in most cases, there are other components in the system masking those differences.
Exactly none of these double blind test take into account if the equipment is a limiting factor in perceiving a difference. Particularly speaker performance. The whole audio chain must be capable. Most listeners may not have equipment up to task but you can not say it's not possible to perceive a difference with good enough equipment. Due to this high resolution is no benefit to the masses at this time. If we are very lucky even your typical consumer big box store equipment may be capable in the future. Not to mention also is the recording of high enough quality in execution. if you are only interested in producig product that's good enough for the masses and not the highest quality then do not worry about this debate.
I hear bats when they pinpoint me, and I easily hear 20khz at high volume, 18.5 in normal volume, Its almost impossible to tell in a normal setup because If Im not mistaken it depends on your interface, the converter at 44.1khz is not the same than 48 right ? Anyway in my hearing, 96khz is like the speakers have more range, have like more width, but not in a stereo standpoint, in a 3D standpoint, the stereo is the same but have like more lattitude, but for me 96khz in my setup is harder to mix cause I can hear more information in the high end, wich make more frequency to mix, and high end is really hard to mix when you hear it fully, and I don't have the level equipement I need to monitor it greatly, when I have like little clothe over my ear, I can focus more on "wholesome" high end and get maybe faster results cause high end is really hard and exciting for ears, Its like saturation knob on colors, mastering colors in 32 bits need great monitor / color management profile and mastering, while 8 bit colors you can't push things much. But I have great result when I mix in 96khz and export/transcode in 44/48 Its Indeed like a tape roll off in the high end wich make it really satysfying.
I can hear a difference between 44.1 and 88.2 but only with certain plugins, particularly those that tend to alias at lower sample rates. But give me the same audio track at two different sample rates? No way in hell.
I work in 96 - but end up exporting to lower sample rates of course. Can someone clear this up for me? I didn't quite get him... What's the best way forward in terms of removing this aliasing while keeping those high frequencies up to 20khz intact? He said Analog aliasing filter, but then said that they usually have a resonance boost? Just give me a straight forward best practise 😭
Woah, super interesting. So when it comes to anti-aliasing, if monitoring your mix through an interface, is that filter applied to the stereo output you’re monitoring? Because if you are monitoring a darker signal due to the filter, throughout the entire duration of the mix, then its essentially akin to having a roll-off on the master bus. Having a master bus chain that you mix into is a common practice anyways. So if that filter is applied the entire time you’re monitoring the mix, you can just add top end to taste as you would regardless in the mix. Essentially making the roll-off of the filter meaningless… correct? It’d be like old school guys adding a bunch of top end to compensate for the high end tape roll off. Basically, if the filter is present the entire time you’re monitoring, you wouldn’t think “ah man, I can hear the filter”. Instead you’d probably think, “hmm it’s a little dark. Let me add some top end and balance it out”.
It depends. The anti-aliasing filter is more important for audio capture. Once the audio has been captured, the filter should have already been applied, so there is no aliasing. So having a strong aliasing filter isn't as essential for playback as for recording. Accordingly, playback can usually be even more transparent with fewer tradeoffs. But theoretically, yes: Everything we listen through is colored to some degree, and informs all of your choices. That said, the actual D/A part of your DA converter is probably one of the LEAST colored parts of your audio chain, whether yours is very cheap or very pricey. This is not to say that all D/As sound identical. They are just unlikely to be anywhere near the weakest link in your studio. And you think they are a major weak link, you should probably do some blind listening tests to make sure you're not chasing your own tail too much! Hope that helps, Justin
I think it's unnecessary to risk at a ridiculously high sample rate but..A lot of nonlinear processing plugins are not well coded and can add up to produce significant aliasing.... It's a very slight big crush effect to my ears when this happens... Maybe it doesn't ruin a mix but.... Even if it's partly psychological... it's a different feel in my opinion. I don't like bit crushing. It's kind of a pet peave of mine.
Not to spam the comment section, but another way of stating this is that speaker cones don't travel in straight lines back and forth. A pendulum doesn't swing with a compressed waveform either; the information has to be written in there for a distorted waveform playback, because all speakers in all forms are inherently . . . analog.
I have a problem with your frequency visuals. You showed 3 different amplitudes, not frequencies, as they were oscillating at the same rate but with different peaks.
You’re correct that at 10:10 I say “frequency” when I meant to say “amplitude”. (Hey, it’s live!) But immediately after that at 11:36 you get additional frequencies added with the same kind of graph.
A sine wave can be described by a mathematical function. Only two sampled points are needed to reproduce a fully correct representation as there is only one solution and frequency that will pass through those two points for anything below half the sample rate. The problem with most arguments about sample rates higher than 44khz is the simple and sole focus on the human limit of hearing at 20khz frequency. However, there is also a temporal element to music and hearing. How the human hearing system processes sound is very complex and we are still trying to fully understand it in the field of psychoacoustics which is a relatively young science. It's believed that humans can detect temporal differences of 50 microseconds. Many interesting things like frequency masking, how we process localization or the temporal differences of reflected sound, and many more unexpected phenomena are being discovered and studied. From what I have read in a thesis for an audio engineering paper the benefit of higher sample rates may be better reproducibility of transients as well as more effective filtering. Maybe there is some other element yet to be discovered. My own personal judgment based on my own ears and research is that the 96khz sample rate is an improvement over 44khz but beyond that becomes barely if at all perceptible. Having said this, how well the recording was done is the most important factor in sound quality and any perceptibility will be more likely with a very good recording than it will with a mediocre or poor recording.
I’m not in any position to make any doubt of this explanation, but frequencies are one side of it, but what about the overall sound quality? When you said 41 can sound more like tape than 48, you lost me. Not like in trust, but when it comes to understanding and logic. When it sounds more like tape at 41, what about mp3 files? Will they sound more like cassette tapes? For me, probably with less trained ears than the vast majority of your followers, this made some confusion to me, and would be more logical to me if it was explained due to differences between Bit Depths.
I may be wrong but I think you may need to look at simpler explanations on audio basics, audio university is the best to bring simple complex things in audio, this is more for "advance" audio nerds, I may be wrong tho I dont know your level but based on your comment it seemw to me that you didn't understand the concepts he bring by missing little bits of knowledge
Sound quality is great if uncompressed or 320kbs in MP3 pr AAC 240kbps it depends on the codecs, this is compression/uncompress quality, bit depth is dynamic quality, sample rate is frequency range quality. You got outstanding quality starting at 44.1khz 16 bit (if not huge dynamic range) wav, and you could have one of the best master of the world with this configuration, tho we work in 64bit float / 32bit float internally in the DAW
@@DarkTrapStudio I’m not new to this, but according to the comments I have to be lacking something due to earl training in comparison. My comments about bit depth and sample rate could be more precisely. I took it for granted that it wasn’t usual to use MP3 without compressed files, but I may have wrong. My self use wave files on everything from 41/16 and above. But thanks for any advice, and I will give audio university a go.
It depends highly on the actual audio interface you're using. In some situations latency makes a difference, but most of the time it doesn't, at least if your interface has powerful enough DSP to handle monitor mixing an FX while recording.
Sure, that is maybe theoretically possible.... but in practice it doesn't seem to work that way! For example: The minimum playback buffer Pro Tools will allow me to set at 44.1k or 48k is 32 samples. But the minimum it allows me to set at 88.2 or 96k is double that, at 64 samples! So in either case, it works out to be exactly the same (negligible) latency of about 0.6 to 0.7 ms. However, in some cases, it could be worse for the higher sampling rate, rather than the same. If you are using a lot of native plugins, and your computer isn't the fastest, you might be required to set a relatively higher playback buffer still in a higher sample rate, thereby giving you higher total latency in the higher sample rate! That was an issue for me a decade ago when I was recording a lot. It made higher sample rate sessions sometimes have even higher latency rather than the same or lower. Computers are more powerful now, so you are a bit less likely to have it be that bad. But in general, the fastest latency you can get out of 44.1 will be the same as the fastest latency out of 88.2k. The improvements in latency from 44.1 to 48 or 88 to 96 are some minute as to not really be worth mentioning. It amounts to maybe 0.06ms of difference. Hope that makes sense! Justin
In Ableton I use 96khz rate 32 for recording, the difference is like 7 to 4 ms of latency but yeah it depends on the interface, your daw, and processing power and if you got dsp built in
I think 88.2k for recording and mixing is a good compromise. Plug-in aliasing can be pushed up into ultrasonic territory, and when it’s time to mix down to 44.1k, the math is easy-half of the bits are tossed out. I think I could live with the file size increases.
@@SonicScoopto be honest, if all of my plug-ins had built-in oversampling, then I wouldn’t need to go with a higher sample rate throughout: 44.1k would be fine. But some of my plug-ins aren’t, so, at least for now, 88.2k seems to be a safe compromise. Great content, BTW. The clock face depiction of a wave form was extremely insightful.
I got the worst hearing being a similar age to you. I don’t hear it when recording. I do hear a difference when pitch correcting and also a lot of guitar sims start sounding crap in the top end. Fold back distortions I guess and aliasing. I couldn’t blind test two wavs. That’s not what I use higher sample rates for at all.
Check out all our courses here:
►Mixing Breakthroughs: mixingbreakthroughs.com
►Compression Breakthroughs: compressionbreakthroughs.com
►EQ Breakthroughs: EQBreakthroughs.com
►Mastering Demystified: MasteringDemystified.com
Become a channel member here:
►th-cam.com/channels/qEKv3KIZoZrjv2LymD3iMg.htmljoin
Justin, I hope this makes its way to you and finds you well. My name is Lorenz. I’ve been following your channel(s) and have purchased several of your classes for the last so many years. I refer to them often. Thank you for your time and talent. Some time ago, in one of your videos, you mentioned (paraphrasing of course) “people take (or go to school) lessons to learn an instrument but, when it comes to recording/mixing/mastering, it’s not looked at the same way. I’ve always wanted to enroll, not only to learn the proper skills but, to be immersed in the culture. I’m surrounded by creative musicians and artists but, no audio engineers.
As of late, several things have fallen into place and here I go. After 20 years of running my own home studio, today will be my first day in a classroom environment. I would like to believe that you had a small part in making this decision.
I’m 56 years old (57 next month), I’ve spent the last 32+ years being a full-time working musician. I’m a local guy, doing local things. Thank you for everything. Wishing you a safe and wonderful New Year.
Musically yours,
Lorenz
Justin, thanks so much for this video. As an audio engineering instructor, I used to spend a lot of time covering minute details of A/D conversion. The longer I have taught this material, the more I have come to realize that much of this won't impact my students at all! As you say, at the end of the day, it's pretty difficult for the vast majority of people will hear no difference in sample rate differences above a certain point. And, it's far more important to focus on the music, mixing, and session in general. Sometimes it's easy for us to get caught up in concepts that don't matter enough to the end result of our projects.
You are right but with the state of technology today does not take much effort to go to the next level 96khz workflow to remove any doubts.
listened to this podcast in 16k still sounds good
This video makes me soo happy! Thank you for telling everyone what matters.
Awesome to hear, thanks for stopping by!
-Justin
An example to describe sampling graphically but with words could be this:
It is a series of sticks of different heights and evenly spaced, with a hole in the upper part (similar to the eyelet of a sewing needle), through which a “thread” with certain characteristics must be threaded; A very good candidate for the “thread” is the fiber optic cable, which everyone knows cannot be twisted at an angle, no matter how small it may be (because it could break), so it can only be bent forming curves. more or less wide depending on the rigidity that would determine how more or less closed that curve would be. Precisely that rigidity (the tightest curve that we could form) would be the “analogy” of the maximum frequency that we want to record and reproduce (Nyquist frequency).
In conclusion, the sampling is perfect (the bits are another story); With higher sampling rates we only get more bandwidth and it remains to be seen whether reaching ultrasonic harmonics will audibly influence the band that is audible.
Greetings.
After all this wonderful information I still use 32 bit floating/48k. Not for me, …for my cat. Also I used to worry a lot about all of this until the first time I saw some kid playing music through a phone with earbuds. It’s like painting a Rembrandt only to see it in a museum lit by birthday candles.
I was in sound for many years and way back in the day I sat through many an AES presentation on the subject. Even blind listening sessions. I probably still have a box of studies in the attic. Not once did anyone get it right. This video explains the topic perfectly. Now don’t get me wrong, I can hear the difference some of bit depth/sample rates offer, but above 24/48 it gets to be a bit moot at my age.
Also I was a student years ago and one of my teachers was Fred Catero. He was legally blind and I swear he could hear out to 25k. He also memorized his eq etc settings by hearing. He could reset up a session by sound. I know because I was a good session note taker. After he re set up an eq I would confirm it in the notes and was always perfect.
very well framed Justin. This does everything to de-mystify the myths and rumours about sample rates. I enjoyed watching this video very much
Awesome to hear, thanks for tuning in!
-Justin
especially considering most people arent listening to hi res audio with speakers or headphones and playback systems that make it easier to tell the difference. great video, thanks
High Res audio is kind of useless for normal use I find yes
@nicksterj Thanks
Thanks for this, I too in the early days of digital was confused by this. I use 96k as an additional confidence in quality, rightly or wrongly. I also believe that its programme dependent how much difference we may perceive, as you mentioned the more complex the waveform the more you have to rely on the quality of the anti-aliasing filter at lower rates, particularly with the sounds that are saturated. Thanks again.
Justin, you just explain life frequency's in terms Ive never though, Im so happy this means so much to me as everything is vibration and Im studying Life.
Awesome to hear!
-Justin
Justin, I loved this on Spotify. I’ve saved it and will watch this version as I really need the visuals. Thanks so much for your content. 👍
Thank you Justin, this was very informative. Your content is so helpful.
When you said "boomers" talking about angry comments, I suddenly remembered the cassette struggle with regards to treble. Man I don't miss Dolby Noise Reduction LOL
Yo Justin, great video overall. The first bit about different waveforms and their harmonic content was a bit misleading, that's more like a "fundamentals of sound" concept rather than a sample rate concept. But I saw where you took it. You ended up nailing the part about aliasing and LPF Q factor so you're still the Superman of sonics / the Clark Kent of content.
Hi Justin. What is the lowest sample and bitrate combination that you would recommend for basic acoustic recording? Is it 16/44.1 or 24/44.1?
16/44.1 is probably more than sufficient in most cases. However, to keep a larger dynamic "window" for recording at low levels without worrying about the noise floor or clipping, 24 bit is generally recommended.
That said, if you do record at 16 bit, the chance of you "ruining" anything is extremely extremely low.
Theoretically, if you record at absurdly quiet levels at 16 bit you could introduce some additional noise. But even then, it's much more forgiving in that regard than any analog medium, especially once dither is applied.
24 bit is probably overkill for any competent engineer. But most systems can handle it and it gives you extra room for error, so may as well!
That's the long answer. Short answer is that 24 bit / 44.1kHz is still the "standard" for audio production in my book. Going higher than that is fine but not necessary.
As a playback medium, no one on record has ever reliably heard any difference between 16 and 24 bit using properly mastered music in a blind test as far as I am aware of.
Hope that helps!
-Justin
@@SonicScoop - Thank you, Justin. I really enjoy your videos and your insights.
24 bit can maybe be beneficial in classical recordings for example as Justin already mention it in an another episode if I remeber correctly
Loving your work Justin, very insightful. 🤩
Awesome to hear Chris! Thanks for tuning in. Hope to see more of you around.
-Justin
Very interesting video. Thank you. May i ask you a question?
Say you have a mastering chain Comp-> Clip->Limit so three devices that can be affected by aliasing .
Where in this chain would you apply a Lowpass filter , say at 20khz in order to avoind aliasing? "After the Comp"? "After the Clipper" ?
"After Comp+After Clipper“? Or am i seeing it wrong?
As to Mp3 i always find it astonishing how many people claim they hear HUGE differences between 320 and High Res which is still to this day find hard to believe.
You don't need to apply a low-pass filter. The anti-aliasing filter he's talking about is part of your digital audio setup. It's automatically applied because data that tries to record frequencies above half of your chosen sample rate can be ambiguous, and the playback would have no way to discern which possible interpretation was the right one. To ensure those ambiguities never happen, the filter throws out the top half of the sample rate (so 44.1kHz sample rates only record up to 22050Hz).
@@joechapman8208 hi there. Thank you for the explanation. If i am honest i still dont get it. "is part of your digital audio setup.“ Lets say my setup is like the mentioned Comp->Clp->Limit. How do i know there even is a antialiasing filter at all in my devices ? "the filter throws out the top half of the sample rate“ Which filter and where exactly? The comp? the clipper? the limiter? How will i know which one has this filter if at all?
It sounds so general to me.
So i thought it might be wise not to hit them with information above 20khz? Dunno .
This is one of the subjects that confuses me as i have never completely understood it in a way in which i could explain it to someone else,, which means i have not understood it till now.
@@saardean4481 When you set a sample rate and start working in it, every sound is affected by the device's anti-aliasing filter. There are some things like Dither which have to be toggled on or off but this filter is not one of them. Different soundcards/interfaces use slightly different curves, which Justin mentions at one point as a possible difference in sound, but these differences are very slight. One device's filter might curve down to 10000Hz before tapering to no effect while another curves to 15000Hz, but the differences between the two devices at any pitch above 15000Hz would only be fractions of a dB. This is one reason why people might favour a 48000Hz sample rate instead of 41000Hz; this raises the anti-aliasing filter curve a bit higher, hopefully keeping the curve just above the range of human hearing without significantly increasing the system workload.
Every time you have listened to digital audio, you have heard audio with an anti-aliasing filter. The constant presence of these filters in your musical experience means that these filters are barely perceptible, and in fact completely imperceptible by most people.
If the device didn't engage the filter automatically, digital audio production simply wouldn't work: it would be a complete mess, mathematically. The only time a plugin might need to apply an anti-aliasing filter is if it is using an Oversampling mode, in which case the plugin utilises a higher sample rate to improve its processing and then down-samples the results to your project's sample rate again. Oversampling plugins apply their own anti-aliasing filters as part of that process, so you never have to think about those either.
When telecom was going digital a lot of us wondered why a voice channel was 64,000bps. Those that knew pointed us to Nyquist-Shannon theorem.
Thanks. It makes me think of the last Led Zeppelin remasters of 2015.They were available as 96kHz digital downloads.Overseen by Jimmy Page.
He was 71 then, so what was he hearing ?!.
He's 80 now ! I loved those remasters. Haha
This was very illuminating technically as well as in regard to the actual perceivable nature of audio given external influence (physical and cognitive). Thank you, sir 💎
Awesome to hear! Thanks for tuning in. :-)
-Justin
Just wow! Thank you so much for this! I'm a novice / learner and your awesome video has sent me down a Bard-fueled brain-cram.
So to be clear, are you claiming that (practical considerations aside like anti-aliasing and dithering) theoretically just satisfying the Nyquist-Shannon sampling theorem guarantees fully accurate transient reproduction (in addition to frequency / waveform shape)? If so, could you explain that (or provide a link). If not, then, how can you make the (implied) claim that from a theoretical standpoint we should only need a 40 kHz sampling rate to fully reproduce music for humans?
And with the state of technology today does not take much effort to go to the next level 96khz workflow to remove any doubts.
I've always used 44.1. When it comes to super high frequencies I rely on Foobar's spectrogram which I've programmed with a couple dozen colors depicting increments of amplitude. The closer to red, the lower the volume of a frequency, and the closer to blue the louder. It's the best plugin to watch casually OR for mastering, because it's a perfect representation yet colorful and beautiful as it scrolls across the screen. I used to have a short video explaining what it is and how and why to do it, but I hate how I speak on camera so it got deleted LOL *edit: and in the process of typing, I forgot to articulate my point. And so, in the spectrogram, I can see all kinds of things we can't hear. For example I can easily see if a song is a purely digital copy or ripped from vinyl. It's shockingly different looking, actually. When people describe a CD as sterile and the same vinyl as full, and I play both on Foobar and literally see the root frequencies of respective instruments being 4x THICKER on the vinyl version, it becomes difficult to deny. I would say if I DO have a secret weapon, it's that foobar spectrogram and how I have it set up and what I use it for. Sure I have good ears, but the spectrogram doesn't lie. The only unfortunate thing about it is that it only goes down to about 45Hz.
Sure, vinyl can definitely sound different than a digital capture!
However, this is because it sounds further from the source (in a way that is cool!), rather than closer to it.
Also, certain albums have been mastered differently for vinyl vs other formats, so that can be an additional factor on top of it.
-Justin
I have a pretty dumb question, but I need to have this explained to me. What if I wanted to reproduce a 10 KHz square wave? Wouldn't I also have to be able to reproduce all the overtones that helps shaping the square wave to make it distinguishable from a sine wave? Like the following frequencies :
10.000 Hz (fundamental)
30.000 Hz
50.000 Hz
70.000 Hz
90.000 Hz
110.000 Hz
What if I record and play back at 44.100 Hz, will it still sound as much like a saw wave than if I recorded and played it back at 192 KHz? Of course, if it's to be played back at 44.100 Hz, then recording at a higher sample rate wouldn't help in this situation, but would it make a difference if both recording and playback occured at a higher sample rate?
I can definitely hear differences in between 48khz sampling frequency and 192khz. But like you said, the differences we hear are affected by lots and lots of things, and might not even cause by limitation of sampling frequency.
1st I have to mentioning the fact that you had oversimplified the audio chain for something you can or cannot hear. To determine what you hear, not only the whole electronic audio processing chain should be way more complicate (which where I will focusing), but the receiving end (aka biological side) also tends to have unmeasurable effects on the results. For a simple example on the latter, your hearing can be affected by simple internal pressure or external pressure of your body. That's also the reason why high blood pressure people will sometimes hear ringing in ear when his heart pumping fast. And who knows, if earth turns 0.00000005% faster tomorrow then maybe I can hear frequency above 15.5khz.🙃Thus, I don't believe in an experiment that had too many uncalculatable constant. I mean, maybe I will do and enjoying it a little bit, but will refrain making conclusion out of it any time soon.
From electronic audio chain alone, sampling frequency could mean for ADC or DAC, namely recording sampling frequency and playback sampling frequency. For recording frequency (first), theoritically it should be only had minor filter's effects that causing slightly EQ imbalance in top end (like you said), plus some minor losses in tone. Because although sine wave is a sound, but sound is not a simple sine wave. And we certainly don't have infinity sine wave capability in current electronic capability. I think the lesser dot in sampling frequency are meant to tell you that, not that directly connecting those dot and make a 20khz become 2khz, not while it's still fulfiling the criteria of nyquist theoram. So the "radio" sound effect might be partially right, but nothing in tone cannot be fixing by good mix, mastering and post processing, right? But the one i wan't to focus is how electronic's physics work, which lead to the topic of playback sampling frequency.
The playback sampling frequency is the thing that I can definitely say I hear the most differences. For one, many subtle tonal quality in high frequency can now be clearly heard and seperate. And also the insane accuracy of the timing of those high frequency notes.
IMO the 1st one is due to the filtering effect you are mentioning (EQ shift in high frequency), plus reemulate some of the tonal characteristic losses cause by using lesser sine wave to emulate the analog signal originally, and plus the extreme low noise floor and higher sudden transient details of the 24 bit depth.
The reason of the latter (more accurate timing, thus seperationand attack) IMO are due to higher frequency quartz clock being used in the DAC(usually higher frequency = higher precision = lesser margin of error), higher power supplying for the processing chipset (lower latency), less throtle/congestion and etc. Basically when you put the DAC in 48khz sampling frequency, it's like putting your computer in power saving mode.
So, the conclusion is although I do not sure if the differences were caused by sampling frequency limitation, but the reemulate of the DAC signal can definitely sound better by using a higher resolution up sampling's sampling rate. The caveat is it's not actually the definition of high fidelity. Cause the truth is every time when you use any emulation/processing, a portion of the sound characteristic will be changed.
Very thorough, thankyou
I've gotten clicks and extremely short beeps when converting from 88200 or 44100 to 48000 and adding some effects. Could we touch on that? Haven't bothered to reproduce it with the most recent version of my DAW and plugins, or with another interface, and it's been ages.
10:06 - I'd say different amplitudes and different phase, but same frequency.
My dac lets me pick between 2 different anti-aliassing filters, and I can easily tell them apart in a double blind test at 48khz, but at 96khz or higher, I can't. Better placement of sounds in a stereo image, but more distortion, versus a bit more fuzzy placement of sounds, but audible less distortion.... but you'll need a pretty decent setup to hear that clearly, in most cases, there are other components in the system masking those differences.
Exactly none of these double blind test take into account if the equipment is a limiting factor in perceiving a difference. Particularly speaker performance. The whole audio chain must be capable. Most listeners may not have equipment up to task but you can not say it's not possible to perceive a difference with good enough equipment. Due to this high resolution is no benefit to the masses at this time. If we are very lucky even your typical consumer big box store equipment may be capable in the future. Not to mention also is the recording of high enough quality in execution. if you are only interested in producig product that's good enough for the masses and not the highest quality then do not worry about this debate.
Would love to see a vide on Bitrates.
I got you covered right HERE: th-cam.com/video/ShmvwTTHDJs/w-d-xo.html
-Justin
I hear bats when they pinpoint me, and I easily hear 20khz at high volume, 18.5 in normal volume, Its almost impossible to tell in a normal setup because If Im not mistaken it depends on your interface, the converter at 44.1khz is not the same than 48 right ? Anyway in my hearing, 96khz is like the speakers have more range, have like more width, but not in a stereo standpoint, in a 3D standpoint, the stereo is the same but have like more lattitude, but for me 96khz in my setup is harder to mix cause I can hear more information in the high end, wich make more frequency to mix, and high end is really hard to mix when you hear it fully, and I don't have the level equipement I need to monitor it greatly, when I have like little clothe over my ear, I can focus more on "wholesome" high end and get maybe faster results cause high end is really hard and exciting for ears, Its like saturation knob on colors, mastering colors in 32 bits need great monitor / color management profile and mastering, while 8 bit colors you can't push things much.
But I have great result when I mix in 96khz and export/transcode in 44/48 Its
Indeed like a tape roll off in the high end wich make it really satysfying.
at 10:14 the 3 signals do not have a different frequencies. they have the same frequency but a different amplitude and a different phase
I can hear a difference between 44.1 and 88.2 but only with certain plugins, particularly those that tend to alias at lower sample rates. But give me the same audio track at two different sample rates? No way in hell.
I use Decapitator a lot and blind at 44.1 at high levels of distortion it’s a big difference. I can tell 10/10.
I work in 96 - but end up exporting to lower sample rates of course.
Can someone clear this up for me? I didn't quite get him... What's the best way forward in terms of removing this aliasing while keeping those high frequencies up to 20khz intact?
He said Analog aliasing filter, but then said that they usually have a resonance boost? Just give me a straight forward best practise 😭
Woah, super interesting. So when it comes to anti-aliasing, if monitoring your mix through an interface, is that filter applied to the stereo output you’re monitoring? Because if you are monitoring a darker signal due to the filter, throughout the entire duration of the mix, then its essentially akin to having a roll-off on the master bus. Having a master bus chain that you mix into is a common practice anyways. So if that filter is applied the entire time you’re monitoring the mix, you can just add top end to taste as you would regardless in the mix. Essentially making the roll-off of the filter meaningless… correct? It’d be like old school guys adding a bunch of top end to compensate for the high end tape roll off.
Basically, if the filter is present the entire time you’re monitoring, you wouldn’t think “ah man, I can hear the filter”. Instead you’d probably think, “hmm it’s a little dark. Let me add some top end and balance it out”.
It depends. The anti-aliasing filter is more important for audio capture. Once the audio has been captured, the filter should have already been applied, so there is no aliasing. So having a strong aliasing filter isn't as essential for playback as for recording. Accordingly, playback can usually be even more transparent with fewer tradeoffs.
But theoretically, yes: Everything we listen through is colored to some degree, and informs all of your choices.
That said, the actual D/A part of your DA converter is probably one of the LEAST colored parts of your audio chain, whether yours is very cheap or very pricey.
This is not to say that all D/As sound identical. They are just unlikely to be anywhere near the weakest link in your studio.
And you think they are a major weak link, you should probably do some blind listening tests to make sure you're not chasing your own tail too much!
Hope that helps,
Justin
I think it's unnecessary to risk at a ridiculously high sample rate but..A lot of nonlinear processing plugins are not well coded and can add up to produce significant aliasing.... It's a very slight big crush effect to my ears when this happens... Maybe it doesn't ruin a mix but.... Even if it's partly psychological... it's a different feel in my opinion. I don't like bit crushing. It's kind of a pet peave of mine.
How does sample rate effect the stereo field?
It shouldn’t! The sample rate effects the highest frequency that can be recorded or played back. That’s it.
-Justin
Not to spam the comment section, but another way of stating this is that speaker cones don't travel in straight lines back and forth. A pendulum doesn't swing with a compressed waveform either; the information has to be written in there for a distorted waveform playback, because all speakers in all forms are inherently . . . analog.
Answered!
@@SonicScoop lol Don’t feel obligated to always answer me, but thank you Justin.
I have a problem with your frequency visuals. You showed 3 different amplitudes, not frequencies, as they were oscillating at the same rate but with different peaks.
You’re correct that at 10:10 I say “frequency” when I meant to say “amplitude”. (Hey, it’s live!) But immediately after that at 11:36 you get additional frequencies added with the same kind of graph.
A sine wave can be described by a mathematical function. Only two sampled points are needed to reproduce a fully correct representation as there is only one solution and frequency that will pass through those two points for anything below half the sample rate. The problem with most arguments about sample rates higher than 44khz is the simple and sole focus on the human limit of hearing at 20khz frequency. However, there is also a temporal element to music and hearing. How the human hearing system processes sound is very complex and we are still trying to fully understand it in the field of psychoacoustics which is a relatively young science. It's believed that humans can detect temporal differences of 50 microseconds. Many interesting things like frequency masking, how we process localization or the temporal differences of reflected sound, and many more unexpected phenomena are being discovered and studied. From what I have read in a thesis for an audio engineering paper the benefit of higher sample rates may be better reproducibility of transients as well as more effective filtering. Maybe there is some other element yet to be discovered. My own personal judgment based on my own ears and research is that the 96khz sample rate is an improvement over 44khz but beyond that becomes barely if at all perceptible. Having said this, how well the recording was done is the most important factor in sound quality and any perceptibility will be more likely with a very good recording than it will with a mediocre or poor recording.
Neat contribution 👍🏾
I always work at 44.1
Atmos is a trash Room Reverb effect
on top of that,
Dolby wants to put the same effect on ALL the tracks you play? Trash is Trash.
24/ 48.8 for life. Is this a war crime? :)
Not a problem by me!
-Justin
Audiophiles are slightly dump. All that matters is physics.
I’m not in any position to make any doubt of this explanation, but frequencies are one side of it, but what about the overall sound quality? When you said 41 can sound more like tape than 48, you lost me. Not like in trust, but when it comes to understanding and logic. When it sounds more like tape at 41, what about mp3 files? Will they sound more like cassette tapes? For me, probably with less trained ears than the vast majority of your followers, this made some confusion to me, and would be more logical to me if it was explained due to differences between Bit Depths.
MP3 can have any sample rate.
I may be wrong but I think you may need to look at simpler explanations on audio basics, audio university is the best to bring simple complex things in audio, this is more for "advance" audio nerds, I may be wrong tho I dont know your level but based on your comment it seemw to me that you didn't understand the concepts he bring by missing little bits of knowledge
Sound quality is great if uncompressed or 320kbs in MP3 pr AAC 240kbps it depends on the codecs, this is compression/uncompress quality, bit depth is dynamic quality, sample rate is frequency range quality.
You got outstanding quality starting at 44.1khz 16 bit (if not huge dynamic range) wav, and you could have one of the best master of the world with this configuration, tho we work in 64bit float / 32bit float internally in the DAW
@@DarkTrapStudio
I’m not new to this, but according to the comments I have to be lacking something due to earl training in comparison. My comments about bit depth and sample rate could be more precisely. I took it for granted that it wasn’t usual to use MP3 without compressed files, but I may have wrong. My self use wave files on everything from 41/16 and above.
But thanks for any advice, and I will give audio university a go.
@@oh515 MP3 is indeed a compressed file, again I may be wrong or misunderstood your comment.
I am 29 and I hear only 14kHz xDD
You are probably not alone!
-Justin
You did forget one additional benefit of higher sample rates: lower latency while recording.
It depends highly on the actual audio interface you're using. In some situations latency makes a difference, but most of the time it doesn't, at least if your interface has powerful enough DSP to handle monitor mixing an FX while recording.
Sure, that is maybe theoretically possible.... but in practice it doesn't seem to work that way!
For example: The minimum playback buffer Pro Tools will allow me to set at 44.1k or 48k is 32 samples. But the minimum it allows me to set at 88.2 or 96k is double that, at 64 samples!
So in either case, it works out to be exactly the same (negligible) latency of about 0.6 to 0.7 ms.
However, in some cases, it could be worse for the higher sampling rate, rather than the same.
If you are using a lot of native plugins, and your computer isn't the fastest, you might be required to set a relatively higher playback buffer still in a higher sample rate, thereby giving you higher total latency in the higher sample rate!
That was an issue for me a decade ago when I was recording a lot. It made higher sample rate sessions sometimes have even higher latency rather than the same or lower.
Computers are more powerful now, so you are a bit less likely to have it be that bad. But in general, the fastest latency you can get out of 44.1 will be the same as the fastest latency out of 88.2k.
The improvements in latency from 44.1 to 48 or 88 to 96 are some minute as to not really be worth mentioning. It amounts to maybe 0.06ms of difference.
Hope that makes sense!
Justin
In Ableton I use 96khz rate 32 for recording, the difference is like 7 to 4 ms of latency but yeah it depends on the interface, your daw, and processing power and if you got dsp built in
I think 88.2k for recording and mixing is a good compromise. Plug-in aliasing can be pushed up into ultrasonic territory, and when it’s time to mix down to 44.1k, the math is easy-half of the bits are tossed out.
I think I could live with the file size increases.
Sure! If you at all feel like you will benefit from higher sample rates, 88.2k seems like it's probably the most sensible one to me.
-Justin
@@SonicScoopto be honest, if all of my plug-ins had built-in oversampling, then I wouldn’t need to go with a higher sample rate throughout: 44.1k would be fine. But some of my plug-ins aren’t, so, at least for now, 88.2k seems to be a safe compromise.
Great content, BTW. The clock face depiction of a wave form was extremely insightful.
This is actually completely intuitive for the completely non visual lifelong sound nerd. Saw tooth is harmonics bruh duh
I got the worst hearing being a similar age to you. I don’t hear it when recording. I do hear a difference when pitch correcting and also a lot of guitar sims start sounding crap in the top end. Fold back distortions I guess and aliasing. I couldn’t blind test two wavs. That’s not what I use higher sample rates for at all.