Testing audiophile claims and myths

[ab initio]

  I keep getting resistance to my "180 degree 2nd harmonic phase shift is equivalent to polarity inversion" argument for my test tone
 
how about this visualization  - I add a phase dependent offset (magneta) to the dynamic sum (sum in blue, magenta is "0" line for blue) to show the +/-90 plots mirrored about the x axis so polarity should be clear by eyeball
 
I move both the fundamental and the 2nd phase to center the +/-90 sum's peaks on the y axis - "48" in the algebra pane is just the value of alpha when I did the screenshot
 
but the trick is the relative phase of the 2nd harmonic is shifted a total of 180 degrees relative to the fundamental over the course of the animation - showing that it results in polarity inversion
 
we do all agree that green and red are opposite polarity?
 

 

 
membership at diyAduio isn't onerous - is  free, so far I get <1 official email from the site per month - you don't have to show your email address in your public profile
 
it is about designing, building, measuring - some even claim to listen, too - several industry "names" in both electronics and loudspeakers regularly participate
 
the headphone subforum isn't a scratch on head-fi though - even compared to just the diy section here - but there are more amp than cable threads at diyAudio

Wait, why did you add an offset? Shouldnt the dc component be zero?
 
Cheers

 

[jcx]

the varying vertical offset is just a visual aid I hoped would make the sine harmonic addition picture easier to read - less lines crossing over to keep track of - our eyes are really happier looking at a mirrored figure/seeing the inverse symmetry when the "fold" doesn't have overlap, is based off the mirroring line
 
green, red should be very easily seen to be mirrored about the x axis == inverses, inverted in polarity waveforms just from shifting a sine and its 2nd harmonic
 
obviously the audio doesn't want the offset - see my earlier Spice sim waveform picture - which didn't seem to convince everyone

 

[analogsurviver]

I listen almost exclusively to acoustic music... classical and jazz. Much of it isn't multi-miked. I've never been able to discern absolute phase. I've been told now that hearing it on speakers is impossible. I have some first class headphones right now, and I've listened to binaural recordings on them. They sound very lifelike, but still I doubt that I would be able to discern absolute phase differences.

I'm asking clearly how I can set up a test where I can reproduce your claims of being able to easily hear differences in absolute phase and you're giving me unclear answers full of anecdotes and off topic comments on other things. I give up. I think you're making this all up as you go along. You're an interesting conversationalist, but I don't think you have ever attempted to verify any of your subjective impressions with objective testing.

In other parts of head fi, rambling impressions are fine. But in sound science we get to put impressions to the test to understand how things really work, nott just how we think they work. If you ever get interested in doing that, we're here to join in and help.

To be blunt how I found out absolute polarity is audible:
 
1. DC coupled high voltage electrostatic headphone amplifier with matching headphones
2. AGI 511 preamp - although not DC, it is VERY near - to arround 0.01 ( one hunreth of a Hertz ) - 3 dB
3. Various phono cartridges ( for one reason or another, it is possible to get 50/50 cartridges either in correct polarity or inversed polarity in the same batch - and they do sound different when obeying their colour code; if wired so that all produce the correct polarity, altrhough slight differences remain, it is no longer "two camps" of sound)
4. Some jazz close miked recordings , particularly on ECM label
 
Then, in for repair came a tube preamp - with comparatively to AGI VERY limited bass response. It no longer sounded "true" - no matter what. And all of a sudden, this polarity thing became next to inaudible. 
 
Then I hooked up my oscilloscope and listened to the portions of music that sonically differed the most - with both preamps. I found AGI could sustain the HIGHLY assymetric waveform from a tuba - where bandwidth limited tube preamp produced almost symmetrical output. Over the AGI, reversal of absolute polarity
( "mechanical", two parallel output posts, one wired correctly and another with reversed polarity , NO unnecessary electronics involved ) clearly produced different sound according to which output post headphones were connected. With tube preamp - all but inaudible; with everything else exactly rhe same.
 
I highlighted the LF extension of electronics; it IS important for absolute polarity, as if too low, it is easily concluded "polarity does not matter".
 
After that, I grew more attentive to this and can now hear it even with not soooo extended LF response as in AGI preamp feeding DC coupled power/headphone amp; but were it not for so good system as described above, I may well be convienced that absolute polarity audibility is a fairytale. Trouble is, the recording chain also has to be at least in the similar ballpark - ask MANY sound engineers,  they will confess to the filtering below XY Hertz to improve "clarity" and give reproduction side an easier time. 
 
It is all interelated; assume this, assume that - and it can no longer be "assumed" the recording will be of true high fidelity. That is why the likes of Sheffield, Miller & Kreisel, Reference Recordings, etc have been rebuilding the electronics of their otherwise specified microphones - it would NOT sound that good with a stock mike.
Same for the entire electronics chain from the mike to the recorder, whatever that might have been. And of course, they did not use stock "recorder" either... - enough is to see liner notes on any of these extraordinarily good recordings.
 
With sound, one can spill all the ink in the world and still not convience the Doubting Thomases; my expereince  from audio fairs, dealer's etc is a VERY simple one;
if it sounds, behind a half closed door, good enough to arouse my curiosity to investigate further by taking the whole demo, it IS good.  Seldom I get to regret the time spent at such demoes.
 
I have written this before, but will repeat it. I have a friend, actually a fellow student, whom I used to be sporadically meeting over the years/decades. I was always "audiothis/audiothat" - and it was trough his one ear in, out trough another. UNTIL he heard the proper demo of "my audio". Now, no matter how wild or outlandish claim I might lay - he simply says it must be true - because every other claim he has heard me to lay proved to be true in sound.  
 
No paper/computer/science "proof" can be more conviencing - but I ALWAYS welcome any help in establishing what helps further reducing the gap between live and recorded/reproduced sound. Trouble is also the following; HOW to measure something that is an order (or more ) better in magnitude than the measuring instruments?
 
Even more so - does anybody know of an audio analyzer capable of assesing the performance of the DUT - IN REAL TIME, for multiple parameters, with
real music and not technical signals ? That would be awesome - but probably the cost would be prohibitive for the military, let alone audio use.

 

[EthanWiner]

I keep getting resistance to my "180 degree 2nd harmonic phase shift is equivalent to polarity inversion" argument for my test tone

Maybe I didn't understand because I'm not a math type. I got the file you sent, and I see that you simply inverted the polarity. So now we're on the same page.

For the benefit of others here, as I said in my email back to you: The tonality of the normal and inverted sections sounds identical to me, but the pitch seems to go very slightly flat on portions 2 and 4 compared to 1 and 3. I know the frequencies don't really change because I looked at an FFT, and I also created an inverted version myself from the source. So I now agree there is a difference! I think it's probably a psychoacoustic effect, but it's there. At least for me.

So what do you hear as the difference, pitch or timbre?

--Ethan

 

[bigshot]

You've stacked up so many variables there, I don't know how you can possibly isolate any one thing being responsible what you're hearing... swap in this amp and that headphone and that phono cartridge and this specific recordings... Chaos. No way to know what is causing what you hear. I'm positive now that there's no way to verify what you're saying, because you haven't set up your own controls enough to know what's going on. The big red flag for me would be the phono cartridge. I'd try to get that out of the chain to make sure I'm not just hearing acoustic noise and misinterpreting it as being part of the signal.

In any case, if you have to jump through that many hoops, it doesn't matter anyway. It isn't something that can be heard during normal music listening with a normal rig. Perfect sound is achievable without having to make it that complicated.

 

[analogsurviver]

You've stacked up so many variables there, I don't know how you can possibly isolate any one thing being responsible what you're hearing... swap in this amp and that headphone and that phono cartridge and this specific recordings... Chaos. No way to know what is causing what you hear. I'm positive now that there's no way to verify what you're saying, because you haven't set up your own controls enough to know what's going on. The big red flag for me would be the phono cartridge. I'd try to get that out of the chain to make sure I'm not just hearing acoustic noise and misinterpreting it as being part of the signal.

In any case, if you have to jump through that many hoops, it doesn't matter anyway. It isn't something that can be heard during normal music listening with a normal rig. Perfect sound is achievable without having to make it that complicated.

It is SIMPLE. I do change one thing at a time - not many. For example - only preamp in an otherwise same system gets replaced at one time. Or cartridge known to have different absolute polarity ( in lack of proper test record, run-out groove with the "click" each revolution will do just fine observed on a scope ) 
 
But I will try as many options as available. And see if there is any pattern to what can be heard.
 
I have to do the normal work, like editing the recordings etc first, but time permitting will repeat that test from quarter of a century ago - with another preamp with limited LF extension vs the same type of AGI ( I was stooooopid enough to sell my first sample and was recentish lucky to get another ) and the same record. And hopefully I will be able to catch it on photo from the oscilloscope and record everything to DSD.
 
And I promised pics of square wave trough many digital devices and by "popular demand" will include some of the LP/cartridge too - this will be done first.
 
One unfortunately has to go trough that many hoops - if cartridge manufacturer screws up polarity, if recording engineer/mastering guy chops off the lows, if electronics limits bass, etc - HOW can waveform of reproduced music possibly resemble the real thing ? And we have not arrived at the end transducer, be it headphone or speakers, yet. So yes, it IS daunting, it IS easy to conclude "absolute polarity does not matter" under - call it that way - SNAFU conditions of an average recording trough an average system. But after things get, one by one, under control,  till each and every one IS under control, much more lifelike sound emerges. THAT is why I press forward  - because the whole chain, from the microphone capsule to headphone or speaker has to be put right - omit the last step and 90+% of all other previous steps taken will bring - next to nothing.
 
If this game is played right, it does not necessarilly mean it has to cost an arm and a leg either. But , of necessity, it absolutely has to be THOROUGH. It costs nothing more or less to wire the microphone/cartridge/whatever correctly - but the awareness that it does matter ( and why )  is mandatory.

 

[bigshot]

I asked for a way to discern "clearly audible" absolute polarity differences and you listed a laundry list of specific equipment that was required. You said you swapped amps and it wasn't discernable any more. Then you threw in having to know the recording chain as well. You're wiggling all over the place here. I think you're just looking for weird hoops to jump through to justify a guess about a purely subjective impression. I don't think that polarity really has anything to do with it.

This isn't how we work here in Sound Science, I'm afraid. But feel free to have fun with all the technical gymnastics.

 

[jcx]

as I said, I don't have the musical training to use the terms with certainty - was biased against describing it as a "pitch" change since we all know pitch is just the periodicity/frequency of the fundamental, right? - but maybe frequency isn't the whole story in pitch perception
 
if musicians do perceive the polarity difference as a small pitch shift then the obvious next step is to see if a real frequency change can make the test tones sound alike by compensating for the apparent pitch shift with polarity
 
or see if you could accent a musical instrument by giving the player the option of swapping polarity on individual notes

 

[ab initio]

  the varying vertical offset is just a visual aid I hoped would make the sine harmonic addition picture easier to read - less lines crossing over to keep track of - our eyes are really happier looking at a mirrored figure/seeing the inverse symmetry when the "fold" doesn't have overlap, is based off the mirroring line
 
green, red should be very easily seen to be mirrored about the x axis == inverses, inverted in polarity waveforms just from shifting a sine and its 2nd harmonic
 
obviously the audio doesn't want the offset - see my earlier Spice sim waveform picture - which didn't seem to convince everyone

Okay, my misunderstanding! Thanks!
 
Cheers

 

[EthanWiner]

I tried mixing in a pure 200 Hz sine wave with both the original and reversed versions, to see if that countered the apparent pitch shift. The result was inconclusive. It sort of did and sort of didn't. The shift is so slight it's barely audible anyway. I'll think about this more and play with it more when I have a chance. I also noticed the effect seems to happen only with headphones.

--Ethan

 

[analogsurviver]

I tried mixing in a pure 200 Hz sine wave with both the original and reversed versions, to see if that countered the apparent pitch shift. The result was inconclusive. It sort of did and sort of didn't. The shift is so slight it's barely audible anyway. I'll think about this more and play with it more when I have a chance. I also noticed the effect seems to happen only with headphones.

--Ethan

It is easier to hear this effect with headphones - but is audible, at somewhat lower/harder to hear level. with speakers that do not have crossover troubles. Best to use are full range speakers, be it electrostatic or dynamic variety. It also requires listening room with reasonably "symmetrical acoustics" - if reflections off the walls differ significantly, the effect will be masked by them. However, it is possible to achieve the conditions the effect is audible with speakers - at this level, the listener will be rewarded with very realistically reproduced sound.
 
Hard, but not impossible to achieve.

 

[analogsurviver]

  as I said, I don't have the musical training to use the terms with certainty - was biased against describing it as a "pitch" change since we all know pitch is just the periodicity/frequency of the fundamental, right? - but maybe frequency isn't the whole story in pitch perception
 
if musicians do perceive the polarity difference as a small pitch shift then the obvious next step is to see if a real frequency change can make the test tones sound alike by compensating for the apparent pitch shift with polarity
 
or see if you could accent a musical instrument by giving the player the option of swapping polarity on individual notes

This can be achieved in strins by bowing - bow up or bow down does sound slightly different and this was/is being  used by several composers and conductors
 
.http://www.enjoythemusic.com/magazine/rickerinterview/ricker9.htm

 

[ab initio]

I played around a bit with relative phase of a fundamental (120Hz) and its second harmonic. I created a few audio files that have the two frequencies in phase (0 degrees), shifted by pi/2 (90 degrees), shifted by pi (180 degrees, opposite polarity to 0 degrees), and shifted by 3pi/2 (270 degrees, opposite polarity to 90 degrees).
 
I've uploaded the files to google drive for public consumption:
polaritytest_000
polaritytest_090
polaritytest_180
polaritytest_270
 
You can use these wav files to test to see if you can hear a difference in polarity/phase! Here's a plot of what each wavform looks like (pretty much the same as jcx's animation):
 

Notice that the blue and red curves (0 and 180)  are asymmetrical about their mean (0). In this case we could define a polarity where we call the blue curve (0) the positive  wave because it achieves higher postive values while we call the red curve(180) the negative  wave.
 
If you look at the green and cyan curves (90 and 270), you can see that they have symmetry about their mean; however, they asymmetric in time! These curves are *almost like* a low-order sin series approximations of sawtooth waves. In this case, the green curve is like a sawtooth wave that ramps down, while the cyan curve is like a ramp-up sawtooth wave as depicted at the top of the wikipedia article linked above.
 
 
It's quite easy in foobar to ABX yourself to see if you can hear the change in polarity:

1. compare polaritytest_000 against polaritytest_180

 
Or a flip in time

1. compare polaritytest_090 against polaritytest_270

 
I did the polarity test on 000 vs 180:

Spoiler: Warning: Spoiler!

foo_abx 1.3.4 report
foobar2000 v1.2.3
2014/03/28 01:07:45

File A: C:\scratch\polaritytest_sample000_120Hz.wav
File B: C:\scratch\polaritytest_sample180_120Hz.wav

01:07:45 : Test started.
01:08:11 : 01/01  50.0%
01:08:21 : 02/02  25.0%
01:08:33 : 03/03  12.5%
01:08:49 : 04/04  6.3%
01:09:06 : 05/05  3.1%
01:09:27 : 06/06  1.6%
01:09:33 : 07/07  0.8%
01:09:49 : 08/08  0.4%
01:10:01 : 09/09  0.2%
01:10:28 : Test finished.

 ----------
Total: 9/9 (0.2%)

 
 
Here's a code similar to what I used to generate the test:

Spoiler: Warning: MATLAB or GNU/octave code

fig1=figure;
for i =1:4,
Fs = 44100;
duration = 5; % length of the sound clip to make
t = [1:duration*Fs]/Fs;
f1 = 120; % fundamental at 120 Hz
f2 = 2*f1;  % 2nd harmonic
phi = (i-1)*pi/2; % assign the phase shift
y(i, = 10^(-12/20)*(cos(2*pi*f1*t) + cos(2*pi*f2*t + phi));
fade = min(1, -10* (t.*(t/t(end) - 1)/t(end)));  % creates a quick fade in/out
end,

% plot a snippet of the waveforms
figure(fig1),
plot(t(1:1000)',y,1:1000)');
xlabel('time, s'),
ylabel('amplitude, %FS'),
legend({'0','\pi/2','\pi', '3\pi/2'}),

%play the generated clips back-to-back
wavplay([y(1,.*fade,y(3,.*fade,y(2,.*fade,y(4,.*fade],Fs)
 

 
Cheers

 

[analogsurviver]

  I played around a bit with relative phase of a fundamental (120Hz) and its second harmonic. I created a few audio files that have the two frequencies in phase (0 degrees), shifted by pi/2 (90 degrees), shifted by pi (180 degrees, opposite polarity to 0 degrees), and shifted by 3pi/2 (270 degrees, opposite polarity to 90 degrees).
 
I've uploaded the files to google drive for public consumption:
polaritytest_000
polaritytest_090
polaritytest_180
polaritytest_270
 
You can use these wav files to test to see if you can hear a difference in polarity/phase! Here's a plot of what each wavform looks like (pretty much the same as jcx's animation):
 

Notice that the blue and red curves (0 and 180)  are asymmetrical about their mean (0). In this case we could define a polarity where we call the blue curve (0) the positive  wave because it achieves higher postive values while we call the red curve(180) the negative  wave.
 
If you look at the green and cyan curves (90 and 270), you can see that they have symmetry about their mean; however, they asymmetric in time! These curves are *almost like* a low-order sin series approximations of sawtooth waves. In this case, the green curve is like a sawtooth wave that ramps down, while the cyan curve is like a ramp-up sawtooth wave as depicted at the top of the wikipedia article linked above.
 
 
It's quite easy in foobar to ABX yourself to see if you can hear the change in polarity:

1. compare polaritytest_000 against polaritytest_180

 
Or a flip in time

1. compare polaritytest_090 against polaritytest_270

 
I did the polarity test on 000 vs 180:

Spoiler: Warning: Spoiler!

foo_abx 1.3.4 report
foobar2000 v1.2.3
2014/03/28 01:07:45

File A: C:\scratch\polaritytest_sample000_120Hz.wav
File B: C:\scratch\polaritytest_sample180_120Hz.wav

01:07:45 : Test started.
01:08:11 : 01/01  50.0%
01:08:21 : 02/02  25.0%
01:08:33 : 03/03  12.5%
01:08:49 : 04/04  6.3%
01:09:06 : 05/05  3.1%
01:09:27 : 06/06  1.6%
01:09:33 : 07/07  0.8%
01:09:49 : 08/08  0.4%
01:10:01 : 09/09  0.2%
01:10:28 : Test finished.

 ----------
Total: 9/9 (0.2%)

 
 
Here's a code similar to what I used to generate the test:

Spoiler: Warning: MATLAB or GNU/octave code

fig1=figure;
for i =1:4,
Fs = 44100;
duration = 5; % length of the sound clip to make
t = [1:duration*Fs]/Fs;
f1 = 120; % fundamental at 120 Hz
f2 = 2*f1;  % 2nd harmonic
phi = (i-1)*pi/2; % assign the phase shift
y(i, = 10^(-12/20)*(cos(2*pi*f1*t) + cos(2*pi*f2*t + phi));
fade = min(1, -10* (t.*(t/t(end) - 1)/t(end)));  % creates a quick fade in/out
end,

% plot a snippet of the waveforms
figure(fig1),
plot(t(1:1000)',y,1:1000)');
xlabel('time, s'),
ylabel('amplitude, %FS'),
legend({'0','\pi/2','\pi', '3\pi/2'}),

%play the generated clips back-to-back
wavplay([y(1,.*fade,y(3,.*fade,y(2,.*fade,y(4,.*fade],Fs)
 

 
Cheers

Just downloaded first three samples - for the fourth, I would have to create an acount with Google. Also just installed ABX comparator in foobar 2000. 
 
I have to learn to use the ABX comparator properly for blind testing etc ( ANYTHING but computer geek myself ) - but the first thing that comes to mind is that one track sounds mellow and the other sharp - depending how you queue them in test. They also differ in appearent loudness - the sharper is a bit louder than the mellow one. Or something in this direction. I will listen more properly after I figure out how to use ABX properly - but the differences among first three tracks is clearly audible to me.
 
I listened with JVC HA-S500 headphones.

 

[jcx]

I think you should want an octave or two higher fundamental frequency - by 100 Hz our hearing is dropping by 20 dB of sensitivity - and especially with loudspeakers the cone excursion becomes higher at lower frequency too, both increasing the possibility of nonlinearity in the speaker
 
zip should do a OK job on compressing simple continuous tones, Flac too of course