Can You Trust Your Ears? By Tom Nousaine

[j_j]

CD players generally do a pretty good job of that. If you stick to mainstream recordings in mainstream formats, you are pretty safe.

I've seen a few where the filter that is supposed to cut out the delta-sigma noise was pretty weak, and a couple DAC's that had 300+kHz tones leaking out. This is, to use a very simple term, "unacceptable".

I have yet to know where the tones came from. But there's an example of something that needed to be filtered out.

 

[Don Hills]

I recall seeing a website somewhere with the results of passing a simple tone sweep through many different DACs. Many of them performed surprisingly poorly. If they can't get that right, they'll likely perform poorly for more complex signals too.

 

[amirm]

I recall seeing a website somewhere with the results of passing a simple tone sweep through many different DACs. Many of them performed surprisingly poorly. If they can't get that right, they'll likely perform poorly for more complex signals too.

I can try to duplicate that. What was the frequency range of the sweep and sampling rate. Do you remember?

 

[Fitzcaraldo215]

Would this be a (good?) reason for limiting/eliminating ultrasonic frequencies (>20kHz) to the speakers?

Personally, I think you would be overreacting to insufficient evidence out of fear, uncertainty and doubt. Again, I have seen no measurements of systems including the speakers where this is shown to be a problem with hirez recordings or where tweeters have been blown.

 

[Fitzcaraldo215]

It is not just theoretical. It shows up in the professional literature. I've encountered situations when it happened with gear that I was using. I've seen where it happened to others or was likely to have.

It definitely happens in some real world systems that well-meaning people assemble for their own use. Try the test tones from my post of 10:03 am EDT today. Note that they aren't even truly ultrasonic. I suspect my "keys jangling" sample can activat this problem, if and when it exists.

.

I am perhaps not as well educated as you on this phenomenon. If it shows up in the professional literature, as you say, can you direct me to any discussions of it, preferably with measurements? If it "definitely" shows up, those measurements should abound. Yet, I have not seen them.

What would I be listening for in your keys jangling sample? The Circle of Confusion regarding recorded sounds vs. how they should sound might seem to interfere.

 

[j_j]

I recall seeing a website somewhere with the results of passing a simple tone sweep through many different DACs. Many of them performed surprisingly poorly. If they can't get that right, they'll likely perform poorly for more complex signals too.

Yep.

Amir, do you have access to Matlab?

I can tell you how to generate some stationary signals that also break DACs.

 

[j_j]

Personally, I think you would be overreacting to insufficient evidence out of fear, uncertainty and doubt. Again, I have seen no measurements of systems including the speakers where this is shown to be a problem with hirez recordings or where tweeters have been blown.

Well, you know, I have seen cases where 22 and 23khz tones have produced a 1kHz tone that decentered a well-known tweeter enough to mush up the voice coil former and create buzz.

I doubt it was published, and yes, everything was within continuous power limits.

And I still stick to ribbons nowadays. There is a connection here, you betcha.

 

[amirm]

Yep.

Amir, do you have access to Matlab?

I can tell you how to generate some stationary signals that also break DACs.

Sure. Just give me the prescription!

 

[j_j]

If it shows up in the professional literature, as you say, can you direct me to any discussions of it, preferably with measurements? If it "definitely" shows up, those measurements should abound.

This is old hat for tape decks. Seriously. This is nothing new at all, there are reports from 1940's. New equipment is better. "better" is not perfect. You're asking for people to "report" phenomena that have been known about for 60 or so years.

 

[j_j]

Sure. Just give me the prescription!

See your "conversations", i.e. "pm's"

 

[DonH56]

Me too, please? I have Matlab...

 

[Cosmik]

Personally, I think you would be overreacting to insufficient evidence out of fear, uncertainty and doubt. Again, I have seen no measurements of systems including the speakers where this is shown to be a problem with hirez recordings or where tweeters have been blown.

Couldn't it be seen as the other way round? i.e. defaulting to high resolution recordings instead of the cheaper, simpler, lower-bandwidth alternative, despite there being no evidence that it sounds better, is the reaction to FUD? The happy CD user is the supreme FUD rejector!

 

[j_j]

Me too, please? I have Matlab...

Look in www.aes.org/sections/pnw for meeting recaps for the sampling rate conversion tutorial. There is a "buzz tone" that puts all of the first-order IM at the same (low) frequency. Which is "icky". but it shows up the problems.

Sometimes if you look at the spectrum you'll also find halfband filters, regularity problems, ...

 

[DonH56]

Ah, I probably have something in my files already to do that, but will look at the site... Frequency binning is a test problem as well as a problem in the real world.

I keep thinking I should renew my AES membership. Then I think, yah, whilst typing at 8:30 pm, waiting for some test results and reviewing a datasheet update. Work is dominating Life lately.

I won't touch the "regularity problems"...

Thanks!

 

[amirm]

Me too, please? I have Matlab...

Here is the link: http://www.aes-media.org/sections/pnw/pnwrecaps/2016/jjsrc_jan2016/

The PPT link is in the middle of the page.

Here is the code:

clc
clear all
close all

len=2^18;
fs=48000;

xt(1:len,1:2)=0;

for ii=179:487:(len/2)
    phi=2*pi*rand();
    xt(ii,1)=1/ii*(cos(phi)+i*sin(phi));
    phi=2*pi*rand();
    xt(ii,2)=1/(ii+2)*(cos(phi)+i*sin(phi));
end

xt(len:-1:(len/2+2), 1:2)=conj(xt(2:(len/2),1:2));
x=ifft(xt);

x=x/(max(max(x)))*.99;

wavwrite(x,fs, 16, ‘bzzzt’)

---

Wavwrite is deprecated in my version of Matlab so I used: audiowrite('buzz.wav', x, fs)

 

[amirm]

And the spectrum of the created file:

And the log version:

 

[Blumlein 88]

And the spectrum of the created file:

How about a copy of the wav for downloading for those of us not doing matlab?

 

[amirm]

Here you go: https://www.dropbox.com/s/3z75ikm67onhyph/buzz_01.flac?dl=0

 

[j_j]

Oh, and

DO NOT LISTEN TO THIS SIGNAL OR PUT IT THROUGH SPEAKERS. IT IS A TEST SIGNAL. EVEN AT LOW LEVELS IT SOUNDS HORRID.

Note also you can loop this signal by simply putting the end against the beginning. You don't need to overlap-add or window because of the way it's generated. You can make at least 4 copies, and capture that much from the DUT.



Also, when you analyze the results of this signal, use at least a 2^20th length FFT. No, that's not too big, and you will have a clear look at what your system is doing.

 

[Arnold Krueger]

I am perhaps not as well educated as you on this phenomenon. If it shows up in the professional literature, as you say, can you direct me to any discussions of it, preferably with measurements? If it "definitely" shows up, those measurements should abound. Yet, I have not seen them.

This is a standard audio test called CCIR or CCIF IM (the same organization had two names over the life of the standard, it is that old) IM distortion. It was more commonly used to test audio gear in the past. Analog tape machines like to make lots of it (nobody was able to disabuse them of this audible fault which is one of may reasons why they have become disused except as boutique items), but they are far from being unique. The CCIR's name on it might confer the impression that there is no serious question about its applicability in learned professional circles.

What would I be listening for in your keys jangling sample? The Circle of Confusion regarding recorded sounds vs. how they should sound might seem to interfere.

IM is pretty distinctive to trained ears. But it is far from sounding exactly the same every time. I think of it as a sort of white sound, as in white noise.

Please download "addistortion.zip" from http://www.audiosignal.co.uk/freeware.html and apply its contents to various .wav files including my "keys jangling" in various ways and degrees to hear more what added nonlinear distortion sounds like.

Here is a link to a downloadble CCIF IM distortion test file: https://hydrogenaud.io/index.php?action=dlattach;topic=114871.0;attach=11908 . JJ's cautions apply in spades to this file. It can be used to roast tweeters to a fine state where the magic smoke escapes, if you are fast enough to catch it!

Cooledit Pro and Audition users have a command called Edit, Mix/Paste with the Modulate option which can be used to add various orders of nonlinear distortion to a file. Copy the file and mix it with itself as needed to obtain increasing orders of distortion. Add the distorted file back to the original file in various amounts using Edit, Mix/Paste with the Overlap option to obtain various amounts of distortion.