or Connect
Head-Fi.org › Forums › Equipment Forums › Sound Science › Why would 24 bit / 192 khz flac sound any better than 16 bit / 44.1 khz flac if both are lossless (if at all)?
New Posts  All Forums:Forum Nav:

Why would 24 bit / 192 khz flac sound any better than 16 bit / 44.1 khz flac if both are lossless (if at all)? - Page 7

post #91 of 385
Quote:
Originally Posted by ralphp@optonline View Post
 

I think that you misunderstood me. One of the reasons that people claim that recordings having a sampling rate above 44.1kHz are needed is that some people can perceive energy/sound above the 20khz threshold and by having a recording with only frequencies above 20kHz would show that this is either true or false. My guess is that no human will be able to hear these frequencies, which is not something the people pushing high resolution would like to have proved.

 

Next up: the myth about higher sampling rates producing smoother, i.e. more accurate, sine waves at all frequencies but especially at or near 20kHz. Oh wait that one's probably been bandied about in the High-end Audio Forum section. I forgot that this is the Sound Science section.


My argument against this test is that it may not be representative of real-world acoustic content.  For instance you can you high frequency sine waves all you want, this testing may be limited in use. Basically, it's a similar type of argument that we are not sensitive to the specific phase of the sine wave, but it doesn't mean that phase response is not important.

post #92 of 385
Quote:
Originally Posted by riverlethe View Post


Then why state it in a public forum? :P

Seriously, what sort of test did you do?

 

As I mentioned in my previous post I compared 16/44 to 24/48. I used same quality 24/96 source with large dynamic range to generate 16/44, 16/44 dithered, 24/48 and  16/44 upscaled to 24/48. I used decent quality conversion software with floating point resolution so the conversion noise is well below the 24 bit noise floor.   I did 24/48 so that I go through a similar conversion process from 24/96 as 16/44 (not identical, of course). The upscaling test is primarily to circumvent any issue that *may* arise with my DAC handling of different sampling rates (in other words, it also goes through similar conversion process rather than introduce one by the DAC that I can't quantify).

post #93 of 385
Quote:
Originally Posted by Digitalchkn View Post
 

 

As I mentioned in my previous post I compared 16/44 to 24/48. I used same quality 24/96 source with large dynamic range to generate 16/44, 16/44 dithered, 24/48 and  16/44 upscaled to 24/48. I used decent quality conversion software with floating point resolution so the conversion noise is well below the 24 bit noise floor.   I did 24/48 so that I go through a similar conversion process from 24/96 as 16/44 (not identical, of course). The upscaling test is primarily to circumvent any issue that *may* arise with my DAC handling of different sampling rates (in other words, it also goes through similar conversion process rather than introduce one by the DAC that I can't quantify).

 

Was the test sighted, blind, ?  What software did you use?  How many iterations?  How did you volume match?

 

Genuinely interested in your methodology - of the actual test itself.

 

For the record - I have performed similar - using ABX software (so completely blind) - multiple iterations (usually 15-20 attempts at a time).  Tracks are volume matched to 0.1 dB.

 

Test results show I can't tell the difference - however my ears are reasonably aged (47), and I know my hearing is a lot less than perfect (one of the drawbacks of age ;)

 

In the past when we've asked others to do this - the people who have actually performed proper blind abx tests haven't been able to tell differences either.  The ones that claim to be able to tell are usually the ones who will only do it sighted, and who won't produce abx logs sadly.

 

Please note - I'm not trying to spark a war here.  I am genuinely interested.

post #94 of 385
Quote:
Originally Posted by nanaholic View Post
 

 

It doesn't really help to have all the high quality gear in the world when the music industry goes back to being lazy/careless and making poor masters.  No amount of high-end gear can fix clipping for example, so all those songs that are mixed too hot is ruined forever and no file format or bitrate will fix it.

That happened/happens a lot. Many producers don't know or don't care that when recording at levels of 0dB or over it on digital equipment results in distortion (clipping). Analog equipment is more forgiving.

post #95 of 385
Quote:
Originally Posted by Brooko View Post
 

 

Was the test sighted, blind, ?  What software did you use?  How many iterations?  How did you volume match?

 

Genuinely interested in your methodology - of the actual test itself.

 

For the record - I have performed similar - using ABX software (so completely blind) - multiple iterations (usually 15-20 attempts at a time).  Tracks are volume matched to 0.1 dB.

 

Test results show I can't tell the difference - however my ears are reasonably aged (47), and I know my hearing is a lot less than perfect (one of the drawbacks of age ;))

 

In the past when we've asked others to do this - the people who have actually performed proper blind abx tests haven't been able to tell differences either.  The ones that claim to be able to tell are usually the ones who will only do it sighted, and who won't produce abx logs sadly.

 

Please note - I'm not trying to spark a war here.  I am genuinely interested.

 

I am a bit younger (35) but I can tell my hearing isn't what it used to be either.

 

I used the ABX foobar2000 plug in to do double blind tests I think maybe 10 comparisons. The volumes were kept identical -- the difference is only related to the sampling rate/bit depth conversion.

 

I don't recall exactly my source material but remember chosing some acoustic material with wide spectrum (harmonics in some passages close to 30KHz), large dynamic range, (read: little or no compression) and very pronounced 3D soundfield (in other words, not point sources with synthesized reverb but naturally captured acoustics).

 

The effect was that more of 3D perception of the soundstage. In other words, the 16/44 slightly flattened the soundstage.  24/48 made it more 3-dimensional.  The difference is very subtle, of course.

 

On the material I used, the difference between "truncated" 16/44 and 24/48 was more noticeable.The dithered version of 16/44 was much more of a challenge, just on the border of being barely noticeable.  Note that both versions were upscaled first.

 

I didn't test 16/44 vs 24/96 in this case because I thought that to be an unfair comparison.

post #96 of 385
Quote:
Originally Posted by Digitalchkn View Post

I am a bit younger (35) but I can tell my hearing isn't what it used to be either.

I used the ABX foobar2000 plug in to do double blind tests I think maybe 10 comparisons. The volumes were kept identical -- the difference is only related to the sampling rate/bit depth conversion.

I don't recall exactly my source material but remember chosing some acoustic material with wide spectrum (harmonics in some passages close to 30KHz), large dynamic range, (read: little or no compression) and very pronounced 3D soundfield (in other words, not point sources with synthesized reverb but naturally captured acoustics).

The effect was that more of 3D perception of the soundstage. In other words, the 16/44 slightly flattened the soundstage.  24/48 made it more 3-dimensional.  The difference is very subtle, of course.

On the material I used, the difference between "truncated" 16/44 and 24/48 was more noticeable.The dithered version of 16/44 was much more of a challenge, just on the border of being barely noticeable.  Note that both versions were upscaled first.

I didn't test 16/44 vs 24/96 in this case because I thought that to be an unfair comparison.

Ten comparisons of 16/44 "truncated" vs 24/48 and 16/44 dithered vs 24/48? And you got 7/10 right in each run?

Here's a study with dozens of participants and 554 trials:

http://www.mixonline.com/recording/mixing/audio_emperors_new_sampling/
post #97 of 385

It would be most helpful to post some samples (with a maximum length of 30 seconds) of the actual files that were compared with the ABX plugin, and for converted files preferably also the original version. This way, others can try to reproduce the results with ABX testing, and/or find potential errors (like bad conversion, unmatched levels, etc.) easier.

post #98 of 385
Quote:
Originally Posted by riverlethe View Post


Ten comparisons of 16/44 "truncated" vs 24/48 and 16/44 dithered vs 24/48? And you got 7/10 right in each run?

Here's a study with dozens of participants and 554 trials:

http://www.mixonline.com/recording/mixing/audio_emperors_new_sampling/

 

Good enough for me to be convinced. Not planning to release a report to AES anytime soon.

post #99 of 385
Quote:
Originally Posted by Digitalchkn View Post

Good enough for me to be convinced. Not planning to release a report to AES anytime soon.

Well, perhaps you'll forgive me if I'm not the least bit impressed.
post #100 of 385
Quote:
Originally Posted by riverlethe View Post


Ten comparisons of 16/44 "truncated" vs 24/48 and 16/44 dithered vs 24/48? And you got 7/10 right in each run?

Here's a study with dozens of participants and 554 trials:

http://www.mixonline.com/recording/mixing/audio_emperors_new_sampling/

 

Thanks.  Really enjoyed reading that article.

post #101 of 385
Quote:
Originally Posted by riverlethe View Post


Well, perhaps you'll forgive me if I'm not the least bit impressed.

 

You are most certainly forgiven.  It's a test for me, in my home setup, under reasonably know conditions, with best that I have available. It's not meant to be a be-all, end-all universal conclusion to the age-old question of "24 bits vs 16 bits".  The objective is not for me to convince the rest of the world... I get no finance gain from it.

post #102 of 385
Quote:
Originally Posted by Digitalchkn View Post

 

I am a bit younger (35) but I can tell my hearing isn't what it used to be either.

 

I used the ABX foobar2000 plug in to do double blind tests I think maybe 10 comparisons. The volumes were kept identical -- the difference is only related to the sampling rate/bit depth conversion.

 

I don't recall exactly my source material but remember chosing some acoustic material with wide spectrum (harmonics in some passages close to 30KHz), large dynamic range, (read: little or no compression) and very pronounced 3D soundfield (in other words, not point sources with synthesized reverb but naturally captured acoustics).

 

The effect was that more of 3D perception of the soundstage. In other words, the 16/44 slightly flattened the soundstage.  24/48 made it more 3-dimensional.  The difference is very subtle, of course.

 

On the material I used, the difference between "truncated" 16/44 and 24/48 was more noticeable.The dithered version of 16/44 was much more of a challenge, just on the border of being barely noticeable.  Note that both versions were upscaled first.

 

I didn't test 16/44 vs 24/96 in this case because I thought that to be an unfair comparison.

 

Thanks.  I have to take you at your word - but as Steve requested, some of the samples you used would be nice to look at.  I'll remain skeptical (basically you appear to be one "anomaly" amongst the rest of us ordinary folk - and I don't mean that to sound derogatory).

 

Can I suggest (in all seriousness) that you contact someone like Ethan Winer (who we see from time to time on these forums) to set up another test?  Assuming you can do this consistently and that there is nothing wrong with the files or testing method - this could be important actually establishing that there is an audible difference.  You would be helping a far wider community.  Plus I for one would be fascinated knowing that it can actually be done.

 

It would also make Neil Young's current crusade a lot more believable.

post #103 of 385
Quote:
Originally Posted by Brooko View Post
 

 

Thanks.  I have to take you at your word - but as Steve requested, some of the samples you used would be nice to look at.  I'll remain skeptical (basically you appear to be one "anomaly" amongst the rest of us ordinary folk - and I don't mean that to sound derogatory).

 

Can I suggest (in all seriousness) that you contact someone like Ethan Winer (who we see from time to time on these forums) to set up another test?  Assuming you can do this consistently and that there is nothing wrong with the files or testing method - this could be important actually establishing that there is an audible difference.  You would be helping a far wider community.  Plus I for one would be fascinated knowing that it can actually be done.

 

It would also make Neil Young's current crusade a lot more believable.

 

I surprised this is generating so much interest.  Unfortunately I didn't keep a very good record of the testing when I did this few months back (I don't even remember exactly when I did it).

It would be interesting to see this kind of test on a bigger scale.

post #104 of 385
Quote:
Originally Posted by Digitalchkn View Post

I surprised this is generating so much interest.  Unfortunately I didn't keep a very good record of the testing when I did this few months back (I don't even remember exactly when I did it).
It would be interesting to see this kind of test on a bigger scale.

Well, your claim probably wouldn't qualify for the $1 million JREF prize, but it certainly contradicts the conventional wisdom. The article I linked to talks about just such a test. I don't think any others have been done since then.
post #105 of 385
Quote:
Originally Posted by Digitalchkn View Post
 

 

I surprised this is generating so much interest.  Unfortunately I didn't keep a very good record of the testing when I did this few months back (I don't even remember exactly when I did it).

It would be interesting to see this kind of test on a bigger scale.

 

Check that link that Riverlethe left (http://www.mixonline.com/recording/mixing/audio_emperors_new_sampling/)

 

This sort of testing has been done on a large scale.  That's why I was so interested in your result.  If you can consistently identify the differences using different samples - then your results are important.  But it would have to be redone with independent verification (which is why I suggested contacting Ethan).  At the very least - contact him.  I'd imagine he would be interested .......

 

When was the last time you tested?

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › Why would 24 bit / 192 khz flac sound any better than 16 bit / 44.1 khz flac if both are lossless (if at all)?