The myth of high-res audio

nsaspook

Joined Aug 27, 2009
13,081
6 channels, WOW... I bet that makes a difference
That's mainly a demo. I have some 24/192 recordings of Yes that are truly fantastic. I'm not a nut about hi-res audio but listening to music is more than just bits. It's a mixture of the mental and physical. Thinking it should sound better actually makes it sound better to you inside your head (the only place it matters to you). Viewed purely with math and electronics in a blind study the difference should be (and usually is) undetectable but that's not how we enjoy music. We enjoy music knowing the quality of the instuments it's played on. That knowledge extends to our electronic devices just like it does with a great artist using great instuments to make music. I know the time and effort it took to master those hi-res recording and even if I know it would 'measure' the same at my ears at 16/44 with a lab quality microphone knowing it was actually 24/192 makes it 'sound' better.
 

atferrari

Joined Jan 6, 2004
4,764
We enjoy music knowing the quality of the instuments it's played on.
So, replacing the front panel of whatever you are using to reproduce a program, I could induce you to a miserable condition. Very bad!

Obviously you are not eligible for blind tests.
 

nsaspook

Joined Aug 27, 2009
13,081
So, replacing the front panel of whatever you are using to reproduce a program, I could induce you to a miserable condition. Very bad!

Obviously you are not eligible for blind tests.
I don't think I would be miserable but if someone painted my speakers pink I might be less happy about my music if I could see them during a listening session.
 

WBahn

Joined Mar 31, 2012
29,978

Thread Starter

ErnieM

Joined Apr 24, 2011
8,377
I don't think I would be miserable but if someone painted my speakers pink I might be less happy about my music if I could see them during a listening session.
Back in the 80's I took a good pair of AR speakers, painted them black, dyed the grill fabric black and in large stencil lettering wrote "THIS IS NOT ART" across the face of each one.

I found it to be a significant improvement overall.
 

cmartinez

Joined Jan 17, 2007
8,218
It also offered a very powerful physical demo that dispelled several residual myths I didn't even know I was still harboring.
:eek: That statement, coming from you, makes me realize that this subject is much deeper than what is covered on the video.
 

wayneh

Joined Sep 9, 2010
17,496
I watched the video and was able to follow most of it. Very interesting. But one thing that always eludes me is how the analog signal is re-created from the samples. How is the interpolation done? The video clarifies that it's not a stair step, but what is it? A straight line point-to-point wouldn't work, although it would likely be better than a stair-step. Does the D-to-A converter try a number of sine wave combinations to produce a "best fit" solution? I can imagine doing this but it cannot afford too many iterations without running out of time.
 

WBahn

Joined Mar 31, 2012
29,978
I watched the video and was able to follow most of it. Very interesting. But one thing that always eludes me is how the analog signal is re-created from the samples. How is the interpolation done? The video clarifies that it's not a stair step, but what is it? A straight line point-to-point wouldn't work, although it would likely be better than a stair-step. Does the D-to-A converter try a number of sine wave combinations to produce a "best fit" solution? I can imagine doing this but it cannot afford too many iterations without running out of time.
Different DACs do different things. Some are zero-order hold while others are linear-interpolating. Usually you have a reconstruction filter that bandlimits the output. If the input and output are properly bandlimited, then the output will reproduce the input quite handily.
 

Thread Starter

ErnieM

Joined Apr 24, 2011
8,377
I don't see how the output of the DAC can be anything but a stair step, just as the output of the lo pass could possibly be steps.

I'm going to have to watch that part again to see what he really says.

Edit: at 7:32 he sez the stair step is "not a finished conversion and not the signal that comes out"

So there is (or could be) a stair step somewhere in there at some intermediate point.
 
Last edited:

wayneh

Joined Sep 9, 2010
17,496
I'm under the impression that that's exactly what's happening, a linear interpolation between adjacent points. Why wouldn't that work?
The error I picture in my head is a sine wave sampled on either side of the peak. If you just draw a line that interpolated from point to point, you chop off the peak in between.

I think that part of the answer is that, because you're sampling so fast, that type of error becomes inaudible. But in his demonstration he shows how even a "slowly" sampled wave can be reconstructed with good accuracy. That's the part I don't get. It seems like you have to assume a sine wave at a single frequency in order to fill in the blanks.
 

nsaspook

Joined Aug 27, 2009
13,081
Back in the 80's I took a good pair of AR speakers, painted them black, dyed the grill fabric black and in large stencil lettering wrote "THIS IS NOT ART" across the face of each one.

I found it to be a significant improvement overall.
I've got a set of old AR-2a speakers for my shop stereo.
They still have great bass response and the cabs are built like a Brick*House.
 

cmartinez

Joined Jan 17, 2007
8,218
The error I picture in my head is a sine wave sampled on either side of the peak. If you just draw a line that interpolated from point to point, you chop off the peak in between.

I think that part of the answer is that, because you're sampling so fast, that type of error becomes inaudible. But in his demonstration he shows how even a "slowly" sampled wave can be reconstructed with good accuracy. That's the part I don't get. It seems like you have to assume a sine wave at a single frequency in order to fill in the blanks.
Well, to me it seems inevitable that some sound quality will be invariably lost, due precisely to the finite amount of data used to represent any waveform. Of course, the larger the number of data the more accurate the result. But since we're accommodating the number of samples to best match a a waveform with a maximum frequency of about 40 Khz, then that "chop" (is it the same as "clipping"?) will be negligible. Am I making sense?
 
Top