Is cable sound real? A more holistic approach trying to track it down.

[pkane]

Clock drift is not an issue as the DAC and ADC are in one single device fed from the same internal independent master clock (as it's USB asynchronous isochronous mode).
Gain drift is, and there block-averaging will help reduce it (though @pkane might implement correction of slow gain drift in DeltaWave, also he might be able to undo simple -- and microscopic, in this case -- linear transfer function differences coming from different cable capacitances and different time-of-flight values).
Heavy block-averaging has its own problems, though. It also reduces loosely correlated random/stochastic effects, while not as much as uncorrelated noise but still to significant amounts so those might go unnoticed unless we also inspect the noise distribution, sample per sample. An example for this is the excess current noise from some resistors, this can best be isolated by inspecting the noise distribution in the residual (compared to an equal but less noisy resistor), where one can find a modulation of the noise with momentary signal level.

I used DeltaWave for some analysis a while back on a few balanced interconnects I had on hand. There were measurable null differences, although @ -100dB or lower:

https://www.audiosciencereview.com/...es-make-a-difference-a-null-test-result.7738/

DeltaWave has progressed since then, so now variable group delay and frequency response errors can be measured and corrected. I've just added a more perceptually-weighted metric to DeltaWave that should help with determining audibility of the difference. I would guess that anything with an RMS null of -100dB is extremely unlikely to be audible under normal listening conditions. Since my audio interface (Apogee Element24) isn't as low noise and distortion as the RME, it would be interesting to see your tests, and run the captures through the latest version of DW

 

[pma]

I would guess that anything with an RMS null of -100dB is extremely unlikely to be audible under normal listening conditions. Since my audio interface (Apogee Element24) isn't as low noise and distortion as the RME, it would be interesting to see your tests, and run the captures through the latest version of DW

Rather -50dB than -100dB. And effective dynamic range of the ear when listening to music sample is only about 30dB, which is moving up and down with average signal level, but instantly it is only about 30dB. There is no chance to hear any -100dB signal in presence of 0dB signal.

 

[pkane]

Rather -50dB than -100dB. And effective dynamic range of the ear when listening to music sample is only about 30dB, which is moving up and down with average signal level, but instantly it is only about 30dB.

I'm with you on this one But maybe others don't believe this and can try @KSTR 's test.

 

[mhardy6647]

None of those.

I'm more into over-engineered tool watches and vintage.

At this moment, I'm wearing a Sinn.

See? You'd make a good audiophile!
(and -- in all seriousness -- I make that observation 100% free of snark)

Fairchild. Western Electric. Early Altec, RCA, & JBL. The hifi equivalents of your watch preferences.
Not McIntosh, Luxman ("the McIntosh of Japan"), or even Audio Research or conrad-johnson.



k13_0034 by Holger Barske, on Flickr
k13_0023 by Holger Barske, on Flickr
etf19_052 by Holger Barske, on Flickr

 

[KSTR]

It would be a good idea to include a ‘sanity check’. Obtain at least one cable that’s objectively, and audibly, inferior: highly inductive, or something like that. You could either fabricate this yourself or stare at one of Audioquest’s more expensive models for a couple of minutes so it breaks.

This will mean that the test contains at least one presentation that should elicit a positive discrimination and provide a baseline.

Yes, positive and negative controls of various kinds are part of the plan. Its probably hard to find a really broken cable bad enough so I will have to make one that does harm to the signal. There are ways....

 

[KSTR]

yes we need yet another cable tread on the internet as much a hole in the head .

"Is cable sound real? A more holistic approach trying to track it down."

Even the title is slightly biased ? try track what down now ?

Feel free to suggest a better title, but it shall be catchy.

 

[KSTR]

As a general note as I might have not been clear on that: I am personally 100% in the realist/objectivist camp. I don't believe in cable sound other than LCR-induced (pathological cases set aside), have neither experienced it nor measured anything substantial so far but I've also never used anything else than garden-variety cables myself. But my current personal stance is irrelevant here. 15 years ago I strongly denied impact of absolute polarity, and phase response in general, now I know better.

 

[solderdude]

About 30 years ago I built a null tester.
Testing run of the mill cables (interlinks of about a meter length) and some higher priced (but not high-end) cables I found all the interlink cables I tested nulled perfectly.
So what went in came out. I cannot see how a boutique cable could do better than no change at all.
The only way would be if it became 'worse' but somehow 'sonic-ally preferred'.

For speaker cables things differ. Length and resistance mattered for nulling with real speaker loads (as expected)
Length becomes an issue when nulling. Due to speed differences phase shifted for the highest frequencies and as it was analog couldn't correct this.
Such phase differences result in amplitude differences which, without analysis of input and output signals can not be differentiated from upper treble roll-off. Digital this should be no problem (for Paul's software) I reckon.

 

[UliBru]

Today I have run a simple test with my RME Fireface UC between output1 and input1.
I have used a short symmetric cable, a short asymmetric cabke, a longer asymmetric cable (5m) , and a longer symmetric cable (5m).
The cables are all plugged in by some mono jack plugs though.

The test signal is a multitone signal, here is a comparison between 64bit float and 24bit PCM (TPDF dithered).
The signal contains sine waves of about 1/6 octave distance. Not exact 1/6 to avoid harmonics falling on a spectral line.
The signal length is 2^20 samples, the FFT without windowing looks like

Clearly the dithering adds some noise, the level is anyway low.

Now the recorded signals are averaged about 6 recordings. Here a comparison between short symmetric and long asymmetric

The long cable has a higher noise floor.

A comparison between the short symmetric and the short asymmetrc cable

Here the short asymmetric cable also shows a higher noise floor. But it also behaves different in comparison to the long asymmetric cable.

And here the comparison between the short and long symmetric cable

There are still differences but very subtle.

The test has been checked to be repeatable. So the same cable always gives the same result.

Of course it is questionable if such a low noise floor can be perceived. But anyway it is not a null test. There is no perfect null with analog playback and recording.

 

[pma]

The low frequency part looks like flawed. Window? Or rather clipping?

This is a cheap SB USB X-Fi HD.

 

[egellings]

I'm all for enjoying an expensive cable if that is what you fancy. I just take issue with claims of sonic superiority being the reason for the cable choice, when in reality the buyer got it simply because it looked cool or or was found to be appealing for some other non-sonic reason.

 

[UliBru]

The low frequency part looks like flawed. Window? Or rather clipping?

This is a cheap SB USB X-Fi HD.

No window (or rectangle or uniform window).
I guess it is an artefact of just cutting the measurement to the length of 2^20 samples like the original signal. The signal length is anyway influenced by the brickwall filters.

 

[KSTR]

KSTR Cable Test #01 (analog XLR loopback) is online

 

[pma]

KSTR Cable Test #01 (analog XLR loopback)

Hmmm ..... you do not want any comments in the test thread, understood.

They are either slight phase shifts due to different cable R-C AND/OR voltage induced into the loop with finite CMR diffamp elements. Interesting, however I am not sure how universal.

 

[andreasmaaan]

Feel free to suggest a better title, but it shall be catchy.

"Is cable sound real? Striking the right cord"

(Embarrassingly, this was not the first or even only pun that crossed my mind.)

KSTR Cable Test #01 (analog XLR loopback) is online

Listened over EQ'd HD6xx, could not hear a difference. Using an ABX comparator, thought I may be able to discern the point at which a changeover from A to B occurred, but scored no better than chance after half a dozen trials and then gave up.

Also performed a null-test in Deltawave, but will not discuss that here I guess?

 

[pma]

Phase shifts only or slightly different propagation delay. So the samples are in ns/10ns shifted.

 

[pkane]

"Is cable sound real? Striking the right cord"

(Embarrassingly, this was not the first or even only pun that crossed my mind.)



Listened over EQ'd HD6xx, could not hear a difference. Using an ABX comparator, thought I may be able to discern the point at which a changeover from A to B occurred, but scored no better than chance after half a dozen trials and then gave up.

Also performed a null-test in Deltawave, but will not discuss that here I guess?

Using DeltaWave is cheating when doing ABX tests Better not bring up your results until others had a chance to do a blind test.

 

[pkane]

Today I have run a simple test with my RME Fireface UC between output1 and input1.
I have used a short symmetric cable, a short asymmetric cabke, a longer asymmetric cable (5m) , and a longer symmetric cable (5m).
The cables are all plugged in by some mono jack plugs though.

The test signal is a multitone signal, here is a comparison between 64bit float and 24bit PCM (TPDF dithered).
The signal contains sine waves of about 1/6 octave distance. Not exact 1/6 to avoid harmonics falling on a spectral line.
The signal length is 2^20 samples, the FFT without windowing looks like

View attachment 108785
Clearly the dithering adds some noise, the level is anyway low.

Now the recorded signals are averaged about 6 recordings. Here a comparison between short symmetric and long asymmetric
View attachment 108791
The long cable has a higher noise floor.

A comparison between the short symmetric and the short asymmetrc cable
View attachment 108792
Here the short asymmetric cable also shows a higher noise floor. But it also behaves different in comparison to the long asymmetric cable.

And here the comparison between the short and long symmetric cable
View attachment 108801
There are still differences but very subtle.

The test has been checked to be repeatable. So the same cable always gives the same result.

Of course it is questionable if such a low noise floor can be perceived. But anyway it is not a null test. There is no perfect null with analog playback and recording.

Yeah, that averaged spectrum looks suspect in the lower frequencies. Here's my own multi-tone test of Apogee Element24 with a 2m no-name XLR cable and a 10-run average:

For comparison, a 20ft/6m microphone cable. Same setup:

 

[andreasmaaan]

Using DeltaWave is cheating when doing ABX tests Better not bring up your results until others had a chance to do a blind test.

Yes, I thought as much

 

[KSTR]

@pma , @UliBru

;-)