AES meta analysis on audibility of hi-res

[hvbias]

Paper freely available: http://www.aes.org/tmpFiles/elib/20160629/18296.pdf

 

[amirm]

Thanks! I read through parts of it and it is very interesting. I will read through and summarize later.

 

[krabapple]

As with every such report of a 'small but statistically significant' effect (like Meridien's) that increases only with 'extensive training', a question to keep forefront in mind is, how does this effect compare with the claims for audibility of hi-rez commonly, and loudly, made by and championed by audiophiles, listening sighted, with typically 'self-training' at best, to all manner of recordings labelled 'hi rez' even if they have no 'hi rez' content?

Another question is, how sure can we be that the cause is indeed *hi rez* itself? Particularly here, a meta-analysis, where data from various and setups are crunched? The author himself quite rightly admits that the causes of this 'perceived effect' are unknown.

 

[amirm]

As with every such report of a 'small but statistically significant' effect (like Meridien's) that increases only with 'extensive training', a question to keep forefront in mind is, how does this effect compare with the claims for audibility of hi-rez commonly, and loudly, made by and championed by audiophiles, listening sighted, with typically 'self-training' at best, to all manner of recordings labelled 'hi rez' even if they have no 'hi rez' content?

We tested audiophiles against our trained listeners at Microsoft and with super rare exceptions, they did poorly and no different than average public. So even if formal studies show ability distinguish between formats, I am afraid most audiophiles will have trouble doing so which I think is your point.

 

[BobShermanEsq]

We tested audiophiles against our trained listeners at Microsoft and with super rare exceptions, they did poorly and no different than average public. So even if formal studies show ability distinguish between formats, I am afraid most audiophiles will have trouble doing so which I think is your point.

How did you determine these people were audiophiles, and how were the trained listeners trained?

 

[amirm]

How did you determine these people were audiophiles, and how were the trained listeners trained?

By what they ate for lunch. The trained ones had hot dogs and the audiophiles, shrimp.....

 

[dallasjustice]

I didn't see this thread. I started another one. Sorry.

I think it's interesting that the author says more study could be done regarding how filter and equipment can influence the result. I think it would interesting to know whether offline upsampling has an effect and if so whether that effect is different from genuine hi-res recordings.

Meyer and Moron took it on the chin.

 

[BobShermanEsq]

By what they ate for lunch. The trained ones had hot dogs and the audiophiles, shrimp.....

Exactly what I thought. Good try though... Might have been better results if they had healthier diets.

 

[amirm]

Meyer and Moron took it on the chin.

They should have. It was amature hour.

 

[Phelonious Ponk]

Trained listeners are, I believe, trained to listen for specific artifacts. Of course if that's true, it has little to do with listening to music. And whatever it is that differentiates hi rez from CD...do we even know what it is? That it's a good thing? I guess it doesn't matter much with the world of recorded music completely dominated by CD-quality and less. Perhaps we should be thankful if we can't hear it.

Tim

 

[amirm]

Trained listeners are, I believe, trained to listen for specific artifacts.

It is that but it is also listening for lost detail, high frequency accentuation, pre-echo (harshness of transients), ability to use tools to isolate potential problem areas, etc. It is the difference between a driver and a mechanic listening to engine noise.

Note that this is specific to this domain. Training for room/speakers is a different animal (linear versus non-linear effects).

 

[Cosmik]

I am currently listening to a pair of DSP active speakers that I have been setting up 'from first principles'. The sound with 44.1/16 is rich, gorgeous, colourful, detailed, with a rock solid soundstage, deep bass etc. etc. - and I am just an idiot; I presume that Kii Threes or Beolab 90s would wipe the floor with even this sound. I can't help but think that everything regarding hi res, MQA, apodising filters, de-blurring etc. which is then played through a single amp with passive speakers is getting the priorities rather arse about tit.

 

[krabapple]

I didn't see this thread. I started another one. Sorry.

I think it's interesting that the author says more study could be done regarding how filter and equipment can influence the result. I think it would interesting to know whether offline upsampling has an effect and if so whether that effect is different from genuine hi-res recordings.

Meyer and Moron took it on the chin.

I hope that's a typo

Reiss devotes a whole section this paper to M&M 2007. He's not the first to have at it -- and it's interesting how M&M 2007 is seen as the finding that *must* be taken down -- but the critiques against M&M are often incredibly disingenuous. Example: audiophiles were forever touting the great sound of SACDs and DVDA's, regardless of whether they were 'pure' hi rez (e.g., source was an old analog tape recording; mastering involved lowpass filter; etc). And they were forever attributing that great sound to the fact that the SACD and DVDA are 'hi rez'. So it's a bit rich to move the goalposts and discount M&M results because they didn't use only 'pure' hi rez releases.
(M&M themselves came to a wholly reasonable conclusion: when 'better sound' is reported compared to a previous release, it could well have been due to different mastering -- better sourcing, different EQ, compression etc -- rather than higher sample/bit rates per se)

Reiss also offers this gem, which takes that goalpost-moving to further heights: "the encoding scheme on SACD obscures frequency components above 20 kHz and the SACD players typically filter above 30 or 50 kHz". So now DSD itself is disqualified? Hilarious!

M&M took audiophiles at their word, and tested *their* claim, which is: the audible difference between Redbook and hi rez isn't just real, it's *obvious* (to audiophiles). A claim that Reiss's metaanalysis as well indicates is nonsense.




.

 

[krabapple]

They should have. It was amature hour.

Yet how much ink has been spilled desperately trying to bury their (almost certainly true) results since 2007?

Face it Amir; 'hi rez' is basically a marketing scam, at the consumer end. 99.9% of reports -- including every single one touting 'OMG night and day' difference, a la TAS/Stereophile -- are simply bias/mastering choices at work.

 

[amirm]

Yet how much ink has been spilled desperately trying to bury their (almost certainly true) results since 2007?

Bad science is bad science Steven. If we don't dismiss it even if the outcome is to our liking, we lose the whole mission. In law, bad evidence is thrown out even if everyone thinks the defendant is guilty. If we were running a drug trial and the people running it didn't bother to make sure that the control was the control and the drug was the drug, no respected journal would accept their paper. Yet the president of AES allowed this to get published in their Journal. We need to speak out about that so that it doesn't cheapen the entire universe of research published in J AES.

Now if they had just published it in their user group report, that would have been fine. But it should have never risen to the J AES and folks who reviewed that work should have been more trained in the field, as young as it was then.

 

[Liberal Democracy]

Bad science is bad science Steven. If we don't dismiss it even if the outcome is to our liking, we lose the whole mission. In law, bad evidence is thrown out even if everyone thinks the defendant is guilty. If we were running a drug trial and the people running it didn't bother to make sure that the control was the control and the drug was the drug, no respected journal would accept their paper. Yet the president of AES allowed this to get published in their Journal. We need to speak out about that so that it doesn't cheapen the entire universe of research published in J AES.

Now if they had just published it in their user group report, that would have been fine. But it should have never risen to the J AES and folks who reviewed that work should have been more trained in the field, as young as it was then.

Weird science don't always turn out bad...,

 

[BobShermanEsq]

Bad science is bad science Steven. If we don't dismiss it even if the outcome is to our liking, we lose the whole mission. In law, bad evidence is thrown out even if everyone thinks the defendant is guilty. If we were running a drug trial and the people running it didn't bother to make sure that the control was the control and the drug was the drug, no respected journal would accept their paper. Yet the president of AES allowed this to get published in their Journal. We need to speak out about that so that it doesn't cheapen the entire universe of research published in J AES.

Now if they had just published it in their user group report, that would have been fine. But it should have never risen to the J AES and folks who reviewed that work should have been more trained in the field, as young as it was then.

It is refreshing that you agree people should speak out when they see bad science.

 

[Liberal Democracy]

It is refreshing that you agree people should speak out when they see bad science.

Bob, if you keep grinding that axe there will be no blade left to really hurt us with

 

[BobShermanEsq]

Bob, if you keep grinding that axe there will be no blade left to really hurt us with

I was just trying to saying something nice Tom.

 

[Liberal Democracy]

I was just trying to saying something nice Tom.

I detected a hint of sarcasm Bob, but forgive me if I was wrong following the well established narrative you have cast both here and else where I think it was a fair assumption on my part.

Still it's all rather either way.