Anyone actually done abx testing with hi res music?

[yepimonfire]

I've always been a critic of it, since 44.1khz can accurately reproduce all the way up to 22khz (especially with modern oversampling DACs), and no music has a dynamic range greater than the 96dB offered by 16bit audio. Out of sheer curiosity, I decided to purchase coldplay's a rush of blood to the head, Nirvana's never mind, and Mussorgsky's pictures at an exhibition, as performed by the mariinsky orchestra, all files were either 24/96 or 24/192. To make abxing simpler, I used my DAW to resample all files to 24/44.1 using a 512pt sinc, which is extreme overkill, and a much higher precision than used for down sampling final CD masters, then resampled them back to 24/96 with the same precision.

I loaded the files up for abxing in foobar, I picked "clocks" from the coldplay album, "the hut on fowl's legs", from pictures at an orchestra, and " in bloom" from never mind. I used foobar's wasapi output plugin, to ensure the audio to the receiver wasn't tainted by windows audio engine in any way.

With clocks, in a 16 trial test, I managed to get 13/16, the parts that gave it away were the cymbal's and snare, in bloom, 11/16, again, the percussion was the major difference, the hut on fowl's legs, 8/16. A second trial with the hut on fowls legs was performed, because at that point I figured the poor results were due to my ears being tired from back to back testing. The second test, I managed 14/16. The attack of the cymbals and drum again, gave it away. This is the same difference I noticed with Tidal's mqa streaming.

Curious as to what exactly might be the cause of the difference, I did a bit of researching. Apparently, 44.1khz is unable to properly capture or reproduce transients and impulses without audible temporal smearing, due to the fact that our ears are able to resolve a timing difference of as little as 6 ms. Yamaha wrote an article on the limitations of digital sampling, stating that in order to properly capture transients and impulses in the audible range, a minimum of 166khz was required (assuming a 6ms response). I managed to dig up a bunch more info, but I want to first hear from other's whether or not you've had a chance to abx test hi res, and if so, what were the results? I was very surprised, as I went into this thinking I'd probably get a 50% or less result, basically due to guessing.

Sent from my 5065N using Tapatalk

 

[everettT]

I'm more interested in listening than testing in high resolution music the most important is the original file recording. Up and down sampling on the final mix can result, in theory, errors.

 

[PENG]

I agree, with everett, that errors might have resulted from the conversions. From my experience (just by listening), the original recording/mastering process determines sound quality in the end, but I do believe CD quality is the point of diminishing return, if not the final limit. I have done enough comparisons to focus on recording quality, not the format, bit rate, bit depth and sampling frequency.

 

[yepimonfire]

I'm more interested in listening than testing in high resolution music the most important is the original file recording. Up and down sampling on the final mix can result, in theory, errors.

It's unlikely, especially with the level of precision used in the sample rate conversion. If studio m,aster conversions (typically done using 192pt sinc) truly caused errors in the 44.1khz redbook CD, then that would further prove that 44.1 is not as good suitable as a format, since it is a bad idea to record, mix, and master at anything below 96khz.

I may repeat the test with the regular 16/44.1 CD version.

 

[Bucknekked]

From my experience (just by listening), the original recording/mastering process determines sound quality in the end, but I do believe CD quality is the point of diminishing return, if not the final limit. I have done enough comparisons to focus on recording quality, not the format, bit rate, bit depth and sampling frequency.

Having spent far too much time recently reading and doing my own little bit of listening tests (nothing scientific, just my own comparisons), I agree with your sentiments. The original recording is the fat part of the curve : if its good, the format just doesn't matter. If its not, nothings going to save it.

If studio m,aster conversions (typically done using 192pt sinc) truly caused errors in the 44.1khz redbook CD, then that would further prove that 44.1 is not as good suitable as a format, since it is a bad idea to record, mix, and master at anything below 96khz.

"then that would further prove that 44.1 is not as good ...as a format". If the point of the excersize is to show that CD quality audio (44.1) isn't as good as HD options, this ground has been covered many times theoretically by smarter folks than I and in practical testing by more disciplined folks than I.

As I read the material available, I would guess (a guess) its 90%/10% or something like that in favor of 44.1 has it covered verses the 10% opinion that says HD presents something "more" and "better".

If yepimonfire hears a difference in the HD materials and likes the difference and is willing to spend the money for it, then I'm 100% supportive of that and give him two thumbs up. Yepimonfire asked when he started this thread if anybody did some listening and did they hear a difference. For me, the answer is yes I did some listening tests, and no I didn't hear any difference that would say an HD format sounded different or better.

For anyone that wants to invest in HD versions of audio, I would say go ahead and enjoy your experience.
I am not knocking HD audio, or yepimonfire, or anybody's desire to hear the very best versions of tunes possible. Its too nice a day outside for anything disagreeable.

 

[everettT]

Without correcting your post, I'd suggest you look at some pro audio forums such as

http://duc.avid.com

In recording and mastering there so many variables. One engineer may prefer to master in the analog domain so they are going da/ad then add the various conversion rates, should they dither, at what point in the mastering does the bit rate conversion occur? What software is used? The topic is never cut and dry. Check this page out

http://src.infinitewave.ca

Think of the process like room correction software, each manufacturer uses different algorithms to try to achieve the same result, but as we know they each vary.

I wouldn't get hung up on the numbers so much.

It's unlikely, especially with the level of precision used in the sample rate conversion. If studio m,aster conversions (typically done using 192pt sinc) truly caused errors in the 44.1khz redbook CD, then that would further prove that 44.1 is not as good suitable as a format, since it is a bad idea to record, mix, and master at anything below 96khz.

I may repeat the test with the regular 16/44.1 CD version.

 

[PENG]

Having spent far too much time recently reading and doing my own little bit of listening tests (nothing scientific, just my own comparisons), I agree with your sentiments. The original recording is the fat part of the curve : if its good, the format just doesn't matter. If its not, nothings going to save it.



"then that would further prove that 44.1 is not as good ...as a format". If the point of the excersize is to show that CD quality audio (44.1) isn't as good as HD options, this ground has been covered many times theoretically by smarter folks than I and in practical testing by more disciplined folks than I.

As I read the material available, I would guess (a guess) its 90%/10% or something like that in favor of 44.1 has it covered verses the 10% opinion that says HD presents something "more" and "better".

If yepimonfire hears a difference in the HD materials and likes the difference and is willing to spend the money for it, then I'm 100% supportive of that and give him two thumbs up. Yepimonfire asked when he started this thread if anybody did some listening and did they hear a difference. For me, the answer is yes I did some listening tests, and no I didn't hear any difference that would say an HD format sounded different or better.

For anyone that wants to invest in HD versions of audio, I would say go ahead and enjoy your experience.
I am not knocking HD audio, or yepimonfire, or anybody's desire to hear the very best versions of tunes possible. Its too nice a day outside for anything disagreeable.

Actually I do purchase those HD versions from time to time but my purchase decisions were still based on perceived recording quality, not because of the "resolution" specs.

 

[3rdeye]

If studio master conversions (typically done using 192pt sinc) truly caused errors in the 44.1khz redbook CD, then that would further prove that 44.1 is not as good suitable as a format, since it is a bad idea to record, mix, and master at anything below 96khz.

Why is it a bad idea to record, mix, and master below 96 kHz?



Sent from my ONEPLUS A3000 using Tapatalk

 

[everettT]

Why is it a bad idea to record, mix, and master below 96 kHz?

It's not necessarily, but the math (algorithm) gets more complex. If your end goal is 44.1 you would record, for example, 88.2 as the math is a much simpler conversion. I've had self made digital recordings @48hz on dat and have heard artifacts that are not on the original after down conversion for cd. I use adobe audition, formally cool edit pro, and for the money I'd have assumed that it could handle it, not always and have had good and bad conversions of the same recording

Sent from my ONEPLUS A3000 using Tapatalk

It's not necessarily, but the math (algorithm) gets more complex. If your end goal is 44.1 you would record, for example, 88.2 as the math is a much simpler conversion. I've had self made digital recordings @48hz on dat and have heard artifacts that are not on the original after down conversion for cd. I use adobe audition, formality cool edit pro, and for the money I'd have assumed that it could handle it, not always and have had good and bad conversions of the same recording

 

[3rdeye]

It's not necessarily, but the math (algorithm) gets more complex. If your end goal is 44.1 you would record, for example, 88.2 as the math is a much simpler conversion. I've had self made digital recordings @48hz on dat and have heard artifacts that are not on the original after down conversion for cd. I use adobe audition, formality cool edit pro, and for the money I'd have assumed that it could handle it, not always and have had good and bad conversions of the same recording

Maybe my question wasn't clear enough. I was curious why (from a listening POV) you wouldn't want to go below 96 kHz.

Are you saying that recording below that threshold translates into less music fidelity?

Sent from my ONEPLUS A3000 using Tapatalk

 

[yepimonfire]

It's not necessarily, but the math (algorithm) gets more complex. If your end goal is 44.1 you would record, for example, 88.2 as the math is a much simpler conversion. I've had self made digital recordings @48hz on dat and have heard artifacts that are not on the original after down conversion for cd. I use adobe audition, formality cool edit pro, and for the money I'd have assumed that it could handle it, not always and have had good and bad conversions of the same recording

Its not just about sample rate conversion that's the issue. In the same way MP3 at 320kbps throws inaudible information out, so do lower sample rates and bit depth precision's, though mp3 is a more extreme example. There may be little to no audible difference in the final down converted CD, but if you record something at 16/44.1, apply a bunch of processing or dsp, mix and remix it, then finally master it, it comes out sounding worse than if you started with a higher sample rate and bit depth. Doesn't have to be 96khz, could be 88, 192 etc.

Sent from my 5065N using Tapatalk

 

[Bucknekked]

Its not just about sample rate conversion that's the issue. In the same way MP3 at 320kbps throws inaudible information out, so do lower sample rates and bit depth precision's, though mp3 is a more extreme example. There may be little to no audible difference in the final down converted CD, but if you record something at 16/44.1, apply a bunch of processing or dsp, mix and remix it, then finally master it, it comes out sounding worse than if you started with a higher sample rate and bit depth. Doesn't have to be 96khz, could be 88, 192 etc.

Sent from my 5065N using Tapatalk

yepimonfire:
If I understand what I've read in other posts on sound production (vs. re-production) and what you're saying here, there is need and usefulness in having data and sampling far in excess of human hearing because in the sound production process there is stuff going on which makes use of the information. I don't pretend to understand the nuances, but I have seen this described in enough detail that I think it must be valid.

If one is involved in the production of music, one may have the need for vastly different requirements. And since the industry as a whole, recording engineers get together at conferences and such, deems it a solid practice, it must be worth it and we benefit from it.

Where I think people get a little confused is trying to bring that same need required in sound production with end user sound re-production systems. I think the sentiment is something like this "if the recording engineer was using 192kb sampling in the recording process, I want and need that same capability at home because then I can hear all that the engineer wanted me to hear".

 

[3rdeye]

yepimonfire:
If I understand what I've read in other posts on sound production (vs. re-production) and what you're saying here, there is need and usefulness in having data and sampling far in excess of human hearing because in the sound production process there is stuff going on which makes use of the information. I don't pretend to understand the nuances, but I have seen this described in enough detail that I think it must be valid.

If one is involved in the production of music, one may have the need for vastly different requirements. And since the industry as a whole, recording engineers get together at conferences and such, deems it a solid practice, it must be worth it and we benefit from it.

Where I think people get a little confused is trying to bring that same need required in sound production with end user sound re-production systems. I think the sentiment is something like this "if the recording engineer was using 192kb sampling in the recording process, I want and need that same capability at home because then I can hear all that the engineer wanted me to hear".

And this is what I'm trying to make sense of. From the best that I've read, our listening range capacity isn't influenced by what engineers mix at, so it would seem to me that the only discussion worth having is what re-production format permits for us to take advantage of the broadest listening range. Or am I missing something?

Sent from my ONEPLUS A3000 using Tapatalk

 

[Bucknekked]

And this is what I'm trying to make sense of. From the best that I've read, our listening range capacity isn't influenced by what engineers mix at, so it would seem to me that the only discussion worth having is what re-production format permits for us to take advantage of the broadest listening range. Or am I missing something?

Sent from my ONEPLUS A3000 using Tapatalk

3rdeye
I think you have it exactly in the right light. Sound production makes use of a wide variety of tools and processes that make use of the extra information and bits and such. When they are finishing the production process and creating a CD for sound re-production, they are using enough sampling and bit rate in the finished 44.1 product that we are getting everything they intended.

Remember too, the 44.1 format for CD's was chosen by folks who understand both sides: production needs and re-production needs. There's always going to be discussion, and I think its a healthy one, about formats above CD quality for home re-production systems. The way to decide for yourself is either read a bunch of technical doc's, or, just get some HD music in a format you can consume and do a nose to nose comparison with music you already own on CD. Let your own ears and brain decide.

You can't go wrong with your own comparisons. If you hear a difference and want to pursue HD. more power to you and you've found another way to enjoy your tunes. If you don't hear any difference, well, you can be happy with your collection and spend more time listening and not testing.

 

[yepimonfire]

yepimonfire:
If I understand what I've read in other posts on sound production (vs. re-production) and what you're saying here, there is need and usefulness in having data and sampling far in excess of human hearing because in the sound production process there is stuff going on which makes use of the information. I don't pretend to understand the nuances, but I have seen this described in enough detail that I think it must be valid.

If one is involved in the production of music, one may have the need for vastly different requirements. And since the industry as a whole, recording engineers get together at conferences and such, deems it a solid practice, it must be worth it and we benefit from it.

Where I think people get a little confused is trying to bring that same need required in sound production with end user sound re-production systems. I think the sentiment is something like this "if the recording engineer was using 192kb sampling in the recording process, I want and need that same capability at home because then I can hear all that the engineer wanted me to hear".

While the ability to store ultrasonics is something a higher sampling rate can do, that is not the only benefit to it. 44.1khz wasn't chosen because it was determined to be adequate, it was sort of a historical accident, mainly involving compatibility with the size of CD and video recording of the day. The antialiasing/reconstruction filters required for 44.1khz mangle transients and smear the temporal resolution. 44.1khz is not high enough to accurately capture or reproduce the impulse of things like a cymbal strike, which shows up on a spectrum analysis as ultrasonic information. Fourier transformation into the frequency domain only applies to sine waves, music is a complex wave, not pure sinewave tones, and the ear is more sensitive to temporal information than frequency.

Sent from my 5065N using Tapatalk

 

[Bucknekked]

While the ability to store ultrasonics is something a higher sampling rate can do, that is not the only benefit to it. 44.1khz wasn't chosen because it was determined to be adequate, it was sort of a historical accident, mainly involving compatibility with the size of CD and video recording of the day. The antialiasing/reconstruction filters required for 44.1khz mangle transients and smear the temporal resolution. 44.1khz is not high enough to accurately capture or reproduce the impulse of things like a cymbal strike, which shows up on a spectrum analysis as ultrasonic information. Fourier transformation into the frequency domain only applies to sine waves, music is a complex wave, not pure sinewave tones, and the ear is more sensitive to temporal information than frequency.

It is entirely possible these things you mention are all true, but they may also make no difference at all in the audible end product of a CD. If they do indeed make an audible difference, then specific examples would be handy so I can hear the difference. A piece of music and a specific place to listen.

I am not challenging your data or your explanation. I will accept it at face value as true. I have no reason not to accept your observations since you seem to have a pretty firm grasp on the topic. But can I , or anybody else, hear an audible difference based on a specific instance?

 

[j_garcia]

It is entirely possible these things you mention are all true, but they may also make no difference at all in the audible end product of a CD. If they do indeed make an audible difference, then specific examples would be handy so I can hear the difference. A piece of music and a specific place to listen.

I am not challenging your data or your explanation. I will accept it at face value as true. I have no reason not to accept your observations since you seem to have a pretty firm grasp on the topic. But can I , or anybody else, hear an audible difference based on a specific instance?

Problem here IMO, is losing sight of the music. It doesn't really matter if we can detect those differences, since the point is to listen to something that hopefully elicits a feeling not a sine wave. That doesn't mean it doesn't matter what we listen at or on, clearly it does, but there is a point at which that fine detail doesn't really add anything meaningful to the experience. There are times when I want to sit and listen to all that detail on the best recordings I have, but I don't need that same level every time I am listening because I am not critically listening.

 

[3rdeye]

Problem here IMO, is losing sight of the music. It doesn't really matter if we can detect those differences, since the point is to listen to something that hopefully elicits a feeling not a sine wave. That doesn't mean it doesn't matter what we listen at or on, clearly it does, but there is a point at which that fine detail doesn't really add anything meaningful to the experience. There are times when I want to sit and listen to all that detail on the best recordings I have, but I don't need that same level every time I am listening because I am not critically listening.

At the consumer level, I think it totally matters. As a music consumer, there isn't any value in what you can't detect.

That's like people who pay for the highest speed of internet and only stream Netflix. There is a "floor" or minimum speed that you must have in order to have a functional experience of no buffering. Outside of that, there is not an "enhanced" experience in any way by having more speed because Netflix doesn't show better imagery or sound as speed increases.

Same thing here. We can't hear outside of established hearing ranges. Those mastering methods aren't going to make our limited human ears hear outside of 20 Hz to 20 kHz.

Sent from my ONEPLUS A3000 using Tapatalk

 

[Irvrobinson]

While the ability to store ultrasonics is something a higher sampling rate can do, that is not the only benefit to it. 44.1khz wasn't chosen because it was determined to be adequate, it was sort of a historical accident, mainly involving compatibility with the size of CD and video recording of the day. The antialiasing/reconstruction filters required for 44.1khz mangle transients and smear the temporal resolution. 44.1khz is not high enough to accurately capture or reproduce the impulse of things like a cymbal strike, which shows up on a spectrum analysis as ultrasonic information. Fourier transformation into the frequency domain only applies to sine waves, music is a complex wave, not pure sinewave tones, and the ear is more sensitive to temporal information than frequency.

Your post is filled with misinformation.

A 44.1KHz sampling rate is adequate for accurately reproducing any frequency up to 22KHz, which is beyond normal human hearing. For anyone over 40 years old, substantially beyond. Your statement about music being "complex waves", and FT being applicable only to pure sine waves is just plain incorrect.

You're also wrong about the Redbook standard being some kind of historical accident. (The Compact Disc specification was originally printed with a red cover, so it was called The Redbook.) The 44.1KHz sampling rate was chosen due to compatibility with the recordable bandwidth of video recorders of the day (late 1970s), because by leveraging video recording hardware Sony and Phillips (the two companies driving the Redbook standard) could get to market more quickly and cheaply, while still providing a 0-20KHz frequency response, which was the objective. 44.1KHz also provides a narrow guard band for the audio spectrum with the digital filtering necessary to remove aliasing for samples above one half the sampling frequency, which is 22.05KHz for CD audio.

The word width to represent amplitude (a sample is just a word containing the amplitude; the frequency is calculated in reference to a clock) was a point of contention between Sony and Phillips. Phillips thought a 14bit word width would be sufficient for CDs, since the resulting ~80db native dynamic range was about equal to the best dolby-enabled analog tape recorders of the day. Sony insisted on a 16bit word width to make CD audio truly superior to anything in analog, and perhaps to reset the playing field for the Phillips TDA-1540 14bit DAC that was already under development, and supposedly ahead of Sony's DAC effort. (Cynic that I am, I tend to believe the latter theory.)

Word width is where digital audio representation strategies do differ, because the greater the word width the more headroom an engineer gets during the recording process before digital clipping occurs. (Digital clipping (running out of amplitude bits) is catastrophic, and results in gross distortion, unlike analog clipping, which is soft and progressive.) So for recording engineers you would want, say, a 20-24bit word width for recording, and then for mastering (creating the final version of the recording for CD audio) you would cleverly truncate the extra bits in software, and reposition the amplitudes of the samples for maximum utilization of the 16bit CD words. But for audio playback, word widths greater than 16bits are just marketing nonsense, in the if-16-is-good-24-must-be-better category of foolishness.

As for aliasing filters "mangling" the audio band, that's bullshit. With modern digital filters and 8x or more over-sampling rates the reconstructed CD audio band is pretty much perfect down to 90db+ below the fundamental amplitudes of the frequencies, which are reproduced perfectly from 0-20KHz.

And the ear is more sensitive to temporal differences than frequency differences? Seriously? Who says that? MQA marketing?

 

[Bucknekked]

Your post is filled with misinformation.

I knew I did not have the technical chops to write a proper reply.
You do. Thank you for saying what I was unable to say in proper tech speak.
In the end, my original observation stands to those who make such arguments about CD's being inadequate for some technical reason : show me a piece of music where I can hear the difference. And, as j garcia so aptly commented, is the difference one that matters?

irvrobinson, that was a good post.