Some commonsense

[tmtomh]

14 hours ago, jabbr said:

 

The real point however is the mechanism by which the 24/192 data stream is converted to 16/44 ... using a sample rate converter! Are these all the same? Are they all of the same quality?

 

Hell no!

 

Please don't suggest that all sample rate converters produce the exact same result....

 

 

See above. PCM *can* be lossless when use to hold the original data. As you very well know, it's the process that converts one PCM format to another that is lossy. Perhaps by y'all definition a perfect sample rate converter would produce lossless CD but I still prefer to have the source audio in as close to the mastering format as I am able to obtain, thank you.

 

I really wish you would fully read the comments you are responding to. I explicitly stated that the one aspect where a CD can be considered lossy is when there's a non-integer sample-rate conversion, as from 96k or 192k to 44.1k. There the audio data in the audible range is necessarily altered by the fact that 44.1 doesn't divide evenly into 96 or 192. 

 

Moreover, you have deflected from the point of the comment you are responding to here - namely, you have neglected to acknowledge that PCM downsampling is not like JPEG compression, and that by extension PCM downsampling is  not lossy in the way that mp3/AAC perceptual encoding is. In this regard I'd remind you that the subject of this thread is about common sense, and your argument here is not a common-sense argument - as @mansr so aptly put it just above, "Insisting that anything with less than infinite bandwidth and precision should be called lossy only serves to muddle the distinction between on the one hand plain sampling, the accuracy of which is known upfront, and on the other hand perceptual coding where the accuracy varies wildly depending on the signal content."

 

To your second point here in response to mansr, PCM sample-rate conversion is not inevitably lossy as you claim. To the contrary, integer conversion perfectly preserves the digital encoding in the audible range. A 22.05kHz tone will be sampled 8 times in a 176.4kHz data file, 4 times in an 88.2kHz file, and 2 times in a 44.1kHz file. Since basic digital sampling theory tells us that 2 - and only 2 - samples are necessary to perfectly reconstruct the signal, and that additional samples are unnecessary and make zero difference in the resulting analogue waveform, integer downsampling is completely irrelevant when it comes to the ability to reconstruct audible-range signals. Therefore integer downsampling is lossless, so long as the downsampled sample rate is high enough to encode the highest frequencies audible to humans, which 44.1k is.

 

Now, if you still would rather have the original, higher sample-rate 88.2k or 176.4k file than a downsampled CD version, I have no problem with your preference. But that preference is in no way, shape, or form evidence that PCM sample rate conversion is inevitably lossy.

[mansr]

3 minutes ago, tmtomh said:

I explicitly stated that the one aspect where a CD can be considered lossy is when there's a non-integer sample-rate conversion, as from 96k or 192k to 44.1k. There the audio data in the audible range is necessarily altered by the fact that 44.1 doesn't divide evenly into 96 or 192. 

That is incorrect. Any sample rate conversion with a rational ratio preserves to an arbitrarily high precision all content below half the lower of the sample rates. It doesn't matter that the sample points are moved in time. There will of course be some alteration in the transition band of the filter, but this applies equally to integer ratio resampling.

[jabbr]

39 minutes ago, tmtomh said:

In this regard I'd remind you that the subject of this thread is about common sense, and your argument here is not a common-sense argument - as @mansr so aptly put it just above, "Insisting that anything with less than infinite bandwidth and precision should be called lossy only serves to muddle the distinction between on the one hand plain sampling, the accuracy of which is known upfront, and on the other hand perceptual coding where the accuracy varies wildly depending on the signal content."

 

Here is a common sense position: I cried a little inside when I read the nytimes article that states that the original Coltrane Impulse masters were destroyed in the Universal fire.

 

 

You apparently couldn't care less because you consider the CDs definitive. Compared to the masters, I consider the CDs (that I have) lossy. The fire was a huge loss.

[tmtomh]

22 minutes ago, mansr said:

That is incorrect. Any sample rate conversion with a rational ratio preserves to an arbitrarily high precision all content below half the lower of the sample rates. It doesn't matter that the sample points are moved in time. There will of course be some alteration in the transition band of the filter, but this applies equally to integer ratio resampling.

 

Thanks for the correction - that further strengthens the point, then, that resampled PCM, e.g. from high-res down to redbook, is lossless.

[tmtomh]

4 minutes ago, jabbr said:

 

Here is a common sense position: I cried a little inside when I read the nytimes article that states that the original Coltrane Impulse masters were destroyed in the Universal fire.

 

 

You apparently couldn't care less because you consider the CDs definitive. Compared to the masters, I consider the CDs (that I have) lossy. The fire was a huge loss.

 

That's a common-sense position. But your supposition that I "couldn't care less" about the loss of these masters is, franky, moronic. I sy moronic because the linked article - and once again, are you even reading the things you link to and respond to? - is about a fire that consumed mostly analogue master tapes, which has nothing to do with what we are discussing.

 

Moreover, wanting high-res digital masters to be preserved is a strong preference I absolutely share with you - but your and my shared preference has absolutely nothing to do with whether or not CDs are lossy. You have claimed they are; it has been demonstrated clearly that they are not; and yet you refuse simply to acknowledge that.

 

Finally, it must be stated that in the case of digital masters that exist in 16-bit, 44.1kHz format - which to my understanding is the vast majority of digital mixdown masters produced up until the late '80s/early '90s - a CD is in fact a bit-perfect digital copy that not only is lossless compared to the original, but also contains the exact same data as the original in the exact same resolution. In fact, one could reasonably speculate that a 30 year-old, well-manufactured CD might today be a more reliable storage medium for that 16/44.1kHz digital data than a 16/44.1k master that exists in the form of a U-Matic tape.

[Ajax]

18 hours ago, Paul R said:

 

Just a question - does your Benchmark internally upsample every input? If so, that is probably the biggest reason why the two files sound alike. My old Benchmark upsampled all inputs to 100khz, which gave the effect of making all the inputs sound the same or at east very similar. .

 

Not sure if the new models do that or something similar.  

 

There are very good sounding DACs that resample everything to DSD and I cannot usually tell the difference between a red book and 24/96k file on those DACs either. 

Hi Paul,

 

You are correct in that the Benchmark upsamples everything.

 

The Benchmark DAC 1 (now over 10 years old and replaced by the DAC 2 and DAC 3 series) resampled all frequencies to 110khz. Here is a link to an article by John Siau giving the reasons behind such an obscure number..

 

https://benchmarkmedia.com/blogs/application_notes/13127453-asynchronous-upsampling-to-110-khz

 

The upsampling frequency was subsequently increased to 210khz in the DAC 2 and DAC 3 as the chip was changed from an AK type to ESS. Some more information here.

 

https://benchmarkmedia.com/blogs/application_notes/inside-the-dac2-part-2-digital-processing

 

 

 

 

[Paul R]

1 hour ago, tmtomh said:

 

Thanks for the correction - that further strengthens the point, then, that resampled PCM, e.g. from high-res down to redbook, is lossless.

 

You both really need to specify the conditions you are making your statements under. 

 

Nyquist sampling can provide, for our intents and purposes, a reproduction that is audibly perfect.  Is it objectively perfect? No, because infinite samples would be required for perfect reproduction. Does it matter in a practical sense? Almost certainly not. 

 

Is Decimated PCM lossless? In one sense, of course not. Data from the original file is not in the decimated file.  In another sense, the sound we will reproduce from the decimated file will be audibly the same. Unless of course, there was any signal above 22khz. That will be lost in the SRC. 

 

It seems impossible this argument is still going on.

 

Common sense says that the higher sample rate file gives you a choice and thus, even if there is no audible difference to you between 24/192k and 16/44.1k,  preserve the choice for the future. There is no longer any justification in terms of size or bandwidth for 16/44.1k. I do not know of any professional studios recording in 16/44.1k these days, and not many amateurs do either. 

 

Otherwise, you may put yourself into a position like those folks who ripped their CDs to MP3 then threw away the CDs. They argued there was no possible audible difference between MP3 and CD quality. There probably are even a few of those arguments preserved here somewhere. 

 

 

 

 

[tmtomh]

35 minutes ago, Paul R said:

 

You both really need to specify the conditions you are making your statements under. 

 

Nyquist sampling can provide, for our intents and purposes, a reproduction that is audibly perfect.  Is it objectively perfect? No, because infinite samples would be required for perfect reproduction. Does it matter in a practical sense? Almost certainly not. 

 

Is Decimated PCM lossless? In one sense, of course not. Data from the original file is not in the decimated file.  In another sense, the sound we will reproduce from the decimated file will be audibly the same. Unless of course, there was any signal above 22khz. That will be lost in the SRC. 

 

It seems impossible this argument is still going on.

 

Common sense says that the higher sample rate file gives you a choice and thus, even if there is no audible difference to you between 24/192k and 16/44.1k,  preserve the choice for the future. There is no longer any justification in terms of size or bandwidth for 16/44.1k. I do not know of any professional studios recording in 16/44.1k these days, and not many amateurs do either. 

 

Otherwise, you may put yourself into a position like those folks who ripped their CDs to MP3 then threw away the CDs. They argued there was no possible audible difference between MP3 and CD quality. There probably are even a few of those arguments preserved here somewhere. 

 

 

 

 

 

No disagreement with any of this. But at the risk of being pedantic, I do feel compelled to note that the argument between jabbr and I is going on because one of is is specifying the conditions under which we are making our statements, and the other is not. I've stated repeatedly that downconversion does remove data (sample). I've stated repeatedly my personal preference for preservation of full/original resolution master sources. I have noted, repeatedly, that redbook downsampled from higher-resolution originals is indeed lossy but that IMHO - and I have always noted IMHO for this piece - it is unproductive and misleading to speak of redbook as lossy in the same way we speak of perceptual encoding (mp3, AAC and so on ) as lossy.

 

No one has to agree with my position here. But the argument persists not because jabbr is disagreeing with my claims, but rather because he is steadfastly ignoring most of them.

[PeterSt]

3 hours ago, mansr said:

Can we please stop this nonsense of calling PCM lossy?

 

👍

[jabbr]

20 hours ago, mansr said:

PCM is lossless in the sense that it captures perfectly any signal within the well-defined bounds determined by sample rate and bit depth. This is in stark contrast to perceptual coding systems that discard a little bit here and a little bit there according to their psychoacoustic models of what is or isn't audible.

 

Again, its not PCM which is itself lossy, rather the process used to convert one PCM format (bitdepth-samplerate) into another. A perfect infinite length stream could be Fourier transformed and a perfect brickwall filter applied but no stream is infinite and no brickwall filter is perfect.

 

So ... the decision to use 16/44.1kHz is based on a "psychoacoustic" model of human hearing that claims that since the cochlea does not respond to tones above ~20 kHz, than bandlimiting the signal to 22 kHz will capture everything that is heard. This is an assumption folks. No need to argue whether it is a correct assumption but nonetheless it is based on a model of human hearing. 

 

Arguably the cochlea and human hearing system is better modelled with wavelets and so wavelet compression has been implemented. Is that "psychoacoustic" because different math is used (e.g. wavelet vs fourier?)

 

The point is that all of these formats whether 16/44 PCM or MPEG or AAC are based on human hearing assumptions (admittedly some assumptions e.g. 20 kHz upper limit, are more substantiated than others) and all of these formats are based on assumptions about what is necessary to represent audio.

 

My definition of losseless conversion between format A and format B is very precise:

There must exist a pair of transforms such that given a file f, when transformed and then inverse transformed, results in a bit identical result.   e.g.

f = T-1(T(f)) where

fB = T(fA) and

fA = T-1(fB)

 

So of course each format is lossy in a different way, and some are too lossy, and some losses aren't audible. Sure. Nonetheless downconverting a master to redbook/CD is typically lossy except where the source data was 16/44.1.

[mansr]

7 minutes ago, jabbr said:

Again, its not PCM which is itself lossy, rather the process used to convert one PCM format (bitdepth-samplerate) into another. A perfect infinite length stream could be Fourier transformed and a perfect brickwall filter applied but no stream is infinite and no brickwall filter is perfect.

You can get as close to perfection as you desire by using a sufficiently long filter.

 

7 minutes ago, jabbr said:

So ... the decision to use 16/44.1kHz is based on a "psychoacoustic" model of human hearing that claims that since the cochlea does not respond to tones above ~20 kHz, than bandlimiting the signal to 22 kHz will capture everything that is heard. This is an assumption folks. No need to argue whether it is a correct assumption but nonetheless it is based on a model of human hearing.

Human hearing has nothing to do with it. PCM at 8 kHz sample rate is lossless for frequencies below 4 kHz. It won't suddenly start discarding content at 2 kHz. This is guaranteed by design. A perceptual coder, on the other hand, makes no guarantees whatsoever. A signal component of any frequency and amplitude might be discarded, regardless of the sample rate, if the perceptual model in use tells it to.

 

I really don't understand your insistence over this. The term "lossless" has a widely accepted meaning. You are attempting to redefine it. Why? What do you possibly hope to accomplish?

[Paul R]

10 minutes ago, mansr said:

You can get as close to perfection as you desire by using a sufficiently long filter.

 

If by “long” here you are referring to the number of taps (i.e. delay),  there are operators that come into play that limit that. If you mean something else, please explain. 

 

What is your opinion on filters designed specifically for high res material vs those designed only for 16/44.1k?

 

[mansr]

13 minutes ago, Paul R said:

If by “long” here you are referring to the number of taps (i.e. delay),  there are operators that come into play that limit that. If you mean something else, please explain.

How about you, for once, explain what you mean instead of making vague allusions?

[mansr]

14 minutes ago, Paul R said:

What is your opinion on filters designed specifically for high res material vs those designed only for 16/44.1k?

I have no idea what you are talking about.

[mansr]

I took a 48 kHz music file and resampled it to 96 kHz and back 16 times using SoX at the highest quality setting. The spectrogram of the resulting difference looks like this:

 

As expected, there is slight change in the transition band of the filter. Below this, the level of the difference is about -150 dB RMS, -140 dB peak. In other words, the "loss" is confined to a few LSBs for all content below a certain frequency. Using higher precision arithmetic would reduce this further.

 

Music and difference files attached.

file1.flac file2.flac diff.flac

[danadam]

19 hours ago, esldude said:

Would be nice to post a few of those dithered 8 bit files for people to hear.

1 minute of Blågutten (from Quiet Winter Night album) available from 2L Test Bench. Because sox can't apply most of it different noise-shaping filters to 96 kHz material I took 96 kHz version from that page and converted it to 24 bit / 48 kHz and to 8 bit / 48 kHz with different options as follows:

-- 01. Blagutten 24 bit.flac
sox -V 2L-087_stereo-96kHz_06.flac "01. Blagutten 24 bit.flac" trim 0 62 fade 0 0 2 rate -v 48000

-- 02. Blagutten 8 bit no dither.flac
sox -V -D 2L-087_stereo-96kHz_06.flac -b8 "02. Blagutten 8 bit no dither.flac" trim 0 62 fade 0 0 2 rate -v 48000

-- 03. Blagutten 8 bit TPDF.flac
sox -V 2L-087_stereo-96kHz_06.flac -b8 "03. Blagutten 8 bit TPDF.flac" trim 0 62 fade 0 0 2 rate -v 48000 dither

-- 04. Blagutten 8 bit improved-e-weighted.flac
sox -V 2L-087_stereo-96kHz_06.flac -b8 "04. Blagutten 8 bit improved-e-weighted.flac" trim 0 62 fade 0 0 2 rate -v 48000 dither -f improved-e-weighted

-- 05. Blagutten 8 bit Shibata.flac
sox -V 2L-087_stereo-96kHz_06.flac -b8 "05. Blagutten 8 bit Shibata.flac" trim 0 62 fade 0 0 2 rate -v 48000 dither -f shibata

 

01. Blagutten 24 bit.flac 02. Blagutten 8 bit no dither.flac 03. Blagutten 8 bit TPDF.flac 04. Blagutten 8 bit improved-e-weighted.flac 05. Blagutten 8 bit Shibata.flac

[fas42]

14 hours ago, sandyk said:

Besides which,whether you like it or not, every format conversion like you are suggesting does result in some audible degradation, and is NOT a LOSSLESS process like many may wish to believe. Even doing the conversions  with different S/W such as Audacity or Sound Forge will result in small audible differences between the conversions.

This even applies to different S/W performing a simple conversion from 24/192 .aiff to 24/192 .wav files.

 

Alex, as mansr has been pointing out, format conversions to a precision level way beyond what can be considered audibly significant are quite easily achieved. The differences that are audible are all about the hardware being used for replay, and the precise manner of the playing of the tracks - tiny, usually ignored factors, or ones that are considered to be of minor importance are where the meat is, and vastly more relevant to the perceived SQ.

[sandyk]

Frank

 I don't give a damn what Mansr and yourself are saying,.

Along with many other members , and millions of people worldwide who put their money where there mouth is and prefer to purchase high resolution PCM and DSD in preference to RBCD, I can hear a marked improvement with well recorded 24/192  with genuine musical content to >55kHZ, to the down converted 16/44.1 version.

You probably believe that all Bit Perfect Music Software sounds the same too ?

 It no more sounds the same than S/W used for example to convert 24/192 .aiff to 24/192 .wav

 Recently I assisted Barry Diament to test 4 different Software versions for the conversion from .aiff to .wav for his recently released Kay Sa album, and the results from all 4 sounded a little different, with both Barry and myself agreeing on which version sounded best.

With the selected version, Barry described it as " more getting out of the way"

 I now have all 5 previous High Res albums from Barry after conversion  from the .aiff of the original made using the new S/W , and the reconverted albums sound markedly better than those already in my possession.

 

 As  Paul pointed out, many people regret converting their music to .mp3 and disposing of the original CDs too , based mainly on claims made at the time by people like yourself and others like Mansr..

 

Alex

[Paul R]

1 hour ago, mansr said:

How about you, for once, explain what you mean instead of making vague allusions?

 

How about you for once, being clear about one of your pronouncements? What do you mean by “long” in your reference to digital filters? 

[The Computer Audiophile]

6 minutes ago, sandyk said:

Frank

 I don't give a damn what Mansr and yourself are saying,.

Along with many other members , and millions of people worldwide who put their money where there mouth is and prefer to purchase high resolution PCM and DSD in preference to RBCD, I can hear a marked improvement with well recorded 24/192  with genuine musical content to >55kHZ, to the down converted 16/44.1 version.

You probably believe that all Bit Perfect Music Software sounds the same too ?

 It no more sounds the same than S/W used for example to convert 24/192 .aiff to 24/192 .wav

 Recently I assisted Barry Diament to test 4 different Software versions for the conversion from .aiff to .wav for his recently released Kay Sa album, and the results from all 4 sounded a little different, with both Barry and myself agreeing on which version sounded best.

With the selected version, Barry described it as " more getting out of the way"

 I now have all 5 previous High Res albums from Barry after conversion  from the .aiff of the original made using the new S/W , and the reconverted albums sound markedly better than those already in my possession.

 

 As  Paul pointed out, many people regret converting their music to .mp3 and disposing of the original CDs too , based mainly on claims made at the time by people like yourself and others like Mansr..

 

Alex

You know you can just put two tracks in Audacity and flip the phase on one to see and hear the differences between the two right?

[sandyk]

10 minutes ago, The Computer Audiophile said:

You know you can just put two tracks in Audacity and flip the phase on one to see and hear the differences between the two right?

 

 Like many other members here, I prefer to use my ears , not look at instrumentation beforehand to tell me how something should sound.

 Neither do I accept that the players used with Audacity or Sound Forge  sound as revealing as JRiver 25 when playing from System  Memory.

[The Computer Audiophile]

20 minutes ago, sandyk said:

 

 Like many other members here, I prefer to use my ears , not look at instrumentation beforehand to tell me how something should sound.

 Neither do I accept that the players used with Audacity or Sound Forge  sound as revealing as JRiver 25 when playing from System  Memory.

You could still use your ears by saving the difference file and playing it in the player of your choice. 

[Ralf11]

or play each file into a different ear

[fas42]

38 minutes ago, sandyk said:

 

 Like many other members here, I prefer to use my ears , not look at instrumentation beforehand to tell me how something should sound.

 Neither do I accept that the players used with Audacity or Sound Forge  sound as revealing as JRiver 25 when playing from System  Memory.

 

Audacity works well enough to be able to pick differences that are audible - I've mentioned a number of times a simple technique that makes pinpointing audible variations quite straightforward, that I've used for years. Roughly align the the two tracks if necessary, select a promising area in the clips of a few seconds or less, and solo one track on repeat - a monotonous pattern of sound is created, like a mantra; as soon as you've tuned into that pattern, switch to the other track, using solo again. It's immediately obvious that the second pattern has very different qualities - if in fact there's an audible variation; going back and forth makes it easy to confirm  that a true difference exists.

[tmtomh]

1 hour ago, sandyk said:

Frank

 I don't give a damn what Mansr and yourself are saying,.

[...]

 

Alex

 

Then why bother responding? And why expect any level of civility or respect in others' responses to your comments? Why participate in a discussion forum if you don't give a damn what others say?