Testing audiophile claims and myths
Oct 11, 2020 at 9:38 AM Post #14,131 of 15,839

KeithPhantom

100+ Head-Fier
Joined
Apr 3, 2020
Posts
398
Likes
221
Location
USA
I have a vinyl rip which has been recorded as 24bit/192 kHz audio.
Why does it have better sound quality than the CD version?
Because of the LP source or the higher resolution?
“Better sound quality” is subjective. What do you mean by “better resolution”? Could you provide objective data supporting your argument?
 
Oct 11, 2020 at 7:18 PM Post #14,132 of 15,839

old tech

500+ Head-Fier
Joined
Jul 8, 2015
Posts
872
Likes
440
Location
Sydney, Australia
I have a vinyl rip which has been recorded as 24bit/192 kHz audio.
Why does it have better sound quality than the CD version?
Because of the LP source or the higher resolution?
Most likely different mastering than the CD version which has a sound that suits your individual subjective preference. CD has higher resolution than vinyl so a rip into a 24/192 container won't make any difference.
 
Oct 12, 2020 at 4:45 AM Post #14,133 of 15,839

hakunamakaka

500+ Head-Fier
Joined
Mar 30, 2020
Posts
749
Likes
375
Location
Lithuania, Vilnius
Most likely different mastering than the CD version which has a sound that suits your individual subjective preference. CD has higher resolution than vinyl so a rip into a 24/192 container won't make any difference.


Is it a complex process ? Why some rip’s sounds low quality ? I’m not as good as describing sound, but can give few youtube examples
 
Oct 12, 2020 at 7:23 AM Post #14,134 of 15,839

magicscreen

Head-Fier
Joined
Aug 29, 2018
Posts
93
Likes
26
Location
Budapest
Is it a complex process ? Why some rip’s sounds low quality ? I’m not as good as describing sound, but can give few youtube examples
You need a high quality turntable+cartridge+phone amplifier+ADC for make a high quality vinyl rip.
First, youtube already has low sound quality. You will never hear a real vinyl rip on youtube.
Second, there are a lots of vinyl rip there like this: TEAC TN300
This turntable is a crap. It is only good for destroying your precious vinyl discs.
 
Oct 12, 2020 at 3:13 PM Post #14,135 of 15,839

bigshot

Headphoneus Supremus
Joined
Nov 16, 2004
Posts
24,207
Likes
5,284
Location
A Secret Lab
Some albums sound best on CD. Others sound best on the first LP release. Still others sound best on SACD or Blu-ray. It has nothing to do with the format. All of these formats are capable of sounding very good. The reason they are different is because of the degree of care taken when they mastered the music to disc and the condition of the masters. A recording made in 1970 might have sounded great then, but the master tape might have gotten beat up in the intervening years. Or perhaps the record label paid to have the recording restored in recent years and the newer version sounds better.

There is no one answer to "which is best?" It all depends on the particular circumstances with that particular album. If you ask serious collectors who have heard multiple releases, they can usually steer you towards the best.
 
Oct 12, 2020 at 6:48 PM Post #14,136 of 15,839

colonelkernel8

Headphoneus Supremus
Joined
Dec 30, 2006
Posts
2,951
Likes
263
Location
Minneapolis
Some albums sound best on CD. Others sound best on the first LP release. Still others sound best on SACD or Blu-ray. It has nothing to do with the format. All of these formats are capable of sounding very good. The reason they are different is because of the degree of care taken when they mastered the music to disc and the condition of the masters. A recording made in 1970 might have sounded great then, but the master tape might have gotten beat up in the intervening years. Or perhaps the record label paid to have the recording restored in recent years and the newer version sounds better.

There is no one answer to "which is best?" It all depends on the particular circumstances with that particular album. If you ask serious collectors who have heard multiple releases, they can usually steer you towards the best.

It's this in a nutshell. More often than not, there are terrible masters (mixed way too hot) for rereleases on CD and digital formats that were much better during the original LP release, but this is just a rule of thumb and I think modern sound engineers and mixers are getting the picture as mainstream consumer audio equipment increases in performance that they need to in some ways return to how they mixed music before radio.
 
Oct 15, 2020 at 4:33 AM Post #14,138 of 15,839

dazzerfong

1000+ Head-Fier
Joined
Jul 5, 2014
Posts
1,396
Likes
650
Location
Sydney
Minimal rules for a successful blind test?
How many tries?
How many positive tries?
For example 5 positive out of 5 tries = 5/5.
5/5?
10/10?
9/10?
8/10?

Depends on what you want to define as 'successful'. For science, it's usually a p = 0.05 (lower the better).

To determine that, use something called a binomial table. Or you use an online calculator:

https://stattrek.com/online-calculator/binomial.aspx#:~:text=What is the cumulative binomial,example of a cumulative probability.

If you're curious, binomial table looks like this:

1602752329894.png


So, take for example you want 20 trials. Assuming probability of each guess is 50%, if you guessed 14 correctly, that's a p = 0.037, or 3.7% you randomly got that result.

Then you have something called effect size, which is a whole can of marbles. Basically, if you can 'pass' the 0.05 benchmark, you can then determine if it's a large effect or small. To make it very simple, basically the larger the pass percentage (taking into account your sample), the greater the effect.

Summary: doesn't really matter unless you're trying to publish your results.
 
Last edited:
Oct 15, 2020 at 9:23 AM Post #14,139 of 15,839

KeithPhantom

100+ Head-Fier
Joined
Apr 3, 2020
Posts
398
Likes
221
Location
USA
Depends on what you want to define as 'successful'. For science, it's usually a p = 0.05 (lower the better).

To determine that, use something called a binomial table. Or you use an online calculator:

https://stattrek.com/online-calculator/binomial.aspx#:~:text=What is the cumulative binomial,example of a cumulative probability.

If you're curious, binomial table looks like this:

1602752329894.png

So, take for example you want 20 trials. Assuming probability of each guess is 50%, if you guessed 14 correctly, that's a p = 0.037, or 3.7% you randomly got that result.

Then you have something called effect size, which is a whole can of marbles. Basically, if you can 'pass' the 0.05 benchmark, you can then determine if it's a large effect or small. To make it very simple, basically the larger the pass percentage (taking into account your sample), the greater the effect.

Summary: doesn't really matter unless you're trying to publish your results.
That will only determine correlation between two variables of the hypothesis testing is isolated and done as the sole method of verifying existence of causation. Repeatability is extremely important, specially with different sample sizes, test conditions, and the such.

Also, p=0.05 only takes account of +- <= 2 SD/ 5% SE (~95% CI), not being enough for fully isolate the presence of outliers in some stringent testing. Sample size is also to be considered, since binomial and t distributions tend to be closer to the normal distribution when large sample sizes are used and more accurately represent the cummulative probability distribution. With statistics, determining causation is a science and an art, because multiple testing has to be provided, being this repeatble and analyzed by external parties as well.
 
Oct 15, 2020 at 4:44 PM Post #14,141 of 15,839

dazzerfong

1000+ Head-Fier
Joined
Jul 5, 2014
Posts
1,396
Likes
650
Location
Sydney
That will only determine correlation between two variables of the hypothesis testing is isolated and done as the sole method of verifying existence of causation. Repeatability is extremely important, specially with different sample sizes, test conditions, and the such.

Also, p=0.05 only takes account of +- <= 2 SD/ 5% SE (~95% CI), not being enough for fully isolate the presence of outliers in some stringent testing. Sample size is also to be considered, since binomial and t distributions tend to be closer to the normal distribution when large sample sizes are used and more accurately represent the cummulative probability distribution. With statistics, determining causation is a science and an art, because multiple testing has to be provided, being this repeatble and analyzed by external parties as well.

And that's why I said this only matters if you're trying to get stuff published - if so, it gets complicated very quickly. You will then have correlation coefficients to determine the level of repeatability, effect size to determine magnitude of said effect.

95% CI is the standard in general: you start with that and based on how confident you want to be, you adjust how many standard deviations you want to be within. Considering audio is hardly a crucial thing to get more than 95% confident, I think p = 0.05 for audibility is reasonable. For drugs, maybe not.

Throwing things at potential testers in a way to dissuade testing in my opinion is not the right way, especially if said things are statistical tricks. Start small and simple, and when you're intending to publish, up the rigour then. Pointless to do it before, especially if one's trying to see for themselves.
 
Last edited:
Oct 15, 2020 at 5:36 PM Post #14,142 of 15,839

KeithPhantom

100+ Head-Fier
Joined
Apr 3, 2020
Posts
398
Likes
221
Location
USA
And that's why I said this only matters if you're trying to get stuff published - if so, it gets complicated very quickly. You will then have correlation coefficients to determine the level of repeatability, effect size to determine magnitude of said effect.

95% CI is the standard in general: you start with that and based on how confident you want to be, you adjust how many standard deviations you want to be within. Considering audio is hardly a crucial thing to get more than 95% confident, I think p = 0.05 for audibility is reasonable. For drugs, maybe not.

Throwing things at potential testers in a way to dissuade testing in my opinion is not the right way, especially if said things are statistical tricks. Start small and simple, and when you're intending to publish, up the rigour then. Pointless to do it before, especially if one's trying to see for themselves.
It was just clarifying because scientific testing can be too strict at times for the regular John. And I agree, he should stick to something simple that works.
 
Oct 15, 2020 at 5:54 PM Post #14,143 of 15,839

dazzerfong

1000+ Head-Fier
Joined
Jul 5, 2014
Posts
1,396
Likes
650
Location
Sydney
It was just clarifying because scientific testing can be too strict at times for the regular John. And I agree, he should stick to something simple that works.

That's why scientific testing is in the realm of scholars and academics, and everyone else has everything else. I mean, hell, if you're doing it for your own benefit, loosen up the confidence intervals for all I care - not exactly proclaiming it as fact to anyone else.

I think the best compromise between the realm of publications and the average Joe is binomial probability for ABX. Simple enough to grasp, and no real calculations to do besides looking up a table.
 
Oct 15, 2020 at 7:19 PM Post #14,144 of 15,839

old tech

500+ Head-Fier
Joined
Jul 8, 2015
Posts
872
Likes
440
Location
Sydney, Australia
When I do my testing (between equipment or to determine different masters) I keep it simple. Level matched and double-blind with a couple friends or colleagues. Not sufficient for a peer reviewed paper but good enough for me. It's amazing the amount of times the test exposes the perception tricks of the brain.

Btw, most of my needle-drops were done in the 1990s, making 16/44 CD versions of my LPs. This was before the days of consumer software and were done through a colleague in a recording studio using my turntable. To date, no-one has been able to reliably pick the CDs apart from the LPs when playing on the same turntable.
 
Last edited:
Oct 16, 2020 at 4:13 AM Post #14,145 of 15,839

hakunamakaka

500+ Head-Fier
Joined
Mar 30, 2020
Posts
749
Likes
375
Location
Lithuania, Vilnius
When I do my testing (between equipment or to determine different masters) I keep it simple. Level matched and double-blind with a couple friends or colleagues. Not sufficient for a peer reviewed paper but good enough for me. It's amazing the amount of times the test exposes the perception tricks of the brain.

Btw, most of my needle-drops were done in the 1990s, making 16/44 CD versions of my LPs. This was before the days of consumer software and were done through a colleague in a recording studio using my turntable. To date, no-one has been able to reliably pick the CDs apart from the LPs when playing on the same turntable.


Turntable setup can change the sound a lot. Good phono preamp itself can create such a large difference that you may think that you got different set of speakers. If you are aiming for very clean, hi-res sound don't even bother with turntables, but if that's not the case, good turntable can shift your music to a new extent especially with instrumental and vocal centric music(brazilian,afro,jazz, organic grooves, opera etc..)
 

Users who are viewing this thread

Top