Sample rates above 44.1 are actually bad for tone?

[MDAguy]

So in my never ending desire to learn the ins and outs of anything to do with my audio hobby, and especially in the digital realm where I have absolutely little to no background, I came across this little gem on YouTube... Assuming this guy is right, and his science seems solid to this novice, there is actually a reason NOT to record at above the Nyquist-Shannon sample rate.

Thoughts?

 

[Blumlein 88]

Okay just looked at two minutes. One thing that confuses people is when he looks at 100 hz and then 10,000 hz samples. The sine wave looks all messed up at 10 khz. Unfortunately while almost all editing software shows it this way that is not the waveform you get. Adobe Audition will show you the actual waveform you get. I don't have it or I'd show a screenshot. The theorem's work and as long as you have just over 2 samples per wave you can correctly reconstruct a smooth waveform.
EDIT:Okay I see now he does go and explain the truth about it.

If you took a DAC, and fed it into an analog o-scope even those higher frequencies come out as smooth sine waves.

Spend time viewing and re-viewing this until it all makes sense.

Among other things he is using purely analog scopes and frequency analyzers to show what the result of digital conversions are.

 

[MDAguy]

super ironic that you posted that video as a response while I was actually watching THAT video...

man, the science behind digital to analogue / analogue to digital is truly fascinating... I can see how someone might make a living with this.

 

[Blumlein 88]

Okay, so the posted video is actually very good. It looks like everything he is showing you is correct. You often hear pros says 96 khz is still worthwhile because of processing. But modern Daw effects offer that internal over-sampling which makes it a non-issue. Even the very inexpensive Reaper offers that. But overall a good video here.

Thanks for posting it.

Also nice to see someone with a product to sell, showing you how it really works and explaining why you don't need more sampling rate instead of spreading FUD to get you to buy more gear or software.

 

[MDAguy]

The learning curve has been steep so far... and mind blowing as well.. to essentially be told that the capability of a top of the line DAC, or for that matter many other electronic audio components is being wasted on the reality that the human ear (even a perfect one) can't hear it????

It's like having a. 1500 horsepower Buggati and living in Manhattan...

 

[RayDunzl]

Using Audacity.

Top: 18,492Hz tone at 44.1kHz sample rate, showing the sample values.

Copy that to the second track.

Resample that data to 352,800Hz sample rate.

This is my "poor man's" method to see what the analog wave would look like coming out of the DAC, even if using the original sample rate for conversion at the DAC:

 

[solderdude]

You often hear pros says 96 khz is still worthwhile because of processing.

Probably they mean the 24 bit depth that goes with the 96kHz which gives much better resolution and they can record well below 0dB which is handy when mixing to have overhead at that point in the process.
24/48 will not be much worse than 24/96

 

[Χ Ξ Σ]

The learning curve has been steep so far... and mind blowing as well.. to essentially be told that the capability of a top of the line DAC, or for that matter many other electronic audio components is being wasted on the reality that the human ear (even a perfect one) can't hear it????

It's like having a. 1500 horsepower Buggati and living in Manhattan...

Congratulations on your revelation!
Next, send your dCS, Accuphase, and McIntosh to Amir for measurements, and you will be one step closer to the truth.

 

[voodooless]

Assuming this guy is right, and his science seems solid to this novice

He is correct and the science is solid.

there is actually a reason NOT to record at above the Nyquist-Shannon sample rate.

But he never said anything about recording. This is purely about music production and the use of very specific filters (and there are of course many more of these) that indeed create aliasing artefacts. He actually mostly complained about CPU load when running on a higher sampling rate, something that nowadays is not really that much of an issue with 32+ core CPU's on the market. The tradeoff here is to use the filters "HQ" option vs raising the total sampling rate frequency of the project. Obviously the filter options is most often the better choice since it will be specifically tuned for the purpose. I'd also agree with the 48 kHz argument. It gives you just a bit more headroom.

Do we need more: probably not. You better make sure you have a competent sound engineer... looks like that is far more valuable than having a project run at 192 kHz.

It's staggering BTW how many times audio get's resampled nowadays! With things like this you could easily run with dozens of sample rate converters.

.. and then come the audiophiles with their very special magical upsampler for the last leg to reveal the final bit of detail and lift the dreaded veil from their systems

 

[Pluto]

It's an awfully long-winded video that tells you little more than one of the oldest metaphors in all of natural philosophy: the wider you open the window, the more sh*t gets in.

When you are producing music, you can do (more or less) whatever you want to obtain the required result. But it always pays to filter everything much above twentysomething kHz for final delivery. Whether that filtering is proactive (if you are delivering at a sampling rate ≥48kHz) or an inherent part of the process (when you are converting to ≤48kHz) doesn't really matter.

If you have analysed a great many so-called "hi-res" recordings – as I have, it is remarkable just how many seem to have nothing but random rubbish above 20kHz and while this garbage is not usually, in itself, audible, it cannot do any good if the opportunity for it to intermodulate (and, thereby, create yet more rubbish that might be audible) comes along.

 

[threni]

He is correct and the science is solid.

He also wrote this (archive.org link because the original no longer exists - perhasps he was run out of town by subjectivists...):

https://web.archive.org/web/20180216031756/https://xiph.org/~xiphmont/demo/neil-young.html

 

[BDWoody]

super ironic that you posted that video as a response while I was actually watching THAT video...

man, the science behind digital to analogue / analogue to digital is truly fascinating... I can see how someone might make a living with this.

You are right about that. If there's magic in audio, it's in sampling theory.

Reconstruction with no loss of information.
Math is beautiful...

 

[voodooless]

Whether that filtering is proactive (if you are delivering at a sampling rate ≥48kHz) or an inherent part of the process (when you are converting to ≤48kHz) doesn't really matter.

The whole point of the video is that is matters a whole lot when you do what kind of filtering!

I'm curious though: how did @MDAguy come to the conclusion that is in the topic title after watching that video?

 

[Pluto]

it matters a whole lot when you do what kind of filtering!

His examples were largely based on the use of plug-ins designed to create massive non-linearities, so it comes down to hearing what you like and liking what you hear.

If you like the sound, it's art. If you don't like it, it's distortion

So the argument stands: do whatever you want in production but ultimately delivering more than the human hear can assimilate will only end in tears.

 

[MDAguy]

Congratulations on your revelation!
Next, send your dCS, Accuphase, and McIntosh to Amir for measurements, and you will be one step closer to the truth.

I would love to see them measured, but that being said, Stereophile had done measurements on the dCS stuff and they measure very, very, well... doesn't mean I can hear the difference, but they're certainly much better than a cheap DAC or CD player objectively speaking. High resulting systems aren't a bad thing, just because we can't hear them... I know it sounds stupid, but it's not a very different logic from many luxury purchases... a jacket is a jacket, whether it's cashmere or polyester.. but some want the luxury of cashmere.

Plus there's the whole quality of the case work electronics, US, UK, and Japan made vs China and such things that matter to me... I don't mind the smaller bang for the buck, as I can afford the buck. (Data Conversion Systems) or dCS has been at the forefront of Analogue to Digital conversion for 3 decades now, and pioneered 24 bit systems in the late 80's that were used by well known studies. They have a very long and impressive history in digital and analogue conversion (to and from).

In the end, It's the same logic on driving a car that can do 192 mph even if the fastest I can drive it is 65 (legally). I want this stuff for it's cool factor as much as anything else... whatever my 50 year old ears can hear the difference is almost unimportant, so long as the sound is amazing, which it is.

No regrets

here's the in-depth measurements on the Rossini w/ Clock.

https://www.stereophile.com/content/dcs-rossini-player-rossini-clock-measurements

 

[voodooless]

I'm curious though: how did @MDAguy come to the conclusion that is in the topic title after watching that video?

I still would like to know your train of thought here Worst thing that could happen is that we all learn some more

 

[MDAguy]

I still would like to know your train of thought here Worst thing that could happen is that we all learn some more

The title of the video, which was essentially a rhetorical question ...

Intermodulation and harmonics issues were clearly higher with no returned benefit with the higher sample rates.

 

[voodooless]

The title of the video, which was essentially a rhetorical question ...

I'm talking about your title, not the one from the video . No sample rates is by definition bad, nor are higher once worse. In any case, you cannot generalise what is said in the video to any audio file at a specific sampling rate. As I said before: this is about production, not playback.

 

[MDAguy]

I'm talking about your title, not the one from the video . No sample rates is by definition bad, nor are higher once worse. In any case, you cannot generalise what is said in the video to any audio file at a specific sampling rate. As I said before: this is about production, not playback.

I'm well aware that a good recording matters more than how it's transported to my HiFi system, what I don't get is that in the video, he proceeds to show that sample rats of 96, and 192 are potentially detrimental to tone due to intermodulation and harmonics ... how does that not line up with my title? Happy to be wrong btw....

 

[voodooless]

what I don't get is that in the video, he proceeds to show that sample rats of 96, and 192 are potentially detrimental to tone due to intermodulation and harmonics ... how does that not line up with my title?

Because it is generalising, and the video is very specific about certain types of filters. If you use those at a higher sample rate without the HQ option, you will get a better result. If you use the filters HQ option however, that will give you an even better result while leaving the sampling rate low because they use an 8x oversampling internally (of 4x in case of one of the other filter options).

So you cannot generalise this to meaning that higher sampling rate audio files have more intermodulation and harmonic distortion. That is simply not the case!