Ultimate listening test ... test your ears and audio chain

[Barleywater]

Bandwidth of DUT is greater than 20-20kHz. Lot's going on <20Hz. Resolving IMD to few Hz likely unmasks the differences better.

All the ELF appears as tiny spread in spectrum peaks not resolved at standard sampling rates with small FFT, like 65k bins; such as limit with Cool Edit that I like to use.

I aligned tracks, and paired right channels of green and blue, re-sampled to fs 375Hz from 48kHz.

Here is 4k FFT spectrum averaged across entire sample widow:



And 32kFFT from near center of sample window:



Blue tracks have increased low frequency content.

Applying same treatment to source data may reveal which, if either is closer.

Which sounds better is likely source and listener dependent.

 

[Pano]

Do you actually hear that increased LF content?

 

[heb1001]

Do you actually hear that increased LF content?

It's pretty easy to hear if you just play the file of the differences between the two tracks amplified by about 30dB 🙂

I couldn't hear it in situ though.

 

[CFT]

Hehe who knows. I know tube well enough to choose solid state as my amplifier of choice. Try listen to SYMEF. No matter what you do, solid state have strength tube amp cannot match. And vice versa.

What about SYMEF Jay?

 

[Jay]

You never know, JAy. I've been in blind tests where I identified the tube amp by its smooth sound, lack of harshness and superb ambiance, as compared to the SS amp with greater dynamics, more kick but more grain and fatigue. Classic qualities of both types, easy to hear.

Only to find out at the end of the test that I got it exactly backwards. 😀

What about SYMEF Jay?

It is designed to mimic the transfer curve of a triode. Ask OnAudio about it.

 

[Barleywater]

Do you actually hear that increased LF content?

Directly hearing, and perceptual alteration aren't the same thing.

The picture with smoother plot suggests increased super low bass; the second picture with big FFT indicates significant quantities of spectral difference components, potentially from across entire spectrum.

Example:

2700.0Hz top track, 2700.6Hz in lower track. At fs 48kHz two peaks show very little separation. Sum of two waves produces single narrow peak with 65k FFT for generating spectrum.

If 2700.6Hz is -36dB to 2700Hz, peak separation requires much bigger FFT; this holds when the two are summed. Summed result looks like 2700Hz with peak to peak ripple of 0.3dB.

If real differences are due to aliasing, IMD, jitter, the low frequency spectrum will change.

Brainwave synchronization with LF modulation in audio is documented; shifts in perceptual state occur.

With turntable resonances due to tone arm, eccentric record centering (even center hole that drops onto spindle v snug) are sources of periodic LF that modulate onto entire pass band. For many this may be the magic of vinyl.

In case of electronics, circuit and component specifics are at work.

 

[heb1001]

I tried again over lunch with my mp3 player and the wav files.

The only point of note I can find is the distortion (clipping?) when he sings "I don't care what you do at home would you care to explain" which is the same in both samples as far as I can tell.

I can't reliably detect any differences.

The trumpet? reliably sounds better in whichever one I play loudest.

Now I'm just waiting to get humiliated by the results of the poll 😀

 

[PMA]

So why was one file bigger than the other .... ???

Different length of silence before music starts and after it stops.

 

[heb1001]

Directly hearing, and perceptual alteration aren't the same thing.

The picture with smoother plot suggests increased super low bass; the second picture with big FFT indicates significant quantities of spectral difference components, potentially from across entire spectrum.

If you listen to the difference file, it just sounds like a quiet version of the original with significant eq to shift energy to the bass. It's quite clear, easily clear enough to make out the words so I doubt it's introducing a lot of spectral difference components.

I was looking for it to present as a difference in the low frequencies or perhaps a slight extra echo but I couldn't hear either.

 

[fas42]

I'm getting very suspicious that some form of modulation of the track(s), using its own waveform as the input signal, has taken place - if one attempts to null by applying a straighforward linear factor to one version and diff'ng, out pops the original signal again! Hmmm ...

 

[heb1001]

I was looking for it to present as a difference in the low frequencies or perhaps a slight extra echo but I couldn't hear either.

The difference is quite mono. So it might present as a slight difference in channel separation.

 

[PMA]

The only situation when differing leads to almost 'null' results is with the same files or files mathematically changed in amplitude only. Any, even slightest phase shift leads to no-null result and just only one 'zero' frequency.

Let's wait for Mooly wake up and posting the poll results. I already know the results, but waiting for Mooly to announce it officially.

 

[Mooly]

I'm awake... just 😀

Here's a shot of the poll.

 

[Mooly]

A zillion posts since last night 😀

The poll (which is now closed) is attached to the thread as normal and should be viewable to all at the top of every page. So it was a close run call and votes were spread fairly evenly although green nudged ahead.

 

[PMA]

I would like to thank everyone who have participated in the poll, everyone posting their comments and I would especially like to thank Mooly for his tremendous help and enthusiasm.

Here is the blue and green DUT.

 

[kgrlee]

I suppose I should put my $0.02 worth in to bolster my pseudo DBLT guru facade .... 🙂

As this is
- a paired test (only 2 'presentations')
- no ABX controls in place
- very small differences in level & response (under 0.1dB) and hardly anything in THD ..

assuming no gross differences like clipping ..

I expect purely random preferences from most people and that my true golden pinnae would probably say they can't hear any difference or they don't believe they can pick it up reliably.

Some of us forget this last result is perfectly valid and often the most important of all.

I'm amused at the efforts to find some difference between the files by differencing, analysing etc.

Of course, this is where Mooly says everyone among the 100 listeners picked the same 'presentation' and shows me up as a fraud 🙂

Any statisticians here ready to give us the probability that whatever result Mooly reveals is pure chance? Serious question and you gotta consider what is the actual proposition being tested.

I hope we repeat the test with the same 2 'presentations' preferably as ABC.
______________

Barley, how did you do your resampling to fs 375Hz?
______________

DAMN! Not quick enough!

 

[PMA]

Richard, thanks. Results and HW already revealed 😉. The non-direct path 'won'.

 

[PMA]

Yet not shown - step response of the 'green' DUT. Upper freq corner (-3dB) close to 1MHz.

 

[kgrlee]

ROTFL ... hope I got the 5LA correct! 😀

Thanks to PMA & Mooly for this great listening test which I still hope they will repeat with ABC ... perhaps sneaking in another OPA instead of just green & blue ... Gotta keep da listeners guessing until they give up guessing and just trust their ears 🙂

Got a list of comments from the various listeners who thought they heard a difference? No need to name names cos gotta persuade them to do the repeat test.

 

[PMA]

I assume you have noticed that the blue was a direct D/A -- A/D path, no OPA inserted, just 35cm of coaxial cable 😀
So the test was a cable vs. preamplifier (with cables).