24bit vs 16bit, the myth exploded!
Mar 11, 2021 at 5:05 PM Post #6,076 of 6,480

bigshot

Headphoneus Supremus
Joined
Nov 16, 2004
Posts
23,413
Likes
4,936
Location
A Secret Lab
I did a listening test comparing direct DSD to 16/44.1 and couldn’t discern any differences. If you’d like to try that yourself, I can tell you how to set that up.
 
Mar 11, 2021 at 5:21 PM Post #6,077 of 6,480

theaudiologist1

100+ Head-Fier
Joined
Sep 23, 2019
Posts
223
Likes
27
Location
Uranus
I did a listening test comparing direct DSD to 16/44.1 and couldn’t discern any differences. If you’d like to try that yourself, I can tell you how to set that up.
I already tried that. To me and my DAC, PCM and DSD sound different and DSD is always quieter, but not necessarily better. Both are equal quality-wise.
 
Last edited:
Mar 11, 2021 at 6:19 PM Post #6,078 of 6,480

old tech

500+ Head-Fier
Joined
Jul 8, 2015
Posts
827
Likes
413
Location
Sydney, Australia
I already tried that. To me and my DAC, PCM and DSD sound different and DSD is always quieter, but not necessarily better. Both are equal quality-wise.
Do you try level matching, ie either raising the volume of the DSD or lowering the volume of PCM so they are equal and then blind testing them? All the controlled tests I've seen don't find a difference in normal listening.
 
Mar 11, 2021 at 8:21 PM Post #6,079 of 6,480

Don Hills

100+ Head-Fier
Joined
Jun 13, 2013
Posts
406
Likes
203
"HiFi Video" recorders did. They recorded frequency modulated audio with rotating heads. But at that time digital audio was already upcoming.

They also had to use a DBX style compression / noise reduction system, because of the noise introduced by the head switching at the end of each scan. It wasn't a problem for video because the head switch was hidden in the flyback for each frame. Most decks had a tracking control knob, if you maladjusted it you could often hear a "purring" noise as the signal dropped out of the FM limiting.
 
Mar 11, 2021 at 8:51 PM Post #6,080 of 6,480

theaudiologist1

100+ Head-Fier
Joined
Sep 23, 2019
Posts
223
Likes
27
Location
Uranus
Do you try level matching, ie either raising the volume of the DSD or lowering the volume of PCM so they are equal and then blind testing them? All the controlled tests I've seen don't find a difference in normal listening.
I did. I increased the volume of my AMP so that the DSD files are the same volume as the PCM files. Keep in mind even PCM files upsampled to DSD have a lower volume (I gave it a -6dB gain). Overall the DSD sounds flatter and more "analog" while the PCM has more "oomph" and the instruments are a bit more "in your face". I did this test with Audirvana.
 
Last edited:
Mar 11, 2021 at 10:33 PM Post #6,081 of 6,480

bigshot

Headphoneus Supremus
Joined
Nov 16, 2004
Posts
23,413
Likes
4,936
Location
A Secret Lab
I already tried that. To me and my DAC, PCM and DSD sound different and DSD is always quieter, but not necessarily better. Both are equal quality-wise.

Equal quality is equal. You can always turn up the volume a hair. If you properly level match, all differences will disappear. High bitrate lossy, CD quality sound, HD Audio and DSD are all capable of being audibly transparent. To human ears, they should all sound the same unless you're doing something wrong.

There are four components to a controlled listening test. I did all three of these when I compared DSD to 16/44.1:

Level matching: Both samples were set to an identical level. Humans tend to perceive slightly louder sounds as sounding better, even when the louder sound is the exact same quality. Even differences as little as a half a dB can make a difference. You can't just ride the volume control. It has to be calibrated carefully.

Direct A/B switched: Human auditory memory is very short. For similar sounds it can be as short as a couple of seconds. If you want to compare similar sounds, they need to be racked up through a switch box, playing the exact same music in sync so you can just flip a switch back and forth and directly compare them.

Blind Testing: Expectation bias is real and it is unconscious. You can't will it away. The way to eliminate it is to focus entirely on the sound and remove any other identifying information about the samples.

Multiple Tests: There is always an element of dumb luck or chance involved in picking one sample or another. To eliminate that, you do multiple rounds of test and average them to see if it reaches the level where it can't be blind chance.

Do all four of these, and you will know to a high degree of certainty what the truth is. I've done this. I know. That is step one. Step two is making your results repeatable for other people. Take care of step one and then we can talk about step two.
 
Last edited:
Mar 12, 2021 at 3:35 AM Post #6,082 of 6,480

118900

1000+ Head-Fier
Joined
Mar 9, 2009
Posts
1,016
Likes
895
Equal quality is equal. You can always turn up the volume a hair. If you properly level match, all differences will disappear. High bitrate lossy, CD quality sound, HD Audio and DSD are all capable of being audibly transparent. To human ears, they should all sound the same unless you're doing something wrong.

There are four components to a controlled listening test. I did all three of these when I compared DSD to 16/44.1:

Level matching: Both samples were set to an identical level. Humans tend to perceive slightly louder sounds as sounding better, even when the louder sound is the exact same quality. Even differences as little as a half a dB can make a difference. You can't just ride the volume control. It has to be calibrated carefully.

Direct A/B switched: Human auditory memory is very short. For similar sounds it can be as short as a couple of seconds. If you want to compare similar sounds, they need to be racked up through a switch box, playing the exact same music in sync so you can just flip a switch back and forth and directly compare them.

Blind Testing: Expectation bias is real and it is unconscious. You can't will it away. The way to eliminate it is to focus entirely on the sound and remove any other identifying information about the samples.

Multiple Tests: There is always an element of dumb luck or chance involved in picking one sample or another. To eliminate that, you do multiple rounds of test and average them to see if it reaches the level where it can't be blind chance.

Do all four of these, and you will know to a high degree of certainty what the truth is. I've done this. I know. That is step one. Step two is making your results repeatable for other people. Take care of step one and then we can talk about step two.
I saw an interesting video where using an app someone played two files on opposite phases. If they cancel each other out and you hear just silence or hiss they are the same. Can’t remember the name of the app but he used it to show that 24 and 16 bit recordings are identical
 
Mar 12, 2021 at 3:37 AM Post #6,083 of 6,480

bigshot

Headphoneus Supremus
Joined
Nov 16, 2004
Posts
23,413
Likes
4,936
Location
A Secret Lab
That is called a null test. It's a great way to test to see if things are the same.
 
Mar 12, 2021 at 8:33 AM Post #6,084 of 6,480

old tech

500+ Head-Fier
Joined
Jul 8, 2015
Posts
827
Likes
413
Location
Sydney, Australia
I saw an interesting video where using an app someone played two files on opposite phases. If they cancel each other out and you hear just silence or hiss they are the same. Can’t remember the name of the app but he used it to show that 24 and 16 bit recordings are identical
This one you mean?
 
Mar 12, 2021 at 10:35 AM Post #6,086 of 6,480

Sebasistan

100+ Head-Fier
Joined
Nov 30, 2017
Posts
293
Likes
232
Location
Chicago
That’s it!
Is the wrong takeaway from this to say "well, he does prove that higher bit depths have the same "amount" of music in them, BUT he also proves higher bit depths ARE superior because they have less noise in comparison!"?
 
Mar 12, 2021 at 10:45 AM Post #6,087 of 6,480

118900

1000+ Head-Fier
Joined
Mar 9, 2009
Posts
1,016
Likes
895
Is the wrong takeaway from this to say "well, he does prove that higher bit depths have the same "amount" of music in them, BUT he also proves higher bit depths ARE superior because they have less noise in comparison!"?
The only thing I took away was what I originally stated, ie that by cancelling each other out when played back in opposing phases proves that no additional info is there. Not sure what else he tried to imply.
 
Last edited:
Mar 12, 2021 at 12:38 PM Post #6,088 of 6,480

Sebasistan

100+ Head-Fier
Joined
Nov 30, 2017
Posts
293
Likes
232
Location
Chicago
The only thing I took away was what I originally stated, ie that by cancelling each other out when played back in opposing phases proves that no additional info is there. Not sure what else he tried to imply.
I also wonder if this can be applied to cd vs. vinyl vs. mp3.
 
Mar 12, 2021 at 1:51 PM Post #6,089 of 6,480

VNandor

500+ Head-Fier
Joined
Oct 17, 2014
Posts
577
Likes
233
I also wonder if this can be applied to cd vs. vinyl vs. mp3.
It could be easily applied to cd (lossless) vs. mp3. However this test is meant to check if the compared signals are different and since pretty much everyone knows lossy has to be different compared to lossless (that's why it is called lossy in the first place) the test wouldn't reveal anything that we didn't know in the first place. CD vs. vinyl would be harder to do because it would require some actual hardware that can do the phase inversion, gain match, and summing in the analog domain.
 
Mar 12, 2021 at 2:40 PM Post #6,090 of 6,480

bigshot

Headphoneus Supremus
Joined
Nov 16, 2004
Posts
23,413
Likes
4,936
Location
A Secret Lab
A null test wouldn't be good for testing lossy vs lossless. You know lossy is missing information, and that is going to show up. However due to masking and other psychoacoustic techniques, the material that is missing isn't audible. So even though it would show up in a null test, it probably wouldn't make any difference in normal listening.

CD vs vinyl would be difficult, if not impossible to null because vinyl has so much speed fluctuation.

Edit: Oops! I just found out that Vnandor said the exact same thing. Bread and butter!
 

Users who are viewing this thread

Top