More network switch tests

[March Audio]

I have performed a number of tests on switches in the linked thread below, specifically looking at jitter.

 

https://www.stereonet.com/forums/topic/572402-jitter/

 

Just got my hands on an ""Audiophile Network Switch"" and have quickly run some basic tests.  Set up is the same as the earlier tests, using an RPi as a Roon end point.  Wired ethernet either directly from the office Asus mesh router/switch or this plus the Audiophile switch.  The RPI feeds a Topping DX7pro DAC.  DAC balanced output is measured.

 

 

 

 

Looking at broadband noise, jitter and phase noise there is zero difference direct fed from the Asus or when fed through the audiophile switch.

 

Asus

 

Audiophile switch

 

Both overlaid

 

 

 

The second test I have run is using a piece of software called Deltawave. 

 

https://deltaw.org/

 

This software compares two music files for differences.  It loads the actual waveform, time aligns it, and then performs a null.  If the files are identical the waveforms will completely cancel out and the null will be completely silent.  It provides a measurement of how well the files null measured in dB, and can export the difference file for you to listen to.  So this isnt some kind of nebulous / arbitrary measurement, this is looking at the actual waveform that comes out the DAC.  It is what you directly listen to after being fed through your amp.

 

I recorded a Vivaldi track (16bit 44khz) played first through the Asus switch then followed by the audiophile switch. I then ran them through Deltawave.

 

The null was excellent at 98.6dB and playing back the difference file, it was  inaudible.  The null wasnt higher simply because of the inevitable variability between 2 replays and recordings.  You get similar results if you make repeated recording through the same switch.

 

EDIT: Please see the re-test several posts down where null increased to 114dB.

 

 

 

 

You can listen to the difference null file by downloading this .mp3 file

 

https://1drv.ms/u/s!AnQ0c7fb_4zLhHT6dwjQELij6ecH?e=zRaRyB

 

I will perform some further testing, but this is yet another confirmation that switches make no difference to the sound or performance of well designed DACs.  The waveform coming out the DAC is unchanged.  Therefore it will sound exactly the same.

 

Edited August 23, 2023 by March Audio

[Guest]

I can hear the track playing in the difference null file so it's not inaudible. If the differences are truly random then how can I hear an intelligible song?

[March Audio]

28 minutes ago, Deepthought said:

I can hear the track playing in the difference null file so it's not inaudible. If the differences are truly random then how can I hear an intelligible song?

 

Only if you turn the volume right up. The null level is beyond 16 bits.

 

Remember that the null process isn't perfect.  It has to try and time align the the waveforms that have been through replay and re digitisation.

 

This is extremely difficult and an imperfect task which leads to some "leakage".

 

Edited August 17, 2023 by March Audio

[Guest]

So the null test is imperfect? Doesn't mean that we can't rely of the test to prove anything?

 

[March Audio]

25 minutes ago, Deepthought said:

So the null test is imperfect? Doesn't mean that we can't rely of the test to prove anything?

 

 

The null, even though imperfect, shows the difference to be below audible levels.  You will get the similar levels even if you make multiple recordings and comparisons through the same switch.

 

Any technical test or measurement you try demonstrates the same.  The noise, jitter and phase noise test really do show no difference.

 

But don't worry.  Working on a controlled listening test.

Edited August 17, 2023 by March Audio

[Addicted to music]

3 minutes ago, Deepthought said:

So the null test is imperfect? Doesn't mean that we can't rely of the test to prove anything?

 

 

There are allowable tolerances to every process that you have to account for, thats in every product you buy, we don't live in a perfect world.

 

The FTT overlay plot puts the final nail in the coffin.   

[roh008]

I think failing anything else - it's a great place to start, takes time and effort and the equipment - thanks for doing it!

[roh008]

Come to think of it - that's the start and the end of it now right?

 

What's left to 'debate'? or a more polite way - what more can be tested?

 

If I crank up the volume sure I can hear something - but essentially you demonstrated neglible difference.

[March Audio]

19 minutes ago, Addicted to music said:

The FTT overlay plot puts the final nail in the coffin.   

 

We are measuring below -160dB.  The AS didn't show any improvement to noise, jitter or phase noise, which are where the improvements are claimed to be made.

[March Audio]

29 minutes ago, roh008 said:

Come to think of it - that's the start and the end of it now right?

 

What's left to 'debate'? or a more polite way - what more can be tested?

 

If I crank up the volume sure I can hear something - but essentially you demonstrated neglible difference.

 

Whatever is tried, technically, no mechanism or significant differences can be found.

 

The ultimate final test is a controlled listening test.  This will take a while to set up.

Edited August 17, 2023 by March Audio

[roh008]

4 minutes ago, March Audio said:

Well the ultimate test is a controlled listening test.  This will take a while to set up.

well - for those that would want to go to that stage sure, but, the input files are so marginally different it would be for those that have most likely purchased this product and are looking to confirm their choice? Otherwise I'm not sure why you would based upon the evidence so far. Maybe that's just me though

[March Audio]

14 hours ago, roh008 said:

well - for those that would want to go to that stage sure, but, the input files are so marginally different it would be for those that have most likely purchased this product and are looking to confirm their choice? Otherwise I'm not sure why you would based upon the evidence so far. Maybe that's just me though

 

I do agree, however we have some members that are adamant these switches make a significant difference.  That there is some mechanism we dont understand and cant measure.

 

So its to gather objective evidence of what is actually happening audibly, and if there is anything more to this beyond expectation bias.

Edited August 18, 2023 by March Audio

[MLXXX]

1 hour ago, March Audio said:

You can listen to the difference null file by downloading this .mp3 file

 

https://1drv.ms/u/s!AnQ0c7fb_4zLhHT6dwjQELij6ecH?e=zRaRyB

 

(Hosted on OneDrive,  so a Microsoft account is needed to download the file.)

 

There are recognisable sounds of classical music (the Vivaldi), amidst some burbling noises, if the file contents are boosted by 76dB to bring them up to full level.

 

47 minutes ago, March Audio said:

You will get the similar levels even if you make multiple recordings and comparisons through the same switch.

 

This is a very important point @March Audio has made. It needs to be borne in mind when interpreting recordings that differ only very slightly.

 

 So for example if comparing device performance with power cord A connected vs the same device with power cord B connected you might find a null that although inaudible is non-zero.  However you'd very possibly get an inaudible non-zero null of similar magnitude if comparing device performance with cord A on separate occasions.

 

You need to be able to show that the difference you've found in comparing A and B is significantly greater in magnitude than when comparing A with A, or B with B.

 

*   *   *

Having mentioned the above, I am a little surprised at the extent to which the residual is -- at times --  readily recognisable as music. I note the 0.608 sample phase offset reported by the DeltaWave analysis software. Generally speaking, if a sample timing offset is not compensated for when comparing two audio files* then the imperfection in timing alignment for the trial null will tend to leave wispy treble more prominent than low frequencies.  (For my ears  the residual music sounds thin [lacking in bass] but seems to include mid-frequencies of comparable prominence to higher frequencies.)

 

Listening to the residual file as it progressed it occurred to me that there was very possibly some small drift in the clock rate(s). Such drift in absolute clock rate could be attributable to the playback device, the recording device, or even to both.  Minor clock drift like that would be well, well below audibility, but a nuisance if striving for an extremely deep, sustained null.

 

I have three Panasonic consumer grade cameras I use for recording video. I find that two of them can stay in audio sync with each other to the nearest 30fps half frame for 2 hours or more when capturing a live concert. The third camera has a slightly different clock rate and will  consistently drift slightly in its audio capture relative to the other two. 

 

__________

* That is if the two audio files are time aligned in a linear audio editor to the nearest audio sample time, the default method and close enough for  most practical purposes; rather than being resampled so that the reconstructed waveforms lie directly on top of each other independent of the arbitrary phase discrepancy that arises with asynchronous ADC capture.  Note: I am not actually suggesting this refinement be attempted in this case, (if it hasn't already been done!),  unless it is thought that the relevant clock rates remained exactly the same.  (I've never attempted it myself other than somewhat indirectly by resampling both audio files to a very high sample rate so that nudging them along the time line for time alignment purposes could be done in smaller time increments.)

[March Audio]

Mea culpa.  I wasnt entirely happy with the null results and was expecting better.

 

So I re ran the test.  It appears that I used too long a file which exacerbates the clock drift between DAC and ADC and alignment of the waveforms.  Shortening the track yielded the results I was expecting.

 

We now see a null depth of 114dB.

 

 

 

the difference recording mp3 is here:

 

https://1drv.ms/u/s!AnQ0c7fb_4zLhHZ4zpArf4i7FKuy?e=KgyRb2

 

I also amplified this by 48dB, so 251 times louder:

 

https://1drv.ms/u/s!AnQ0c7fb_4zLhHXXTm5Pe7mBorj5?e=fH2vZX

 

 

 

Edited August 23, 2023 by March Audio

[March Audio]

3 minutes ago, MLXXX said:

 

(Hosted on OneDrive,  so a Microsoft account is needed to download the file.)

 

There are recognisable sounds of classical music (the Vivaldi), amidst some burbling noises, if the file contents are boosted by 76dB to bring them up to full level.

 

 

This is a very important point @March Audio has made. It needs to be borne in mind when interpreting recordings that differ only very slightly.

 

 So for example if comparing device performance with power cord A connected vs the same device with power cord B connected you might find a null that although inaudible is non-zero.  However you'd very possibly get an inaudible non-zero null of similar magnitude if comparing device performance with cord A on separate occasions.

 

You need to be able to show that the difference you've found in comparing A and B is significantly greater in magnitude than when comparing A with A, or B with B.

 

*   *   *

Having mentioned the above, I am a little surprised at the extent to which the residual is -- at times --  readily recognisable as music. I note the 0.608 sample phase offset reported by the DeltaWave analysis software. Generally speaking, if a sample timing offset is not compensated for when comparing two audio files* then the imperfection in timing alignment for the trial null will tend to leave wispy treble more prominent than low frequencies.  (For my ears  the residual music sounds thin [lacking in bass] but seems to include mid-frequencies of comparable prominence to higher frequencies.)

 

Listening to the residual file as it progressed it occurred to me that there was very possibly some small drift in the clock rate(s). Such drift in absolute clock rate could be attributable to the playback device, the recording device, or even to both.  Minor clock drift like that would be well, well below audibility, but a nuisance if striving for an extremely deep, sustained null.

 

I have three Panasonic consumer grade cameras I use for recording video. I find that two of them can stay in audio sync with each other to the nearest 30fps half frame for 2 hours or more when capturing a live concert. The third camera has a slightly different clock rate and will  consistently drift slightly in its audio capture relative to the other two. 

 

__________

* That is if the two audio files are time aligned in a linear audio editor to the nearest audio sample time, the default method and close enough for  most practical purposes; rather than being resampled so that the reconstructed waveforms lie directly on top of each other independent of the arbitrary phase discrepancy that arises with asynchronous ADC capture.  Note: I am not actually suggesting this refinement be attempted in this case, (if it hasn't already been done!),  unless it is thought that the relevant clock rates remained exactly the same.  (I've never attempted it myself other than somewhat indirectly by resampling both audio files to a very high sample rate so that nudging them along the time line for time alignment purposes could be done in smaller time increments.)

 

Please check the updated null recording

[MLXXX]

16 minutes ago, March Audio said:

Please check the updated null recording

Yes, much more impressive this time. I needed to boost the basic mp3 file by 92dB to bring it up to a full level.  

 

This time no recognisable music. Just burbling sounds.

[March Audio]

5 minutes ago, MLXXX said:

Yes, much more impressive this time. I needed to boost the basic mp3 file by 92dB to bring it up to a full level.  

 

This time no recognisable music. Just burbling sounds.

 

Yep, lesson learnt about Deltawave 

 

so you had to amplify it 40000 times !

Edited August 17, 2023 by March Audio

[Guest]

You should post this on the ASR Forum Alan. They'd love this definitve proof to drive the final nail in the subjectivist coffin.

Edited August 17, 2023 by Deepthought

[Guest]

7 hours ago, MLXXX said:

Yes, much more impressive this time. I needed to boost the basic mp3 file by 92dB to bring it up to a full level.  

 

This time no recognisable music. Just burbling sounds.

Fixed it! 

Edited August 17, 2023 by Deepthought

[Addicted to music]

12 minutes ago, Deepthought said:

You should post this on the ASR Forum Alan. They'd love this definitve proof to drive the final nail in the subjectivist coffin.

 

ASR has already done it:

 

https://www.audiosciencereview.com/forum/index.php?threads/uptone-audio-etherregen-switch-review.10232/

Edited August 17, 2023 by Addicted to music

[Guest]

I assumed that Alan's latest measurements showed  some new proof of the lack of efficacy of the device. 

[Guest]

It must be hard being a manufacturer when people post these sort of threads on the internet. But I guess we need to silence those pesky subjectivists once and for all! 

Edited August 17, 2023 by Deepthought

[The Mad Scientist]

1 hour ago, Deepthought said:

You should post this on the ASR Forum Alan. They'd love this definitve proof to drive the final nail in the subjectivist coffin.

Alan is banned from ASR - Amir doesn't play nicely if you disagree with him, even if it's done in the most gentlemanly of fashions.....

Edited August 17, 2023 by The Mad Scientist

[The Mad Scientist]

1 hour ago, Deepthought said:

I assumed that Alan's latest measurements showed  some new proof of the lack of efficacy of the device. 

They do, at least to the most bone-headed subjectivist. Rather than just measuring SINAD and saying the etherregen makes no difference (the die-hard subjectivists then argue that not everything that can be measured counts and not everything that counts can be measured), Alan has captured audio samples with and without the etherregen and generated a null test file which irrefutably shows that there really is no audible difference.

Edited August 17, 2023 by The Mad Scientist

[Guest]

46 minutes ago, The Mad Scientist said:

Alan is banned from ASR - Amir doesn't play nicely if you disagree with him, even if it's done in the most gentlemanly of fashions.....

Yes, I've since read that -https://audiosciencereview.com/forum/index.php?threads/the-real-problem-of-march-audios-sointuva-wg-review-measurements-and-reinforcements-with-klippel-device.35255/post-1229017