Some commonsense

[mansr]

11 hours ago, Rexp said:

Most CD Players sound bad to me and I've never heard a good sounding Jazz CD, who is at fault? 

Could it be that you don't like jazz?

[mansr]

20 minutes ago, tmtomh said:

The difference between a sound wave and a radio wave, for example, is just the frequency.

The primary distinction is actually the medium. A sound wave is a mechanical vibration travelling through a solid, liquid, or gas. A radio wave, on the other hand, is a perturbation in the electric and magnetic fields. As you point out, there is in fact considerable overlap between the frequency ranges commonly associated with sound and radio.

[mansr]

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.

[mansr]

13 hours ago, Jud said:

Since there are vanishingly few ADCs/workstations that don't decimate a sigma-delta modulated form to some (other, if you consider the original SDM form to be PCM) form of PCM, and this is not mathematically lossless, we are left in nearly all instances with a non-perfect capture, mathematically speaking.  Not saying anything about the impact on human perception of the signal, as I imagine with good filtering there's very little or no perceptual effect.

Sigma-delta is an implementation detail. My Tascam ADC (PCM4220 based) captures input signals up to around 60 kHz limited only by the analogue noise level. Above 60 kHz, the modulator noise starts becoming noticeable (though obviously not audible). If recording at 96 kHz sample rate, the output is thus indistinguishable from that of a hypothetical flash ADC with at least as good precision.

 

Can we please stop this nonsense of calling PCM lossy? Everybody knows and agrees that a given PCM format will have some limitations. That is not the point. 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.

 

By referring to PCM as lossy, you are also inviting the converse, i.e. bogus claims of losslessness from the likes of MQA, while additionally making these claims that much harder to refute.

[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.

[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?

[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

[mansr]

2 hours ago, PeterSt said:

At least the DXD from 2NL does not make use of filtering at all (unless they changed their mind by now).

At the very least, the ADC will use a filter in its internal resampling from the raw sigma-delta output to DXD. If you were referring to filters in the analogue front-end of the ADC, there probably isn't one.

[mansr]

10 hours ago, Rexp said:

Will there be information present on a 24/96 master that is not present on the CD, yes or no? 

If the 96 kHz master has spectral content above 22 kHz, it will obviously not be on the CD. Nobody has ever claimed otherwise.

[mansr]

39 minutes ago, Jud said:

Yes, there are any arbitrary number of mathematically lossless conversions that can be made between AIFF, ALAC, FLAC, and WAV at a given sample rate, for instance. Conversion between sample rates isn't a mathematically lossless operation, though as you've pointed out it can be done so as to be "perceptually lossless."

No, that is not what I'm saying. Sample rate conversion can be done with arbitrarily high precision. Pick any non-zero number, and the error can be made smaller. The band limiting filter merely needs to be made sufficiently close to the ideal sinc filter. It might require 1024-bit precision and 10 billion filter taps, but it can be done.

 

In practice, the SRC precision need only be high enough to fully utilise the resolution of the target format, typically 24 bits. Taking into account realistic DAC noise levels, the requirements can be relaxed further still.

[mansr]

1 hour ago, Jud said:

Would be interesting to see one of the Native DSD files recorded live to DSD256 compared to the same file decimated to Redbook resolution with a filtering chain that might typically be used for mass market product.

Compared how?

[mansr]

1 minute ago, Jud said:

Except we were talking about a definition of "lossless." I would guess you could make a damn fine mp3 as well.

No, not without limits. MP3 is 320 kbps at most. This alone limits what it can encode. That's basic information theory. The particulars of the encoding method further restrict the achievable accuracy. A pure tone at 500 Hz will probably be represented fairly well, a 5 kHz tone less so. This is because the format inherently quantises high frequencies with lower precision. That is a choice made by the designers of the format based on models of human hearing. Yes, that is perceptual coding.

 

Optimally encoding the full information content in a typical CD quality music file requires roughly 500 kbps, a bit more or less depending on the type of music. An MP3 encoder thus must discard nearly half the information in order to fit the remainder below the 320 kbps limit. Beyond a few basic assumptions, nothing in the format spec dictates which parts of the input get discarded in the encoding process. It would be perfectly valid, for instance, to apply a bandpass filter retaining only the 500-1500 Hz range. You probably wouldn't like the result. A better encoder is more selective, discarding primarily signal components deemed inaudible according a perceptual model.

 

The important point here is, in an MP3 encoding something is always lost, and there are no guarantees whatsoever. PCM, meanwhile, promises to preserve every frequency below Nyquist with the accuracy afforded by the bit depth of each sample value. Need higher frequencies, increase the sample rate. Need better accuracy, increase the bit depth. The only limitations are of a practical nature, such as storage space and computing power. For any given PCM encoding, say 44.1 kHz 16-bit, it is precisely defined what can and cannot be represented.

 

Getting back to the specific term "lossless," it is not normally applied to raw data formats such as PCM. Rather, the lossless/lossy distinction is used when characterising compression algorithms. FLAC is lossless, whereas MP3 is not. Both operate on PCM data. Similarly, PNG and JPEG are both compression methods for raster image data, one lossless and the other not. When FLAC and PNG are termed lossless, there is no implication that the files encode every shred of information hitting the microphone or camera lens, only that upon decoding the uncompressed form of the data will be recreated exactly.

[mansr]

1 hour ago, Jud said:

As you did in your prior comment, comparing the initial file to the sample rate converted one. I'd like to see something quite typical of the usual filtering chain used.

How would you carry out this comparison. You can't, for a multitude of reasons, do a sample by sample subtraction between a DSD file and a downsampled version of the same.

 

1 hour ago, Jud said:

Since we're talking about common sense, I'd like to get an idea of what's common.

I'm not sure what's common, but it sure ain't sense.

[mansr]

2 minutes ago, jabbr said:

True or false:

a) Downconversion from 24/192 to 16/44.1 is frequently lossy?

vs

b) Downconversion from 24/192 to 16/44.1 is always lossless?

 

My position, according to the common definition of “lossy” is that (a) is true and (b) is false. 

Why are you even asking that question? Sample rate conversion isn't considered a compression algorithm (unless your name is Bob Stuart).

[mansr]

35 minutes ago, Jud said:

How odd. All I'm asking is for you to perform an operation similar to the one you posted about in comment #65 in this thread, and I get a lot of handwaving about how impossible it is.

Let me try again. In the test you're referencing, I started with a 48 kHz, 24-bit file, resampled it to 96 kHz and back to 48 kHz a number of times, the end result being a 48 kHz 24-bit file, same as the original. The difference is then calculated by simply subtracting corresponding samples in the original and the processed files.

 

You are asking me to start with a DSD256 file and downsample it to some normal PCM rate. Then what? The sample rates differ, so subtracting samples is not possible. Convert the downsampled file back to DSD? Then we'd be comparing sigma-delta modulators more than anything else, and besides, DSD being a 1-bit format means arithmetic (such as subtracting one sample from another) isn't possible.

[mansr]

49 minutes ago, Jud said:

Yes, downsample to 16/44.1, using filtering that would be typical in the industry for mass market music.

I don't have the (expensive) DAW software likely to be used by commercial studios. Would you accept that SoX resampling is probably of similar quality?

 

49 minutes ago, Jud said:

Then straight to the spectrogram comparison, as I imagine all the bouncing back and forth was to prove a point that I didn't need proved the first time, much less again.  What I'd like to get is some idea of how much if at all the spectrogram of a typical Redbook product differs from the SDM'd signal that was the initial stage in the digital recording process.

The purpose of the back and forth bouncing was to show that the resampling, even after many iterations, causes only a barely noticeable change in the passband. What makes you think it would be any different with a DSD source?

[mansr]

6 minutes ago, Jud said:

It depends on the parameters used. Can you adjust the parameters to result in something like the "half-band filters" sometimes mentioned on the forum?

Sure, I can do that.

 

6 minutes ago, Jud said:

Only the quality of the filtering used - see above.

5 minutes ago, Jud said:

No, not nearly. Only that this is somewhat similar to the initial SDM form of the signal in the workstation, ADC, whatever, before decimation to the final product.

Most studios probably run the ADCs at 48/24 or 96/24, using the chip's internal decimation filter. Are you looking for an illustration of how these perform compared to high-quality software algorithms? 

[mansr]

7 hours ago, Jud said:

Yes. I'd like to know what the level of measurable difference, if any, would be for a typical recording chain between the original signal and a Redbook result (Redbook because I'm not interested at the moment in mp3 or other lossy compressed formats).

What do you mean by original signal? Typical recordings are multi-tracked at 44.1/48 kHz with a substantial amount of processing and mixing before a CD release eventually pops out. Comparing any individual microphone feed to the final CD wouldn't make any sense. It's not supposed to be the same.

[mansr]

22 minutes ago, manisandher said:

Here are two 16/44.1 captures, from the analogue output of my DAC. One is the DXD file (downloaded from 2L) played back natively (no upsampling, filtering or SDM) and the other is the 16/44.1 file (downloaded from 2L) upsampled to 352.8 using XXHighEnd's AP filter (but still no SDM).

 

2L-092, Jan Gunnar Hoff, Living, track 01:

 

https://drive.google.com/open?id=1T3rUw-f42ef5xWlXkj4IiR5pgLdQyoYX

 

https://drive.google.com/open?id=188f-KoMAkg6-krSEei19VVaoe4DA8qC_

 

I believe there is an audible difference between these two captures; not as much as when the DXD and 16/44.1 files are played back directly, but it's still there. (I actually did 3 quick blind comparisons and identified the DXD capture correctly each time - not scientific, but then again, even 10 blind ABXs wouldn't convince some people around here 😉.) The DXD file played back natively has more clarity and is more strident. Just more realistic. And yet, taking a cursory look at their respective spectra, they look pretty much identical to me.

One of them clearly has rolled-off highs. This is the difference of the spectra:

 

The recordings contain mostly noise above ~10 kHz, which is why the difference also turns into noise there.

 

Which DAC did you use for this? What does the "AP" filter do?

[mansr]

17 minutes ago, manisandher said:

A couple of quick questions:

 

1. So, there's a maximum of 0.4dB difference at ~11kHz - am I reading this right?

Right.

 

17 minutes ago, manisandher said:

2. Would you expect me to be able to hear this difference?

Assuming you can indeed hear a difference, this is as likely as any.

[mansr]

1 hour ago, jabbr said:

The statement above belies a fundamental difference between your and my desired music and philosophy. No doubt we will never agree. I’m fact it is exactly my desired audio experience to listen to vocals sounding as close to the mic feed as possible. For me it is supposed to be the same (or at least give me as close to that illusion as is possible)

It's not about what I or you want. Jud specifically said he was interested in "typical" recordings. In a typical music release, the final master is not intended to sound the same as the microphone feed. That's just how it is, like it or not.

[mansr]

1 hour ago, Jud said:

Original SDM signal in the ADC/workstation immediately upon initial digitization before decimation. My notion is the live DSD256 recording would serve as a very rough approximation.

Most ADCs use a multi-bit sigma-delta A/D converter. The data it produces is then digitally processed on-chip to produce PCM or DSD output. The 1-bit modulator providing the DSD output may well cause at least as much "damage" as the decimation filters used for PCM.

[mansr]

18 minutes ago, Jud said:

Certainly a possibility, though one would hope the folks doing the recording for the labels sold on NativeDSD would be sufficiently finicky to use a reasonable modulator.

The modulator is whatever the chip designers at TI, AKM, or wherever concocted.

 

18 minutes ago, Jud said:

If you have a closer proxy to suggest from which you can derive a spectrogram (potentially playable on a consumer DAC - I have some thoughts about other things I might like to do to compare the files), of course I'd be interested.

I'm still struggling to understand what exactly you're looking to achieve. We already know that the noise level in a 24-bit recording is determined by microphone preamp, not the ADC. What does it matter if a couple of the lowest bits are altered by the decimation filter?