24bit vs 16bit, the myth exploded!
Aug 17, 2021 at 11:49 AM Post #6,331 of 6,480

gregorio

Headphoneus Supremus
Joined
Feb 14, 2008
Posts
4,125
Likes
2,244
Gregorio, you really didn't understand what I mean because you don't seem to consider what happens in the time domain.
In my last response to you I stated "It has been explained to you that sound waves loose high/ultrasonic frequency over distance/time ..."and previously I stated that "what we sample is amplitude over time" so clearly I am considering the time domain and your statement is FALSE!!
To understand what a waveform is we not only need to consider which signals are audible, but we also have to look at their position in the time domain.
Again, what we sample is amplitude over time, that's it, nothing else!!
Frequencies are not random, but their position in the time domain is random.
So going back to my previous post, what you're effectively saying is that "yes a violin placed on a stage will start (randomly) playing itself but not random frequecies, it will always sound like a violin, not a piano or the spice girls. You still confident in that are you?
This is what makes the waveform more complex than what most people think.
No it's not, most people do not think a violin could randomly start playing itself. I don't really know what most people think but I don't need to, the complexity of the waveform can easily be analysed. And don't forget, much of this isn't new science it's some of the oldest, most well established science we have. Pythagoras discovered the mathematical relationship of music notes/harmonics around 2,500 years ago!
Then, we have to consider that frequencies in the waveform are not really audible.
Some are, some aren't. If none were audible music wouldn't exist, we wouldn't hear pitch or in fact any sound at all!
They are not what we use to call sound but only information about sound.
That obviously cannot be correct because sound is defined by it's frequency content. If a sound has no frequency content then it isn't sound!
This information has to be computed by the ear to produce actual sound.
Have you been reading too much metaphysics? Somehow I don't think so. According to you, sound is produced mathematically and randomly, now you're saying the ear produces the sound. So the human ear must be a random mathematical machine! Great, why stop at sound science when we can completely mangle the science of biology and anatomy as well! :) And, how can recording work? If it's the ear producing the actual sound then we would have to put our microphones next to the ear rather than next to the musical instrument that's being played. If there's lots of people listening to a concert, who's ears do we record? How much confidence do you have in your assertion? :)

G
 
Aug 17, 2021 at 11:57 AM Post #6,332 of 6,480

71 dB

Headphoneus Supremus
Joined
Sep 17, 2017
Posts
1,697
Likes
561
Location
Helsinki, Finland
71 dB, why do you suggest 60 KHz for signal filtering, and not 80 KHz ou 120 KHz, for instance ?
Is there a good reason for choosing this frequency ?
About 60 kHz is the "optimal" sample rate for digital audio, because it allows "relaxed" anti-alias filtering, but isn't unnecessarily high. Higher sample rate than that doesn't give any kind of benefits. Using 44.1 kHz and 48 kHz sample rates forces the use of steep anti-alias filters, which isn't really a problem and also file sizes are smaller.

A digital consumer audio format of 54 kHz/13 bit would have almost the same bitrate as 44.1 kHz/16 bit. Anti-alias/reconstruction filtering would be "easier" and 13 bit dynamic range is enough especially if shaped dither is used and since 54 kHz sampling rate has bandwidth up to 27 kHz, most of the shaped dither noise energy could be pushed in the 20-27 kHz frequency band. However, we have the formats we have and fortunately they allow extremely good sound quality.
 
Aug 17, 2021 at 12:01 PM Post #6,333 of 6,480

71 dB

Headphoneus Supremus
Joined
Sep 17, 2017
Posts
1,697
Likes
561
Location
Helsinki, Finland
What about recording dolphins or bats?
That's different. That's scientific work. Also, if you want to use ultrasonic sounds in your music played back at lower speed to make them audible for humans high sample rates are beneficial, but consumer audio is a different story.
 
Aug 17, 2021 at 12:01 PM Post #6,334 of 6,480

audiokangaroo

Head-Fier
Joined
Jul 26, 2021
Posts
70
Likes
2
Location
France
What about recording dolphins or bats?
Did you notice that the transfer of the audio content of real to real tape onto 44.1/16 results in a very significant loss of audio information ?
If you want to make and opinion you can the visit the youtube channel called MarioPindaBR. This guy has remastered old tape recordings (some are from very well kown artists to DSD and 96 KHz PCM. Even through youtube transcoding, most DSD masters sound impressive and much better than 44.1/16 or even 96/24.
 
Last edited:
Aug 17, 2021 at 12:17 PM Post #6,335 of 6,480

audiokangaroo

Head-Fier
Joined
Jul 26, 2021
Posts
70
Likes
2
Location
France
About 60 kHz is the "optimal" sample rate for digital audio, because it allows "relaxed" anti-alias filtering, but isn't unnecessarily high. Higher sample rate than that doesn't give any kind of benefits. Using 44.1 kHz and 48 kHz sample rates forces the use of steep anti-alias filters, which isn't really a problem and also file sizes are smaller.

A digital consumer audio format of 54 kHz/13 bit would have almost the same bitrate as 44.1 kHz/16 bit. Anti-alias/reconstruction filtering would be "easier" and 13 bit dynamic range is enough especially if shaped dither is used and since 54 kHz sampling rate has bandwidth up to 27 kHz, most of the shaped dither noise energy could be pushed in the 20-27 kHz frequency band. However, we have the formats we have and fortunately they allow extremely good sound quality.
This sounds like old Dan Lavry's litterature.
Do you have any scientific element to decide when an anti-alias filter becomes relaxed in terms of corner and slope ?
 
Last edited:
Aug 17, 2021 at 1:24 PM Post #6,336 of 6,480

KeithPhantom

100+ Head-Fier
Joined
Apr 3, 2020
Posts
385
Likes
203
Location
USA
This sounds like old Dan Lavry's litterature.
Do you have any scientific element to decide when an anti-alias filter becomes relaxed in terms of corner and slope ?
Filter design is based on a tradeoff between steepness and frequency response. You can add ripple to the mix if you want. Assuming a stopband of 22.05 kHz, you can have perfect filtering in terms of steepness, but you have to reduce the upper limit of your bandwidth; that means that you will start filtering a bit before 20 kHz in order to be at least -100 dBFS at 22.05 kHz. This filter may be audible if you start too early to someone with good ears, so there is another approach. The other approach is to have a linear frequency response until the end of the passband, but relax the steepness of the filter by missing by a bit the 22.05 kHz target. Usually, filters that do this have their stopband at 24 kHz.

Even though that we have this filter design, we can also oversample before filtering, thus relaxing the requirements of the filter. The more filtering you need, the higher the order of the filter.
 
Aug 17, 2021 at 1:36 PM Post #6,337 of 6,480

bfreedma

The Hornet!
Joined
Feb 3, 2012
Posts
2,748
Likes
1,846
Did you notice that the transfer of the audio content of real to real tape onto 44.1/16 results in a very significant loss of audio information ?
If you want to make and opinion you can the visit the youtube channel called MarioPindaBR. This guy has remastered old tape recordings (some are from very well kown artists to DSD and 96 KHz PCM. Even through youtube transcoding, most DSD masters sound impressive and much better than 44.1/16 or even 96/24.
At this point, I actually hope you are trolling. Assessing DSD vs. 44.1 via YouTube streams? We can’t possibly need to discuss why that isn’t viable, can we?
 
Aug 17, 2021 at 1:57 PM Post #6,338 of 6,480

71 dB

Headphoneus Supremus
Joined
Sep 17, 2017
Posts
1,697
Likes
561
Location
Helsinki, Finland
Aug 17, 2021 at 2:04 PM Post #6,339 of 6,480

audiokangaroo

Head-Fier
Joined
Jul 26, 2021
Posts
70
Likes
2
Location
France
Filter design is based on a tradeoff between steepness and frequency response. You can add ripple to the mix if you want. Assuming a stopband of 22.05 kHz, you can have perfect filtering in terms of steepness, but you have to reduce the upper limit of your bandwidth; that means that you will start filtering a bit before 20 kHz in order to be at least -100 dBFS at 22.05 kHz. This filter may be audible if you start too early to someone with good ears, so there is another approach. The other approach is to have a linear frequency response until the end of the passband, but relax the steepness of the filter by missing by a bit the 22.05 kHz target. Usually, filters that do this have their stopband at 24 kHz.

Even though that we have this filter design, we can also oversample before filtering, thus relaxing the requirements of the filter. The more filtering you need, the higher the order of the filter.
I read that steep analogue filters could have an impact on phase inside the audible band. Do you think it's true ?
 
Last edited:
Aug 17, 2021 at 2:08 PM Post #6,340 of 6,480

audiokangaroo

Head-Fier
Joined
Jul 26, 2021
Posts
70
Likes
2
Location
France
At this point, I actually hope you are trolling. Assessing DSD vs. 44.1 via YouTube streams? We can’t possibly need to discuss why that isn’t viable, can we?
Do the experience. It may look surprising, but in spite of the youtube transcoding and compression I can clearly hear the difference between DSD and 96/24, at least
when the original analogue master was not or only a little processed. Processing can make things at little blurred.
By the way, I believe that Youtube can go up to 192/24 and their compression process is very clean.
You can also listen to samples from DSD on highdeftapetransfers.com.
 
Last edited:
Aug 17, 2021 at 2:09 PM Post #6,341 of 6,480

71 dB

Headphoneus Supremus
Joined
Sep 17, 2017
Posts
1,697
Likes
561
Location
Helsinki, Finland
At this point, I actually hope you are trolling. Assessing DSD vs. 44.1 via YouTube streams? We can’t possibly need to discuss why that isn’t viable, can we?
Yeah, the insanity level is astonishing.
 
Aug 17, 2021 at 2:14 PM Post #6,342 of 6,480

VNandor

500+ Head-Fier
Joined
Oct 17, 2014
Posts
571
Likes
226
Hey guys, why argue? Measure

I used リライト from Asian Kung Fu Generation (2016 Version), 96 KHz 24 bit. I changed the bit rate to 44.1 KHz and changed it back to 96 KHz and let Audacity analyze the difference

Here is the result



The biggest peak/difference is right after 20KHz, but as we are all able to see, there is a difference in all other frequencies too.

There is a difference, audacity says so. It exists. There is nothing to argue if it exists or not, its there and its measurable.

You can argue if it does matter, but not if its there.
Could you upload both the original and your downsampled version somewhere? A short clip like a minute or so is already enough if you don't want to upload the whole files. I could go into details of why the graph doesn't show what you think it shows (it's not precisely the spectrum of the difference) but I would like to try it for myself with your samples before.
 
Aug 17, 2021 at 2:22 PM Post #6,343 of 6,480

71 dB

Headphoneus Supremus
Joined
Sep 17, 2017
Posts
1,697
Likes
561
Location
Helsinki, Finland
I read that steep analogue filters could have an impact on phase inside the audible band. Do you think it's true ?
Not only steep ones. All analog filters have an impact on phase because they are minimum phase filters. In case of anti-alias filters the phase distortion within audible band is largest near the cut off frequency (20 kHz) and gets smaller toward lower frequencies. Phase distortion caused by minimum phase filters is "natural" and doesn't sound that bad so typically it is not very harmful.

The only way to use analog filters without phase distortion is running the signal thru the filter twice, the second time reversed! This cancels the phase distortion.

Digital filters can be designed to be linear phase meaning they don't produce phase distortion.
 
Aug 17, 2021 at 2:46 PM Post #6,344 of 6,480

audiokangaroo

Head-Fier
Joined
Jul 26, 2021
Posts
70
Likes
2
Location
France
Yeah, the insanity level is astonishing.
Why do you think that Youtube sonic quality should be much inferior to Apple Music, for instance ?
They are both audio streaming other IP. They both can reach 192/24. Youtube uses a little compression, but this is not very audible. They have made a long way since
5 or 6 years ago, when they had a more or less trashy sound.
 
Aug 17, 2021 at 3:51 PM Post #6,345 of 6,480

KeithPhantom

100+ Head-Fier
Joined
Apr 3, 2020
Posts
385
Likes
203
Location
USA
I read that steep analogue filters could have an impact on phase inside the audible band. Do you think it's true ?
They affect phase due to they being minimum-phase, but there are other bigger issues that affect phase. Enclosures such as rooms and headphones affect phase response due to cancellations and resonances, and these are orders of magnitude greater than anything an analog filter can do. Not only that, we are pretty insensitive to absolute phase shifts, so the audibility factor still needs to completely be assessed.
 

Users who are viewing this thread

Top