Audio Test of Power Conditioner by Audioquest

[amirm]

Hello you all. Audio quests of power conditioner product engineer, has written this "white paper," on differences such products make to audio fidelity: https://www.audioquest.com/resource/1138/Power-Demystified-whitepaper-8-23-18.pdf

In the paper, there is some kind of null test conducted by Bernie Grundman at his mastering house. I have read it a few times but still can't figure out what the test is. Usually in a null test, the lower the null, the better it is. Yet the claim is the reverse here. The say their power conditioner produces the loudest difference signal! At least that is my read of it.

Please right click on the above link and download it. Then open it in Adobe Reader and give it permission to play that multimedia content. Then you can listen to the three tracks.

I am not interested in the rest of the paper as far as filtering, surge protection, etc. My interest is this one specific test of fidelity difference.

From the bit I can figure out, it is a simple play a file through a DAC and then digitize it using an ADC. Samples are aligned visually instead of synchronizing them. Whether Bernie is running the same master clock so this was not needed, is an unknown.

I like to replicate the test so if we can figure this out, we would make progress. If not, we can circle back to the company and have them explain. Just want to make sure it is not just me that can't figure out what is going on.

As an aside if their claims are true, it is a shocker that Bernie's' mastering system needs a line conditioner to play and capture just 16 bit audio without loss!

 

[PierreV]

In the paper, there is some kind of null test conducted by Bernie Grundman at his mastering house. I have read it a few times but still can't figure out what the test is. Usually in a null test, the lower the null, the better it is. Yet the claim is the reverse here. The say their power conditioner produces the loudest difference signal! At least that is my read of it.

Same here. From their hardware description, that's just it. Run one system on AC, record. Run one system on conditioner, record. Spot the difference. Seems fairly easy to replicate by anyone with that conditioner, a decent ADC and, Deltawave.

But... but... it also seems you essentially replicated it in advance in your initial test and the AP did not show a difference...

 

[Blumlein 88]

Yes, the assumption is conditioned power IS different than untreated power. So the bigger the left over from a difference between treated and untreated the more improvement the AC Conditioner has accomplished. An awful lot in that assumption.

Of course finding a difference does indicate the AC conditioner is doing something.

 

[Sir Sanders Zingmore]

The say their power conditioner produces the loudest difference signal!

If their claim is that the device makes a difference surely they want the null test to show a (big) difference?

 

[amirm]

Of course finding a difference does indicate the AC conditioner is doing something.

Or they had clock drift and that is all we are hearing.

 

[Blumlein 88]

Or they had clock drift and that is all we are hearing.

Yes, they mention aligning to the nearest sample. That is nowhere near good enough. They don't give us enough info to know, but Deltawave and these files should let you figure it out. Or maybe they are using locked clocks as likely in a studio everything is locked.

I'm not getting the files to download. Let me try my Windows laptop.

 

[Blumlein 88]

Okay listened to these. Way too loud and clear to be true. They've either been amplified or this is a large difference the result of timing mismatch. It has the tinny uptilted tell tale property of something with a timing mismatch. The residual of those has a 6 db/octave uptilt. This sounds that way as I have that recording.

They are telling you the fine detail you are hearing is what is lost. They used some acoustic guitars probably thinking the uptilt won't be noticed as much by most people.

 

[pkane]

Hello you all. Audio quests of power conditioner product engineer, has written this "white paper," on differences such products make to audio fidelity: https://www.audioquest.com/resource/1138/Power-Demystified-whitepaper-8-23-18.pdf

In the paper, there is some kind of null test conducted by Bernie Grundman at his mastering house. I have read it a few times but still can't figure out what the test is. Usually in a null test, the lower the null, the better it is. Yet the claim is the reverse here. The say their power conditioner produces the loudest difference signal! At least that is my read of it.

Please right click on the above link and download it. Then open it in Adobe Reader and give it permission to play that multimedia content. Then you can listen to the three tracks.

I am not interested in the rest of the paper as far as filtering, surge protection, etc. My interest is this one specific test of fidelity difference.

From the bit I can figure out, it is a simple play a file through a DAC and then digitize it using an ADC. Samples are aligned visually instead of synchronizing them. Whether Bernie is running the same master clock so this was not needed, is an unknown.

I like to replicate the test so if we can figure this out, we would make progress. If not, we can circle back to the company and have them explain. Just want to make sure it is not just me that can't figure out what is going on.

As an aside if their claims are true, it is a shocker that Bernie's' mastering system needs a line conditioner to play and capture just 16 bit audio without loss!

They claim that the loud difference in the null file represents information that was not detectable in a null file that was produced without their power-cleaning equipment. In effect, claiming that their power conditioner is restoring missing music. What they don't tell you or maybe don't realize is that aligning on a sample boundary introduces a large error into the null calculation. This error sounds like music because you are subtracting misaligned music files which no longer cancel each other out. Proper sub-sample alignment and any clock drift must be taken care of when computing a null, otherwise, the error will be loud and sound like the original music.

 

[Pdxwayne]

@amirm, you just tested the Audioquest power conditioner. You still have it? If you have it still, you can easily do recordings and use deltawave to check.....

 

[sandymc]

They claim that the loud difference in the null file represents information that was not detectable in a null file that was produced without their power-cleaning equipment. In effect, claiming that their power conditioner is restoring missing music. What they don't tell you or maybe don't realize is that aligning on a sample boundary introduces a large error into the null calculation. This error sounds like music because you are subtracting misaligned music files which no longer cancel each other out. Proper sub-sample alignment and any clock drift must be taken care of when computing a null, otherwise, the error will be loud and sound like the original music.

That's correct. Unless they have a way of ensuring precise sync, and I mean down to at least nS levels, which they don't mention, all they're measuring is the effects of sample timing differences on the DAC reconstruction filter and the ADC anti-aliasing filter.

 

[pkane]

That's correct. Unless they have a way of ensuring precise sync, and I mean down to at least nS levels, which they don't mention, all they're measuring is the effects of sample timing differences on the DAC reconstruction filter and the ADC anti-aliasing filter.

Yep. Here's an example of a null file / error spectrum from a recent DAC comparison. With only a sample-boundary alignment, as in the paper:

And now, with subsample alignment and clock drift correction (using DeltaWave):

Same two files nulled. Night and day difference, I would say

 

[Pdxwayne]

Yep. Here's an example of a null file / error spectrum from a recent DAC comparison. With only a sample-boundary alignment, as in the paper:
View attachment 145467

And now, with subsample alignment and clock drift correction (using DeltaWave):
View attachment 145469

Same two files nulled. Night and day difference, I would say

I thought all devices were the same and the differences were where those devices plugged in.

I really don't know how many samples they collected to write the paper. But, if there are enough samples taken and the difference were consistent based on just where the devices were plugged, wouldn't this an interesting observation?

 

[pkane]

I thought all devices were the same and the differences were where those devices plugged in.

I really don't know how many samples they collected to write the paper. But, if there are enough samples taken and the difference were consistent based on just where the devices were plugged, wouldn't this an interesting observation?

If you record two consecutive loopbacks even using the same devices, there's little chance that they will align exactly on a sample boundary or that there will be no clock drift, like you saw with the D30pro over USB. Since the method to produce the null waveform was flawed, the result cannot be used to demonstrate what they claim.

 

[Pdxwayne]

If you record two consecutive loopbacks even using the same devices, there's little chance that they will align exactly on a sample boundary or that there will be no clock drift, like you saw with the D30pro over USB. Since the method to produce the null waveform was flawed, the result cannot be used to demonstrate what they claim.

What would be a better method?

Let's say I do the same kind of tests using optical chain. For example, the optical chain:

Node2i optical -> DAC -> ADC -> Laptop

Then, capture playbacks with everything plugged to wall. Then capture playbacks with everything plugged to power conditioner.

Would this method more appropriate?

 

[pkane]

What would be a better method?

Use DeltaWave

 

[Pdxwayne]

Use DeltaWave

So it doesn't matter the chain, just the software to do null.

: )

 

[pkane]

So it doesn't matter the chain, just the software to do null.

: )

Both are important if you want the result to be meaningful.

 

[sandymc]

What would be a better method?

Let's say I do the same kind of tests using optical chain. For example, the optical chain:

Node2i optical -> DAC -> ADC -> Laptop

Then, capture playbacks with everything plugged to wall. Then capture playbacks with everything plugged to power conditioner.

Would this method more appropriate?

You need to have a mechanism for (a) syncing the A/D and D/A clocks, and (b) syncing the conversion start, because clocks are typically a multiple of the conversion time. Professional test environments typically have the DACs and ADCs built into one unit, running off common clock and control circuits to allow for this.

 

[Blumlein 88]

I suspect the files in this paper may have come from this Furman test in this video. Same technique, with more detail on gear used and the fact Diffmaker was used for the nulling. This used the same exact music file which is why I highly suspect this same info from Furman was used in the AQ paper.

Even here they are playing fast and loose and deceptively so with the results. They are keying on the dynamic range output of Diffmaker (when I think they are using correlated nulls). Plus they are showing their units vs the wall AC has the greatest dynamic range vs other companies products. The only thing is if true it means Furman puts out a signal more like the mains AC than other units. In other words it makes the least difference.

 

[Descartes]

AudioQuest are one of the most dishonest companies making do much false advertising! Dealers loves them as the margins are extremely high free money!