Jitter Correlation to Audibility
May 6, 2014 at 6:41 PM Post #286 of 361
That sounds like even MORE fun!
 
May 6, 2014 at 7:35 PM Post #287 of 361
  just don't sue me if you have to turn it up to >90 dB SPL to overcome room noise from hiding the low tone


What? I can't hear you.
 
Oh, by the way, my lawyer should be contacting you soon....
wink.gif

 
Cheers
 
May 6, 2014 at 8:49 PM Post #288 of 361
 
So what if I made a sound clip with a 500 Hz square wave at -86 dB super imposed onto 6--10 kHz noise with RMS at -6dB. Is that the situation you are describing?
 
Cheers

Here's m-code:
 
Code:
 % sampling frequency f=44.1e3; t=[1:10*f]/f; % create the signal at 0dB FS ys = 2*double(sin(2*pi*500*t)>=0)-1; % create gaussian noise (white) at 0dB RMS ynn=randn(size(t)); % FFT stuff for bandpass filtering the noise  NFFT = length(ynn); N=NFFT; n = [1:NFFT];  Fnn = f*(n-1)/NFFT;  Fnn((n-1)/NFFT >= 1/2) = Fnn((n-1)/NFFT >= 1/2) - f;    Ynn = fft(ynn',NFFT, 1)/N; % bandpass filtering from 6kHz--10kHz Flow = 6000; Fhigh = 10000; Ynnf = Ynn; Ynnf(Fnn<=Flow & Fnn>=-Flow)=0; Ynnf(Fnn>=Fhigh | Fnn<=-Fhigh)=0; ynf = ifft(Ynnf*N,NFFT,'symmetric')'; % synthesize the two signals, the squarewave at -92 dB FS and the bandlimited noise at -12 dB FS dBs = -92; ys = ys * ((10^(dBs/20)) /std(ys)); dBn = -12; ynf = ynf * ((10^(dBn/20)) /std(ynf)); y = ys + ynf ; % plot waveform and spectrum figure(1); pwelch(y,[f*2],[f],[],f); set(gca, 'Xscale', 'log'); figure(2); plot(t, y); xlabel('Time, [s]'); ylabel('Amplitude, [-]'); % note, the noise energy is distributed across 6--10kHz... one must integrate the power to get the total noise energy. % check dB level disp([num2str(20*log10(std(y))), 'dB']) % >> -12dB % write to audio file wavwrite(y,f,'test_500Hz-92dB_8kHzNoise-12dB.wav') wavwrite(ys,f,'test_500Hz-92dB.wav') wavwrite(ynf,f,'test_8kHzNoise-12dB.wav') [ys_import, Fs, nbits] = wavread('test_500Hz-92dB.wav'); figure(3); pwelch(ys_import, f*2, f, [], f); % end code

Figure 1. Spectrum of  square wave (-92dB) + noise(-12 dB)
 

Figure 2. Snipit of the waveform of  square wave (-92dB) + noise(-12 dB)
 

 
Figure 3. Spectrum of just the square wave (-92dB) audio file.
 
Here are the files created (in the spoiler):
GUYS, WARNING: DON'T BLOW YOUR HEARING OUT TESTING THESE TRACKS. AFTER LISTENING TO THE SQUARE WAVE TRACK, MAKE SURE YOU ADJUST YOUR VOLUME BEFORE LISTENING TO ANYTHING ELSE. ALWAYS START WITH THE VOLUME DOWN AND THEN TURN IT UP FROM THERE.
Just the 500 Hz square wave @-92dB
test_500Hz-92dB.wav
 
 
Just the 6--10kHz noise @ -12dB
test_8kHzNoise-12dB.wav
 
The whole test signal
test_500Hz-92dB_8kHzNoise-12dB.wav
 
The ABX test to do would be trying to differentiate just the noise vs the noise+signal.
 
Maybe someone who makes cool color-coded spectrograms wants to plot the spectrum of these files?
 
Cheers
 
May 6, 2014 at 9:36 PM Post #290 of 361
  Are you sure you got the right files? They all look like the same spectrum to me...
 
 
http://cdn.head-fi.org/4/45/455ab1b7_audacity_spectra.jpeg


Uh-oh. Good catch. let me fix them.
 
Cheers
 
EDIT: ha! bug in the code. I wrote the noise+signal out to all of the file names. Doh!
 
EDIT2: Fixed (I hope). Let me know if there are outstanding issues remaining. Thank you cjl
 
Aug 22, 2014 at 3:02 PM Post #292 of 361
what part of human hearing threshold do you not understand? - especially the part where the threshold is much higher for bass frequency
 
you can also reason from masking which follows an ~ 4th order curve so you wouldn't expect to be able to hear anything much below -80 dB below the high tone SPL at 1/10 of the high tone frequency which takes you below our ~3kHz lowest noise threshold dip
 
the masking cuve @ (1/30)^4 gets you down to -120 dB of the 12 kHz tone at 400 Hz - much below that and our noise threshold is rising quickly with lower frequency
 
Aug 22, 2014 at 3:49 PM Post #294 of 361
search should find noise threshold curves, how much worse they are for headphone listening due to microphonics of having them coupled to your muscle tremors, pulse
 
these should be basic 1st things to look for in any question of "is this audible"
 
 
I didn't like the Wikipedia charts - try: https://www.etymotic.com/publications/erl-0096-1997.pdf
 
Aug 22, 2014 at 4:27 PM Post #296 of 361
  I'm trying to understand jitter better and so far I thougt that it's not the same as noise.
I also asked because the guy who measured this says that it's audible.


the main problem is that we tend to call jitter any delay in the signal at any time along the audio chain, so the magnitude, type and effects can be all and nothing. still one thing is always true, if in the end there is a delay in the signal, what will count mostly is it's size compared the frequency of the played sound. high frequency being faster, the delays will always start manifesting there. what could be half the period of a 20khz signal will be peanuts on a 50hz wave.
 
Aug 22, 2014 at 4:43 PM Post #297 of 361
  I'm trying to understand jitter better and so far I thougt that it's not the same as noise.
I also asked because the guy who measured this says that it's audible.

 
Jitter is a frequency modulation of the original signal. The result is side bands being produced in the presence of a signal. The magnitude of the side bands are directly related to the mangitude of the original signal and the characteristics of the jitter. The addition of frequency content that was not part of the original signal but that is proportional to the the source signal amplitude is called distortion. Hence, jitter manifests itself as distortion. 
 
THD+N plots tell you what the combined distortion products + noise is for a given input signal; hence, they include all effects from noise, harmonic distortion, and jitter. You will notice that identifying jitter on these plots is extremely difficult because it is often swamped out by other factors.
 
Cheers 
 
Aug 22, 2014 at 4:46 PM Post #298 of 361
Jitter in the degree it exists in even the cheapest home audio equipment is inaudible. Most of the time, it's many many times below the threshold of audibility.
 
Aug 22, 2014 at 6:33 PM Post #299 of 361
 
THD+N plots tell you what the combined distortion products + noise is for a given input signal; hence, they include all effects from noise, harmonic distortion, and jitter. You will notice that identifying jitter on these plots is extremely difficult because it is often swamped out by other factors.

 
That's good to know, thanks!
 
Aug 22, 2014 at 10:32 PM Post #300 of 361
  I'm trying to understand jitter better and so far I thougt that it's not the same as noise.
I also asked because the guy who measured this says that it's audible.


The vast majority of audio enthusiasts don't actually know what jitter is. But what they are told is that the product they are evaluating doesn't have it and that competing products have a detrimental amount of it. Some audio enthusiasts stop there and feel smart about themselves and their informed product selection. Others demand to be shown exactly how jitter affects output. Again they are presented with competing graphs - they are told that some look worse and some look better. They are told not to think about it too much, because it's far too complex to explain, but to use their ears. After all, what do we really know about jitter? It's like ****ing magnets, how do they work?
 
Yet there was still hope. A brave clan of enterprising young minds began secret investigations into the matter in the early twentieth century. Toiling away in secret labs of major telecom players of yore, your Ma Bells and your DARPAs, they discovered the secret lair of the jitter beasts. They fought valiantly and by the late 70s they thought they had defeated jitter through the advent of phase-locked loops. Alas, while they might have run the beast out of town, it was beaten and not dead. Jitter crawled its way to a deep dark cave to recuperate. There it was found by a group of trolls, some were withered husks of researchers who felt uncompensated and struck out on their own, others were trinket merchants looking for something shiny to peddle to the unwashed masses.
 
These trolls began to tell tall tales of the jitter beast having grown stronger than ever, of it hiding beneath the noise floor of the distortion graphs just waiting for an opportunity to strike unsuspecting listeners with its dreaded digititus. In truth the beast had licked its wounds and surrendered to a subsistence of scraps - a bad USB implementation here or an unstable crystal oscillator there. It no longer wished to face the fury of thousands of engineers grinding it down into the nether depths of Brownian noise.
 
Yet the villagers were scared, for each successive generation had been told of the monster by their parents. That if they were naughty and spent less than a thousand dollars on their transport the jitter would unleash the fury of a thousand Death Magnetics upon their ears. So the youngsters thought themselves wise and purchased jitter insurance from the elder trolls. Each one promised a lower period than the next - a nanosecond? Why you can get it down to a picosecond! A picosecond? Why with our current technology we can go femtoscale!
 
And so it was, perhaps in the same spirit of their right to bear arms that audio enthusiats purchased mighty DACs and Cables to whip the beast into submission (if they ever ran across it). Some would even hold championship contests where contestants would be sent into the field to see who could survive the beast's onslaught and be declared winner of the Snipe Hunt. The flaxen-haired golden-eared champions would pass their judgments onto the masses as sponsors of the order of the trolls, directing the purchasing decisions of puny losers.
 
Sadly the brave audio engineering knights had grown old and weary, they did not wish to fight a beast they knew had been slain, nor did they wish to interrupt or spoil the fantasy worlds of bright-eyed youngsters. Here we are -- everyone has heard of jitter but no one wishes to dust of the archival memoranda of the knights, for they know that as long as they spend enough they remain safe.
 
Even now I fear having told you of this ancient order, for I am but a simple bard who can't afford the Kingdom's tax on having an opinion. Even in this tucked away ghetto of sound science I fear the trolls will find me. Please, if I am no more after this post remember my tale.
 

Users who are viewing this thread

Back
Top