Null test of analog interconnects

[esldude]

Comparing analog interconnects to see if they are different. Null test in this case. Play music with one Audioquest Diamond X3 on the left channel, and a free in the box with a DVD audio cable on the right channel. Same file sent through both channels of a computer sound card then looped back to the input of the same sound card on the same computer.

Once the track was recorded one channel was inverted. Then normalized both channels as one channel was about .1 dB more sensitive. Combined the out of phase channels and obtained a null. If the pure silver Audioquest was better than the cheap cable then some part should fail to null out.

The nulled result was at about -110 dB at 20 hz and fell to around -120 dB for most of the range. The source was a ripped file from a Wilson CD of Hyperion Knight playing Gershwin. Solo piano music. I am attaching a 15 second segment of the nulled result after it was amplified by 50 dB digitally. It is a zipped wav file 24 bit 96 khz.

You actually can hear a hint of a few piano notes if you listen closely. 50 dB quieter they are of course not audible. You also could amplify this file another 25 dB or so and have a better chance of hearing it though this will be a loud noisy file to listen to over your system. You also can reduce the amplitude by 50 dB to get what the original null file was.

I did this with a few other good recordings, some more complex or louder. This one is actually the only one that didn't null completely out so that something is left once you amplify it. The others are simply gone. I suspect it still is actually a very minor channel imbalance causing what is left as it sounds the same either with mismatched cables or matching cables on both channels. Rather makes one think there is no difference in these audio interconnects.

Now, I think the AQ cable is an excellent sounding cable. In sighted listening at least my perception is it sounds better. Does it? Is it really any different when you get null results like this? I do believe for one cable to sound different than another it must pass the waveform differently. The null result appears to argue that at least within the audio band there actually is no difference that can possibly be audible.

[One and a half]

Fascinating!

Can you the same recording again, but just record 'nothing', no music. If there is any noise present from one channel to the other, the recorded file will have some content. Cables do make a difference, perhaps it's at the very low bottom end where the differences could be.

[sandyk]

Null Test.
What a loaded couple of poll questions !!!
I seriously doubt that most people who do not tick the option "The null result indicates your perception is creating the difference."
completely agree with the other option.Methinks you are being plain mischievous Mr. esldude.(grin)
Null tests also used to be used to prove that all amplifiers sound the same. Obviously there is something going on, more than likely at a lower level than being used,as "One and a half" suggests. MANY members , including senior members have reported hearing differences between USB cables for example.
Alex

[esldude]

Fascinating!

Can you the same recording again, but just record 'nothing', no music. If there is any noise present from one channel to the other, the recorded file will have some content. Cables do make a difference, perhaps it's at the very low bottom end where the differences could be.

If I understand your suggestion, to record only silence and see what is there, I have done that. You end up with a pretty even noise floor around 123 db down or there about. Basically the residual noise in the electronics I am using for this. If one does this using front panel jacks instead of plugging directly into the sound card on the rear you get a good bit of garbage at various frequencies. The noise floor in places is raised 20-25 dB, with a few frequencies higher. I assume this is from the wire running right through the middle of the computer where so much electrical noise is located. So that spurious noise from inside the machine is much much larger than any measured difference in the interconnects.

[esldude]

Null Test.
What a loaded couple of poll questions !!!
I seriously doubt that most people who do not tick the option "The null result indicates your perception is creating the difference."
completely agree with the other option.Methinks you are being plain mischievous Mr. esldude.(grin)
Null tests also used to be used to prove that all amplifiers sound the same. Obviously there is something going on, more than likely at a lower level than being used,as "One and a half" suggests. MANY members , including senior members have reported hearing differences between USB cables for example.
Alex

Not being all that mischievous. I do hope to get some stimulating discussion here. Maybe learn something new.

So have any ideas how something lower in level than this could be heard reliably? How close do two signals have to be before we count them the same?

I believe 0 dB is said to be right near where the thermal noise of the atmosphere would be heard if hearing were much more sensitive. Is something going on below this? Seems hard to imagine how that would work.

[One and a half]

If I understand your suggestion, to record only silence and see what is there, I have done that. You end up with a pretty even noise floor around 123 db down or there about. Basically the residual noise in the electronics I am using for this. If one does this using front panel jacks instead of plugging directly into the sound card on the rear you get a good bit of garbage at various frequencies. The noise floor in places is raised 20-25 dB, with a few frequencies higher. I assume this is from the wire running right through the middle of the computer where so much electrical noise is located. So that spurious noise from inside the machine is much much larger than any measured difference in the interconnects.

Yes, without expensive lab instruments it's difficult to filter where the noise is coming from. The computer is great noise generator!

Certainly neither cable should be allowing any differences for tonality, or act like a filter. The null test takes into account amplitude, I just wonder how well frequency time shifts are taken into account. Time shifts occur with the geometry of the cabling, composition of metals and the joints and all.

[Julf]

Certainly neither cable should be allowing any differences for tonality, or act like a filter. The null test takes into account amplitude, I just wonder how well frequency time shifts are taken into account.

As long as there is some signal to compare, the null test will show the effect of time shifts. But if there is no signal (only silence), there of course won't be anything to shift...

[mayhem13]

You'll burn for this Esldude!

[Brian A]

Very cool thread. Will follow with interest. (NO, I am not voting your two options.)

[esldude]

Maybe a few good questions.

What would a null test not show that still would be audible? I don't really know of anything if the null is deep enough. Am I missing something on this?

Also, how deep a null would it take to convince you there is no difference that can be heard? Is there anyone who would argue with 120 dB nulling? I believe 60 dB nulls with a musical signal are actually going to be very hard to hear. I think 80 dB nulls probably are enough as a practical matter to declare something functionally identical. Anyone have some good info to add on how much is enough?

[Julf]

What would a null test not show that still would be audible? I don't really know of anything if the null is deep enough. Am I missing something on this?

Oh, ye of no faith! Open your closed mind, and hear! Don't you realize that there might be something like audio homeopathy at work here! Somewhere down in that low-level noise there is the diluted remains of that half bit that makes a difference.

You are also not taking into account the genetic memory of the air particles. Because your null test never passes through a loudspeaker into the air, it doesn't activate the memory of the original sound that the air particles still carry with them.

But most importantly, of course you don't detect a difference, because you are trying to detect one. As modern quantum music dictates, you can't both measure a sound and enjoy it at the same time - the moment you measure it, it looses it's animal magnetism, and you end up with a dead cat - and PETA won't be pleased!

[mitchco]

Null test?!? Now you are talking my language esldude! Have you thought about using Audio DiffMaker as another null test tool? One of its features is being able to “boost” the level of the difference track relative to the reference track. This is how one can easily tell if the difference is audible or not relative to the program level. See image below. It is educational to see how much boost (in dB) is required for the difference to become audible. Most people may be surprised how much boost is required to hear the difference. I zipped up the help file and attached it. There are a number of sample test procedures, not one for cables, but one for investigating the sounds of different capacitors. If nothing else, makes for an interesting read.



Even in the quietest home listening room, the ambient noise is likely to be around -60 dB SPL or a bit lower. There is a spec for this: NC - Noise Criterion So any difference signal below this level will likely not be audible due to the ambient noise of the room masking the signal. With headphones, a slightly different story.

Another point to consider. The null test shows the absolute signal difference between two waveforms – including any timing changes, sloped waveforms, whatever. I performed extensive testing using Audio DiffMaker to compare FLAC vs WAV, two bit-perfect music players, and 16/44 vs 24/192.

In the case of the comparing two bit-prefect music players, I measured the difference at the analog outputs of my Lynx DAC so I could capture any D/A artifacts, jitter, power supply noise, analog amplifier anomalies, anything and everything. I was able to null the signal to -90 dB. So relative to the ambient room noise and certainly relative to the program material, inaudible. Even wearing headphones, cranked to the max, can barely hear the difference signal on its own.

My view is that room acoustics has more effect on SQ than any other parameter in the signal chain, by orders of magnitude. When I first tried digital room correction (in my case Audiolense) I demoed it on a cheap old Soundblaster card in a gaming rig with Logitech G51 computer speakers. It made a huge difference. Even my wife (who is not an audiophile, but lets me get away with taking over the living room with all my gear – I am a lucky guy!) was able to tell the difference, it was night and day – no correction, they sound tinny. With correction, wow they sound great, how did you do that!

[mayhem13]

Null test?.

My view is that room acoustics has more effect on SQ than any other parameter in the signal chain, by orders of magnitude. !

Another Heretic.....you too sir will burn in the eternal flames of objectivity!

I've been preaching this stuff since joining CA and what do I get...........nuttin but grief i tell ya!

[esldude]

Hey, Mitchco,

Have been reading with interest your blogs and about that software. Have downloaded it, but not yet gotten around to using it. One purpose of doing this cable test was to see how the results compare with that software. Thanks for all the info it will be helpful. Also looks like it does more than I knew Diffmaker could be used for.

As for Mayhem13, do you keep an official CA heretic list?

And Julf, I don't know I guess you are invoking the Heisenberg feline uncertainty principle and the feline's name is Schrodinger. Is the cat's meow audible in the box and which cable do you need to hear the difference?

[Jud]

What would a null test not show that still would be audible? I don't really know of anything if the null is deep enough. Am I missing something on this?

*If* you were failing to measure something significant (and I'm not saying you are), then of course the fact of a null in what you did measure wouldn't matter. I have to watch the Keith Johnson video again to try to understand some of what he's saying about system effects and when and where they show up, to see what, where, and when he's saying the effects of things like cables can and should be measured. I suppose there is also the question about whether your testing platform is tending to submerge in background something that would show up more clearly with more quiet/sensitive/sophisticated apparatus.

Also, how deep a null would it take to convince you there is no difference that can be heard? Is there anyone who would argue with 120 dB nulling? I believe 60 dB nulls with a musical signal are actually going to be very hard to hear. I think 80 dB nulls probably are enough as a practical matter to declare something functionally identical. Anyone have some good info to add on how much is enough?

Heh, well, as you know, Keith Johnson and others have made remarks about artifacts one would ordinarily think were inaudible, even below the noise floor, affecting the listener's experience. Do I believe it? I dunno. I mean, I remember reading good cogent arguments that non-linearities at frequencies way beyond audible would be insignificant in amps, then KOJ goes and designs amps with flat response out to mHz and they just sound better than anything else I've heard. And probably the most intuitively outlandish contention I ever remember hearing in audio was the green pen thing, then KOJ goes and shows not only that it's true (that it has an audible effect), but why. So he has this track record of saying these things, being able to measure them and confirm their effects on the sound, building equipment based on these principles, and lo and behold the stuff sounds great. Though it may be tough to swallow some of it, it's darned hard to bet against him. (And of course throughout this entire discussion and thought process I'm conscious of the fact that the idea of me evaluating KOJ's claims is pretty laughable.)

[Julf]

I've been preaching this stuff since joining CA and what do I get....

A membership card?

[Julf]

Is the cat's meow audible in the box and which cable do you need to hear the difference?

Depends on how much the cat costs.

[esldude]

I agree that Spectral amps sound wonderful. But is it the actual extra response or something else? Demian Martin has said here (he was the designer of their early amps) that the extra wide frequency response wasn't a goal. It was a side effect of accomplishing other things they thought important. Maybe he could comment on what they specifically worked toward to get the desired sound quality. My guess is they wanted the slew rates to be very high.

KOJ's comments about the green pen remember were that it made a difference just not to the good side. Basically he seemed to be saying the painted edges either unbalanced or at least weighted discs so the servo's worked harder and caused measurable increases in jitter. Lots of early players has weighted damper discs or glued on sorbothane rings etc. Probably the same effect.

The thing that is so compelling about a null test is you don't even need to know what the effect is. If you don't get a good null, you will have to figure out what is different. With a good null, where is the difference? The signal on cables is varying voltage and current. There isn't anything else. A varying current flow of electrons with the varying pressure of voltage. Once you null below the inherent thermal noise I don't see there can be anything. Once below certain levels above that you likely are into an inaudible range. Nulls are not an accidental result either, it takes very little to corrupt them.

I wish I knew in more detail what KOJ was saying about audible stuff below the noise floor. I think he had mind digital noise floors. Like with CD being 16 bit yet you can hear things below the 96 dB noise floor once you add dither. You can hear some 20-25 dB maybe in some ranges 30 dB into noise. HDCD was making use of this. It used dither, it used companding where it amplified lower level signals on command to give a synthetic range of an extra 4 bits. Turned it off when the level was sufficient, and was basically undoing compression applied during the recording. It also used more than one filter type depending on transient information which switched depending on program material. When all put together it squeezed a 20 bit result (or close to it) from a 16 bit signal. But saying your nifty patented digital process works more or less like a digital DBX range expander doesn't sound so sexy. (yes I know that is an oversimplification of HDCD). I think he had that way of looking at things from his background with tape and LP.

Bob Stuart was a little less mystical about some things he thought worth doing. I always wondered why they had pre-emphasis as an option on CD's. Well it is a little like an RIAA curve on LP's. Done well you can gain the equivalent benefit of a couple bits or more effective resolution improvement.

[sandyk]

"KOJ's comments about the green pen remember were that it made a difference just not to the good side.
Basically he seemed to be saying the painted edges either unbalanced or at least weighted discs
so the servo's worked harder and caused measurable increases in jitter.
Lots of early players has weighted damper discs or glued on sorbothane rings etc. Probably the same effect.

Hi Dennis.
I would have thought that reducing vibration with a disc spinning around at a rate of knots, would reduce vibration, and hence "Jitter"

"Bob Stuart was a little less mystical about some things he thought worth doing.
I always wondered why they had pre-emphasis as an option on CD's. Well it is a little like an RIAA curve on LP's.
Done well you can gain the equivalent benefit of a couple bits or more effective resolution improvement."

A little like an RIAA curve on a LP ? Just what we need ! (wink)
Let's dumb down CD to -60dB (or worse) to make it sound more like vinyl!
Perhaps we could find a way to make the CD's read head vibrate in time to the loudspeakers too,
to give a little more warmth to what is claimed by some to be a cold and sterile presentation, (grin)
Kind Regards
Alex

[mayhem13]

As for Mayhem13, do you keep an official CA heretic list?
?

Pure sarcasm on my part. If I'm the only heretic.....so be it.

[The Computer Audiophile]

Another Heretic.....you too sir will burn in the eternal flames of objectivity!

I've been preaching this stuff since joining CA and what do I get...........nuttin but grief i tell ya!

Ha! I love it -> the eternal flames of objectivity!

Anyway, very good thread!

[esldude]

"
Hi Dennis.

"Bob Stuart was a little less mystical about some things he thought worth doing.
I always wondered why they had pre-emphasis as an option on CD's. Well it is a little like an RIAA curve on LP's.
Done well you can gain the equivalent benefit of a couple bits or more effective resolution improvement."

A little like an RIAA curve on a LP ? Just what we need ! (wink)
Let's dumb down CD to -60dB (or worse) to make it sound more like vinyl!
Perhaps we could find a way to make the CD's read head vibrate in time to the loudspeakers too,
to give a little more warmth to what is claimed by some to be a cold and sterile presentation, (grin)
Kind Regards
Alex

Odd you mention that Alex.

A couple nights ago, I took a 192/24 recording of some classical music for cello, knocked it down 50 dB in Audacity. Pulled up the RIAA curve, inverted that curve, and then played the result into the phono input of a very good vacuum tube pre-amp. The source was my old Dell laptop headphone out. The result was surprisingly satisfying. I don't think anyone would have noticed anything remiss. Only listened briefly, but the good quality of that recording came through quite clearly. Now had I only added some surface noise, rumble and acoustic feedback I could have turned that digital piece into a true 'reference recording'.

[xyz]

Why do a null test at all? If you are trying to evaluating cables, why not just listen to them as they serve no other purpose? Also, with RIAA EQ curves, I'm not sure what the goal is beyond using it for vinyl. Unless I'm mistaken, the reason for running the signal through the EQ curve is to restore the origional signal. The lower the note, the wider the groove on the record. If the groove becomes too wide, the stylus can't pick up all the information on the record. The solution is to roll the lower frequencies off too keep the groove smaller. Then, the phono preamp runs the signal back through the EQ curve to restore it to flat. What is to be gained?

[esldude]

Why do a null test at all? If you are trying to evaluating cables, why not just listen to them as they serve no other purpose? Also, with RIAA EQ curves, I'm not sure what the goal is beyond using it for vinyl. Unless I'm mistaken, the reason for running the signal through the EQ curve is to restore the origional signal. The lower the note, the wider the groove on the record. If the groove becomes too wide, the stylus can't pick up all the information on the record. The solution is to roll the lower frequencies off too keep the groove smaller. Then, the phono preamp runs the signal back through the EQ curve to restore it to flat. What is to be gained?

Because sighted listening has limitations. You listen to them, and they sound different. Yet the only difference may be in the listener.

Yes, the RIAA curve was a method for putting sound through a limited medium at higher fidelity. My purpose was just a goof. Just try it and see what happens. Surprisingly good with all the stuff going on actually. Also shows the versatility of digital manipulation with very little cost in terms of fidelity. And how well worked out the very good version of reproducing LP through a phono stage can be done.

[esldude]

Used the files I have gotten in this cable null test in the Audio Diffmaker software from Liberty instruments. The one Mitchco has made good use of in some of his testing.

Over 20 hz to 20 khz the Hyperion Knight recording used above had a correlated null depth of 97 dB using the raw files. Using files I normalized in Audacity increased that to 98 dB null depth.

Same test over some other tracks I recorded gave results from 95 to 98 dB correlated null depth according to Diffmaker.

As a test I compared a couple files to themselves and Diffmaker reported 300 dB nulls in both cases.

I also recorded silence for 10 seconds at two different times. Though low in level, the random fluctuations I would expect not to exhibit a great null depth. Sure enough Diffmaker reported a correlated null depth of only 1 dB for these two files.

There is much this software can do. In this comparison at first I did not do a frequency sweep of the AD/DA chain I am using. I then did that which generates an EQ file for use later by Diffmaker. Repeating the tests with EQ of my particular play/record chain improve correlated null depth by a decibel or two in each set of files I compared.

I can see a number of good uses for this software. For instance it can do nulls over restricted frequency ranges. One might track down where something is not the same using that feature. I haven't even looked at the help files linked to by Mitchco, so I likely have missed a number of things. Still this is good, useful software that isn't hard to use. It is set up so you can hear the difference result and amplify the difference to make hearing what is there easier. All very nice for null testing.

I am a bit bothered that running the same test over again gives a different result. But running the same files several times the results differ by less than a decibel at worst.

With nulls of 98 dB or so with 16 bit source material one has to think the interconnects used were responsible for no difference that should be audible.