Why 24 bit audio and anything over 48k is not only worthless, but bad for music.

[chef8489]

I accidentally fell on a video that proves my point from #3544:


Here professional sound engineer proves with experiments that any manipulation / mixing of a digital source at 44.1 kHz is introducing significant audible distortion. You can go to the video for the explanation, I don't want to repeat. He does, and encourages everyone, to always mix strictly at 96 kHz all the way to the final master. Only then, if necessary, the final downsample to 44.1 kHz could be made.

And to that I will add, if the final master is 96 kHz, and the DAC chip requires at least 96 kHz anyway, and we know from his experiments that any sound manipulation on 44.1 kHz audio adds distortion - what's the point to ever downsample the final audio to 44.1 kHz at all? Disks are large, Internet is fast, just stream 96 kHz on Tidal or Qobuz and enjoy HiRes, because it is almost always (I won't claim always) audibly better. It's time for CD format to die, there is no reason for it to exist.

I don't understand why you keep posting in here as you clearly don't understand anything of this. You sure don't understand the video you accidently stumbled upon as it doesn't prove the point you keep stating.

 

[71 dB]

I accidentally fell on a video that proves my point from #3544:


Here professional sound engineer proves with experiments that any manipulation / mixing of a digital source at 44.1 kHz is introducing significant audible distortion. You can go to the video for the explanation, I don't want to repeat. He does, and encourages everyone, to always mix strictly at 96 kHz all the way to the final master. Only then, if necessary, the final downsample to 44.1 kHz could be made.

And to that I will add, if the final master is 96 kHz, and the DAC chip requires at least 96 kHz anyway, and we know from his experiments that any sound manipulation on 44.1 kHz audio adds distortion - what's the point to ever downsample the final audio to 44.1 kHz at all? Disks are large, Internet is fast, just stream 96 kHz on Tidal or Qobuz and enjoy HiRes, because it is almost always (I won't claim always) audibly better. It's time for CD format to die, there is no reason for it to exist.

I spend 25 minutes to watch that video. It does not prove your points. This thread is about consumer formats. 24 bit is overkill in consumer formats. As little as 13 bits would be enough. 96 kHz is overkill too. In studio things are different. 24 bit is beneficial (You can make stunningly good sounding music with just 16 bit, but it is harder and more difficult, so why do it that way when 24 bit is available?). In the video the 96 kHz has been chosen as the workflow and some people do that, but as professional sound engineers such as Gregorio have stated, 44.1 and 48 kHz are the most common project sample rates depending what kind of production it is (music or video).

The point made in the video was using 44.1 kHz project sample rate + oversampling plug-ins can end up causing more CPU load than having a 96 kHz project without the need of oversampling plug-ins. I suppose the nature of the project dictates how this is (any views about this Gregorio?). Anyway, if someone can make the workflow better using 96 kHz because of how their PC handles CPU load caused by plug-ins, then no problem, but we consumers don't need more than 44.1 kHz and 16 bit. The video seems to agree with that!

The anti-alias distortion demonstrated in the video is far from "significant audible!" There is not even sound samples in the video! The aliased harmonics have quite low level. He talks about cumulated distortion due to running the plug-ins many many times whatever that means. I mean, do you decapitate the sound 100 times?

Is 44.1 kHz really wasn't enough, EVERYONE would be using higher samplerates when mixing music. That is not the case, not even close. A lot of sound engineers agree 44.1 kHz is enough and they can't all have bad ears, can they? I mix my music using 44.1 kHz and I have NEVER run into it not being high enough, but maybe my ears are just garbage. Same with my beloved CD collection. I would never replace it to some streaming services with whatever selection of music never knowing when the music is taken away because of some stupid copyright issue.

 

[VNandor]

I don't know at which point showing that a nonlinear plugin with no oversampling will cause aliasing became this:

Here professional sound engineer proves with experiments that any manipulation / mixing of a digital source at 44.1 kHz is introducing significant audible distortion.

I also have no idea at which point "if its sounds great to you and there are no problems, I'm not going to say your mix is bad because you didnt mix it in 96k" became this:

He does, and encourages everyone, to always mix strictly at 96 kHz all the way to the final master. Only then, if necessary, the final downsample to 44.1 kHz could be made.

 

[Ghoostknight]

I accidentally fell on a video that proves my point from #3544:


Here professional sound engineer proves with experiments that any manipulation / mixing of a digital source at 44.1 kHz is introducing significant audible distortion. You can go to the video for the explanation, I don't want to repeat. He does, and encourages everyone, to always mix strictly at 96 kHz all the way to the final master. Only then, if necessary, the final downsample to 44.1 kHz could be made.

And to that I will add, if the final master is 96 kHz, and the DAC chip requires at least 96 kHz anyway, and we know from his experiments that any sound manipulation on 44.1 kHz audio adds distortion - what's the point to ever downsample the final audio to 44.1 kHz at all? Disks are large, Internet is fast, just stream 96 kHz on Tidal or Qobuz and enjoy HiRes, because it is almost always (I won't claim always) audibly better. It's time for CD format to die, there is no reason for it to exist.

also certain plugins react heavly on resampling (as 44.1khz is insufficient near nyquist), thats why some have oversampling built in, tho you can somewhat circumvent this running your session in the highest possible samplerate

also some enginneers made it clear on youtube that preferably you run your session always in the source material sample rate.... just to avoid unnessecary down and upsampling

imo many things get way oversimplified in this subforum, there are real world differences SPECIALLY on recording/mixing/mastering

 

[gregorio]

Here professional sound engineer proves with experiments that any manipulation / mixing of a digital source at 44.1 kHz is introducing significant audible distortion.

That’s a lie, he absolutely did NOT prove that any manipulation of 44.1kHz introduces audible distortion! He demonstrated that using a specific distortion plugin (Decapitator) and feeding it a high level 8kHz signal that you get “significant audible distortion” (plus some aliasing at 20kHz). Well duh, why the h€ll would anyone apply a distortion plugin not to introduce any audible distortion, are you completely nuts? You think we add EQ plugins because we don’t want any EQ and compression plugins because we don’t want any compression? Do we also put turbos on car engines because we want them to be naturally aspirated and add Coke to Rum because we don’t want any Coca Cola? Clearly you have not understood the point he was making, or that it’s an edge case use!

He does, and encourages everyone, to always mix strictly at 96 kHz all the way to the final master.

He explains why he personally always mixes at 96kHz, which is that he apparently uses a very atypical distortion plugin that does not oversample, he does not want to use a DAW such as Reaper that has a built-in oversampling function or a different distortion plugin that does oversample and apparently he uses that plugin on many dozens of channels, with a significant amount of HF content. That is an extremely edge case scenario and contrary to yet another false assertion on your part, he does NOT “encourage everyone to strictly mix at 96kHz all the way” in fact he states pretty much the exact opposite at about 23:00! (Edit: As @VNandor quotes).

I’ll ask again: Are you really this deluded and ignorant or are you just trolling?

The point made in the video was using 44.1 kHz project sample rate + oversampling plug-ins can end up causing more CPU load than having a 96 kHz project without the need of oversampling plug-ins. I suppose the nature of the project dictates how this is (any views about this Gregorio?).

The points he made seem a bit bizarre: He mentioned using a high shelf to boost freqs higher than 20kHz but there are no use cases for that. He mentioned that using several oversampling plugins on a 2006 6 core Pentium could cause CPU load issues, and sure, I won’t disagree with that but I don’t know anyone in a professional/commercial studio who uses such an ancient computer, in fact no current DAW/software will even run on that CPU. You can buy a Mac for about $1,400 with 12 cores, each of which is probably around two orders of magnitude more powerful than a Pentium core, that can probably run 100 oversampling plugins without issue.

He mentions (or implies) using 100 instances of Decapitator, which is thoroughly bizarre. Decapitator isn’t a particularly good distortion plugin to start with and I can barely even imagine using more than a dozen or so distortion plugins on the same mix, maybe a couple of dozen in an extreme case and even with heavily distorted music genres around 6 or so would be typical, so 100, very strange/atypical! Certainly using over 100 plugins is pretty common, I’ve done mixes with several hundred plugins but not 100 non-linear plugins that would benefit from oversampling.

He also mentioned good and bad oversampling/downsampling algorithms in plugins, again, *sometimes* an issue in 2006 but not for many years. Definitely seems like he has a bit of an agenda, which is somewhat justified if it’s based on experience going back to 2006. It would be interesting for him to do a mix at 48kHz with oversampling distortion plugins, the same mix at 96kHz and do a DBT between them.

To answer your question, there might be cases where one is using such a high number of oversampling plugins that CPU load would be less to just use a 96kHz sample rate. It would be pretty rare though, most likely case would probably be using a lot of soft-synths that oversample, although some would downsample 96kHz because some patches employ audible aliasing as an effect. There’s not a definitive answer but the circumstances where it would be beneficial are a lot rarer than they were 20-25 years ago.

Absolutely none of the above is applicable to @Androxylo or any other consumers though, unless they’re playing back audio with a great deal of HF content and slathering it with distortion using the Decapitator plugin!

G

 

[2leftears]

Disks are large, Internet is fast, just stream 96 kHz on Tidal or Qobuz and enjoy HiRes, because it is almost always (I won't claim always) audibly better. It's time for CD format to die, there is no reason for it to exist.

Well that may be your preference, but I still prefer CD (or DRM-free downloads).

Streaming doesn't work for me as a service model; there is too much I don't like about it, especially the commercial model behind it and a lack of access control.

 

[Androxylo]

Two more videos:



Here Sound Engineer tells a story about some music processing project that was run at some unspecified frequency and produced an audibly bad result. Then he repeated the same processing at 172 kHz for the whole project, and the result was perfect. He said that he did repeated A-B listening testing and also used those audio clips as his demo of why the sound processing must be done at 96 kHz or higher.



Here are two important points I will narrow it from this video: first they both agree that all music mixing must be done at either 96 kHz or 172 kHz and gave various real life examples of why it's bad not doing that. Second they discussed the various upsampling algorithms and stated that the proper upsampling requires infinite computing resources, and that any real world upsampling is a compromise between quality and speed.

Let's put consumer format aside for a moment, do we all agree at this point that all music is actually recorded with sample frequency above 1 MHz and requires to never go below 96 kHz for the proper mixing and mastering? Are there still any questions here?

 

[gregorio]

Here Sound Engineer tells a story about some music processing project that was run at some unspecified frequency and produced an audibly bad result. Then he repeated the same processing at 172 kHz for the whole project, and the result was perfect. He said that he did repeated A-B listening testing and also used those audio clips as his demo of why the sound processing must be done at 96 kHz or higher.

He states that he always runs his converters at 48kHz and that in some cases higher sample rates are worse. He also states that aliasing with some plugins can be an issue and at about 4:00 states the easiest way to overcome this is to enable oversampling in those plugins. He NEVER states that “sound processing must be done at 96kHz or higher”, enough with the lies!

Here are two important points I will narrow it from this video: first they both agree that all music mixing must be done at either 96 kHz or 172 kHz and gave various real life examples of why it's bad not doing that.

You mean; here are two points I’m going to invent and then lie that both video participants agree with them! Did you not even look at the conclusion, which was about when oversampling is beneficial and when it’s not even needed and no mention at all of all music mixing at higher than 48kHz.

Second they discussed the various upsampling algorithms and stated that the proper upsampling requires infinite computing resources, and that any real world upsampling is a compromise between quality and speed.

Again, yet another lie. They did NOT state that proper upsampling requires infinite computing resources, they stated that absolutely perfect upsampling would but that with limited resources something like a windowed sinc function is virtually/almost perfect (proper upsampling), and in fact there are various different ways of achieving “proper upsampling”.

Let's put consumer format aside for a moment, do we all agree at this point that all music is actually recorded with sample frequency above 1 MHz and requires to never go below 96 kHz for the proper mixing and mastering?

Pretty much all digital audio is recorded with a sample frequency above 5MHz these days, although it contains no audio content anywhere near the MHz range because there are no studio music microphones that have much response beyond about 60kHz and the vast majority roll-off around 20kHz or lower. Proper mixing and mastering requires a sample rate of 44.1kHz (NOT 96kHz) and certain non-linear plugins should oversample. There are some instances where a higher sample rate can in fact be beneficial, some time-stretch or large pitch shift operations for example. Absolutely none of this has anything to do with consumers however!!

Are there still any questions here?

Yes, the one you refuse to answer, are you really that deluded and ignorant or are you just trolling?

G

 

[castleofargh]

Two more videos:



Here Sound Engineer tells a story about some music processing project that was run at some unspecified frequency and produced an audibly bad result. Then he repeated the same processing at 172 kHz for the whole project, and the result was perfect. He said that he did repeated A-B listening testing and also used those audio clips as his demo of why the sound processing must be done at 96 kHz or higher.



Here are two important points I will narrow it from this video: first they both agree that all music mixing must be done at either 96 kHz or 172 kHz and gave various real life examples of why it's bad not doing that. Second they discussed the various upsampling algorithms and stated that the proper upsampling requires infinite computing resources, and that any real world upsampling is a compromise between quality and speed.

Let's put consumer format aside for a moment, do we all agree at this point that all music is actually recorded with sample frequency above 1 MHz and requires to never go below 96 kHz for the proper mixing and mastering? Are there still any questions here?

You managed to both misinterpret completely the first video you posted, and move the goalpost to another plan of reality by pretending that the oversampling of a DAC and some distortion DSP that probably doesn't oversample , have similar issues/needs. And while doing it, you managed to choose a video making arguments directly opposing your view on sample rate playback. You're clearly your own biggest enemy.

Again, IDK if it's stupid trolling or a level of tunnel vision rarely achieved, but I'm going to stop wasting time by not watching the 2 new videos you've linked, as I bet they also don't mean what you think they mean.
Maybe stop posting until at least you know what you're trying to argue about? Just an idea.

 

[gregorio]

I'm going to stop wasting time by not watching the 2 new videos you've linked, as I bet they also don't mean what you think they mean.

I’ve seen the second one before and actually I did find it interesting but your bet would be entirely worth it!

Much of what they’re discussing is theoretical or edge case at best, even to a sound engineer. The example of using a saturation plugin on cymbals for instance might be something a programmer might consider but wouldn’t make much sense in practice. The example of TP half a dB above 0dBFS also wouldn’t have any audible effect, DAWs are all 64bit float these days, so no clipping and a TP limiter will handle it no problem. They did state that what they’re discussing would be inaudible, although if a bunch of unlikely planets all aligned there might be a (extremely rare) circumstance where it might be audible.

So, it’s only of passing interest even to sound/music engineers, as many/most engineers don’t want or need to go into such “under the hood” detail, because if the mix sounds how you want it to, no one knows or cares if there’s audible aliasing or a lack of it. It’s only a minority of engineers like me who finds this sort of “under the hood” stuff interesting and it certainly has absolutely nothing whatsoever to do with consumer playback!

G

 

[Joe Bloggs]

I accidentally fell on a video that proves my point from #3544:


Here professional sound engineer proves with experiments that any manipulation / mixing of a digital source at 44.1 kHz is introducing significant audible distortion. You can go to the video for the explanation, I don't want to repeat. He does, and encourages everyone, to always mix strictly at 96 kHz all the way to the final master. Only then, if necessary, the final downsample to 44.1 kHz could be made.

And to that I will add, if the final master is 96 kHz, and the DAC chip requires at least 96 kHz anyway, and we know from his experiments that any sound manipulation on 44.1 kHz audio adds distortion - what's the point to ever downsample the final audio to 44.1 kHz at all? Disks are large, Internet is fast, just stream 96 kHz on Tidal or Qobuz and enjoy HiRes, because it is almost always (I won't claim always) audibly better. It's time for CD format to die, there is no reason for it to exist.

Not going to watch 24 minutes trying to fish out what is inevitably going to be pablum. At least provide a bookmark.

edit: it's not pablum, it just doesn't have anything to do with your argument. Unless you use DSPs in your plugin chain that introduce harmonic distortion deliberately. Do you?

 

[71 dB]

I reply to you only now, because my Sunday evening was spent with snooker: Kyren Wilson vs Judd Trump.

The points he made seem a bit bizarre: He mentioned using a high shelf to boost freqs higher than 20kHz but there are no use cases for that. He mentioned that using several oversampling plugins on a 2006 6 core Pentium could cause CPU load issues, and sure, I won’t disagree with that but I don’t know anyone in a professional/commercial studio who uses such an ancient computer, in fact no current DAW/software will even run on that CPU. You can buy a Mac for about $1,400 with 12 cores, each of which is probably around two orders of magnitude more powerful than a Pentium core, that can probably run 100 oversampling plugins without issue.

I think I missed the high shelf boost part myself. I wasn't paying attention enough, but this of course doesn't change anything. We have tons of CPU power today, of course, but in the past things were different. then again, limitations are often the source of creativity. I started making computer music with Sharp MZ-821 in the 80s, then Amiga 500 + Protracker in early 90s, a half-ass 266 MHz Pentium PC in 1998 and now I using a 2014 Mac Mini + Garageband + Audacity (for mixing). I make zero money on the music I make. I have no talent to create anything that might interest even a small group of people let alone millions of people. This is my hobby. Even talentless people need to create something to stay sane (especially in the World as it is today). I write Nyquist plugins for Audacity. Currently I am testing and developing a plugin to high-pass filter side-channel in a way that is more sophisticated ("binaural"). Just highpass filtering side-channel makes the bass approach mono, but even headphones allow about 3-4 dB* of ILD below 500 Hz. That's why instead of highpass, it might be better to use a shelf filter that leaves some channel difference at the lowest frequencies. The use of minimum phase filter create ITD that approximated real life values. However, designing the filtering so that both ILD and ITD profile is close to natural values in coarse manner is something I am working on...

Limitations are familiar to me. The need to work with them is part of the charm. The amount of hardware and software other people use is "insane" to me. I don't like the philosophy that only rich and fortunate people can create music. Just use whatever you have and learn to take 100 % out of it.

* If the sound source is a few feet from head and the azimuth angle 𝛉 is around 90° and altitude is small. For sound sources more distant natural ILD is smaller. The effect of distance to the ILD can be approximated with 20*log ((D+d)/D)), where D is the distace of the sound source to the closer ear and D+d is the acoustic distance to the other ear that comes from Woodworth, Schlosberg and Kuhn: d ≈ r * (𝛉 + sin 𝛉) for high frequencies and d ≈ 3 * r * sin 𝛉 for low frequencies where r is the "head radius", about 8.5 cm.

He mentions (or implies) using 100 instances of Decapitator, which is thoroughly bizarre.

It is. Maybe he works with music genres that are based on excessive use of decapitators?

Decapitator isn’t a particularly good distortion plugin to start with and I can barely even imagine using more than a dozen or so distortion plugins on the same mix, maybe a couple of dozen in an extreme case and even with heavily distorted music genres around 6 or so would be typical, so 100, very strange/atypical! Certainly using over 100 plugins is pretty common, I’ve done mixes with several hundred plugins but not 100 non-linear plugins that would benefit from oversampling.

I have been thinking about this over the years. To avoid the need to oversample, these decapitators should process low and high frequencies differently. If you limit the hard clipping action to say frequencies below 2 kHz, the distortion harmonics can't fold that strongly. For higher frequencies high ratio compression with fast attack and release. Overlapping the frequency regions make the change from hard clipping to compression smooth.

He also mentioned good and bad oversampling/downsampling algorithms in plugins, again, *sometimes* an issue in 2006 but not for many years. Definitely seems like he has a bit of an agenda, which is somewhat justified if it’s based on experience going back to 2006. It would be interesting for him to do a mix at 48kHz with oversampling distortion plugins, the same mix at 96kHz and do a DBT between them.

People get used to doing things in a certain ways. At least he acknowledges the way he does things is his way and other people may do things differently.

To answer your question, there might be cases where one is using such a high number of oversampling plugins that CPU load would be less to just use a 96kHz sample rate. It would be pretty rare though, most likely case would probably be using a lot of soft-synths that oversample, although some would downsample 96kHz because some patches employ audible aliasing as an effect. There’s not a definitive answer but the circumstances where it would be beneficial are a lot rarer than they were 20-25 years ago.

Yeah, it seem his thinking is based on the past.

 

[HedleyD]

So for the layman here, I have a hiby R5 gen2 and Hidizs mp145s. Should I be getting hi-def files or not. I see this thread was started a long time ago, what's the latest? Please keep it simple. Appreciate it!
@Joe Bloggs what say you?

 

[sander99]

Should I be getting hi-def files or not.

Sometimes a better master is used for one format compared to another, so that can be a good reason to select one format over another in individual cases.
But there is no audible difference in the formats themselves. [Edit: except with too low bitrate lossy of course.]

 

[Golden Ears]

Well if you go and find out people who test and repair cartridges they will explain how a cartridge can produce sounds with much higher sample rates. In the end if your sample rate was infinite all you would do is recreate analog.

The key is dealing with noise. If the noise in the analog signal is overwhelming..well then a lot of the finer reproduction is masked. Digital has a shitload of noise- but we try to keep that way out of the audio passband with steep filters. STEEP FILTERS ARE NOT GOOD FOR SOUND. So if we upsample - even if like in 512 DSD we are just adding blank bits atop a 128 DSD bitstream.... then it allows for a less steep filter.

Sounds more accurate. It's not the sample rate that really changes the music quality ...it's to a large degree the sharpness and phase shift of the filters.

After a certain lower sample rate you start to lose detail which the lossy algos think is meaninglessly masked.. Like if a cat jumps on your bed at the time of a thunderclap it's ok to toss out the cat in the recording... or is it???..depends on how close you are to the cat and your eardrum. Is it needed?

Well unfortunately IMHO these well meaning compression algos often do take away some of the timbre and musicality of the instruments. But if you listen on a Best Buy system that can't properly reproduce these sounds, you won't miss them if they are deleted because they sounded kinda wrong anyhow- because they might have phase amplitude relationships caused by the cheaper speaker drivers that sounded bad.

But when you delete them... a part of the most beautiful part of music is lost. Forever- because you can not upsample it back once its gone.

Now put that original uncompressed lossless file at high rate and run a 125 MP3 file . The MP3 will sound "cleaner" as it has a lot missing but certainly not better with woodwinds, sax, stringed instruments...etc... You aren't going to notice much of a difference listening to Nine Inch Nails which uses digital etch as part of its musicality. And Tool won't help , nor will a lot of garbage made on DAW using Pro Tools. (Pro Fools).

I'm not trying to say you can't make a great digital recording... it just won't have the details that made great analogue recordings great. So for instance ...Patricia Barbers "Modern Cool" was FULLY digitally recorded (Actually I know something special about the recording that the artist does not know ) . It's a great digital recording whether you care for the music or not. It will sound better as a Redbook recording than most 24/96 and 24/192 and 64 DSD recordings of other artists. and as a 24/96 it sounds better still because they did not mess with the original master.

Why? because of the way it was recorded... how was it exactly done? Well even if we knew how it was done people insist on using their own home brew special sauce and screw it up. If you think some millennial kid who was raised on bluetooth speakers and ear buds is suddenly going to be a better mastering engineer using plug ins in post than a 55 year old guy who has to get the musicians to do their 23rd take at 2am and has to THINK hard to fix the issues he has because he can not fix it in post...... well then hire the cheap millennial. Look at what $100 a night DJs did with their MP3 and powered class D QSC speakers compared to a $1200 DJ with great speakers using vinyl and good class A/B amps. Which party would you rather go to?

Well Digital is that inexpensive but kinda blah experience. That has faster listener fatigue and rarely gets you the full alpha state unlike great immersive Analog.

SO MANY Hi- Rez remasters sound worse as recordings because they weren't mixed as well as the originals even though they had the capability to sound better- and people say...well I didn't like the hi-rez as much...well its not the same recording. It was ****ed with by some twerp. Some good music is a result of dumb luck. What you really want is good music , that was poorly mastered, remastered at high Rez- by the great Mastering people. But it might not sell

"The music systems that do the least enjoyment damage help shape the popular music selected as popular"

SO ..1960's let's imagine . using the common table radio with 2 watt tube amplifier and 4 inch monaural speaker .... you play The Beach Boys wouldn't it be nice at 65db then you play "Intergalactic" by the Beastie boys on the same system. Do you think the Beastie Boys would have ever gotten popular limited to a 4" monaural speaker and 2 watts?

Sometimes it's what's left after the filter that is the most important part.