AIX Records High Rez Test Results

[amirm]

What about those who scored zero? Maybe they have golden ears but preferred compressed to hi-res, thinking that their preference was linked to higher fidelity?

I thought the test was to determine what was high-res. To the extent they got that wrong for whatever reason, then we don't need to test them again.

 

[svart-hvitt]

I thought the test was to determine what was high-res. To the extent they got that wrong for whatever reason, then we don't need to test them again.

If I knew a guy who could guess the colour where the roulette ball falls consistently wrong, I would have a gold mine.

 

[Blumlein 88]

Yes, statistics is hard for some. I agree if you wonder about the 6 for 6 people give them a second chance. But we really don't need to do that. His results are pretty much within expectations for random chance. And yes breaking out the 0 of 6 from those who simply said, I can't hear a difference would be better. Those who didn't hear a difference would actually be expected to average 3 of 6 correct had they answered.

In any case I quickly set up a spread sheet which has 83 lines with 6 choices each. A random number generator decides whether a choice is correct or not. In three runs of that the number of 6 for 6 results were 1, 3 and 1. Number of 0 correct was 1, 1, and 0. Number of 5 of 6 were 7,9, and 8. The number of 1 of 6 was 3, 4, and 10. The middle with 2, 3, or 4 correct is where everything clustered. Out of 83 people sampled this way you would expect a couple or three every time to get all or get none. And outlier results of a few more will happen every so often.

His results show no reason to believe that hirez is audible. Why that is, and whether different conditions would show something can of course always be debated. Especially with online tests like this. The listening gear is completely uncontrolled. It is like I've said before however, if the difference in SD and HD sound was half of the difference in SD and HD video there would be no disagreement at all. The fact this far down the road the answer is ambiguous means at best, let me repeat AT BEST, redbook vs high resolution is a very small, marginal difference that will show up only under the most special circumstances, and even then is small, like a really small difference. So just about anything else matters more. Actual differences sometimes found in redbook vs higher resolution recordings always ends up being because of different mastering. If you stop with the different mastering it simply doesn't matter.

 

[TBone]

I find it incredibly silly, funny and quite sad, that many vinyl experts claim their rips are better represented using higher rez formats.

My advice ... if you can't make your rips sing @16/44 - 24 bits won't "fix" that problem.

 

[Fitzcaraldo215]

I agree hirez vs. RBCD differences are small, but not necessarily of the same magnitude in all cases. Here are some possible scenarios for a comparison of hirez vs. RBCD playback:

1. Native hirez recording played in hirez vs. downsampled to RBCD -

This was Waldrep's test. I believe differences will be very small and detection difficult, as the results suggest.

2. Native RBCD recording upsampled to hirez vs. straight RBCD playback -

I expect small to nonexistent differences, likely even smaller than 1, above. This was not Waldrep's test.

3. Native Analog, presumably tape, recording mastered to hirez vs. independently mastered to RBCD -

Again, I expect small to nonexistent differences as limited primarily by the source material. However, I doubt this happens often, if ever. Usually, if the 2 versions are simultaneously released, the conversion/remastering is done in hirez then downsampled to RBCD, as in 1. Or, the hirez and RBCD versions were done independently at different times, and it is likely there may be differences due to conversion/mastering. Again, this was not Waldrep's test.

4. Native hirez recording played as hirez vs. simultaneous Native RBCD recording played as RBCD, both from the same live session -

Again, not Waldrep's test. This test would be difficult to impossible, as there are two simultaneous masters of the original session material with possible differences other than just sampling rate and bit depth. But, it brings in possible additional sonic benefits from a hirez recording production chain vs. an RBCD chain. Mainly, that might be the A-D in hirez vs. RBCD using full range source material, rather than possibly more limited source material, as in 2. and 3., above.

I have never seen a test like this formally published or comparison samples of it. Anybody? My hypothesis is that this, if properly doable, would exhibit a larger difference than any of the above, including 1., though still likely a small difference, but perhaps a more noticeable one.

I also believe some recording engineers likely have done this, at least informally or partially, including Waldrep. Some of those have decided for sonic or marketing reasons to adopt hirez as a primary distribution format in spite of the marketing and distribution difficulties and niche status it imposes.

Other engineers remain loyal to RBCD as a distribution format, but they see technical and sonic advantage to recording in hirez then downsampling to RBCD as their finished product. Sonically, it would appear that that approach closes any noticeably audible performance gap vs. hirez.

 

[Thomas savage]

I can’t understand the hirez debate, unless there’s some jeopardy in using hirez for playback we can just use hirez for playback or whatever Rez it was mastered in.

There’s unlikely to be much difference, indeed it’s probably the very very last thing to concern yourself with, generally someone’s already likely buggered up the recording way before you get you hands on your desired format so..., who cares.

If there’s a cost difference, buy the one you can afford..

Much more intresting is the audibility of codecs, how good you can make it when you start cutting file size right down. The science and reasearch involved there must be really intresting .

 

[TBone]

Any such noted differences that do exist, if they exist at all, between redbook and higher rez, from native to tape copies, is totally moot.

True native high rez material is rare, and even if it wasn't, mastering influences still anchor any perceived SQ. One could even claim that DSD started out as a scam ... remember when the labels dynamically tilted the results in some of those early dual SACD/CDs comparisons, compressing the CD layer to a greater extent than the SACD layer. Even with that, many people couldn't hear the difference.

 

[Fitzcaraldo215]

I find it incredibly silly, funny and quite sad, that many vinyl experts claim their rips are better represented using higher rez formats.

My advice ... if you can't make your rips sing @16/44 - 24 bits won't "fix" that problem.

There might possibly be some slight advantages to A-D in hirez vs. 16/44k, as JA discusses in the September Stereophile. But, agreed, none in frequency response or S/N for vinyl rips.

 

[TBone]

There might possibly be some slight advantages to A-D in hirez vs. 16/44k, as JA discusses in the September Stereophile. But, agreed, none in frequency response or S/N for vinyl rips.

I use a professional Sony unit that records at 24 bits then truncates down to 16 bits and/or utilizes SBM (Sony noise shaping scheme). SBM is suppose to provide ~20bit resolution to redbook, but I generally avoid the feature.

Sometimes, I think some reviewers record into DSD as to NOT easily share their rips when asked, but that's another rant altogether ...

 

[Fitzcaraldo215]

Any such noted differences that do exist, if they exist at all, between redbook and higher rez, from native to tape copies, is totally moot.

True native high rez material is rare, and even if it wasn't, mastering influences still anchor any perceived SQ. One could even claim that DSD started out as a scam ... remember when the labels dynamically tilted the results in some of those early dual SACD/CDs comparisons, compressing the CD layer to a greater extent than the SACD layer. Even with that, many people couldn't hear the difference.

Well, yes, true hirez is rare in popular genres, jazz, etc. It is still only a niche in classical, but much better represented. Over the past 15 years or so, many new sessions were recorded directly in hirez, particulary on SACD, many in Mch. As I keep saying, I have thousands. Kal does, too.

I think what really happened with some pop SACDs is that they took a popular oldie, long available on CD. Then, they just copied the old CD release to the CD layer and remastered the hirez DSD layer from the session tapes, which were analog or RBCD. So, the mastering was different - better, worse or, often, undetectably - between the two.

That doesn't really seem to happen with classical releases nearly as much, although I generally avoid remasters in the hirez albums I buy, since I prefer Mch. The Mch formats are hirez, even Bluray's common 48k/24. So, naturally, I play them in hirez. No problem. I don't really need to worry about the question of whether RBCD is "just as good".

 

[restorer-john]

In any case I quickly set up a spread sheet which has 83 lines with 6 choices each. A random number generator decides whether a choice is correct or not. In three runs of that the number of 6 for 6 results were 1, 3 and 1. Number of 0 correct was 1, 1, and 0. Number of 5 of 6 were 7,9, and 8. The number of 1 of 6 was 3, 4, and 10. The middle with 2, 3, or 4 correct is where everything clustered.

Ah, a spreadsheet guru. Just what I need. I shall send you a PM.

 

[Blumlein 88]

snippage...........

4. Native hirez recording played as hirez vs. simultaneous Native RBCD recording played as RBCD, both from the same live session -

Again, not Waldrep's test. This test would be difficult to impossible, as there are two simultaneous masters of the original session material with possible differences other than just sampling rate and bit depth. But, it brings in possible additional sonic benefits from a hirez recording production chain vs. an RBCD chain. Mainly, that might be the A-D in hirez vs. RBCD using full range source material, rather than possibly more limited source material, as in 2. and 3., above.

I have never seen a test like this formally published or comparison samples of it. Anybody? My hypothesis is that this, if properly doable, would exhibit a larger difference than any of the above, including 1., though still likely a small difference, but perhaps a more noticeable one.

I also believe some recording engineers likely have done this, at least informally or partially, including Waldrep. Some of those have decided for sonic or marketing reasons to adopt hirez as a primary distribution format in spite of the marketing and distribution difficulties and niche status it imposes.

Other engineers remain loyal to RBCD as a distribution format, but they see technical and sonic advantage to recording in hirez then downsampling to RBCD as their finished product. Sonically, it would appear that that approach closes any noticeably audible performance gap vs. hirez.

Yes there has been a test like this. Concurrent recording in 88/24 and 44/16. Offerings to university students in a blind test were 88 vs 44, 88 vs downsampled to 44. A result for 88 vs downsampled to 44 was positive for a difference. However, 88 vs concurrent 44 gave null results. The conclusion one might come to is the downsampling was audible.

 

[derp1n]

Which only tells you they used a bad resampler.

 

[andreasmaaan]

Yes there has been a test like this. Concurrent recording in 88/24 and 44/16. Offerings to university students in a blind test were 88 vs 44, 88 vs downsampled to 44. A result for 88 vs downsampled to 44 was positive for a difference. However, 88 vs concurrent 44 gave null results. The conclusion one might come to is the downsampling was audible.

Interesting, do you have a link?

 

[Kees de Visser]

Interesting, do you have a link?

AFAIK this is not the same test/study, but it's an interesting (2007-AES) paper about the lo-res/high res comparison to a live analogue source.
http://www.extra.research.philips.com/hera/people/aarts/RMA_papers/aar07pu4.pdf

 

[andreasmaaan]

AFAIK this is not the same test/study, but it's an interesting (2007-AES) paper about the lo-res/high res comparison to a live analogue source.
http://www.extra.research.philips.com/hera/people/aarts/RMA_papers/aar07pu4.pdf

Strange study though, not sure if much can be drawn from the results. What seems particularly weird to me was that apparently they only used the super tweeter for playback of material containing HF content. Is that how your read it?

In condition C1, Sanken CO-100k microphones (flat to 100 kHz) were used to capture the sources, and ribbon super-tweeters to reproduce them. In condition C2, measurement microphones limited to 20 kHz, and standard studio loudspeakers were used, with super-tweeters not active.

Addition of a speaker driver, even if executed very well, will result in erratic off-axis frequency response at the crossover (the higher the crossover point, the more erratic this behaviour will be). Surely this introduced major audible variables that were not controlled for. Or did I misunderstand the experiment?

Also seems strange to ask for a subjective evaluation of which source is most like the reference, without testing to make sure that subjects can actually discern the sources from the reference and each other first.

I'm not much good at statistics. Can anyone explain how statistically significant these results are?

 

[bakker_be]

Why don't we all do our own test? That's what I'm planning to do now I've got a playback chain allowing almost everything to be played "as is". I've downloaded the different versions of "Vision of Her" from https://www.oppodigital.com/hra/dsd-by-davidelias.aspx except for the MQA and using the Foobar2000 ABX-component these will be sent to my 50-year old, (well protected) concert-going ears through Topping D10 --> Topping A30 --> Focal Elear. I'll do 10 iterations for each comparison and present the results here

 

[andreasmaaan]

Why don't we all do our own test? That's what I'm planning to do now I've got a playback chain allowing almost everything to be played "as is". I've downloaded the different versions of "Vision of Her" from https://www.oppodigital.com/hra/dsd-by-davidelias.aspx except for the MQA and using the Foobar2000 ABX-component these will be sent to my 50-year old, (well protected) concert-going ears through Topping D10 --> Topping A30 --> Focal Elear. I'll do 10 iterations for each comparison and present the results here

You need to know that the original contains super-22KHz content and that the different available versions are downsampled from the same hi-res Master using appropriate conversions and filters though.

Best to do something like this with your own recording and downsampling/filtering unless you know exactly what happened in the studio.

 

[JJB70]

I have a lot of respect for Mark Waldrep, given his previous well known opinions on hi-res it displays a lot of integrity for him to publish these results and publicly accept them, especially his own admission. I really enjoy reading the articles on his website, never less than interesting and often genuinely fascinating and informative.

 

[NorthSky]

The guy is a sound explorer, recorder, mixer. He plays with microphones @ the source of the music playing to give us various perspectives of sound reproduction @ the other end of the music chain. He's a digital recording sound engineer.

Me too I have respect for people who explore with sounds and music; from anywhere in the chain...from the artist playing his instrument/voice to the listener playing his album/hires audio phile.