WBF Poll: Which Sounds Better, Digital or Analog?

[rbbert]

As an aside; 'P' values may not be as reliable as many assume. I found this article pretty interesting and sobering:

http://www.nature.com/news/scientific-method-statistical-errors-1.14700

Like any statistical test, it needs to be used appropriately (for the type of data analyzed and the purpose of the analysis). It should be well known by educated people that statistics can be manipulated and misapplied to demonstrate any number of things. Properly used, with its limitations understood, "p" value means a lot.

 

[amirm]

As an aside; 'P' values may not be as reliable as many assume. I found this article pretty interesting and sobering:

http://www.nature.com/news/scientific-method-statistical-errors-1.14700

Very good read and very true. I rant about the same thing in audio when people quote these numbers but they don't even know what they mean or how they are calculated! Here is the article I wrote on that: http://www.madronadigital.com/Library/High Resolution Audio/Statistics of ABX Testing.html

I reference a superb AES paper that digs deep into all the things that can generate invalid 95% confidence intervals and proposes new schemes to get reliable outcomes: "Statistical Analysis of ABX Results Using Signal Detection Theory," Jon Boley and Michael Lester, Presented at the 127th Convention 2009 October 9–12

They say a little knowledge is a dangerous thing. For all the talk about statistical analysis of ABX tests, we don't even seem to have that little bit of knowledge!

 

[jazdoc]

Very good read and very true. I rant about the same thing in audio when people quote these numbers but they don't even know what they mean or how they are calculated! Here is the article I wrote on that: http://www.madronadigital.com/Library/High Resolution Audio/Statistics of ABX Testing.html

I reference a superb AES paper that digs deep into all the things that can generate invalid 95% confidence intervals and proposes new schemes to get reliable outcomes: "Statistical Analysis of ABX Results Using Signal Detection Theory," Jon Boley and Michael Lester, Presented at the 127th Convention 2009 October 9–12

They say a little knowledge is a dangerous thing. For all the talk about statistical analysis of ABX tests, we don't even seem to have that little bit of knowledge!

Reminds me of Andrew Lang's classic insult "He uses statistics as a drunken man uses lamp-posts... for support rather than illumination."

 

[amirm]

Reminds me of Andrew Lang's classic insult "He uses statistics as a drunken man uses lamp-posts... for support rather than illumination."

That's one of the most clever lines I have heard .

 

[microstrip]

Like any statistical test, it needs to be used appropriately (for the type of data analyzed and the purpose of the analysis). It should be well known by educated people that statistics can be manipulated and misapplied to demonstrate any number of things. Properly used, with its limitations understood, "p" value means a lot.

The easiness of performing statistical analysis using software tools is a blessing and a curse - anyone can generate beautiful figures, even ignoring what is behind them. We often debate preference, that is intrinsically a statistical concept in audio, even if we debate just our own preference at a precise moment, and must be debated so. I have seen many SPSS users (Statistical Package for the Social Sciences is a well known software package often used by medical science researchers) who could not answer simple questions such as what is a probability distribution!

We should not mix the statistical limitations of data analysis (most of the time due to sample size) with the intrinsic limitations of the method of performing the tests, our big source of disagreement in WBF.

 

[amirm]

The easiness of performing statistical analysis using software tools is a blessing and a curse - anyone can generate beautiful figures, even ignoring what is behind them. We often debate preference, that is intrinsically a statistical concept in audio, even if we debate just our own preference at a precise moment, and must be debated so. I have seen many SPSS users (Statistical Package for the Social Sciences is a well known software package often used by medical science researchers) who could not answer simple questions such as what is a probability distribution!

Just to add to this, before I wrote my article, I thought for sure there would be dozens of bloggers, forums threads, etc. that would have talked about how statistics of ABX testing is computed. But I could not find a single one despite searching and searching. Even on HA forum where that is their core edict, there was not one post describing it. Worse yet, people were confusing confidence interval with percentage right answers! This is why I wrote that article. I think it is the only documentation on how the stats are computed and pitfalls in interpreting them.

 

[esldude]

Just to add to this, before I wrote my article, I thought for sure there would be dozens of bloggers, forums threads, etc. that would have talked about how statistics of ABX testing is computed. But I could not find a single one despite searching and searching. Even on HA forum where that is their core edict, there was not one post describing it. Worse yet, people were confusing confidence interval with percentage right answers! This is why I wrote that article. I think it is the only documentation on how the stats are computed and pitfalls in interpreting them.

Well I think using 3 sigma as a standard would be a step in the right direction. I would agree with Fisher's original idea, that 95% confidence levels indicates something worthy of further scrutiny. Until it reaches another SD however, I wouldn't get too bent out of shape about something. Meridian's test of their filters didn't quite reach that. Despite using sub-standard dither and much steeper than normal filtering. Still worth a second look, but not enough on its on in my estimation.

I agree general discussions of interpreting statistics of listening tests is rather stunted in the audio world. But the whole blind testing paradigm is stunted in the mainstream audio world.

I would surely like to see a group of 'high end' reviewers, show they can differentiate DACs in a statistically significant manner when the listening is not sighted.

 

[JackD201]

Just to add to this, before I wrote my article, I thought for sure there would be dozens of bloggers, forums threads, etc. that would have talked about how statistics of ABX testing is computed. But I could not find a single one despite searching and searching. Even on HA forum where that is their core edict, there was not one post describing it. Worse yet, people were confusing confidence interval with percentage right answers! This is why I wrote that article. I think it is the only documentation on how the stats are computed and pitfalls in interpreting them.

I asked a former member here (since banned) many times. I never got an answer.

 

[RadioWonder]

Audio memory form past years is an iffy slippery little bugger, but I believe I have surpassed my excellent analog vinyl system from the 1980's with my present computer based system...
I started listening to vinyl in the late 1950's and in later years got my first belt driven Rek-O-Kut Turntable and my last TT was a vacuum SOTA turntable...
Of course you fellows are way beyond minor things like that...

 

[RogerD]

There's nothing like a difference of opinion ^^ I use to think analog...now firmly digital.

 

[NorthStar]

I think the sound of analog is better. For the reason why we can read more at here What are the Beautiful Secrets Behind the Vinyl Records?

I hope when reading this article we will understand the reason why.

That was a fun read.

 

[morricab]

But that is precisely the question it does NOT ask. According to the question asked I find analog superior to digital in sound, from the experiences I personally have, but that is irrelevant for how I identify myself. I'm a digital-only guy. There you have it. (I did vote for digital, because that's where all the music is.)

But Al, the vote has nothing to do with "where all the music is". Besides there is more than enough good music on analog to last several lifetimes of continuous listening. So your statement is a bit nonsensical.

Amir, by his own admission he thinks analog sounds better so it is really 40 in favor of analog, Al just wants to bend the question in a very strange way to justify why he has no analog rig.

 

[Al M.]

But Al, the vote has nothing to do with "where all the music is". Besides there is more than enough good music on analog to last several lifetimes of continuous listening. So your statement is a bit nonsensical.

Amir, by his own admission he thinks analog sounds better so it is really 40 in favor of analog, Al just wants to bend the question in a very strange way to justify why he has no analog rig.

Well, this thread is two years old. Rather than indulging you with directly replying to your post, I will point out that now, two years later, I am not so sure anymore that analog in priniciple sounds better than digital. I wouldn't affirm the opposite either. There can be found examples in both media that are exemplary. One of my major complaints about digital, the reproduction of tenor and baritone saxophone, has ceased to be one after hearing digital that excels in it just like analog can.

I do, however, continue to hear analog reproduction that is transcendent. Just last night in Peter A.'s system, which I may report on at some point.

 

[Gregadd]

What does digital sound like? I have never heard it.?

 

[Gregadd]

For the reasons stated, i can't simply vote for what I perceive as 'utmost fidelity'. Other factors play into this. So no, this is no 'simple poll' for me, and just like for this, for most matters in life there are no simple polls anyway.

The question is also to which degree clicks and pops intrude into the perception of fidelity. Even the best turntables don't make them go away even though perhaps they may minimize their impact.

To the extent possible a good rcm dispatches "clicks and pops."

 

[treitz3]

A good RCM and steam cleaning of the LP's will get rid of an estimated 90-95% of the snaps, crackles and pops. Those that remain are at a DB level severely less than before, even on new LP's. Equipment for the playback reproductive effort aside, if you have the best going in? You can expect the best playback effort one can get, regardless of playback gear.

"You can't make chicken soup out of chicken s**t. (droppings)"

BTW, nice article Alex! Thank you sir.

Tom

 

[Don C]

Today the analog LP Vs digital debate is ridiculous.

You need both stereo LP for 1957 to mid 80's music, and digital for newer music.

Either can give state of art reproduction under the right conditions IMO, YMMV, etc.

 

[spiritofmusic]

Don, I'm a big fan of analog w 2.5k lps, but it's incorrect to say you NEED both
Tidal will take care of 90% of most people's needs
Now if Tidal didn't cover Golden Age catalog, I'd agree w you
But other than pretty obscure genres, Tidal pretty comprehensive
And as a massive convert to digital off Tidal thru a SOTA server like SGM, I'd have less issues relying on digital 100% if I had to

 

[Don C]

Don, I'm a big fan of analog w 2.5k lps, but it's incorrect to say you NEED both
Tidal will take care of 90% of most people's needs
Now if Tidal didn't cover Golden Age catalog, I'd agree w you
But other than pretty obscure genres, Tidal pretty comprehensive
And as a massive convert to digital off Tidal thru a SOTA server like SGM, I'd have less issues relying on digital 100% if I had to

I am not sure that Tidal streaming is superior to CD/SACD on a player.

Anyway, I feel it is best to playback older analog recordings Via LP, and newer digital recording Via CD/SACD, keeping the recordings in the original issued format with no conversions.

You need both to cover both the analog and digital music timeframes in the best light IMO.

I have 2K LPs and 500 CD/SACDs.

 

[PeterA]

Today the analog LP Vs digital debate is ridiculous.

You need both stereo LP for 1957 to mid 80's music, and digital for newer music.

Either can give state of art reproduction under the right conditions IMO, YMMV, etc.

This is true, I guess, if you want access to all of this music. Some people don't need to hear it all. I understand why some would want to optimize one format and focus on only that if it is their priority and they want to optimize their resources.

Don, have you ever directly compared the same recording on both LP and CD in the same system? If so, do you prefer one to the other. If so, can they both be considered state of the art?