For the part at 8:36, as long as the sample rate is at the Nyquist frequency and the signal was properly low-passed to less than half of that, any samples between peaks are automatically reconstructed properly; no need to sample at the peaks. They figured this out decades ago. Thus, a 96kHz sample rate is going to give you pretty much whatever 192kHz would give you, including a gentle low-pass slope, but require only half the storage. As of several years ago, 192kHz was problematic from a sampling standpoint as detailed in a couple of white papers by audio engineer Dan Lavry, which you can find if you look them up. I've not seen anything on that since, so it may still be an issue.

Truly sorry, but this is another example (between a multitude) of misunderstanding the Nyquist-Shannon theorem. When it comes to sampling process, it doesn't matter if the samples coincide or not with the more "relevant" points of the wave. Usually, they not. The tbeorem demonstrates once tbe Nyquist criterion is satisfied, the resulting function is one and only the one possible that corresponds to the original. This function is perfectly reversible, whilst the following process (quantification) is not, independently of how many bits you use to represent every sample. Thus, the only reason to use high sampling rates is no other than, up to date, to build an efficient low-pass filter of 96 dB/octave slope, needed to remove frequencies beyond 20 kHz up to 44,1 kHz with no group phase rotations and other issues inside the audible band is possible, but extremely difficult and expensive. The solution to this problem is using higher sampling rates, not because it will increase the frequency audible detail but because the low-pass filter will have a gentler and less intrusive slope, nothing else. Bit depth is a different matter. Any bit you add, 6 dB in S/N ratio and dynamic headroom you upgrade. Considering the quantification error is equal to + -1/2 the significance of one bit, it's quite obvious that the more bits, the better, assuming this "amplitude" resolution is native, of course.

@3phaseBerlin

5 жыл бұрын

please dont keep on repeating that age old pro cd standard propaganda .. for sure the niquest theorem ist true regarding the band limited reproduction of the band limted signal..but it never stated that this is lossless in relation to the band unlimited signal... Again there is a reason for pro audio standards, and that reasons are real professional sound engineers that wanted it like that..and an industry that likes to build new studios once in a while.. But for sure.. low level cymbals do sound dreadfull on cd standard recordings ... A good reason to update to the double samplerate and 24bit depth, just to get it right in the 10-15 k area..

Likely explanation: Your audio interface works better at 192khz, perhaps due to poor filter in the converter?

@daxarinn

4 жыл бұрын

Boom. Finally someone who understands sampling rate and AD convertion.

Hi, DSP programmer here. No such thing as "sample synchronization," each sample point is its own discrete point in time.

I'm willing to accept that you can hear a difference between 48 kHz and 192 kHz when you're working with your system at those sample rates, but what I'm skeptical of is if it makes any beneficial difference whatsoever when you downsample it to 44.1 kHz which is what most streaming services run at and also what most people's listening devices will be running at. I think you should just do a test where you do one version of a mix in 48 kHz and another version in 192 kHz, and then downsample and render them both to 44.1 kHz 24-bit wav files and do an A/B/X test to see if you really can hear a difference in what is essentially the final product.

do you find that sometimes certain virtual instruments and plugins dont run well at 192? maybe i need to adjust my buffer....

Brother, what kind of dithering are you choosing when its time to export the final audio for release? Whats your full preference on sample rate bit rate dithering etc etc. Thank you!

What you call "sample synchronisation" - or timing, phase of the signal is not a problem in a digital band-limited system (like all systems we have). It's a bit counter-intuitive, but the math determines there's only ONE way of solving the equation, and if we ensure no frequencies outside Nyquist are present in the input - sampling will perfectly store and restore the analogue signal, complete with exact phase and frequency response. Aliasing might occur if the anti-aliasing filter is not removing all high frequency content though, and to some degree it's present in all systems (though in modern systems it's so minute we can't hear it).

@mikewild3550

5 жыл бұрын

Jitter in clock sync causes lots of problems.!

@robertsyrett1992

5 жыл бұрын

Additionally, fidelity on this level only exists when directly comparing two signals in an acoustically treated room coming out of monitor speakers. Even then it is a nuanced distinction and after a few minutes our brain adjusts and the only difference is how quickly you are filling up your hard drive. In the real world 44.1k is even sufficient when anti-aliasing filters are present. I almost feel like this video is @whiteseastudio saying, "Guys the weirdest thing happened, I thought supersampling didn't make a difference but I think I could hear something different. Now I'm gonna draw sine waves on a dry erase board and tell you what this experience made me think about." It's stream of consciousness rant first and scientifically accurate info coming in a far second.

@MDHaughton

5 жыл бұрын

Hi, Unfa. Nice to see you here. Been running Ardour for a while now. ;)

You mention MOTU. Which MOTU interface are you using, do you like it and recommend it (in a setup like yours)?

Really appreciate this, and all your videos!

Hey :) If you want to do your full 48k blind test against 192 without switching your system rates, simply convert the 48k sample to 192 with an algorithm of 4 samples to 1 (smoothing or conversion filtering settings OFF etc.. effectively pitch it down 4 to 1 ! ). Then you have two 192 files, the 192 [aka original 192 tracking and mix output] and the 192 [pseudo 48 mix output - sounds EXACTLY like the 48 and effectively is 48 even though the clock says 192]. The first has many IO channels etc going in and out and processing and bussing and summing at 192.. the other has all been done at 48k but simply every sample is made 4 times longer (before AB playback at 192 which is necessarily clocked 4 times faster). Just get a friend to flip a coin and remame the files A and B and write down in secret which is A and B. 48khz +_+_+_+_ . pseudo 48khz ++++____++++____++++____++++____ . pseudo 48khz @ 192khz playback rate: +_+_+_+_ Cheers!

At about @7:30 you start to veer off into psudeo science and blatant misunderstanding of sampling theory. I hope amateurs don’t want this video and think you know what you’re talking about!

Do you use 192 for Vst instruments too? And if so what are the positives, same as mixing?

What Louis Jans said, please! Could you do an A/B test with something processed at 192kHz and downsampled to 44.1 compared to something captured and processed at 44.1 throughout? This is what would really matter because the "use the sample rate of your final output target" crowd's argument is that more quality is lost in the conversion than is saved by working at the higher sampling rates.

10:02 As for the lower latencies - from my experience it's not that simple. Sure - you have lower latencies with the same DSP buffer size, but you also need to process way more data every second to play audio without dropouts - so I would say that if you can have lower latencies using 192 kHz , you can as well just make your buffer smaller for 48 kHz and have the same latency (or even lower) as your hardware proven it can handle more data per second. Higher sampling rates always mean more work for your hardware, as there's more data to process.

@boothbuster

5 жыл бұрын

Thank you!!! I just replied the same to someone else. You are the first sane comment I’ve seen!

I’m wondering about sample rates for the first time as well. The biggest thing I’m having trouble with is when recording to 16 track tape, when I dump everything to digital, (using the same motu interface) I really feel like there’s a difference in the depth. This also happens when I come 24 channels digital out, through a console, then back in, I feel like the stereo image fades a little compared to what I’m monitoring through the board. This could be something else, but I’m ready to do some trials.

Because sine waves can have any starting phase per sample, phase accuracy is in the _combination of rate and depth_ , that at 48KHz at 24 bits the phase accuracy is already in the trillionths of a second - indistinguishable from 192KHz, in terms human hearing. Accurate filtering/resampling does not affect phase response beyond a hint of distortion in the stop band (the part that's being rejected) and transform rounding errors, which are just as tiny as the phase per sample quantization error.

Nice video, that blind test would definitely be something very interesting to listen.

Hi i wonder if higer sample rates only affect the high frekvence area, will a frekvens below 10k sound the same on 44khz and 192khz ? Also let's say i have audio signal in 1000 or 100 hz will does frekvens be affected by the filter on the converter?

nice, been waiting for this

Hello! I have been enjoying a lot of your videos since I found you a couple days ago. I think your perspective is very good, and I feel like I could learn a lot from you. I'd like to address your question about how a tone at exactly half the sampling rate could be a problem. (I've been experimenting with and studying digital audio informally since about 1993. I challenge everything I hear, and every thought that I have, and I feel like I have a pretty thorough understanding.) While in that exact scenario, you could be right about the phase problem, that exact scenario really never happens. That exact halfway point is where the therom breaks down, and I'm pretty sure that it would be a non issue at frequencies even VERY slightly below that point. If it is an issue at all. It's an interesting thought, but it would never be a problem in practical use. My inclination about your experience at the beginning is that your DAC, when changed to 192Khz, may gain some advantages other than just the sample rate. The wider filters you mentioned can occur in many places along the system when you are running everything at 192, even perhaps effects I imagine! I'm not saying that I don't value 192 Khz, or higher. I love being able to say I have a direct sampled 192Khz copy of the master of 'Hotel California', for example, and even a lot of (ew) DSD files. Just because I love knowing that I have the best case scenario. However, I would be completely lost if asked to identify any differences in sound. If I had a studio, (only a dream, since I have poor health), I would do exactly the same as you! Still, a lot of internal anxiety disappeared when I came to the complete realization that in the final delivery format for the listener there is no actual discernible difference. This allowed me to enjoy music more, without being concerned about numbers on the screen. For example, I have a small Player/DAC from China that has very high end (to me) components in an extremely simple and economical design, (Zishan Z2), which can play huge sample rates, and even DSD natively. When used as a DAC though, it only runs at 48kHz sample rate. This used to bug me, until I realized that the overall quality of the components probably paid off much more than the sample rates above 48kHz, and all of a sudden, I could enjoy things a lot more without getting hung up on something that has no payoff. Delivery formats and studio processing needs are very different of course. Anyhow, keep it up!!

Hello, Really interesting video, but got some questions. Gonna be a bit long but if anyone wants to speak/debate friendly, would appreciate it! For classic, jazz etc I agree 100% to work with high end resolution, cause most of people that misten to are mostly audiophiles and have room/gear to deal with high quality music. But let's focus on electronic music for a bit. - I can understand the way to work with high sample rate when we are making recordings of our analog hardware or instruments. But what's going on when we create music using samples from samples pack we bought that, let's say 99% of the time, are 44.1kHz. To me that's something I don't get. Is there a value to work with higher sample rate on soundcard and daw while we're working with samples with lower sample rate? For re-sampling and time stretching, will it make this kind of audio editing better and cleaner? - What's the real deal to work with such high sample rate while every mastered music will finish as a 48kHz max for a vinyle release? I mean, just a few people in the world can enjoy high quality music, those are people with high end hifi/monitors in an audiophile studio. In clubs that's over compressed, in car that's radio or CD, at home we mostly listen to music on hifi, not everybody have a decent studio to enjoy all details in music. Isn't it better to work with let's say 48kHz to deal with what people will actually really experience? I can get the fact it's better to ear and work details that nobody can get on everyday's sound systems, but out of the studio, that's an absolute different experience. - Dj's plays wave and shitty one mp3, clubs are, sadly, having compressed sound systems. Most of DJ's don't give a damn about volume and play out loud in the red zone, so distortion to the max, whoohoo. Can working with high end master help getting better sounding once file are converted to 44.1kHz? Signal is going down and so details are, even with dithering. - To me mastering is 80% about feelings and 20% sound control/technic, but when I send my tracks to a mastering guy, I don't care about the fact he can sent me wave 24bit 96kHz, I want my track to sound good, what ever the file format. I won't be able to ear frequencies above 18kHz (and I'm sure it's going down to 17kHz now cause I'm getting older haha), people won't be able to ear that background reverb i put at 3:30 to make the snare wider in club contexts or car listening or home listening. I much more am focused on how he will act with me as an artist, with my music and the way he feel it. I want my music to be listenable on every system that mister nobody can get. But some people can listen those high end resolutions and that's a good way to have some engineers that works this way for artists, so that's all fine. These are things I really am wondering cause I'm really interested by mastering work and how everyone deals with it. Greetings from France.

@willb3698

5 жыл бұрын

good read - but Are you mistaking 'mastering' for 'composition' ?

I would run at 192 khz too but my computer/audio interface doesn't handle it well. Also I can't watch youtube videos, for some reason the audio glitches out if I use anything but 44.1. It most likely has to do with the fact that audio coming from online is set to that sample rate, and I'm not sure if there's some way I can change that on my end.

Is 192 more taxing on the cpu and/or audio interface? Does it affect number of tracks/plugins you can use in the DAW?

OK it sounds better and more accurat with a higher sample rate, but when I deliver the end master mix to for example Spotify or for a CD, then I have to use the normal sample rate, and is there then any difference in the final master. I mean I could record with a higher sample rate, but still has to change to a lower normal sample rate. (Or hav I misunderstood this ? ( Best Regards Björn

blind test 48 and 192. would it work if you just compare mixed (rendered) tracks in those sample rates? or live project processing is the key element of the test?

Super leuk! Ondanks dat ik hier een opleiding in heb gehad heb ik dit nog nooit zo goed en helder uitgelegd gezien. Top. Weer wat geleerd vandaag :)

Riemann Function for measuring surface area. Beautiful!

This video discusses the “alignment” issue at high frequencies: kzread.info/dash/bejne/lX2Fm6uRg7jgpbA.html The basic idea is that D/A conversion solves for a unique waveform that can fit a given set of samples, assuming a band-limited signal. It does not simply connect the dots between individual samples. So in theory, the exact location of a transient or waveform peak in relation to an individual sample does not matter as long as the input signal is band limited. Your observations about fold-back artifacts that appear to be a result of the band-limiting itself are really interesting though. I’d love to learn more about that since the implementation details of a band-limiting filter are usually glossed over in discussions of digital audio.

I’m slightly confused. What was the originally recorded sampling rate of the audio in the other project you opened the next day? Are you saying you were up converting to 192khz when you heard this difference?

Could you talk about 24bit vs 32bit in terms of tracking and mixing a project. I'd be interested to hear your thoughts. I'm generally not a very tech minded AE, but I do get the jargon and basics.

@santishorts

5 жыл бұрын

24 bit is fine for recording as you can keep the noise floor down, 32 bit is floating point, is a mathematical thing which is useful for mixing. There are no A/D D/A converters that work at 32 floating point bits, so you are always either listening to 16 or 24 bits, no exceptions ever. 32 bits makes sense for mixing, to have maximum headroom, but that's it, 32 bit files are just a waste of hard drive space.

Ik vind dit een waardevolle video, maar wat gebeurt er als ik mijn audio in Reaper op 192 render en daarna weer in een video bewerkings programma doe dat werkt tot 48000 kHz en daarna doet KZread er ook nog wat mee. Blijft het dan wel goed klinken?😎👍👊

Wouldn't similar, or worse quality loss happen by putting this Hi-Res audio study into a youtube video? I suppose you'd still be able to hear the result and report for us, but how do WE experience your efforts accurately? Thanks

Hi Wytse! I generally enjoy your videos! Here's some terms for you to look up for the technical explanation, which does lack a bit in this video: Shannon theorem, reconstruction filter. Nonetheless: You're right to trust your ears, even if the reason for 192 kHz sounding better is probably the filters in your converter ;) I do think that 192 kHz sampling rate could be beneficial when digitally converting the sample rate to 44.1 kHz for CDs, though. Even if the best way to do this is still to make use of your reconstruction filters and go one round of DA/AD.

I like your Videos! Well explained!

So I'm going to ask a really dumb question the PreSonus 192 advertises operating at this level does that really mean that they can perform at that quality?

Interesting point- frequencies above nyquest can be 'sampled' (if nor filtered) but this is called aliasing distortion. It will interact with the sample rate and create non harmonic products some of which are lower than the original signal. Hence removing signals above nyquest before sampling is so important to avoid audible by products.

@Audio_Simon

5 жыл бұрын

192khz alleviates the importance of the filters thus reducing aliasing distortion (which is non harmonic and horrible sounding if present). This said modern converters do a great job of filtering even at 44.1khz. So the main advantage of 192khz is the ability to time stretch with minimal digital sounding artefacts.

Does this benefit a recording that has to eventually be submitted for streaming or pressed to a CD? Even if the session is 192, it eventually has to go down to 44.1 or 48, right?

what freq did you scream at when you bashed your finger nail?? lol

@MaksKCS

4 жыл бұрын

This isn't frequency... it's sample rate

192kHz 32 float is fine for movies soundtracks and vinyl but workflow has to be modified accordingly treat each entry on the project and freeze save to alt track than disable the plugins after printing but unless you run an external DSP handling unit and avid accelerator cards in the multicore and plenty of ram you will run out of resources, that being said I usually do 32 float 48kHz for cd projects .

i record at 8khz

@MaksKCS

4 жыл бұрын

That's literally what phone call audio is sampled at.

@mthomas1091

4 жыл бұрын

Dixie Normous I do that AND sing my songs slower.

This is all a bit dubious. Not that many people actually hear up to 20 kHz. There’s a steep drop-off generally a few kHz before then. I can’t hear much over 17 kHz myself. (Without turning up the gain to possibly damaging levels.) It was hardly different in adolescence when I first played experimented with this. Interestingly, after that drop-off it flattens out. With extremely high sound levels it’s been shown we can detect 26 or 28 kHz. I don’t know what it sounds like or if you feel rather than hear it, but it has nothing to do with music. Another tidbit is that speakers and other components can have difficulty with ultrasonic signals, leading to audible intermodulation at lower frequencies. If that’s happening, preserving those high frequencies makes quality worse, not better. If you simulate such intermodulation, you can reproduce it at lower sample rates, without causing unpredictable results on other audio systems. Your explanation of sampling shows preconceptions more than anything. You cannot have a signal at the Nyquist frequency itself. Lower frequencies don’t have an ambiguity problem. Any decent audio DAC will reconstruct them just fine. DACs actually tend to run at a few hundred kHz internally regardless of input sample rate, using a digital sinc-like filter to reconstruct the peaks you think are not ‘synchronized’. That nicely works around analog filtering limitations after the DAC itself. If you hook up an oscilloscope, you’ll see nice bandlimited sine waves, certainly no triangle or square waves. There was a video demonstrating this very well, that I can’t seem to find anymore. Edit: kzread.info/dash/bejne/lX2Fm6uRg7jgpbA.html and its corresponding article people.xiph.org/~xiphmont/demo/neil-young.html Latency is a function of both sampling rate and buffer size. You can get about the same latency as 192 kHz at 48 kHz by quartering the buffer. I think you can actually decrease latency a little more because having to produce fewer samples makes it easier to catch up after an impending buffer underrun. Generation loss could be a thing. I don’t know how bad that tends to get. If you apply a gentle low-pass filter many times over, I could see it becoming pretty steep near the cutoff, possibly leading to artifacts if you actually have a signal up there. An even gentler filter would then do better. Definitely do blinded ABX tests and repeat many times to get a more accurate average. In the end it doesn’t even really matter. My biggest gripe with music in the past decades is that our recording consciousness has become about filling everything with sound and overproduction. There’s the clipping and dynamic range compression of the loudness war, but also using other means to generally fatten everything up to the point where nothing stands out anymore, and artificial corrections like Melodyne. A kind of sound I like seems to have died somewhere in the early 90s, or at least gone from common to rare. This is speaking as someone who was a mostly musically ignorant toddler at the time. Can’t be much of a nostalgia thing. My point is that I’ll take something good at CD quality over a numbers game that can theoretically reproduce something bad a little more accurately.

@SimonBlandford

5 жыл бұрын

OK, but consider this : www.ryanschwabe.com/blog/96k

@Carlo24515

5 жыл бұрын

@@SimonBlandford "Admittedly, all of the plugins are character style processors that add harmonics to the signal." So I'm not sure what exactly this guy proved (if anything). He used plugins that introduce harmonics into the audible range by design. He did illustrate that they behave differently at different sample rates, but that doesn't really make an argument for either, so if anything it's probably preferable to use plugins at 44.1khz because that's most likely how they're designed to be used and (the good ones anyways) are probably up sampling internally to prevent aliasing when it's not what is intended. Working at any unconventional sample rate will just result in less predictable behavior in a lot of ways.

what do you think about the new i9 processor by intel for music production. The most powerfull one costs 1900euros, is it worth to buy it for a completly new pc

@R3BBiT

6 жыл бұрын

Wouldn't buy it solely for music production, but if you have money, go for it! I have an i7 6700k and that one is also completely overkill, but you want a processor with many cores.

@MrSkyTown

6 жыл бұрын

I would get a ryzen

Interessant! Ik heb nog een vraag daarover trouwens... als ik kicks maak en ze zijn klaar dan exporteer ik het als wav en op 96khz.. (waarom weet ik ook niet maargoed) en als ik die kicks nu in cubase laad voor een nieuw project dan speelt de kick zich heel sloom af(tis niet meer herkenbaar) hoe kan ik zoiets fixen? Het project waar de kick in gemaakt is is niet meer terug te halen trouwens🙈 Thanks!

@Whiteseastudio

6 жыл бұрын

Cubase in 96kHz zetten, hij speelt hem namelijk nu op 48kHz af, dat is de helft, en dus de helft van de snelheid...

@Thundermasterad

6 жыл бұрын

White Sea Studio ja dacht al zoiets... maar is dit op te lossen? Heb nu alleen de WAV file op 96khz

If your motu has an adat output and you use 8 outputs, it is max 48kHz at default. You could use that for 48kHz A-B .

I have always heard that down converting from 88.2 to 44.1 makes better sense than 48k because is is exactly half and therefor makes for better sound? It’s probably insignificant but it seems to make sense.

pls include also the file size (impact) of finish products as it can (maybe) easily fill ones hard drive, for hobbyists consideration only, of course if you're in a business it won't matter.

What is the point to mix it in 192 khz when after that you have to convert it into 44.1 khz so it can be playable for others . Then you will see that your mix will sound different from what you have been mixed . Correct ?

@troelsknudsen253

5 жыл бұрын

i wish it was true, i would be able to run more plugins and save hd space :D but a lot of plugins do sound pretty different at higher sampling rates. reverb plugins seem to be the most noticeable. they're much smoother in the high end and that seems to correspond quite well with the theory. resampling to 44.1 can create artefacts but you'll still be able to hear the difference in processing across the mix and production (plugin synths can also sound quite different at higher sample rates).

Would be interesting to check with and without plugins. Filters behave better at 192. 5hz highpass filters are nice 😁

You did the two things that I first noticed: better transients and stereo imaging which led to better separation of instruments. A part of it is due to the remaster I suspect and not simply the quality of hi res

44.1 or 48 ..24 bit really is all you need,the advantages with higher sample rates are the EQ plugins which can go up to 384khz,this happens internally usually by upsampling regardless of what sample rate you work at,I think this is where the confusion lies.

Only benefit would be lower latency. The nyquist can be set higher but most converters will only have one or two filters one digital and and analog with fixed frequency tops 20kHz or slightly lower to optimize smooth filter curve. So audible there will be no frequencies recorded above the 24kHz with higher sample rates. It already filtered making sure no aliasing mirroring distortion will occur. Btw nice mirror/aliasing comparison is when you look at car wheel spinning accelerating, your eye can sample up to a freq and the wheel will start rotating backward(mirroring), that point should be Nyquist filtered :P

I'm also pretty interested in the comparison. I've waited to get a project even at 96KHz or higher to mix. As I'm in the box it would be easy to convert all project files to 48KHz and just bounce the mix again and convert the original 96KHz bounced mix also to 48KHz. I'm pretty sure that the results are not identical. But someone needs to deliver a project with 96KHz or beyond to me first...

I discovered this myself like 5mins ago, I had to look it up if there's any others talking about it. I heard exactly what you heard. Note, I listened to my system through a symphony I/O for the first time. It's like seeing a really sharp image for the first time, mind boggling

Thanks for the vid.. What Motu audio interface are you using? I believe my two 8pre's aren't able to go to 192kHz but I'll check.

@Whiteseastudio

6 жыл бұрын

I’m using their AVB range

@georgemickeldotcom

6 жыл бұрын

As an experiment.. I'm going to push my sample rate to the highest level and listen. I've never gone higher than 44.1

And then there's the problem with Plugins that don't run at 192.. How do you work around this issue?

@KiLLACAiN

4 жыл бұрын

Warthog Studios i think they work in mix... think it be a bitch to use in recording

This video was in the dictionary under "placebo effect." Very helpful!

thanks for making the vid.. i will give 192 a try 4 shoooo

I would suggest that a higher samplerate results in a more detailed transient reproduction because it is four times the space.

I support your thinking and ears 100%

The interesting A B testing is how the different mixes will sound after bouncing and converting to 44.1 16bit. That is what really matters.

@edwardfanboy

5 жыл бұрын

I bet the mix at 192 kHz would still sound better after downsampling to 44.1 kHz, because doing the processing at a high sample rate would reduce aliasing noise / folded-back harmonics from plugins with no built-in oversampling.

@DJKroehnadus

5 жыл бұрын

No! Ofcourse it is also interesting how it would sound in 44,1 kHz, but If there is a difference, it would also change the decisions while mixing.

I was using 96khz for all my professional projects for a whiule but went back to 48khz due to the CPU hogness of 96khz plus I bought the Behringer ADA8200 which doesn't properly go above 48khz (the problem with ADAT is its limited bandwidth to 8 channels at 48 and 4 channels at 96 and so on). Most plugins I use can oversample on mixdown which is a compromise between the two. With a better PC and a better interface I may go back to 96k though, the most annoying thing is using samples that are 44.1k or 48k because then they need to be upsampled in real time.

The reality is that musicians can't hear the difference but are afraid that someone else can. At this point in history it is not a big expense to record at 24/96 as most platforms support it. Manufacturers need new snake oil to justify their new products and subscriptions. What a manufacturer can't sell you is musicality and originality. So even the most sophisticate studio will generally produce sub-standard crap even if it has gold encrusted gear. Things that you can actually hear included good pre-amps, good microphones, the sound of a good room, a nicely setup instrument and soulful musicians.

I discovered something similar a few days ago just after updating the driver of my Focusrite Scarlett 18i20. But what happened seems to be a mystery to me. I played back a CD from my computers CD-Drive and the samplerate of the Focusrite was set by default at 48khz/24bit. The driver does not change the samplerate, when using a Windows-Player for playback. (if I use the DAW it does). I don't know, how to describe - the sound was more roomy and more detailed. I switched manually to 44.1kHz/24bit and the sound became more flat and less detailed. But how could this be? The original signal was a 44.1kHz/16bit Redbook-standard file, sent via USB into the Focusrite, converted to 48kHz/24bit and then converted to analog to feed my monitors. What is goin' on there? Mysterious . . .

Im using Mixcraft 8 running at Win10, and Realtek. At the option i have an option to turn on ASIO (Realtek ASIO). From there i noticed if my latency is best at 44.1 and 192, which is 11ms. Other sampling rates just have worse latency, 48, 96, 128... Although i can say it is pretty insignificant. 1-5ms slower. Why, is because buffer size automatically change, scaled with sampling rate. The box displayed the buffer size had greyed out. Some other systems could manually input the buffer size i heard, but not at my case. I wonder if this is related to the lower latency you've said? Working with 192 was...CPU intensive however. I just using 7 tracks and 2-3 Plugins each. It showed 55-72% CPU usage. I know i just using i3-8100 but LOL. Im not trying to say which one better or wrong with 192, im just following.

@okaravan

5 жыл бұрын

Does this Realtek ASIO driver work with any Realtek codec? Where have you found it?

In the realm of video, more pixels only add sharpness and details to the imagery. Yes the viewer can't put their fingers on why they prefer 4k over full HD, But in my own experience I thought I was deluded to grow an affection for 96khz because of that sense of depth, But my pc doesn't seem to like it much! Amazing content as always!

Bonjour If you do your mixing at 192, what happen to the track when you put it at 44 or 48 to deliver the mastering Does the track take any benefit of working at 192? If you hear things that you will not hear after mastering do these things can hinder from deliver a good mix? Merci Very interesting video I am waiting for a A/B test with the same song in different formats

@gherbent

5 жыл бұрын

Some people have a better setup than you. you watch old movies in 4k now because they were filmed on quality film.

But you could do a real life test by mixing in both sampling rates, convert the mixes to MP3 and/or 44.1 and then A/B compare those. Right?

@stonail665

6 жыл бұрын

Louis Jans true

@Nullllus

6 жыл бұрын

Done that. Higher sample rates sounded better.

@TheHirade

5 жыл бұрын

@@Nullllus done it too, and nobody could hear any difference

@feggyo

5 жыл бұрын

Nullll1111 placebo

@the3mu606

5 жыл бұрын

MP3 is a lossy audio format so you can't use it for these kind of critical listening tests. Most audio engineers and people who have experience with critical listening can reliably tell the difference between an MP3 and a WAV when they are played side by side, there is a website you can do this on if you are curious. mp3ornot.com/ Most people don't notice or care, but some people (like mastering engineers who have really well trained ears) can hear the difference %100 of the time in blind tests. Mp3's have weird sound artifacts in the high end that I have heard described as "ultrasonic birdies" (which is pretty funny). In order to compress the file size the MP3 conversion tries to remove everything possible from the file. In doing this it removes stuff that is hard for the human ear to hear and pick up on. A big part of that is the super high frequencies. This is exactly the kind of stuff that someone hoping to hear a difference between 48000 and 192000 would need to hear (the very high frequencies). The MP3 conversion is just throwing any difference you would hear away.

Can I up sample an electronic production done in 44k to 192k??

Would be very interested in your A-B Blind test. You could consider Making the 48khz and 192 Versions and then Upsample the 48khz to 192 and then playback both at 192. If the 48 is really worse you will hear it since theoretically 192 can reproduce 48khz files "for breakfast" . Its not the perfect solution but it is one. Btw the Video comparisson was not that good. Acoustically 48khz is pretty much like 4k video. Then there is 6k 8k etc. Question is , what is the benefit. So if one day we have 20k resolution i fail to see the point and its just there because its possible. In different words , does your studio and production really sound better because you jumped from 48 to 192 or because you got better at what you do over the years? I believe its more likely the second option

I think its great if your system can cope. Like having images in 16/32bit not 8 bit as long as your sources are in that quality (ie. any loops or samples or lower quality inputs might not work) my compositions are not so perfect it matters but pros might benefit recording from scratch. Songwriting is more important, but record the best quality you can in the time you live in.

I'd say just take the originals sample rate. Upsampling won't bring anything except maybe better harmonic translation of plugins that produce frequencies up that high (I don't think many or any do). But interesting content, I'd love if you uploaded the A/B comparison files at some point for us to listen truefully as well!

Amazing video as ever. Just wanted to ask one thing. We audio producers work for consumers and consumers consume audio differently everywhere. How does working on an absolutely high standard audio translate to when the product is delivered on a CD or a mp3 or many sound formats? Does it not get downgraded to 44.100 ultimately? Even on the most popular platform KZread, what is the highest standard that we get to hear? Basically how does a highest quality audio produced gets translated to the end consumer? Thanks n Regards.

@RealHomeRecording2

4 жыл бұрын

Thanks to Amazon music HD setting the trend towards high sample rate music streaming, the questions you have are less of a concern now in October 2019 moving forward! But to answer your question most music streaming services will downgrade to 44.1 kHz. It is my belief that producing music at higher sample rates is beneficial but for me 16-bit 48 kilohertz is all that is needed for consumers.

I absolutely agree, same experience for me!

everyone busting his ass over incorrect analysis when there are still companies out there putting "hd audio" on things these days smh

The only Problem you Facing in digital Audio is the Filter Design of the ad / da Converter. If it Supports only 48 kHz regardless the Material comes with ( But more Resolution is always good), will Sound better instead using 44 kHz because the Filter which is Used has more shit going than at the max. Resolution capable. For example , if youre converter is able to Play 192khz, Listen to a 44 kHz Track , than resample it and Play it again at 192khz. The only difference is the Filter Design Used at 44 and 192 kHz. Every Converter Sound the best at there specific max. Resolution. That is why there is Music out there recorded in fucking 16 Bit But Great Converter at there Time. And for the engineers, you can Mesure that. Take a Square Wave at 44.1 and 192 on a 192 Converter and Look it up on a osciloscope. You will See a difference at the angle of the Square Lines horizontal. At 192 should be strait, at 44.1 for example you will See an angle. Thats the Point wenn the Filters doing shit.

The whole "more sample rate" thing is kinda messy: theorically a bigger sample rate means you don't need oversampling/anti-aliasing to prevent the signal from sounding muddy/harsh, because of the aliasing, but depending on the processing done on the track (plugins or hardward back into the DAW), higher sample rates can make other distortions, mostly if your converters don't behave well with those high sample rates. I suggest you ask Fabien from Tokyo Dawn Labs (maybe in their site or the KVR forum) about this whole thing, also maybe to the guys from Fab Filter, both companies do very well in the anti-aliasing matter.

Since you're running a studio, you're right that there's no downside to using the highest quality. Higher quality can always be down converted, but music recorded at a lower quality can't ever be increased in quality later.

You experienced the early stages of snake oil ;) Add 30 years to your age and imagine where this can lead :)

Now I see that 192kHz provides same noticeable advantages, especially when you have chains of filters and when you mix stereo signal. And that is exactly what any DAW does! The advantages are bigger at higher frequencies. They sould be also hearable at mid-to-high frequencies (guess 1-5kHz), where the sensitivity of our ears is quite good.

I use 96k since plugins/IRs that I use won't go higher in many cases. I do like listening to it better than 44.1k and since I have a powerful enough rig it does not impact my workflow when tracking.

It gets quantized via pulse code modulation respective to bit rate. So the higher the bit rate, the more it gets quantized or as you’d say synced in place. I agree with higher sample rates because of true human auditory survival practices.

You can't take samples of EXACTLY 24 kHz sinewave at 48 kHz sample frequency. Digital audio doesn't "sound" like what it looks like on paper. Visualizing digital audio can be misleading because some aspects of digital audio are a bit counter-intuitive. For example, a 15 kHz sinewave sampled at 44.1 kHz looks very messy and distorted, but the original 15 kHz sinewave can be completely reproduced from these samples with some quantization or dither noise dictated by the bit depth and applied dither method. Band-limitation is the quarantee that messy looking samples become original bandlimited signals. So, don't completely trust your eyes or intuition when dealing with digital audio. Temporal resolution of digital audio isn't dictated only by sample rate, but also by bit depth. This is one of the counter-intuitive aspects of digital audio. The temporal resolution for 44.1 kHz sample rate is not 1/44100 s = 23 µs, but many orders of magnitude higher even at 16 bits, way beyond what our hearing requires. The main reason why different sample rates sound different is that DACs operate differently at different sample rates. If your DAC is optimized for 96 kHz, it will propably sound best at that sample rate. Reconstruction filters have some effect on the perceived stereo image. Is it a standard linear-phase filter or something else? A 44.1 kHz project upsampled to 192 kHz using good sinc sample interpolation should sound identical to a 192 kHz project unless there's infrasonic junk in the 192 kHz project causing audible distortion in the playback gear. The correct sample rates in studios are 44.1 kHz (music) and 48 kHz (video) unless the client wants something higher.

@fluctura

5 жыл бұрын

Thank you I was about to post the same. For anyone seeking more depth of knowledge: Read the first (free) chapter of the book "Designing Audio Effect Plugins in C++" (kindle)

@GoatPepper

5 жыл бұрын

Its a fickle concept to understand. Ive read about this and still have a hard time conceptualizing it visually to understand it better. Now I learn that a sample rates khz isn't measured in fractions of time. I would assume this is true for standard PC processors as well. What I gathered is it takes two computed points to create a wave cycle, and 42.1khz is actually 21.05 Khz. Also before reading this, I thought floating bit points would fix the problem he is talking about....I need to find a good source of info about this, even a few youtube videos about this apearently went over my head.

@RobLocksley

5 жыл бұрын

For now I've given up on finding a complete 'take' on higher sample rates. Depending on the Analog side it can sound better or worse with 96kHz, the electronics affects the sound we actually can hear. Unnecessary high energi in high frequency's 'added' (I'm looking at you SACD) puts unnecessary strain on the electronics leading to affect sound - is there a likewise culprit in multitrack mixing? As mentioned, oversampling is used in many plug-ins to avoid aliasing whilst processing the sound, are we getting that benefit with multitracking in 96kHz? Or are there other benefits to sound processing in out DAW when using higher frequencies that are part of computing the sound in a mix-environment? What about Clocking, does 96 need a better clock not to sound worse than 48, or does it sound better? Bottom line is that the combination of analog and digital electronics will sound and affect sound differently depending on your 'gear'! What happens in the higher regions can affect sound that you actually can hear, sometimes it might not be for the best. In a mixing environment I have not really read any analysis that take into account how a higher sample rate works with plug-ins, mixing and all that jazz. I really should do a more thorough comparison on my system, I'm running at 96kHz basically to feel future proofed if I should return to some projects on a more advanced setup, for some Hardware/software it is worse for some it is better. Since I hear up to 15k I am always 'looking' at the high Fs to not have too much energi in that area, for me it is thinking of others and thinking ahead, what will it sound like for someone with better ears or in the future. And maybe, just maybe, we will get cybernetic ears and hear up to 48kHz in the future and go "Blimey, filter that HF out, now!"

If the nyquist theory didn't work, we wouldn't have DSP at all, simple as that. The "sample alignment" point is not possible because any difference in the phase of the stereo signal that would be cause by "misalignment" would happen in frequencies that are filtered out/outside the range of human hearing. It's really confusing for those without a strong mathematics background (I do not have one personally, and I found it very confusing at first), because it's really easy to make the kind of mistakes you make here. I remember someone saying something along the lines of "Ok, but how do you make a square wave over 15khz, the first harmonic would be outside the sample rate" and being confused until I remembered that any harmonic that could not be expressed in a sample rate such as 44ks would be a sine wave outside the range of human hearing and therefore inaudible. When people say "the theory says this, but in reality it's this" they seem to imagine that theory is wrong or just an approximation of the truth, but that is not really what is happening. There may be circumstances where having a higher sample rate will produce a subjectively better result, but it's not actually due to the higher sample rate, in every case where there is a difference in a blind test, there are other factors that can account for it. It may also be that the higher sample rate is degrading your sound in some way (worse converters, etc.) and those issues are masking other issues in the sound, and this sounds "better" to your ears. One some level, doing whatever sounds best to you is fine, but the sample rate snake oil is fairly irritating because it takes up way too much space, and makes it impossible to collaborate if you don't have an interface that operates at ultra high sample rates. I personally think it's better to find the real thing that's affecting your sound (I personally love the sound of really crappy D/A converters!), rather than wasting data/money on snake oil.

For White Sea Studio, I'm glad to know that this experience put you into the question and I will tell you that this case is even more complicated than you taught, I will point you where you are not rite, first not only frequencies above 1/2 sample rate are affected but also frequencies close to it (on both sides lower and higher) and it will start even at F 1/4 (12KHz) or less and the error will increase gradually with frequency, people affirming the opposite just have no idea of maths and have no adequate equipment to reveal this. Second thing, ideally we should use a very steep filter to filter the artifacts but where graphically it may look well the reality is not that good, in fact, the maths in digital world do not work so well away from our eyes and is better always to use a less steep filter. If you will study maths behind all this processing you will discover and understand many things. I agree music files should have at least 20Khzx4=80KHz sample rate and 20-24 bits definition, not 16bits =1/65536=96db ? A CD format is considered de facto to have a 60db dynamic range in which music producers are forced to compress the sound. Martin Logan's electrostatic speakers can go up to 100KHz and the sound makes difference even when some people affirm that everything higher 20KHz is useless. The sinusoidal signal is not the real world. The science and real experience are not contradictory in the end, we just overlook some thiny things.

8:30 If you add a third sample, you'll see that the signal is already fully retrievable. A higher sample rate wouldn't change that.

As I also go outside the box... Guess I need also to move away from 48Khz to avoid signal bad interpetation from a/d conversion Never thought of that like this, for me it was useless! Thanks to have change my mind!

One thing you may want to consider bro... If your final delivery format is CD, maybe 176.4KHz might be better for you, that way you have integer division mathematics to render the output CD files (176.4 -> 44.1KHz which is divide by 4) rather than smearing from 192KHz -> 44.1KHz (which is divide by 4.353741496598639...) If you're truly on the quest for purity. I would suspect that the Clock Jitter length amount being reduced by 4x would be the reason you and I have both 'heard' the difference between recordings done at 48KHz & 192Khz The stereo image is MASSIVELY affected by jitter... it's INSANE when you A/B different Word clocks driving a A/D-D/A converter.. We did a test with a industry production CD coming through a LYNX Aurora Converter clocking from an APOGEE Rosetta, then flicked the clock source over to a BURL B2 and we all nearly wet our pants in disbelief at the difference! I never would have believed there would be such a difference on those two and just on CD playback. Any ribbon tweeter manufacturer is only too happy to tell you about super-harmonics and sub-harmonics, and the benefits of having audio that way surpasses our 'normal' hearing range. Regarding what you experienced at the start of the video, I personally had this happen with a few country music tracks I was engineering. As a quality courtesy I built the session @ 96KHz/32bit, nothing was immediately noticeable, but throughout the whole project I was thinking; "Everything is sounding really silky today!" I tweaked on to it later when I wasn't able to use some older plugins in the session (as they were 48KHz.

I do understand the physics behind why 192khz is better than 48... however I can say I’ve done lots of A/B testing between the two, and I am one of those that can absolutely say I hear the difference. So instead of waxing on about bandlimiting, oversampling, dithering, blah blah - I work in 96 or 192 because it sounds better to me. Period

Now you have to do a video on buffer size

Your source tracks will need to be recorded at 192, so how will you ever do an A/B without using mic splitters into a different Sound card running at 48? Seeing as must recordings will be converted to a lower sample rate, why not record a band using 192, then get them to 're do the song at 48. Then mix them both down to CD format and see if there's a difference?

You are one of the few with an amazing grasp of sound engineering. Love your channel!

I just discovered a very interesting article on this topic that every audio engineer or listener ought to read. It's very technical, but here's a summary of the relevant info: A 192kHz workflow is not just overkill, it's actually damaging to the audio quality. Sounds counter-intuitive, but here's why. It introduces audio that is way outside the spectrum of human hearing, and that audio causes the kind of "bounce back" artifacts throughout the audio spectrum similar to what linear phase EQs exaggerate. This actually makes a lot of sense. So, while it does decrease latency, this is about the only benefit. It's damaging to the audio, takes up a lot of extra space, and has other detrimental effects that are very well explained in the article. So, it seems that 24 bit 96kHz is really how you can give your clients the highest quality product. Those of us who were using 192kHz workflow are falling into the same snake oil pit that the million dollar home theatre geeks are in. people.xiph.org/~xiphmont/demo/neil-young.html And there are numerous studies and lots of technical information sited in the article that show that 16 bit 44.1kHz is ideal for the final output. The 24bit workflow just gives us a lower noise floor and more headroom when working with the audio. More than we need, really. I've got a lot of places to go retract things I've said about 192kHz workflows. Especially about the stair-step sampling missing transients thing, which is demonstrated to be completely inaccurate in this video demonstration. We all need to watch this: xiph.org/video/vid2.shtml

I'm considering pressing my music to vinyl. Because it's an analog format, it makes sense to me to output the files to send to the vinyl cutter at the highest possible sample rate. However, I've seen videos from one of the more famous professional cutting rooms where they use 44.1kHz systems between their analog mastering desk and the vinyl cutter, which makes no sense to me. There are videos if you search KZread for vinyl mastering, I'm sure.

@deadscenedotcom

6 жыл бұрын

20kHz will get filtered off anyway prior to cutting, so all the precious info above that we cannot hear nor objectively mix anyway will be lost.

you realize that "quality" doesnt come only from using certain sample freq.... The converters do the heavy lifting. Filters, Signal Path, AD/DA cirquits, thats where the quality is....

A vid on 192Khz... Hm. I pause the vid and check the comments first. And yes, they don't disappoint. Well, not in the sense of negative comments and nay-sayers. Higher sampling freqs are still controversial. Some don't say nay because an important bobo said so. Really? How to think for yourself? Others come with theories and formulas. In my opinion that all doesn't mean anything. Try it for yourself; I believe that 90% of the naysayers are discussing something they have no experience with. Same went on with digital vs analog battles. In the past, I did an experiment with recordings on different sample rates; I expected that there would be differences but I was very surprised how big the difference was. 44 vs 96 -> big difference, 96 -vs 192 -> not that big. For me the little diff with 192 didn't convince me to use it; I use always 96 as a minimum. That was like 8 years ago, maybe I should do another experiment with my new converters. Try it, if you like it; keep it, if not, pass. Don't be mean because other people think or hear differently. That nuts! Spent your time with something that has some value. On the other hand: just 49 thumbs down; that is surprisingly low for a vid like this. There is hope for humanity still... And now I will listen to the vid :) Aaaand watched it. Totally agree with it: if you hear it, you hear it. I would have skipped the explanation and theory; that just attracts negative comments. Why trying to theoretical proof what you already have experienced?

@PianoScoreVids

5 жыл бұрын

Thats a cool comment!

@nathan43082

5 жыл бұрын

I am most curious to find out if the problem detailed with 192kHz sampling has been resolved since Dan Lavry wrote about it in his White Papers at www.lavryengineering.com/pdfs/lavry-white-paper-the_optimal_sample_rate_for_quality_audio.pdf and lavryengineering.com/pdfs/lavry-sampling-theory.pdf. I use the Lavry AD11 at 24/96kHz and it sounds phenomenal, but have no way to compare to a 192kHz sampling of the same material.

@ahareally

5 жыл бұрын

If you miss a lot of (which is pretty obvious after watching this video) needed theoretical (academic level) information (misinterpretation included) - your mind is widely open for cognitive bias... which "feels real" - despite it isn't.

I am usually trying to get the best possible material as a starting point for conversion for the multiple target audio players. As a principle. I think in a direct comparison 48/192khz you won't spot a difference. Maybe in the highest frequencies. My 46y/o ears degrade, I know I can't hear 15k any more and I doubt, I'd hear a difference. Maybe, as you assume, in the stereo image. But I don't think so.

@boothbuster

5 жыл бұрын

192 adds on two octaves that no one can hear, basically. And eats up way more disk space. :)

I do think you need to compare the same project/take with 2 ways 48kHz and 192kHz Otherwise it doesn't make any sense. Scientifically it has not any value. Get the same mic, same preamp and the out goes to 2 audio board: the first records 44k1 or 48 and the second to 192kHz and you record the same source. Then make a double blind A/B test to listen to the files.

@Whiteseastudio

6 жыл бұрын

That is one of my plans, as stated around the end of the video, it's just not easy to do this practically... But I will figure out a way!

@Not-Only-Reaper-Tutorials

6 жыл бұрын

You need 2 different computers but the same hardware (2 identical audio IFs) and software in terms of recording and playback. Otherwise HW can significantly color the sound and falsify the result. At the same time, DAW should be exclusively working at the sample frequency of the file. Hence 2 different computers with each one the same DAW with same settings except for the Sampling rate.

@Not-Only-Reaper-Tutorials

5 жыл бұрын

at AES they performed several double blind tests about. No test put in any evidence any audible difference between a 44k1 or 48k and 96k or 192k. However if ever you have a hearing system like a bat ... than you can ... assuming to have enough energetic content after 13-14 kHz on ...

Nice explanation I agree to most of it. Well, the Nyquist theorem is working a little more complicated but ist has it flaws too. But there are some points not mentioned: 1) It all highly depends on the equipment one is using. Listen to music from an average iPod, computer or cheap to medium hifi amp/speaker an you can't really hear the difference between 44.1k and higher. Listening to a very good hifi or high end equipment you will hear of course a difference. But to be fair - things will be expensive and anybody has to decide if it's worth that. 2) everyone is talking about the highest frequencies human ears are able to recognize. As if music is an isolated sine wave! In fact one can hear the difference between a 5k sine wave and a eg. 5k square wave easily. A square wave is a composition of a base sine wave (5k) plus 3x, 5x, 7x etc. harmonic waves. At 5k this means 15k, 25k, 35k harmonics and so on. The difference one can hear (excellent equipment assumed) is the timing or how fast the wave is rising, the slew rate. And this is like one could hear much higher frequencies than 20k even if one cannot hear the isolated frequency itself.

the samplerate limits the minimal possible latency ! the bigest advantage of 192 khz is that you can get a lower latency given a fixed sample buffer size. example: 64word buffer 48khz minmal possible delay adda (1/48000*64*2) =2,6ms the same with 192khz: (1/192000*64*2)=0,66ms with this you can even do vocal rec monitoring in the box. without giving the singer a strange feeling