Does digital audio work like digital images...ie more bits for highend?

Most A/D-converters made in the last 15 years are of the Delta/Sigma type. I don't think any of the major converter-circuit manufacturers (AKM, Texas Instruments, Cirrus Logic, etc.) have used any other approach since the early 1990's (though all modern models use multi-bit modulators instead of 1-bit modulators).

You'll be hard pressed to find modern equipment that doesn't use that type of A/D-Converter.

The you should probably familiarize yourself with The Fourier Theorem as well, since that is fundamental to understanding digital audio:
Any waveform, no matter how complex, can be represented as the sum of various sine waves.

That's why we can accurately describe any waveform that is within the bandwidth.

No offense, but this is so basic that I sincerely hope that you're already familiar with it.

Yes, but the required bit depth of any material intended for playback is related to the intended peak amplitude of the playback.

Allegedly!? 16 bits can provide more than 96dB. With noise shaping and colored dither you can have over 120dB of dynamic range in the most sensitive frequency ranges of human hearing.

If everything is done right and the recording is within the dynamic range that won't happen. It sounds like you're getting distortion so it may be a problem with dither, clipping or some other hardware/software related problem. In a properly dithered signal the quantization error will not be correlated to the signal (distortion), only white noise.

 

Where do the intersample peaks found in the real world fit into this fundamental description?

 

Hi FalloutBoy,

I wasn't referring to sigma-delta.

Consider this: Perhaps I'm already familiar with the theory and interpret it differently than you do.

No offense but I pay more attention to the empirical evidence coming from my loudspeakers than to personal interpretations of theories expounded in internet forums.

I did not say anything about dither when I talked about hearing a coarsening in 16-bit at lower levels? I was talking about straight 16-bit.

I have been posting about my own experience and not about what someone's theory (or another's interpretation of same) claims to be true.

Since it is plain you and I hear it quite differently, let's just agree to disagree.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

 

Hi Grant,

Yes. But I'm open to listening to your perspective and why you might feel differently.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

 

What a great attitude.

 

Hi Doug,

Thank you.

Who among us (aside from politicians and a few shoot-from-the-hip internet audio "experts") does not seek to learn?

Presented with an opposing point of view and a convincing argument, I have no problem in making a 180 degree turn. I look forward to hearing Grant's perspective.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

 

I don't see how intersample peaks have any relevance to it.
Intersample peaks is a practical issue that is easily avoided by using modern converters and modern software.
It also helps to stay away from "loudness mastering" (the biggest problem in audio today).

 

Ok, maybe you could elaborate a bit on how you interpret it. I'm genuinely interested.

Maybe I'll learn something new.

 

The reasoning behind my "chunky" comment was that if a DAC/ADC is restricted to using 0 to 5 volts to operate with (like many other digital-based circuits) then a 24 bit convertor would be designed to break up that 5V input into smaller divisions vs. a 16 bit's convertor which also has only 5 volts to work with.

But yea, if all digital convertors were designed to have access to any power supply voltage they needed (say, zero to 100 volts instead of just 0-5), I can see where a 24 bit's convertor's 16th bit would share the same value as a 16 bit convertor's.

 

Hi FalloutBoy,

You missed my point.
I base my posts on what I hear coming from my loudspeakers when I listen to recordings I make myself of real musical performances and to albums I master for other folks. I base them on direct experience with a broad assortment of some wonderful hardware and software. I do not base them on theories (which ultimately, I find of limited interest and then only when they explain what I hear).

The theories and writing you cite can be interpreted to suggest 16/44 is perfect and nothing more is needed to accurately capture the sound of Life. These interpretations are meaningless to me when the flaws in the very best 16/44 (good as it can be) are plainly audible (to my ears anyway, as well as several folks I know), particularly when compared to decent 24/96 (I won't even compare with superb 24/192, which is what I am lucky enough to be using now).

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

 

I mention intersample peaks because they seem to show how the application of the theorem with real-world converters tends to go awry.

I imagine that I use modern software for recording (will Audiotion 3 do?), and I have noticed intersample peaks on some of my recordings.

There has been a lot of talk lately on several threads here about iZotope's RX software. This has an option in the Preferences > Misc menu where you can opt for the program to show the analog waveform and this often shows peaks where there seem to be none in the digital one.

I agree, but the highest peaks I have seen have been when recording a DSD signal through the analog outs of my Pioneer universal player at 192/24.

 

FalloutBoy, I'm not sure you're aware of who Mr. Diament is. I can assure you Barry is very well aware of the loudness wars, as he's made a career out of avoiding them. His work is some of the finest I've heard, so I tend to side with him on this issue.

 

It's was not really the fault of the older D/A-converters. The waveforms they are asked to recreate exceeds the bounds specified by the systems being used. The levels should not be that high to begin with and with proper mastering it would never have been an issue.

There are much better examples of bad implementations of digital audio, like 1-bit systems.

 

I've never claimed that 16/44.1 is perfect, and I have nothing against using higher bit depths and sample rates.
What I do have issues with is the explanations that people present for the benefits of using them. And I do feel compelled to point out when they are based on basic misunderstandings of digital audio.

 

I think you misread my post. I was not replying to Mr. Diament, but to Mr. Metoo.
And I didn't accuse anyone in particular of "loudness mastering".
I just pointed out that it is a major problem and I think Mr. Diament (and hopefully everyone else) would agree with me on that.

 

Regarding intersample peaks: is this referring to peaks - which AFAIK are also sine waves - located between two sample points?

Because if an ADC "skipped" over such a sine wave, then that simply means its sampling frequency was too low.

BTW: for me anyway, once I saw diagrams of the Nyquist theorem it was much easier to understand.

 

The Moran and Meyer paper has been widely discredited. You need to record in hirez to understand the clear benefits of extra word length and sampling rate.

 

We always have these discussions of Nyquist yet the practitioners here on the board really understand and hear better playback of hirez.

These arguments go nowhere ultimately.

I can only suggest that you keep an open mind and listen to recordings where the hirez and redbook versions of the same album are mastered similarly and play them back. On a decent system, you will hear clear improvement in the hirez playback.

 

44.1/16 is only perfect in the sense that it covers the audio band. However recorded audio has frequencies in it, even if inaudible, that are greater than 22khz, and they must be removed before going to 44.1/16. It is the audible effects of these anti-alias filters that, to me, benefit from the move to 96khz or 192khz as they can be made much less audible. So to me it's not the reproduction of higher frequencies that is the benefit from the higher sample rate, but that the necessary filters can be made less audible by virtue of that higher sample rate.

 

Maybe you could direct me to the relevant papers in the peer reviewed literature?

I know that there are benefits to higher bit depths when recording, and in some cases there are benefits to higher sample rates as well.
For playback the benefits are more questionable, but under some circumstances there are benefits there as well.

I'm not questioning anyones use of high bit depths and sample rates, just the technical explanations they use to explain the benefit of them.

 

If there are benefits to recording in hirez, why would there not also be benefits to the playback of same?

 

I don't believe that 16/44 even does a good job of covering the audio band. There's simply not enough information there imho to capture transients.

 

That's not as much of an issue with modern digital anti-aliasing filters.

Dan Lavry explains the pro and cons of higher sampling rates in this paper:
http://www.lavryengineering.com/documents/Sampling_Theory.pdf

 

Lavry is a smart guy no doubt but my experience on recordings does not match what he makes a case for in theory. He would suggest that 24/176 does not offer up advantages. But in practice that is not the case.

 

Me too, mostly because I'm trying to figure out how there could possibly be other ways of interpreting it...