IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
48kHz vs 44.1kHz, test file wanted
AliceWonder
post Sep 26 2012, 17:48
Post #1





Group: Members
Posts: 119
Joined: 13-July 12
From: California
Member No.: 101393



Does anyone have a link to a 16 bit 48kHz test wav that has audible (ABX) differences when downsampled to 44.1kHz?
Go to the top of the page
+Quote Post
DVDdoug
post Sep 26 2012, 18:01
Post #2





Group: Members
Posts: 2535
Joined: 24-August 07
From: Silicon Valley
Member No.: 46454



I doubt it...

But the real issue is... If you are making an audio CD and you have a 48kHz file, you don't have any choice so don't worry about if someone might be able to ABX the file. The listener will never have a chance to ABX or hear the original 48kHz file anyway (unless the listener is you).

And if you don't have a good reason to downsample, don't do it!

I've done quite a bit of back-and-forth between CD (44.1kHz) and DVD (48kHz). I've never noticed any problems, and I've never bothered with ABX testing.

This post has been edited by DVDdoug: Sep 26 2012, 18:02
Go to the top of the page
+Quote Post
greynol
post Sep 26 2012, 18:12
Post #3





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



QUOTE (AliceWonder @ Sep 26 2012, 09:48) *
Does anyone have a link to a 16 bit 48kHz test wav that has audible (ABX) differences when downsampled to 44.1kHz?

You might find one that exposes a bad resampler, which is really where your question should be focused rather than at the samples themselves.

Otherwise I imagine you ought to try 48 kHz files that only contain information that 44.1 kHz files cannot, which is content at 22.05 kHz and above. Or to be generous, files that only contain information near 22.05 that will give low-pass filters a difficult time (go back and read the first sentence in my reply). The reason I use the term "only" is that files with a fuller frequency response will likely mask the high frequencies, assuming of course that you can actually hear them to begin with.

This post has been edited by greynol: Sep 26 2012, 18:13


--------------------
Placebophiles: put up or shut up!
Go to the top of the page
+Quote Post
AliceWonder
post Sep 26 2012, 18:15
Post #4





Group: Members
Posts: 119
Joined: 13-July 12
From: California
Member No.: 101393



I'm asking because a very qualified programmer who is intelligent swears he has ABX'd differences when going from 2 inch tape.

I think the difference (assuming he is telling the truth) is from 2 different analog to digital conversions, so far he has not offered a 48KHz sample for me to downsample and see if I can hear a difference.
I don't believe there is one, but I could be wrong. Since he (so far) has not provided a test sample, I figured if such a test file existed, people here would now.

Doesn't even have to be music, computer generated wave forms would be fine.
Go to the top of the page
+Quote Post
lvqcl
post Sep 26 2012, 18:21
Post #5





Group: Developer
Posts: 3329
Joined: 2-December 07
Member No.: 49183



+1 +1 -1 -1 : 12 kHz tone (with intersample overs +3dBFS)

...Of course the difference is audible because of clipping, not because of resampling itself.

This post has been edited by lvqcl: Oct 27 2012, 19:21
Attached File(s)
Attached File  12kHz_ISO.flac ( 22.46K ) Number of downloads: 79
 
Go to the top of the page
+Quote Post
greynol
post Sep 26 2012, 21:09
Post #6





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



Clever!

Of course you could just have easily produced a 11.025 kHz tone with the same sequence with a 44.1 kHz sample rate, which sounds fine at its native rate but equally as nasty when converted to 48 kHz.


--------------------
Placebophiles: put up or shut up!
Go to the top of the page
+Quote Post
bandpass
post Sep 27 2012, 05:43
Post #7





Group: Members
Posts: 326
Joined: 3-August 08
From: UK
Member No.: 56644



QUOTE (lvqcl @ Sep 26 2012, 18:21) *
+1 +1 -1 -1 : 12 kHz tone (with intersample overs)

...Of course the difference is audible because of clipping, not because of resampling itself.

The original sample will clip in the reconstruction filter so will also sound bad, but not as bad as the resampled one, which will likely be clipped twice, during both resampling and reconstruction.

This shows why loudness-wars peak-smashing is such a bad idea - what you ultimately hear becomes dependent on the ripping/playback chain.
Go to the top of the page
+Quote Post
AliceWonder
post Sep 27 2012, 09:37
Post #8





Group: Members
Posts: 119
Joined: 13-July 12
From: California
Member No.: 101393



I think this answers my question: http://www.aes.org/e-lib/browse.cfm?elib=14195

I don't have access to the actual paper, but it seems to indicate that using excellent quality audio equipment, people can not distinguish high definition audio from 16/44.1 in ABX testing.

They down-sampled on the spot (without dithering), and kept volume within 0.1dB.

Out of 554 trials, the match was made 49.8% of the time (heads or tails anyone ??)
Go to the top of the page
+Quote Post
alexeysp
post Sep 27 2012, 12:28
Post #9





Group: Members
Posts: 130
Joined: 3-April 09
Member No.: 68627



QUOTE (bandpass @ Sep 27 2012, 07:43) *
The original sample will clip in the reconstruction filter


Here's a snapshot of the output from an ancient ALC201A AC'97 codec reproducing the sample posted by lvqcl, taken with Agilent DSO3152A scope. Can't see any clipping.


Go to the top of the page
+Quote Post
bandpass
post Sep 27 2012, 13:40
Post #10





Group: Members
Posts: 326
Joined: 3-August 08
From: UK
Member No.: 56644



Seems that some do, some don't -- "if mastering engineers monitor their work through converters having the required 3dB of headroom and do not use meters that show intersample peaks, these engineers will be completely unaware of the additional distortion that many consumer playback devices will produce".
Go to the top of the page
+Quote Post
greynol
post Sep 27 2012, 15:27
Post #11





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



Of the literally hundreds of thousands of hard limited tracks with those limited samples being well above -3dBFS and often at -0.1dBFS or even FS, you would think someone would have reported a problem with audible differences due to resampling on this forum as this is the type of thing that people do around here.

This post has been edited by greynol: Sep 27 2012, 16:26


--------------------
Placebophiles: put up or shut up!
Go to the top of the page
+Quote Post
pbelkner
post Sep 27 2012, 16:55
Post #12





Group: Members
Posts: 412
Joined: 13-June 10
Member No.: 81467



QUOTE (greynol @ Sep 27 2012, 15:27) *
Of the literally hundreds of thousands of hard limited tracks with those limited samples being well above -3dBFS and often at -0.1dBFS or even FS, you would think someone would have reported a problem with audible differences due to resampling on this forum as this is the type of thing that people do around here.

Have you ever ask yourself why ITU BS.1770, EBU R128 and the like require inter-sample peak determination?

If you've ever determined inter-sample peaks of those contemporary hard limited (-0.1 dBFS) samples you would have noticed that they typically clip with 1.5 up to 2.5 dBFS.

As we all know this forum doesn't allow us to report whether there are auditable differences or not between the clipping original and the non clipping attenuated version because it is impossible to ABX them (obviously you will always distinguish the two versions because of different loudness and not because of clipping or not. Moreover, most likely you will prefer the clipped version because it is louder fooling your brain it is sounding "better").

This post has been edited by pbelkner: Sep 27 2012, 17:04
Go to the top of the page
+Quote Post
pdq
post Sep 27 2012, 17:09
Post #13





Group: Members
Posts: 3372
Joined: 1-September 05
From: SE Pennsylvania
Member No.: 24233



Surely it would be possible to record the clipped reconstructed waveform, as well as the unclipped version reduced a few dB. Then match their amplitudes and compare them.
Go to the top of the page
+Quote Post
krabapple
post Sep 27 2012, 17:36
Post #14





Group: Members
Posts: 2181
Joined: 18-December 03
Member No.: 10538



QUOTE (pbelkner @ Sep 27 2012, 11:55) *
If you've ever determined inter-sample peaks of those contemporary hard limited (-0.1 dBFS) samples you would have noticed that they typically clip with 1.5 up to 2.5 dBFS.


They can, but *do* they? Who has actually done the determination of this, using contemporary samples, and where is it published?


Go to the top of the page
+Quote Post
greynol
post Sep 27 2012, 17:36
Post #15





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



I don't wonder about recommendations regarding best practice at all, but I certainly don't accept arguments from authority as proof.

The profound lack of objective evidence amidst a staggeringly huge sample base pretty well classifies this particular concern as a non-issue.

...and yes, I have in fact looked at inter-sample overs from this abysmal stlye of mastering, thank you very much.

Finally, pbelkner, if you're trying to say there is no evidence here because it can't be tested (let alone quite easily tested) you need to review my post and how it fits in the context of the overall discussion, as you would be sorely mistaken.

This post has been edited by greynol: Sep 27 2012, 17:50


--------------------
Placebophiles: put up or shut up!
Go to the top of the page
+Quote Post
pbelkner
post Sep 28 2012, 06:59
Post #16





Group: Members
Posts: 412
Joined: 13-June 10
Member No.: 81467



QUOTE (krabapple @ Sep 27 2012, 17:36) *
QUOTE (pbelkner @ Sep 27 2012, 11:55) *
If you've ever determined inter-sample peaks of those contemporary hard limited (-0.1 dBFS) samples you would have noticed that they typically clip with 1.5 up to 2.5 dBFS.


They can, but *do* they? Who has actually done the determination of this, using contemporary samples, and where is it published?

I do it on a regular basis, and you can do it too. Just tag your audio files using R128GAIN and play them back using Winamp in conjunction with the in_ffsox plugin. The in_ffsox plugin displays clipping / headroom of the original as well as of the EBU R128 normalized playback in the properties panel.

This post has been edited by pbelkner: Sep 28 2012, 07:01
Go to the top of the page
+Quote Post
pbelkner
post Sep 28 2012, 07:22
Post #17





Group: Members
Posts: 412
Joined: 13-June 10
Member No.: 81467



QUOTE (greynol @ Sep 27 2012, 17:36) *
I don't wonder about recommendations regarding best practice at all, but I certainly don't accept arguments from authority as proof.

Just to continue with proof by authority and listening to what Christopher “Monty” Montgomery (as far as I can see a well respected authority to this community) has to say on the topic:

QUOTE
Clipping

Clipping is another easy mistake, sometimes obvious only in retrospect. Even a few clipped samples or their aftereffects are easy to hear compared to an unclipped signal.

The danger of clipping is especially pernicious in tests that create, resample, or otherwise manipulate digital signals on the fly. Suppose we want to compare the fidelity of 48kHz sampling to a 192kHz source sample. A typical way is to downsample from 192kHz to 48kHz, upsample it back to 192kHz, and then compare it to the original 192kHz sample in an ABX test [21]. This arrangement allows us to eliminate any possibility of equipment variation or sample switching influencing the results; we can use the same DAC to play both samples and switch between without any hardware mode changes.

Unfortunately, most samples are mastered to use the full digital range. Naive resampling can and often will clip occasionally. It is necessary to either monitor for clipping (and discard clipped audio) or avoid clipping via some other means such as attenuation.

24/192 Music Downloads ... and why they make no sense
Christopher “Monty” Montgomery
March 1, 2012
http://people.xiph.org/~xiphmont/demo/neil-young.html


This post has been edited by pbelkner: Sep 28 2012, 07:26
Go to the top of the page
+Quote Post
greynol
post Sep 28 2012, 07:30
Post #18





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



Once again, you've completely failed to put this into the context of this discussion. This is a discussion about samples that reveal audible differences between sample rates (check the title). It is not about about the evils of hard limiting and normalization without adequate headroom.

If you have samples of real music where different samplerates can be ABXed, now is the time to present them.

This post has been edited by greynol: Sep 28 2012, 07:42


--------------------
Placebophiles: put up or shut up!
Go to the top of the page
+Quote Post
pbelkner
post Sep 28 2012, 07:33
Post #19





Group: Members
Posts: 412
Joined: 13-June 10
Member No.: 81467



QUOTE (greynol @ Sep 28 2012, 07:30) *
It is about about the evils of hard limiting and normalization without adequate headroom.

It's exactly what I'm taking about. Do I need to add that you have to expect each descent DAC to oversample?

Answer to edited version:

QUOTE (greynol @ Sep 28 2012, 07:30) *
It is not about about the evils of hard limiting and normalization without adequate headroom.

As far as I can see Monty's quote is exactly w.r.t. the evils of clipping, in particular when unnoticed happening due to re-sampling.

QUOTE (greynol @ Sep 28 2012, 07:30) *
If you have samples of real music where different samplerates can be ABXed, now is the time to present them.

I never claimed that. My only topic was clipping, and if somebody is going to tell me to demonstrate by ABX that clipping disturbs the sound you can't expect me ... Sorry guys, I'm not going to violate any of the forum's TOS.

This post has been edited by pbelkner: Sep 28 2012, 07:57
Go to the top of the page
+Quote Post
greynol
post Sep 28 2012, 08:03
Post #20





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



QUOTE (pbelkner @ Sep 27 2012, 23:33) *
Sorry guys, I'm not going to violate any of the forum's TOS.

Too late for that, you violated #5 by failing to remain within the vein of the discussion.

I'll gladly participate in a thread about the audibility of inter-sample overs, btw. You can even start it like this one was: by asking if samples exist, rather than by claiming audibility without the objective evidence which you appear unable to provide. wink.gif

PS: Regarding what you're incorrectly calling "proof by authority," it is a logical fallacy, but I'm guessing you already know this by your signature. If you don't you can read about it here:
http://en.wikipedia.org/wiki/Argument_from_authority

This post has been edited by greynol: Sep 28 2012, 17:57


--------------------
Placebophiles: put up or shut up!
Go to the top of the page
+Quote Post
alexeysp
post Sep 28 2012, 10:49
Post #21





Group: Members
Posts: 130
Joined: 3-April 09
Member No.: 68627



I just realized that I didn't check for properly set volume levels in the Windows volume control panel when I tested for clipping yesterday, and both Master and Wave volume were set to 50%. So I repeated the test today with corresponding settings adjusted to maximum. Some distortion now indeed can be seen with the ISO sample. The third and fifth harmonic magnitudes are about 2..3% (-30..35 dB) relative to main.

12 kHz fullscale wave generated by foobar2000 with Wave and Master volume set to 100%:



12 kHz ISO sample at the same settings:



Unfortunately I don't have any other DACs around here to test.
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 26th July 2014 - 13:36