What SAMPLE RATE Should You Record At? | Why HIGHER Can Be WORSE!
Вставка
- Опубліковано 28 лис 2024
- Why recording at 96/192Khz can sound WORSE than recording at 44.1/48, and why audio sample rate bears no relationship to frame rate in video. We bust some myths in an easy-to-digest video that explains the science behind sample rates, aliasing, and bit rate - as well as Mark playing with his new 55" iPad pro!
Many thanks to The Goodbye Look for permission to use their track - check em out!
www.thegoodbye...
If you want a deeper dive into the science, check out this awesome video from plugin developers FabFilter:
• Samplerates: the highe...
Download the test tones at:
www.presentday...
Check out our FREE test master service at:
www.presentday...
Join our new discord server at:
www.presentday...
To receive a 7% discount (and earn us a tiny commission) on a DistroKid subscription and increase your chances of higher ranking on Apple Music and Spotify playlists (assuming Mark has mastered your tracks, that is!) and help use this link:
distrokid.com/...
If you would like to send us new and interesting products for review, or are interested in a collaboration, please email us at:
info@presentdayproduction.com
If you enjoy our content and would like to make a small PayPal donation to the channel (it gets pretty expensive pretty quickly to make these videos!), then we would be eternally grateful, and give you a shout-out in the next video!
www.cosmic-aud...
If you'd like to use Epidemic Sound's extensive library of well-recorded music as a fantastic learning tool, as well as being able to use it in your own content for UA-cam, Facebook or Instagram, follow this link (We earn a small commission which supports the channel):
share.epidemics...
Alternatively, if you'd like to try out a personal Epidemic Sound subscription at a reduced price, follow this link:
share.epidemics...
We are also now Waves Ambassadors and we will be bringing some epic content alongside Waves!
Please follow this link for our PresentDayProduction partnership with Waves plugins (We earn a small commission which supports the channel):
waves.alzt.net/JrqZ47
The ONLY reason I use higher sample rate is for audio stretching. And it’s a rare occasion. Usually some sort of flown in sample that I want to stretch.
Same here. Time stretching at 48kHz is okay if you have very good plugins like X-form for Pro Tools but 96kHz with standard plugins works much better for me. I do audio stretching all the time because I work with vocals groups that aren't necessarily professional.
Using the same principle of Nyquist, isn't better to record at 24bit 96kHz and export the final project as 24bit 48kHz?
Another fantastic video Mark! Marvellous work!
Thank you my friend, much respect!
Warren are you planning to also get a robot in your studio that calls you names? :)
@@iainparamor1482 Eric does that for free!
You can have these Audio Files if you want to see what we're talking about...THERE IS A DIFFERENCE THAT ANYONE CAN HEAR!!!
@@Producelikeapro RoboEric, haha.
Learned more from you in a half hour than from my university audio engineering courses tried to explain over a week. Well explained and thanks for showing examples. People like me learn from seeing and hearing examples not just having definitions thrown at them 🤘🏼
I record audio for film and tv and have been asked to record sound effects such as explosions at higher sample rates as post production want the option for time stretching with no artefacts. But still record 99% of our work at 24/48
As in time stretching while retaining the pitch? How does a higher sample rate result in less artefacts in that situation?
@@dwindeyer Less interpolation.
@@dwindeyer Not retaining the pitch.
That's what I expected from the video to explain. Pretty disappointing.
@@codyrap95 the distortion you get from exceeding nyquist itself is less important than the fact most recording converters just really suck at high sampling rates. I prefer 1080p 60fps over 4k 15fps. There is a sacrifice somewhere
I mostly use 48k at 24 bit. There are instances where I either need higher rates or higher bits. I use higher rates for audio restoration when digitizing and restoration work. It allows me to see all the problems clearly when I have to zoom right in. But it has no effect on sound. For bits higher than 24 on the other hand, I only use to record sound in the field (sound effects). I again, only use it rarely when recording extremely loud sounds (gun shots, cannons, explosives, jet engines, rockets, Formula 1). Higher rates and bits, I assume are also useful for scientists who use sound and acoustics in their studies (bioacoustics, stellar acoustics). But for musical purposes, 48k@24 is perfect.
Can the AD/DA converters you use actually handle the extra dynamic range though (for the examples you gave of reasons to use higher bit depth)?
@@IsaacJDean Steinberg AXR4U can handle it
When doing the 48k vs 96k null test, it would be useful to have a spectrum analyser (Voxengo SPAN is my goto) to show where in the frequency domain the additional harmonics are occurring.
Testing plugins for how well their oversampling algorithms work can be quite revealing - some are awful and the aliasing becomes readily apparent on loud high frequency content (particularly sustained tonal sources like Glockenspiel, Celeste and Triangle etc).
Non-linear processing in the digital domain has definitely improved a lot with higher processor speeds allowing for higher oversampling rates.
Great video, lots of interesting info. I record vocals and other instruments at 32-bit so I don't have to set my levels at all and I never distort a take. I can sing very loud and very soft and never have to worry about my input level and where it is. The files are bigger but I have yet to have a project take up 1gb of audio data. That is with lots of vocal tracks and takes. And my 18TB drives are accommodating to these levels of data. Also, the 32-bit sounds better to my ears.😊
but your converters are still 24 bit, so you can still definitely clip them..
I am using 96KHz for editing purposes when I'm capturing audio or working on a project within a DAW. It's much easier to varispeed a 96K file at extreme percentages without it sounding off. As for capturing audio, it's much easier to edit out crackles and pops on a needledrop or cassette recording than it would be at 48KHz. When it comes to getting stuff out of the box, I always use 48K.
Home at Last sounds fantastic regardless of the sample rate.
I have a music technology degree but have forgotten so much as the years have passed. Your channel is top notch!
Thanks for the video, and thank you for the invitation to disagree in the comments (as long as we include some supporting evidence). I totally agree that 16 bits at 44.1k is perfectly good for music reproduction. However, I have twice recently NEEDED to record at 88.2k. Both times it was because Fiverr session violinists had sent me recordings that were unusable due to their recording hardware not coping with the surprisingly high amount of ultrasonic energy from their violins.
The first session player was using their computer's onboard audio input, and after I heard it I viewed its spectrogram to see what was causing the harsh sound. The aliasing was clearly visible because the violinist's vibrato made the harmonics wiggle quite a lot, and the aliased wiggles were all upside-down. I asked for a revision recorded at 88.2k and then I down-sampled the file to 44.1k on Pro Tools so I could import it into my session. This removed the ultrasonic frequencies digitally and left me with a surprisingly good recording. The second session musician claimed to be using a Focusrite Scarlett 2i2 (I have no way of knowing for sure) and their recording displayed the same upside-down vibrato when recording at 44.1k.
So, the moral of the story is that if the audio interface has not got what it takes to filter out ultrasonic energy, then it is better to record at 88.1k or 96k so that the Nyquist frequency is raised above that energy. Afterwards, you can rely upon digital filtering to remove it while converting to a sensible sample-rate.
Best explanation I've ever seen on this topic.
Thanks, we’re glad you found it useful, it’s a tough topic to cover!
Excelent video! I agree fully that ADDA is always the most comfortable on 48kHz... However, I would like to see your take on the processing sample rates. I've experimented a lot with this (oversampling in the plugins etc.), and from my perspective it does make a lot of sense to have those parts of your chain running in higher sample rates.
We may do a dedicated video for this in particular, however, yes, oversampling in plugins is definitely useful. We found it sounded great on the FabFilter plugins we tested: ua-cam.com/video/bgqWwJ3kip4/v-deo.html . We'll look into oversampling into plugins in particular in a future video. Thanks for your comment!
whats wrong with 44.1 ?
@@eliaszerano3510 sure you pass the nyquist test in human hearing, but you get less artifacts from the plugins at the higher 48k 24-bit.
@@eliaszerano3510 there's nothing wrong with 44.1 if you're recording distorted guitars and drums.
@@caseykittel how about 88.2 then ?
I also record and mix at 48/24. The only reason I would record at a higher rate of 96khz is that my audio interface latency drops down to less than 3ms round trip, allowing me to use VST effects during recording with virtually no latency. Maybe that is a topic you can take on next?
Depending on what daw your using you can fix the latency even better for recording & play back. If you have an option to set the i/o buffer to a higher value for both. This is not to be confused with the buffer latency slider etc…
@@g.m.6417 Not sure I get what you are referencing. And, how a higher buffer size, in the driver or within your DAW can reduce latency.
Think of it this way. If you have boxes lined up and at one end someone fills them with papers and push them to the other side. More boxes (higher buffer size) means it will take longer before the first full box reaches the other end for the papers to be taken out, hence introduce latency. This also depends on how quickly a box is filled and emptied on the other end (CPU power). If only a few boxes, then the CPU may not have enough speed to empty all papers in time for the next full box arriving (crackles and pops).
Hope this simple analogy will help you out.
Thank you, thank you, thank you. Despite having a degree in audio tech and have been recording my own music for the last 25 years, I became almost fixated with the idea of more is better. As a result, I always struggled to optimise my setup to match the continuously evolving marketing requirements and pressure from peers. I think you have just demonstrated in layman's terms the reason why going as high as 96K, 24bit is not only unnecessary but potentially, even harmful. I am now hoping that going back to 48K, 24bit will give my computer another couple of productive years whilst I'll also perhaps enjoy the process more now I won't need to keep optimising my setup for unreasonable specs.
According to Ian Shepherd, dithering actually matters also at when bouncing audio to 24 bits. He points out that audio in a DAW is processed at s higher bitrate (32 bit floating point) than 24 bits, eventhough the material was recorded at 24 bits. Henceforth dithering would be a necessity also when bouncing audio to 24 bits. Please correct me if this is not correct.
you are right
In Electrical Engineering school we learn all about this. Everything you said was spot on. Nice work!
I am 100% in agreement with your summation of 48/24 as this also circumvents any distortion and coloring issues due to the addition of noise that is introduced by the dithering process to a lower sampling rate. Why choose this distortion?
One additional consideration I recommend is to decide before you begin the project whether the final product is destined for video/on line distribution or to create a CD. If a CD, then sample at 44.1/24. If video, 48/24. Avoid dithering, the process deliberately introduces noise to fill in "holes/gaps" at the lower sampling rate.
should render out the difference between the 192k and 44k versions and pitch shift it down. would be interesting to see what that sounds like.
Pitch shifting is interesting for monitoring, it gives you time to think. We've all press halftime in Pro tools.
One factor that would benefit from higher sampling rate is I/O latency, especially for those who rely on non-hardware monitoring while recording. A common example are guitar players recording through IR speaker simulators in the DAW. As long as CPU and HW resources in general allow it, 96Khz usually guarantees the 5ms or less latency most player need to get a realistic response during a recording session, without having to lower the buffer too much and incur in stuttering or artefacts.
Apart from this potential reason for higher sampling rate (which is specific to recording rather than distribution), all the considerations in the video are agreeable, especially since you pointed out the internal oversampling often done by plugins dealing with high harmonic content, such as compression and saturation DSPs. Thanks for the vid, keep up the good work!
Yes... I've found that working at the higher sample rate makes my studio latency much snappier. Should I then be individually re-sampling these files recorded with higher sample rates to the lower rate, or will they sort themselves out when I change the whole project down to 48K for the final mix down?
@@minimalmayhem Not worth it. 96k is good. The fact that it can record frequencies up to 48 K is actually an advantage. There will not be aliasing of the frequencies between 24K and 48k as they are recorded accurately. Aliasing occurs when harmonics above the Nyquist frequency bounce back down below it.
@@minimalmayhem That depends a lot on HW resources. Personally, I do simple stuff for home use, nothing pro or commercial, so working at 96KHz and exporting the mixdown at 48Khz works fine. Some DAWs (Cubase for sure, don't remember Logic) prompt you asking if you want to change the sample rate of all recorded samples when changing the project sample rate after recording, even asking if you want to move or keep samples at their location, so if you want to record at 96+ for low latency, but prefer to mix at 48Khz togo light on CPU/Mem/Storage, that's totally doable. The main limitation is that if want to record new parts after the samplerate switch, you'll have to do it on the new "mixing" setting, with 48Khz and probably lots of plugins loaded, so the I/O latency will be higher.
@@Indrid-Cold Interesting video, but we need to be careful when looking at research data, as it's very easy to be deceived into deducing incorrect absolutes. The capitalisation in "FASTEST transit rate and ANY nerve impulse" appears to suggest that it's humanly impossible to perceive audio latency under 80ms, and we all know that's not the case. Stimuli perception has been studied in depth and there's plenty of interesting literature on it. Perception has been found to be different for visual stimuli (the 80ms Vsauce is talking about, some studies estimate it more around 50ms), touch (approx 50ms with that interesting nose vs toe delta explained in the video) and audio stimuli (approx 10ms), with tests showing substantial differences between casual listeners and musicians. Those times are always averages, not physical constants. Back to audio latency, let's also keep in mind that the I/O latency indicated in the DAW is pure processing time (bufferSize/sampleRate=512/48k=10.67ms), which is only the portion of total perceived latency contributed by DSP. In our machines there's additional buffering and delay happening at drivers layer (interrupts in audio and USB drivers) and can contribute an additional ~10ms, or as low 3ms with latest USB C drivers at 96KHz (again, latency improves with higher sample rate).
@@Indrid-Cold I think you may be mixing up m/s (a speed) and ms (a unit of time).
24/96 is great when you need to pitch/tune vocals or samples; Analog emulation plugins give a much better result espicially at higher freqs (anything that brings saturation), it's very clear with synths or cymbals or overdriven sounds
Actually it’s in the low frequencies that we separate the wheat from the chaff with converter quality and where the higher sample rate converters really shine
@@antigen4 Explain.
@@Wizardofgosz better converters tend to have the BIG payoff in the LOW frequencies at least as much as in the highs ... what more is there to explain?
@@antigen4 let's see some science on this please. I call total BS. Since lower frequencies are not the hard frequencies to reproduce.
Or is this just religion?
@@Wizardofgosz - gotta get up PRETTY EARLY in the morning to fool YOU huh?? haha ... ok don't take my word for it - you wanted the explanation. go listen to some proper converters sometime and let's see what you think. go demo a BURL of Forsell or something in the 20-30K range and compare to an inexpensive soundcard or what have you. this is exactly what i was talking about when i mentioned (in another comment) people dwelling on the superficial ...
Hey guys, mr. Know-it-all here. I've really been enjoying your videos since finding them today. Very informative, very well explained, and the British accent is superior :)
I did disagree with one thing here though where you said that no dither is required when exporting at 24 bit. Since most (all?) DAWs operate internally at 32 or 64 bit floating point, the bit depth is indeed lowered when exporting something to even 24 bit. Of course it's debatable if it's ever audible, but truncation distortion is in fact introduced. And in my opinion, a bit of inaudible dither is better than potential unwanted errors in the audio.
Again, love your videos and it is one of my dreams to work with a team of such nice people some day! :)
Thanks for your great comment! You’re right, but as you said, if it’s audible is another question in itself!
Glad you enjoy our videos, this one is one of our favourites!
@ReaktorLeak You mean that the noise from any component during recording will be acting as dither? That wouldn't work since the noise is printed and "frozen" with the file when recording and is then no longer random. Hence why dither would be needed when rendering/exporting a new file with the material.
@ReaktorLeak When it's recorded, it's printed into the file and then it's the same every time.
There’s so much misinformation about this subject! Thank you for a great video. Also: just because a plugin offers oversampling options doesn’t mean higher is better. The quality of oversampling can vary and often sound worse than defaults. Always trust your ears, not the numbers :)
Yes, that’s a really good point! And ‘always trust your ears, not the numbers’ is so very true. Thanks for your comment!
If you're modeling any kind of non-linear hardware, though, there are benefits to going higher, or at least oversampling those models.
My FIRST video with you... and you ALREADY SOLVE one of my biggest curiosities as I am a NOOB to Audio.
Thank you. And, Happy Holidays!
Thanks for watching! Glad we could help :)
Mark does a brilliant job with this topic. Everything he says squares with the work of Shannon, Nyquist, et al. Mark explains the potential harms of oversampling (faster than at the Nyquist rate).
Wow so much useful information! Especially the AES Test Tone part to put it into perspective.
I did this test myself and I'm shocked that I can clearly hear the unwanted noise in my monitors at 96+kHz for both test tones. Considering I use a typically well regarded audio interface around 250€ this alone definitely convinced me to stay with 48kHz for my recording, mixing and mastering. Thank you for this!
I'll dive deeper into this topic and I'm looking forward to more videos :)
Best regards from Germany
Great stuff. I had already forgotten quantizing, AD/DA conversion and Nyquist and was in the state of "just knowing". This brought back that in simple and easily understandable terms. On top of human hearing only reaching about 20 Hz to 20 kHz, people who the most argue about that stuff are the ones whose hearing often doesn't reach higher than 12 or 14 kHz either, leaving them out of the paradise they're so keen to protect. Not only that, it often feels like many sounds at the top end are nasty unless it's just some echoey shimmery whoosh happening there making the sound and space feel bigger. And many producers love old analog gear like boards that have a sweet top and low end cut. Even better, the mixes often sound better when you cut some of that outer limit of hearing range and beyond rumble and shizzle, making it sound more powerful and focused without adding anything and seemingly not removing anything (a lot of what I said was of course said in the video as well and I seem silly stating it again). Also it was some producer on youtube that showed a plugin that adds tape saturation like distortion to the mastering, might as well use that to get a bit of that warm grit, since it's designed with that goal in mind.
Guitarists also often like to criticize limited bandwiths of old digital gear, and at the same time love using such a tape machine that has a worn out tape and a preamp that heavily cuts high frequencies for smoother top end tone. Weird bunch. I just recently acquired a nice piece of 80's rack effect, a Yamaha SPX90, which had only range up to about 12 kHz and it didn't even have proper AD converter, but a combination of chips doing some kind of trick to digitize the analog signal. The result was a pretty nice sound just because the signal goes through the unit even without effects, even though it is digitized and loses some high frequencies. Just like Echoplex's preamp was used without the tape delay itself, just because it sounded nice. An example of how it is often the case that technical limitations provide happy accidents.
Dither is another funny thing in video context. And gaming context. Playstation 1 had crazy amounts of dithering to make the games look great while saving in technology costs and processing power, allowing more complicated and fantastic games and graphics to exist.
The mention of fps in movies and tv made me remember how some modern shows look weird and unnatural for the technological advancement of increased recorded framerate. I've been looking for a camera that can do video recording as well, and just now realized 60 fps might not even be what I wanted for filming myself. Thanks!
Ultra sonics are definitely generating some extreme Steely Dan sounds in this track :) Thanks for the well explained tech. stuff.
Great in depth video.. thankyou..
There's only a couple of areas you neglected to address, such being the use of higher sampling rates where audio recorded is intended to be repitched in music composition.. once such audio is pitched down by 2 octaves then surely theoretically the stuff we don't want which is floating outside the upper limits of human hearing can be brought down into the audible spectrum.. thus making higher sampling rates more effective at keeping the audible band clear if sound is destined to be mapped across a keyboard of several octave range.
Also though.. as I understand it.. high sample rates are used in such pursuits as Oceanographic Audio Research... But for the same reason.. that there are sounds captured which are of interest to the scientific research, but which lay outside the human audible band.. whereby such sounds are again destined to be repitched in order to bring them into audibility.. By oversampling at 96 or 192kHz, one can again repitch downward by up to 2 octaves without bringing the sampling frequency itself down into the audible spectrum.
People keep saying this nonsense and I have yest to see one single example of this being done anywhere. How can this method even be practical?? How exactly do you use these mythical ultrasonics? I mean you can't HEAR them so how do you even know what you're getting or even IF you are getting when you pitch them down?
@@weschilton
Is not nonsense.. is marine science..
If marine science is using it then it is to some intelligent and practical purpose.
Just because you yourself do not understand something.. or it is outside of your sphere... does not merely make that something 'nonsense'. You only present yourself as ignorant.
As for everything I said about recording samples for intended repitching in musical composition. Such is patently true. It is important to keep the clock signal out of the audio band.
Either you undestand a subject or you do not. Takes no effort at all to call something nonsense, merely because you cannot be bothered to make some effort towards understanding.
Casual and lazy.
Underrated Channel. Great editing and explanation of concepts with the graphics. Keep it up!
Thank you, Aristos!
Very well put video and explanations. Had watched the fab filter too. Have one concern though (although I’m perfectly happy with 48/24), it is that Nyquist theorem applies to periodic signals. Fourier says any complex periodic signal can be decomposed into an infinity of sine waves (but only the fundamental and harmonics in the hearing range need to be taken into account, which rightfully caps the n harmonics to 20-ish kilohertz and therefore the sampling rate to about twice that). The only problem here is that only EDM targeted at Goldfishes can be seen to fulfill the criteria of both Fourier and therefore Nyquist. The very nature of a transient signal (plucking of strings, voice attacks etc) is that you can’t tell much about it in the frequency domain, and therefore the whole mathematics tumbles down because you’re outside of the applicability range and conditions. Still, low-pass filters characteristics are doing a pretty good job at filtering ultrasonics without killing details, and they are known to us in the frequency domain only (gain and phase Bode plots, are equivalent to a frequency sweep). When in need to have a feel for what’s going on in the frequency domain for a non-periodic signal, you can use what is called Power-spectrum density (PSD), wich relates energy content (per time unit hence power) to (pseudo ?) frequencies. But then, it’s only a limited overview and time-response (f(t)) is pretty much all you are left with. In audio, this is you hearing the thing. So there might be a rationale to push for higher sample rates to get more subtle details on highly resolving systems, although my hi-fi, room and listening conditions didn’t quite allow me to hear a difference in such transients between CD quality and 192kHz versions of the same recording. I mean, I think I did (on Pianos, Acoustic Guitars playing classical music in isolation), but I cannot be sure this isn’t my brain tricking me into an “I want to believe” situation. Even though, It would not mean none can hear it on a better system with a better room and better conditions...
As an audio engineer myself, I should point that most of what you said is true, nevertheless, there is some utility in high sample rate recording, specially in audio design and postproduction.
Being so much above nyquist gives pitch headroom to manipulate (lower by several octaves depending the sample rate) audio without losing detail. In that regard ultrasonic information is usefull.
Ummm, WHAT?
@@Wizardofgosz it is a bit tricky to explain, but if you record something @ 48kHz and then you pitch that audio file an octave lower (but in a traditional way, not with a process that adds interpolation like some plugins do) you are basically dividing the number of samples in half, and doubling the duration of the audio file. By doing that the real sample rate of the resulting file would be equivalent to 24kHz, which means that you are now not able to reproduce audio information in the whole audible spectrum do to the Nyquist Theorem, the half of which is the highest frequency you can accurately represent ( in this case 12 kHz). If you record at higher sample rates, even the crazy 384kHz, you should be able to pitch down several octaves before running out of pitch headroom.
Some audio applications could benefit from this fexibiliy.
@@Wawisupreme Did you seriously just say "pitch headroom"??? hahaha!
What's not to understand?
It's a great analogical phrase that simply describes what they are talking about.
I’ve been working in film and music recording my entire life. I’m 58. I absolutely agree with your assessment and thank you for the simple proof of it. Many people need to see this. You did a great job of making it simple.
Thanks for your comment Danny!
This is easily THE greatest explanation on this subject. Ever - and with a great dose of hilarious British humor! Thank you so much guys, we're forever grateful for this. Bless you all.
Thanks R S! Thanks for your comment, I’m glad it was helpful!
I seem to be settling on 24/48, before watching this video, for reasons mentioned in the video, and because I like having creative options, sometimes I like the sound at 48k for virtual instruments and plugins, and sometimes I like the sound at 96K when I use Metaplugin to x2 the sample rate. So, make things easier on the computer, and have more sound options.
Wow. 30 Years recording digital and now I understand it. Kudos to you!
Nice to see this hammered home with some excellent illustrations and science to back it up. I just did my first project in 96k as an experiment and my system just about coped. Its funny how your mind convinces you it sounds better. But maybe I'm just getting better at production and mixing? 🤞
pitch shifting is much smoother at higher latencies. halving the pitch of a 96k file still means its at 48k resolution so its not a grainy and artificial sounding. also the latency at the same buffer size is lower. but i do agree about everything else.
Yes. As an "audiophile" I'm perfectly happy with 16/44, but as a DJ I would much rather have a much higher resolution so I can pitch-shift and beat match without it sounding weird or introducing artifacts.
Yes, in mastering and mixing there is often an advantage at working at 96Khz when you are doing things like pitch shifting, etc., particularly when using plugins that do not do oversampling (Soundtoys, I'm looking at you). But upsampling a 48 or 44.1 file to 96 will give you the same advantage without the risk of distortion, because of the intrinsic bandwidth limit. Live DJing with such effects is essentially equivalent to mixing ITB. In a final master, though, it makes no sense to distribute anything over 48.
I love the comparison between audio and video!! This made both concepts click for me, as I enjoy videography and music production. Thank you 🙏🏾
As a video guy, the comparisons between higher sample rate and frame rate was great!
Cool, thanks Austin!
Being part of the online post production community, I find that there is always arguments about videos running out of sync. And the first comment is always : "what frequency rate did you record your audio?" They have this idea that 48khz mixed with other sample rates - i.e. 44.1khz ) makes the sync drift over time. I have jumped in and pointed out that frame rate has nothing to do with how many times a sound is sampled per second. A second is always a second. I do remember when digital audio started to creep into recording studios that we had to designate a master clock and set devices to be slaves, to stay in sync. So I am now going to save this share link and use this to support the explanation and discussion on sample rates in audio - Would love to hear a more educated explanation of audio drift ( sync ). Thank you for these detailed explanations of sampled audio.
Finally, a definitive answer--that merits a sub. Thanks for the great video. I do have one use-case-specific question in mind: if I were to record a sound (let's suppose the mooing of a cow) with the intention of playback at 0.25 speed (think sound design, not music) would this be a case where recording at 192k would be the preferred method--given that the sound will still have 48,000 samples in each second after it's been stretched to quadruple length?
Is the cow mooing at 96khz? Are there bats in the vacinity that need to be captured alongside the cow? If not, then you don’t need a higher sample rate. If the cow is mooing at, say, 2khz and you slow that down to 0.25, it’s now mooing at 500hz. So a higher sample rate is only useful if there is information higher up in the frequency range you wish to capture. Will you get audible artefacts if you record it at 48khz and slow it down to 0.25? No, because you only need a sample rate of 0.25 - 12khz - to play that back. Hope that makes sense!
To add some juice, i work with ADC and DAC for converting informations and i measured every data you can extract out of an ADC.
Working with a sampling frequency adapted to the bandwidth you work with is the most importan because of folding effect.
In short, it means if you have something going on in another frequency area and without filtering the bandwidth you work with, you will have a copy of this something inside your frequency range.
So filtering is a very important step in the ADC/DAC world.
As well, you want to work with a 24 bit depth but the ADC or DAC effective number bits will be greatly reduced by its internal noise, powersupply noise and all noise added by digital chips sending noise inside the supply chain.
For example, the usual ENOB (effective number of bits) on a 16Bits ADC will be around 12bits (on a very good converter).
Great video !
I agree with the general use of 24-48 for the benefits you mentioned. I can hear the difference between 16-44.1 and 24-48 with certain types of acoustic instrument recording but not generally with pop recordings. The only time I have ever heard the difference between 24-48 and 24-96 was during the recording and playback of classical music in a state of the art recording studio through B&W speakers. Generally just a little extra space around the instruments. Very subtle but was audible in a blind test. Other wise 24-48 is just fine for most pop-folk-jazz recordings. And anything that is over compressed and has no dynamic range then you might as well listen at 16-44.1 or (ARGHHH) mp3 as it generally has no air around the recordings to benefit from higher sampling rates. *** note that the classical 24-48 vs 24-96 listening test was about 12 years ago and I'm not sure I could hear the difference now that I'm 65 but I can still hear the difference between 16-44.1 and 24-48 on most acoustic instrument recordings via A/B comparison. I also notice that after listening for extended periods at 16-44.1 I can feel quite tired whereas when listening for a similar amount of time to analog tape playback I don't feel tired. Digital ear fatigue syndrome ?
Hi, terrific video, many thanks!
My conclusion: 24 bits for sure, 48khz is good for most people/applications, but when you have "know/trusted/great gear" 96khz is worth playing with, especially when using digital plugins behind.
See his point is that no is isn't.
6:40 that 96khz beep had so much more warmth & presence, was it recorded on an Neve 1073 preamp?
Come on Lee, it was an API! You should have heard that! 🤪 The transients!!!
@@PresentDayProduction even though you should have compressed the beep to glue it together ;-)
@@gareth432 It's all about the beep glue.
Ahahahah! Very good!
Omg like butter!
Great video! I never used sample rate above 48 in any of my project only because every time I tested I could not hear the difference, even with colleagues swearing that they could (I always get a vibe of "The Emperor's New Clothes" when people talk about high sample rate). And also I always thought that the file sizes were not practical.
Thanks for confirming that the little voices inside my brain were telling me the truth.
This video is free?! My God the information! Thank you guys! Great content! New Sub!
I 100% agree on the audio arguments, but I'm choosing 96KHz for my new mixing system as a default. Here's why: the customer demand and the difficulty of using multiple sample rates effectively in a hybrid mixing system. The key difficulty here is inserting the analog gear to channels and buses. The workflow needs to be fast so I want to have a script that creates folder track, moves the audio there, and routes the audio through the selected analog gear without any unreliable and time-consuming pinging process. I use a few different converters in the system with of course slightly different latency. It's doable to script the system so that each device connected to a known converter always gets an accurate latency compensation. But this would be a nightmare to do supporting several sample rates since it's already quite a hassle to make it reliably work sample accurately every time. So the workflow is to convert the audio to 96Khz before mixing with an automated script and then the latency compensation matches every time. Here's where the customer demand comes in: I don't want to make the argument to the customers that I can downsample their 96KHz recording down to 48KHz and nothing is lost. If they believe it's better that way then it's better to go with it. And those recordings appear every once in a while. So basically, because of the hybrid approach, I choose to double my processing power to keep a good relationship with some of my 96K customers. Silly, right?
Thanks for your comment, it’s interesting to hear your approach! If you have a workflow that works for you, then great, keep at it! Sample rates are always going to be a very divisive topic, everyone has their own way of approaching it. We love hearing examples like yours.
AMEN!!! Thanks for this, this is what I've known and promoted for ages. 24/48 since 2004. Still holds true today.
Thanks Tommi, and thanks for watching 👍
@@PresentDayProduction it's odd there are many known producers telling people to "back up your material in highest possible frequency", in this context that makes no sense...
Tommi Björk Everyone knows 192khz stores better! Only if you keep your hard drives in the fridge though 😂
@@PresentDayProduction and use the correct cables
This is the best audio production channel on UA-cam. Thank you.
Thanks Simon!
This is by far the best video of the subject I ever saw. Great job.
Many thanks Towi - thanks for watching!
Very nice, great explanation. I am a theatrical sound designer (I record and produce sound effects and voice over bits (often distressed), and edit music, for live theatrical productions.) I started out working in (recording and editing) 1/4" half track, then I switched to minidisk. When I was first working with a DAW, I was rendering my sounds on CD, so I was using 16 bit, 44.1 for my recording. Currently I am rendering sounds to AIFF or MP3 files, I still work in 16 bit 44.1, I don't believe the differences I hear between that and higher resolutions, in my monitors, are noticeable in theatrical playback systems.
Thanks John. 1/4” tape... minidisc... then there was that weird transition period in the late 90s/early 2000s where nobody was quite sure how to get audio from one location to another! Thanks for watching 👍
great! Can't believe you only have 14k followers, it's like a professionally produced television show!
Thanks Gabriel! Thanks for your support, we are working hard on some great new content for next year!
I would literally pay for content like this
In all explanations of audio sampling theory/the Nyquist Theorem that I've seen, a simple sine wave is always used. Would be nice to see it illustrated using a highly complex waveform for a change. I think it would put skeptics in a more accepting state of mind before going into all the other explanations and demonstrations.
I use a focusrite 212 audio interface via a mixer to my phone via a usb adaptor. One thing i have noticed is the focusrite 212 operates at 192 24bit.. fine for recording.. but unfortunately can't adjust the sample rate of the focusrite.. although it is deemed class compliant, i have noticed the differences in social platforms when live streaming.. for example - live streaming to youtube the sound is as you say distorted somewhat.. yet live streaming on tik tok its much better.. live streaming to facebook i have found changes dependent upon their latest updates.. which is frequently. This video is very useful in helping me understand whats going on.. and i have to agree that lower would be better for audio. Trying to figure out now how i could improve my audio between online platforms without using plug ins or Daw etc... as trying to keep my set up minimalist. Hope you could do a video with regards the above for best live streaming audio set ups.. in relation to different social platform requirements & using audio interfaces would be very helpful. Thankyou for the post. Very helpful. 🤗👍
Fascinating stuff beautifully presented. I can 'perceive' a difference between 24/96 and 24/48, but have never been quite sure why until I watched your video! I'll definitely standardise on 24/48 now. I've always used it for video soundtrack work, but will now also use it for my CD output 🙂
Would be interesting to do it in a controlled bind test.
44.1K is most suited for CD, rather than 48K. Better to work with - and export in - what you need; rather than having to get it converted when it’s then burned onto CD
@@PresentDayProduction my CD authoring software automatically converts whatever is thrown at it, and auto-optimises accordingly. Much of my output these days is supplied on DVDs at 24/48 anyway.
Ok. So I've been testing many overtone generating plugins with various sample rates. I came to the conclusion that testing with a full-level linear sine sweep makes it look like a plugin sounds a lot worse than it actually does with realistic sound material. Firstly, high frequencies usually have a much lower amplitude than lower frequencies at equal energy distribution. Secondly, only transients are usually loud enough to mirror back into the audible range.
So, what aliasing actually adds is usually a bit more transient clickiness. That's not necessarily a bad thing. In fact, when I did those tests I often preferred the aliased version, because it sounded more alive. But usually, I couldn't make out a difference. Anyway, there never was any inharmonic harshness, because the amount of aliasing just never was enough. I had to use REAPER's worst sample rate conversion to actually get an unpleasant result.
So after all of this, I'm now working at 44.1kHz with a good consciousness and enjoying the performance and workflow benefits.
I love the video, it's been a dilemma for me for so many years. I've settled to work 48khz/24bit for almost 5 years now. However, there is an aspect you are disregarding and as a mixing engineer i have experienced it first hand. Plugins- Plugins run algorithms. Algorithms processing higher numbers (sets of data inputed) yield higher results (or at least different ones). So running a vocal file recorded at 48khz through a seventh heaven reverb plugin Itb sounds very different than running a 96khz one. Processed reverb might sound smoother or more "lush" simply because the data coming out of it is higher. In photo editing for example, photos are finally viewed on a screen with a maximum size of like 30x25cm, prob jpeg compressed at 1280x720p. But in order to get a perfect edited photo, the original content was MUCH much bigger so editors can zoom in and craft details like eye lashes or pixel editing. The higher the resolution, the more control they get over altering the content. In higher sample rate sessions, especially with plugins processing, the story seem to be similar. Try it with simple waves rbass plugin on a kick, you will have more control with the higher sample rate. With multiple tracks, now you see a difference. Once you finally mix all that down to a 44.1 it just sounds so sharp and tight, just like that big res photo you edited down to jpeg... Try it out and lemme knw - also love from Egypt y'all!!
Thanks for your comment! Most plugins upsample for the reason - check out the excellent video from fabfilter, that explain that part much better than me! 😉 love right back at ya! ❤️
As an "award winning" music producer I agree with you. There's a reason why high profile engineers work at high resolution. That aside, in my line of work music must be the center of attention, a good song overcomes any small technical fail so, to relieve my cpu I work at 24/44.1 ;-) Thanks for the input PDD!
I produce mostly dance music and mixing at 48 coming from 44 I can hear a very clear distinct difference (even before I treated my room).. slighty off topic but I just wanted to share
UPDATE 2 Years on :
Im back at 44k ... I was getting thin sterile sounding mixes at 48 and 96,,, its more clear but doesn't sound like music that is mixed. A bit like watching your favourite film at 60 FPS instead of 24 FPS .. it just doesn't feel "right"
One thing to keep in mind is that for Sound Design, there are some mics that can record up to 100Khz and It is a reason to record at 192Khz to pitch down A LOT a source and still have some crispy high end.
And you still hear only up to 20khz. The point is?
@@tomislavsekerija1957TN If you pitch down a sound a lot, the ultrasonic informations will come to the audible range. If those informations don't exist (typically if you recorded at 44.1kHz), the pitched down sound may miss them to sound good across the spectrum, and it can even cause a sort of ugly distorsion that we have to filter out (the same sort you have when using a bitcrusher effect to artificially reduce the samplerate). In sound design for film or video games, the pitch down effect is very very common. That being said, for most production scenarios I agree with what's being said in the video.
@@Indrid-Cold You already found 2 examples that contradict your first statement. Some crickets also make sounds that go beyond the audible range that can be beautiful when pitched down, and transients from everyday sounds and instruments can go well beyond 20kHz. The fact that you don't find a use in them doesn't mean they're of no interest to anyone (typically audio-naturalists, field recordists, sound designers, sound artists, etc., for study as well as for sound creation). One example of such a microphone would be the Sanken CO-100K, but there are others going up to 50kHz (Sennheiser, Primo).
@@Indrid-Cold Hi Indrid, I'm sorry you took those comments (Paulo's and mine) this way. Maybe this is the problem of UA-cam comments not translating the tone. What I saw in Paulo's original comment wasn't that it contradicted the video's main point (it didn't and didn't claim to). It just answered the video's invite to "comment" and brought further (and interesting) informations. I don't get why you need to feel vexed about it to the point of calling it "silly", "know-it-all" and using sarcasm : it wasn't false, it wasn't insulting, it wasn't aggressive, it wasn't contemptuous, it wasn't written so as to embarrass anyone. From my point of view it doesn't undermine the great quality of this video or this channel, which I just discovered and which I think is awesome.
As for context, I will gladly agree with you on your point, you're absolutely right : in 99% of pop and traditional music genres, these sounds won't be used. However, I stand by what I said earlier. The fact you don't use them doesn't mean no one does (even in the entire music industry). Think of the ambient scene, sound art, movie soundtracks; loads of creative ideas come from niches and other fields.
By the way, this comment is no attempt to antagonize you, I'm just sharing thoughts about music genres I like. I've never tried yet making music with ultrasonic materials, this kinda makes me want to. If I ever make something good out of it I'll post it here.
Peace !
@@barthe7606 using really high frequency for sound design, seems cumbersome, unless it is down-pitching crickets, and similar sounds for the audience that no longer can hear them. So the use for typical listeners, would be quite limited. If that was the kind of sound designer, the original commenter was thinking of, it was a bit unclear, so I get why people would feel sketical to seing such a comment on a video, that points out, that higher sample rates aren't useful for music.
Great video; thanks Mark! I went back to 48/24 after several years at 96K... basically to get more simultaneous channels recording 'live'. There is no difference in quality using good plugins. The exception to 'bigger isn't better' though, is using e.g. Amp Modeling/IR's in a live or live-in-the-studio context... assuming you have enough grunt, higher sample rates will deliver reduced lag to the performer's cue mix and so for recording direct through e.g. BIAS Amp, I have a seperate machine running at 96KHz running the BIAS amp and Torpedo IR sims which *just* lowers the latency enough to make it indistinguishable from IRL with an amp two metres away, and gets that audio back onto t he main machine's interface in almost-real-time to line up with all the other instruments being recorded via 'normal' AD conversion on the main machine at 48/24.
If it's not being recorded live, I obviously *just* record those parts at 96 and import them into the other machine directly with a down conversion to 48. Obviously, I reckon 96 is the go for effects/interfaces that are used live, for similar reasons, but I always keep in mind that 'sound' has 'latency' too... if you are used to playing e.g. 5 metres from your guitar amp on stage, the delay from a 48/24 wireless signal from your guitar to amp and then back to your ears via a digital IEM system is not going to faze you :-)
*I feel like I'm watching the audio engineer equivalent of "Red Dwarf" (I love RD BTW!😏)* 😄👌
Letsall Be-Friends Funny you should say that just as I’m putting the finishing touches on mastering the new Craig Charles (Dave Lister) Trunk Of Funk album! 😎
@@PresentDayProduction Wow! That's amazing! His character was the mentor for all my adult slovenliness! 😅😅
I'll definitely keep and ear out for the release. 🔉🎶👂🕺✨
Fantastic channel BTW! 🙌🙌🙌
I also got the RD vibe and I'm thrilled about it
@@PresentDayProduction amazing!
This is amazing! Great video to show what sample rates really means. I had't no clue :D
If you guys keep this up then you will be Kingz of teh internetz. Excellent channel Bong Friends.
Thanks Big Mac! We’ll try our best to keep it up ❤️❤️
Great video, and excellent explanation of the psychoacoustic's theory involved. While I agree on the whole with this, I would also comment that given the right equipment, and the right listener, Higher sample rates have always sounded more like the 'glue' is working better. A sine wave is only one element, and can not interact with other sounds. Higher resolutions always (to me) had better 'air' and richer bottom end. Just overall better clarity IMO. but this is comparing 16/44.1 to higher rates. My gear (lol, and most likely my ears) doesn't seem to produce any more clarity from 48 upwards. I am sure there is a difference in the air and bottom with the right gear (and ears ;). Cheers.
It all comes down to one thing, what sample rate is the final master going to be? Any conversion in digital audio is less than ideal and you should record at the same rate you are going to playback.
While this was true in the early ages of digital audio, this is not really the case anymore. Sample rate converters have come a long way and produce distinguishable results now.
Brilliantly explained. Sound is measured and then the wave is reconstructed from it.
I really like your video, I remember when the ADAT went to 18 bit recordings and later higher, we did A / B tests and the higher bit rate sounded better, at least we imagened. Today 24 bit for the headroom and the processing of rooms etc. suite me fine. I think the 96 ... khz options are there for marketing reasons only. Thanks for your well produced knowledge video.
When working in programs like Sound Miner or Soundly, for doing sound design, having samples at higher sample rates work very well when doing more extreme pitch downs where you want kkeep that fidelity. Like a lions roar being pitched down for some thunderous sfx. While this would even more introduce aliasing well into the audible range that may not be as big of a concern as the extreme lack of fidelity when doing the same pitch down to a 48k sample. Also that pitched down aliasing may uncover some interesting and maybe pleasant, or deliberately unpleasant information!
Also love this video so much, you guys are incredible and there is so much valuable information in this video
You guys are incredible and I wish your channel every success. Thanks so much for the effort and quality of this video 🙏🏻
Thanks Kristian, we’re glad you enjoyed it! We’ll keep the content coming!
Hello!
Great video.
Though I agree with you that 24-bit 48kHz is “the way to go”.
I can objectively say that I have a 24-bit 48kHz album that sounds great.
I’ve re-exported the album in 16-bit 44.1kHz and I was able to hear the differences in trebles.
That might be down to the sample rate conversion. Not everything does it particularly well! What software are you using?
@@PresentDayProduction,
I used Sony Sound Forge Pro 13 to import, and re-export the tracks at lower resolution.
As the proud recipient of a Music Physics degree, I love this question. You are completely correct..... for 95% of the video. The ultra-sonics CAN and DO interact with each other to create audible waves while beating against each other. This is where I must put forth that I live in "Live World" (part of why I just got around to watching this video) where 24/96 is kind of standard (most of our DACs to get to CAT5 are for 24/96 only) but we do use that ability to "hear" the ultrasonics from the mics, let them bounce off each other and create audible tones in the console, and bounce them out to speakers that only give us up to 22,000 Hz.
The only thing I have a problem with is that you have proved for recorded sound being mastered 24/48 makes the most sense. I agree with almost everything, except that 96 doesn't need to exist, cause it does. Instruments create these sounds, we need to reinforce the WHOLE sound of the instrument, not just what WE hear from it cause that instrument and another will beat together and create beautiful harmonics that get lost with 44.1 or 48. (Again this is all in Live World).
Thanks for your comment Ben. Firstly, I hope the current situation we’re in ends soon and gets you off UA-cam and back into live world! You guys and girls have had it tough this year, and we feel your pain.
I do occasionally master at 96 if the client requests it, and some of my clients swear that 96 sounds better, and that may well be because of the points you raise above. And that’s fine, if it sounds better it is better. The main reason most of my work in mastering is at 44.1 or 48 is because I’d rather handle the conversion from anything recorded and mixed at higher sample rates here, so as I know what it’s going to sound like, and don’t leave it to the streaming services!
I’ve done a fair bit of work in low-mid level live sound but you’re clearly in high-end world. Can you explain to me the reason for wanting to hear the ultrasonics from the mics and create audible tones in the console please? That’s a new concept to me, and I find it fascinating!
Thanks again for your excellent comment, and I hope things can return to at least a little bit of normality again for you soon!
Cheers, Ben!
Mark
i too am interested in learning more on this concept 🤔
I'm a saxophone player. Thinking about how manipulating the overtone series can affect the timbre of the instrument makes me agree that there is inaudible information there but it might be information that is interacting with the audible range's harmonics. Of course that isn't taking into account the science of digital audio. It's just the way that I've pictured a higher resolution giving me a clearer picture.
@@PresentDayProduction Ahoy there Mark! First, thank you for the well-wishes -- being based out of the good ol' U.S. of A. I have a few more decades before this ends, at least the way things are going.
I think you are right. I think for Mastering purposes especially, the difference between 44.1, 48, and 96 are.... negligible at best. I think that pretty much for everything you are doing, the difference is negligible. I would have been interested to see you record the same thing at the different frequencies to see if more changed or not.
Frankly, I see no justification for above 96k. If we give some extra on either side for sensations but not "hearing" sounds we can say that humans are good for 10-22k Hz (and there is some interesting evidence that humans can sense frequencies higher than they can hear). Even so 48k is more than enough for this range of human sensing. So why 96k?
Y'all correctly pointed out, mainly marketing.
In Live World we also argue that having those ultrasonics in our console gives us the options. We have the ability, if your guitar makes a funky harmonic at 27k and another funky harmonic at 31k, with 96k I can hear both and I can hear the beating of them together at 4k which is audible. If I had only 48k that beating might be lost in the DAC.
The other argument thrown around in Live World (which I am not a hundred percent convinced of) is that IF shit goes wrong for a second while computing in the console (lag, or error in memory or what have you) (I have had consoles entirely seize up on me mid show), then having those extra samples the compute can still recreate the waves that are in audible range if a sample or two here or there are dropped.
Again, for what you are doing, I completely agree -- save the processing power. A lot of consoles will downsample for FX making the whole argument moot anyhow (not all, but many). I think that the only reason I would move from 24/96 is if someone comes out with a good peer-reviewed study showing that humans can sense more than the current max known frequency which (according to Physics and Music by White) is about 23k.
That my reasoning. I greatly enjoy talking about this stuff though, and I would love a video about how you guys in recording world deal with this whole stay at home business. How has recording changed, was it already changed, does it piss you off as much as it does me to not have live shows? I would be interested.
Thanks,
Pyke
Thank you Ben Pike; I was thinking of this kind of thing because I have noticed the difference in (air) if you will and this is what had me chasing high bit rates is cheap consumer recorders. Maybe I was hearing things but it’s true about how lower harmonics are affected.
Interesting to note that people are often looking for flavor when drinking alcohol for instance.
The same can be said for harmonic distortion and compression in a playback system. Maybe I do mind the information phase warping as much as the Perceived loss of dynamics which likely could be attributed to a lower noise floor or less perception of limiting on the high frequencies; all I know is that I have heard examples where 96k 24bit 44was far superior to CD @ 44.1 16 bit So at the end of the day it’s bit rate we are looking for.
This reminds me of marking hype over watts in audio instead of efficiency and bandwidth.
Cool stuff, great channel; great comment. 😉
First off, good video on a basic level! However, while your example of higher a sample rate having no effect on a sine wave is correct, that does not apply to a signal containing complex harmonics such as a piano or an orchestra. You're also spot on (partially) about the marketing claims of AtoD converters. While they do deliver adequate sample rates (which as you pointed out may not result in a good product), very few deliver the specified bit depth. Many units advertised as 24 bit deliver only 20. I agree (as many engineer colleagues) that SR's higher than 96 are erroneous when compared to the sonic capabilities of DSD, so 96K for video and 88.2 for CD have been by my default SR's by elimination. Keep up the good work and greetings from across the pond!
Five thumbs up from me. I live in London. Can I be your tea boy? I'm not going to tell you my hearing is shot from too many Zappa plays Zappa gigs because who needs hearing when you have a DAW? And thank you Alan Parsons for my tinnitus.
I genuinely can't hear a thing above 16.5k haha
TL;DR capturing in high sample rates is good because high-res A-D converters can use gentler anti-aliasing filters.
In many discussions of sample rates, advocates of “high” resolutions (including 96k and 192k) will swear that the capture of higher frequencies really do make a difference, and that >20kHz sound can be heard… however, I am not aware of any double-blind studies that have been able to demonstrate this. Having said that, the capture of these high frequencies is not necessarily the chief benefit that one gets from high-res digital audio.
As you mentioned, digital audio recorders must have low-pass anti-aliasing filters in order to remove frequencies above Nyquist, so that the aliasing artifacts are not captured/heard. In order to accomplish this, and assuming a 20k corner frequency of the anti-alias filter, the filter itself must be very, very steep. It would need to essentially mute any audio above 22.05k or 24k (44.1k and 48k sample rates, respectively). A filter that has 72dB per **octave** is considered steep, and those two nyquist limits are a nowhere near an octave away from 20k. Steep anti-alias filter = more phase distortion near the corner frequency.
However, in **properly-designed** high-resolution audio converters, since nyquist is so much higher, the anti-alias filter does not need to be nearly as steep. It only needs to “cut off” frequencies above 48k or 96k (for 96k and 192k A-D conversion, respectively). Less-steep filter means less phase distortion.
In practice, I’ve found that audio captured in higher sample rates tends to have much more precise stereo imaging, and transients are much clearer. I’ve tested this by recording the output of my record player, simultaneously, to two different interfaces - same model, but one set to 44.1 and the other set to 192. When blind testing the two transfers, I was able to tell which one was the 192kHz with program material I knew well almost every time. I wasn’t able to do it as reliably with records I hadn’t listened to much, but after a couple listen-through’s I could pick it reliably almost 75% of the time.
Excellent thanks for this video!
Super interesting point: recording at 192 introduces some distortion, and that's why some folks are thinking it sounds "like analog".
Whereas, in analog recording, tape and hardware would introduce distortion. But this kind of distortion is different than the digital one.
So, recording at 192 it wouldn't give you an "analog" sound 😎
Really good explanation of the quality issues related to various bit depth and sampling rates ...24 48 is certainly my way forward. 👍🦄🙏🎩
THANK YOU!!! been saying this for years!!! it comes down to your DAW's specs and YOUR EARS!!! 24/ 48 is awesome for me . . .
Endgame level explanation. Well done!
Thanks RecordingKC!
Excellent - I am convinced. Absolutely convinced - Will no longer record at any rate higher than 48K. Thanks. Further to recording, there is nevertheless a benefit for non linear plugins, and poorly designed eq's with cramping, to process at higher sample rates, and/or ensure that these plugins have been suitable oversampled internally, and if we use eq's that these eq's do not cramp at nyquist.(the upper limits of 1/2 the sample rate).
🤔 20:25 but could the ultrasonics be issuing secret ‘instructions’ 😉 🤦🏻♂️
Damn, you’ve sussed it!! 😱 If people would stop recording at 192 it would be the end of the current pandemic too!
Well, we know dogs can hear them.... and dogs eat poop. So maybe it’s a good thing we aren’t getting those instructions? ;)
Only the cats can hear it. Messages from the home planet.
Why do you think they're always hanging around in synth studios ;-)
Nice use of the video analogs to illustrate the bad assumptions! Kind regards, Daniel
I've never understood how some people get so hung up on high sample rates, and just a general rabbit hole of spec technical details, but later in a conversation will tell you how great sounding and warm their 1960s psychedelic vinyl records were....which at the time they probably were listening to on a Dansette with a penny taped to the arm.
Dr. Who theme made everyone sit up and take notice and who knows what Derbyshire recorded those little plucks and hits with before stretching them and taping them all together. Pink Floyds Dark Side of the Moon sailed into the FM radio charts in 1973 with single speaker transistor radios in full swing. I guess I could sum it up as, "How I learned to stop worrying and love the audio."
I get the point, but you cannot compare directly the two things...digital is a different beast and it really makes a difference if you understand the technical details behind it
One reason you might want to record at higher sample rates is if you would like to detune down an octave or two. This could be for sampling special sounds for your music or for scientific work. This was easily done on tape, but digitally if you detuned down 2 octaves at 44.1 khz sample rate the final sample rate would be around 11khz which would leave everything under 5.5khz undistorted. Lot's of birds and animals make sounds out of our hearing range and we only hear the carrier frequency.
Try it sometime...record the most boring bird sounds (particularly birds that only seem to make one sound)and drop it down and you will start to hear all the information the other birds are hearing....
Hi. I’ve been wondering how you played back files of different sample rates from the same logic project. Am I missing something here?
With GREAT difficulty... in the end we upsampled the 44.1 and 48 to 96
@@PresentDayProduction what do you use for SRC?
Thank you. On my own I did notice aliasing at higher frequencies so I record at 48 kHz and 24 bits. Now I know why.
Magically delicious fun facts found in this video!
My science teacher used to explain the principle of aliasing with a fan and a strobe. Very clear analogy!
Is a higher sample rate / bit depth beneficial for tempo stretching? The shorter length and higher dynamic range could theoretically maintain the smoothness of the audio as it’s stretched, but if the DAW oversamples prior to the stretch, the outcome should be the same-? I might be putting my ignorance on display here 🤪
It depends on what you’re doing - theoretically, yes, and theoretically, no! The best way is to experiment with different sample rates in your own projects and see what works/sounds best to you. I’ve time stretched vocals 10-15% for remixes at 44.1 or 48khz, and there have been artefacts, but not because of the sample rate, because I should have re-recorded the vocal at the new tempo!
Great video. So well produced with so much accurate info. Thanks- you do the recording world a great service with videos like this.
Question- what about recording DSD256 via appropriate mic’s like Sanken Chromatics of woodwind, vocals and strings? I’ve done null and related testing in our well isolated triple walled recording/live rooms here and have found some useful technical and creative advantages, which I assume you are well aware of. Obviously, we are talking about a much more expensive recording chain and construction standards than many studios can justify, but the differences are very interesting and quite different and very useful.
You deserve way more views 👏 👏
Thanks Ricardo 😎
23:00 That back to back comparison is brilliant
This video offers several misleading comparisons between audio and video, as well as measurably incorrect conclusions, in terms of benefits or lack of, when it comes to choosing sample rates and bit rates.
More specifically, here are two specific examples:
When comparing 24bit/44.1khz recording to 24bit/192khz recording with the same exact (very high end) signal chain, when applying noise reduction (RX7) on the standard sample rate recording, despite significant efforts, the algorithm failed to properly remove the noise without harming the signal, while on the high sample rate recording, successfully and flawlessly removing the noise from the unharmed signal was effortless.
Also, when recording audio on set, using 32bit with floating point, provides an important increase in dynamic range, which in turn provides a helpful safety net when levels are unpredictable (as they often are in a location recording on a movie set).
Good info and I will definitely try 48K. I converted to 96K a few years ago and I haven't had anything weird that I can hear but I will try a project at both and compare. Cheers!
Could be placebo, but I hear 96k responding better to elastic audio, pitch correction, and some plugins.
Higher sample rates can help with some time stretching software, and non linear plugins that don’t oversample (or do but don’t have it turned on!)
If a human had only one ear, I agree, recording at 44.1 or 48 kHz would be sufficient. However, human brain is able to resolve sound resolution for directionality, location like distance and sound character differentiation of 7 microseconds. This is why recording audio in stereo at lower sample rate take that data and lowers the resolution by 4. As the amount of that information encoded in a stereo stream recorded at 44.1 kHz is at 20.8 microseconds. It is much harder to precisely pinpoint the location of a given object in a stereo recording. You can also play with phase shift and other aspects to emulate a 360 degree sound scape using two only speakers, and you'll be able to do so if you record at 196 kHz sampling rate as the positional resolution is now at 5.2 microseconds.
All that you said is correct, as a physicist, I agree with your argument, but that argument is only valid if you have a single source of sound and one ear. If you have two sources of sound and two ears, which also compute the location of the sounds encoded within the stereo stream to locate them, the spatial data of what our ears and brain can decipher needs to be contained in the recording. it's not about the pitch, or frequencies we are trying to create. It's about timing, of when sound is produced at the two sources and it arrives at our ears. When you hit the piano key and the two speakers produce the sound, but having one speaker produce that sound at 5 microseconds sooner than the other speaker to help us locate the location of the piano in the sounds stage, you can only get such small time intervals when you record at 192 kHz sampling rate. I hope this helps bit. This is why some people say that 192 kHz sounds crisper or cleaner or clearer. It's not that they can hear better reproductions of the frequency or pitch or better sound. It's because they can resolve the timing better, which in many times is translated as crisper cleaner sound. And if one covers one ear, the difference between 44.1 or 48 or 96 or 192 can no longer be differentiated. Our eyes do something similar when we look at people while they speak. As we get older we hear less, and looking at someone speak, helps our brain to fill in the sounds we may not hear anymore. Spatial positioning has a similar effect on the sound, this is why some claim they can hear better soundstage, or crisper cleaner better sound. Here's good video on this. Cheers, ua-cam.com/video/r_wxRGiBoJg/v-deo.html
that doesn't matter though, it's still downsampled and interpolated for playback. the null test confirms nothing 'survived'. If our playback medium was 96 then you might be able to prove this in lab conditions.
I think you should have a look at binaural recording istead. Typical recording techniques, does not even attempt to address this issue, it just pretends that arbitrary level mixing of different microphones/recoridng situations and panning, and then playing back through stereo speakers can create a realistic sound-stage. So the reason you don't hear a realistic sound-stage, isn't affected by the sample rate, but by the methods recording and mixing. To address this, I have an idea, of using binaural recording, in a room, but use speakers/resonators/drivers placed in the room, and playing the sounds, at the desired position, so that it will be possible to get the binaural experience, while still using typical recording techniques and sound processing and editing.