Null test of analog interconnects

[esldude]

Actually Alex, you were almost reading my mind. How much difference between crimped end and welded silver and the tiny bit less resistance of silver vs copper. I noticed in the spectrum values, almost all levels show the cheaper interconnect a tiny bit lower in level. I have no highly precise way to measure tiny resistances handy. So using a multimeter and connectors to string a pair together I estimate the difference between the two interconnects in resistance is between .1 and .2 ohms. The numbers fluctuate and I would say .15 ohms though I am not confident that is highly accurate. But the cheaper interconnect was clearly the slightly higher resistance.

So the output impedance of the Tact is 33 ohms, the input impedance of the other device 25,000. Were one to assume .2 ohms more on one interconnect vs another that by itself would result in a difference level of a calculated -102 decibels. That would be if every other parameter is exactly the same and perfect. None of those are likely exactly true, but are more than likely near the case. Even if off a bit, it would appear a considerable amount of the limit to the null depth is simply these tiny resistance differences. That fits in with the amplified residuals sounding flat and even in frequency response. If the resistance difference was only .1 ohms, then it would limit the null to around 108 db. I also tested 6 feet of the cheaper cable. If I add the additional resistance for that, I would expect a null of only 95 db. I think I got 93 or 92 db which is in the general area due to resistance. It doesn't answer why the cable doesn't null deeper to itself. It does seem likely these tiny resistance differences are a quite considerable reason for much of the difference that does exist. Meaning any other qualities of cable construction beyond RCL differences are exceedingly small or non-existent.

This is the final thing. I would say this myth is busted. Lumped RCL values describe interconnect effects. Nothing else to it. The silver and welded connectors did cause a difference. One could use some good solder and about 3 inches less copper to get a better match without the expense of silver. Also points out just how deep even a 'mere' 100 db null is when a tenth of an ohm or two can limit the null being deeper than this.

Been a fun bit of messing about and learned a good bit of odds and ends about such things and the equipment I have. But it is a done deal for me. I won't worry about interconnects anymore. If there isn't any difference in the waveform to hear then you cannot hear it. Pretty simple. I suppose I could go further and see if trimming these things down to the exact resistance deepens the null. At this point I don't think I have enough doubt to bother.

Do I still hear differences? Yep, I think I do. And no they are not really there. Interesting on another level, but it is not the wire that is causing it.

I am sure plenty will not accept this, and that is fine. Someone else needs to do some measuring and show where something is changing the waveforms.

[esldude]

Something I have noticed looking over the data in spreadsheets from the spectrum analysis of these cable differences.

If I had cables that were both completely and fully identical would I get an absolute series of zeroes in the difference file? The answer is no you would not because of thermal noise limitations. If the difference signal was approximately the same as the thermal and other noise limitations of the system for playback and recording you would have to think any difference was very small. Otherwise differences between cables would result in levels above the system noise.

Playing a digital file of silence and recording the result gave a spectrum plot that averaged either -139.8 or -139.9 db over several different silent recordings. When I look at the value for the difference file in cases where music was used over two compared cables I had averages of -138.2 db for a 16 bit source file and -139.4 db for a 24 bit source file. That is awfully darn close to looking like the difference file is very little other than system noise. The tiny remaining extra noise fits reasonably well with the small resistance difference in the cables I talked about in an earlier post.

[Jud]

Hey, Dennis. Been following with great interest, but definitely believe you've ended things a bit early. The primary thing that has not been done is to conduct the test on a cables that sound different and measure different in some aspect commonly accepted to affect sound, and see how deep the null is. One example would be to measure cables coming off the analog output of a DAC when using Toslink and asynchronous USB inputs.

If you get deep nulls even where cables not only consistently sound different but where the system measures different, then I think there are grounds to question what a null result means.

Another interesting test would be what level of noise is necessary to make a signal of a given volume level completely in audible. In other words, how deep into the noise can we hear, or is the noise floor an absolute limit to audibility?

These and other assumptions underlying your stated conclusion do not seem to me to have been tested yet.

[Brian A]

... Been a fun bit of messing about and learned a good bit of odds and ends about such things and the equipment I have. But it is a done deal for me. I won't worry about interconnects anymore ...

Thank you for running the ball down the field as far as you have. You've done some interesting work and you have laid out the territory for others (hopefully) to explore.

It is so rare to have someone actually do the work rather than just rant hypothetically or cite tangential technical literature. Thank you again.

[Paul.Raulerson]

Hey hey - I thought this article might be of interest. It is only tangentially related to the subject matter, but much more related to the reactions of many of us. It is amusing how the pictures in the included graphic are - eh - reflective. (grin)

-Paul

PJ Media » Climate Change: Why Do the Facts Fail to Convince?

[jhwalker]

Hey hey - I thought this article might be of interest. It is only tangentially related to the subject matter, but much more related to the reactions of many of us. It is amusing how the pictures in the included graphic are - eh - reflective. (grin)

-Paul

PJ Media » Climate Change: Why Do the Facts Fail to Convince?

Interesting read - but I must point out that I am *not* balding, nor do I have a beard

[esldude]

Hey, Dennis. Been following with great interest, but definitely believe you've ended things a bit early. The primary thing that has not been done is to conduct the test on a cables that sound different and measure different in some aspect commonly accepted to affect sound, and see how deep the null is. One example would be to measure cables coming off the analog output of a DAC when using Toslink and asynchronous USB inputs.

If you get deep nulls even where cables not only consistently sound different but where the system measures different, then I think there are grounds to question what a null result means.

Another interesting test would be what level of noise is necessary to make a signal of a given volume level completely in audible. In other words, how deep into the noise can we hear, or is the noise floor an absolute limit to audibility?

These and other assumptions underlying your stated conclusion do not seem to me to have been tested yet.

Typically you can hear a good 20-25 db into noise. Now that assumes the noise and embedded signals are loud enough to hear in the first place. In this particular question of cables you are looking at hearing the tiny differences while hearing fairly loud music. Masking effects sometimes mean one signal can mask others just a few decibels lower and often masks quite a bit if it is more than 20 db down from the main signal. That idea is how MP3's manage to work as well as they do. One of the things I was looking at recently, I wish I remembered where, showed violin notes and overtones along with the masking curve it would create. Those notes and harmonics would mask anything more than 15 db below that note which would cover a couple octaves and then some more at lower levels. In this particular area for such a note and any info lower than the note two cables could difference down to only -20 db and you wouldn't hear the difference.

I know it is hard to let go of, but there just doesn't seem to be anything here. It looks to be down around the physical limits of the electronics and properties of sound in air. If there is so close to nothing there you cannot hear it.

Don't know if I sent you the link to PDF, I think I did Barry. It was about how close to the brownian motion of air molecules our ears respond. Nerve firings from the ear do respond somewhat to brownian air motion. It is random and so is the nerve firing. It appears this increases our low level sensitivity in much the same way dither extends noise floors in digital recording. Our hearing as in our perception in the brain ignores this random firing though it primes it for better sensitivity. Somewhere around that 20 db above brownian motion level our nerve firings change from randomness and we begin to perceive. In other words our hearing is not very far from the limits of the physical properties of air itself.

Now think about what a difference signal of -139 db means. It means if you were listening at only 120 db loudness the difference would lie at the random brownian air motion level. And our low level thresholds down there are only when there isn't other louder sound to mask anything. So if you were listening at 120 db you will never hear that tiny difference. It is swamped or at least down among random air motion. Or look at it another way. If two cables can be differenced to -139 db actual levels it is just about the same difference caused by random properties of air when you compare listening now to listening 10 minutes from now without changing anything.

Now the signal levels of these cables didn't null to -139 db. But the resulting difference did lie in that area. If a lot of of your levels are down from maximum levels it takes less nulling to get them buried in the noise which appears to be what is happening. Same effective result in that reproducing these signals any difference in them is lost in low level thermal noise.

As for measuring toslink vs usb and such that is something that could be done. But if the resulting measurements were similar with only tiny differences apparent I might question how audible they were rather than assume it confirms that nulling or difference testing is somehow flawed.

And this is just a repeat of an old, old story line. You can find stuff at least a couple generations back where such testing shows things more similar than listeners supposed. And listeners refused to believe it. As the technology has improved so have the measurements and the depth of the nulling from it. Not much else has changed otherwise. People don't want to believe, they experience the subjective differences and want just a little more precision in the measurements as they are sure it will turn up the readily audible differences.

So I don't think I have stopped short. The subject was analog interconnects. They are similar enough not to worry in most situations. Digital interconnects and clocking are another topic for another thread at least. Or if one questions the veracity of the whole idea of null testing maybe another thread is in order.

[Jud]

Typically you can hear a good 20-25 db into noise. Now that assumes the noise and embedded signals are loud enough to hear in the first place. In this particular question of cables you are looking at hearing the tiny differences while hearing fairly loud music. Masking effects sometimes mean one signal can mask others just a few decibels lower and often masks quite a bit if it is more than 20 db down from the main signal. That idea is how MP3's manage to work as well as they do. One of the things I was looking at recently, I wish I remembered where, showed violin notes and overtones along with the masking curve it would create. Those notes and harmonics would mask anything more than 15 db below that note which would cover a couple octaves and then some more at lower levels. In this particular area for such a note and any info lower than the note two cables could difference down to only -20 db and you wouldn't hear the difference.

Thanks, I really appreciate the good information you continue to provide here. You've done plenty of work on this, and if anyone (including me) has any questions left at this point, it's only fair that they take on the burden. So as I indicate some areas below where I continue to be curious (and that is what it is - please don't take this as any desire to be cheaply contradictory to or unduly skeptical of your good work), it will be up to me or anyone else who thinks these areas might repay additional research to do the necessary work.

One of the things I do wonder about is how far the "masking effect" goes. It seems to me low noise levels are something to be pursued for best sound, and that reliance on the "masking effect" tends to turn that on its head. I wonder, for example, to what extent use of home computers rather than sensitive test equipment tends to mask differences and thus create nulls where differences might exist in an electronically quieter test environment. I also think I detect, rightly or wrongly, some nagging doubts re the "masking effect" in your remark regarding mp3s working "as well as they do," rather than that they are completely indistinguishable from lossless formats.

Don't know if I sent you the link to PDF, I think I did Barry. It was about how close to the brownian motion of air molecules our ears respond. Nerve firings from the ear do respond somewhat to brownian air motion. It is random and so is the nerve firing. It appears this increases our low level sensitivity in much the same way dither extends noise floors in digital recording. Our hearing as in our perception in the brain ignores this random firing though it primes it for better sensitivity. Somewhere around that 20 db above brownian motion level our nerve firings change from randomness and we begin to perceive. In other words our hearing is not very far from the limits of the physical properties of air itself.

I was actually wondering about the volume level of air molecules in Brownian motion, having heard it when the environment is especially quiet. Thanks, for answering before I asked. :-)

As for measuring toslink vs usb and such that is something that could be done. But if the resulting measurements were similar with only tiny differences apparent I might question how audible they were rather than assume it confirms that nulling or difference testing is somehow flawed.

Let me clarify that what I'm after here is a picture of what the analog cables are doing when fed audibly and measurably different signals. Academic work since the 1980s has said jitter differences of 20ps or even less should be audible in the resulting analog reconstruction, and I would suppose the jitter differential between the toslink and async USB inputs of a DAC should be at least that great, likely much greater. There would also be a difference between any electrical interference carried through the system by the USB cable versus none carried over Toslink. So again, to be clear: What I'm after is what a null test shows regarding analog interconnects in situations where the source signals have measurable differences commonly accepted by experts (academics, audio engineers in published literature) to result in audible differences.

And this is just a repeat of an old, old story line. You can find stuff at least a couple generations back where such testing shows things more similar than listeners supposed. And listeners refused to believe it. As the technology has improved so have the measurements and the depth of the nulling from it. Not much else has changed otherwise. People don't want to believe, they experience the subjective differences and want just a little more precision in the measurements as they are sure it will turn up the readily audible differences.

I think there is some merit to the idea that even supposedly "objective" people are quite subjective in the publications and research they choose to credit. How many people on this thread are still questioning the audibility of jitter differentials far in excess of the ~20ps Hawksworth and Dunn found in their AES papers in the early to mid-80s?

So I don't think I have stopped short. The subject was analog interconnects. They are similar enough not to worry in most situations. Digital interconnects and clocking are another topic for another thread at least. Or if one questions the veracity of the whole idea of null testing maybe another thread is in order.

You certainly haven't stopped short in terms of the consistent effort you've given throughout this thread. I will still have questions about how efficacious and reliable null tests are until I can perform or be pointed to the types of tests I describe above, so I can see by actual measurement whether deep nulls are achieved even where there are measurable and audible differences. But at this point that's my problem to research.

[mitchco]

Hey esldude, really nice job. I have spent many hours performing my own null or difference testing on quite a few audio aspects including:

WAV vs FLAC (includes MP3)
JRiver vs JPlay
16/44 vs 24/192
Sonic Signatures: The Art and the Science

Null or difference testing is older than most of us on this forum and has been around since the first amplifier was built. Whenever an analog circuit is biased or a differential pair is matched, that in itself is a null or difference test.

Ever hear of the Audio Mythbusters?



And the files used for this session can be found here: AES Workshop Video Files

Btw, great section on null testing too near the end of the video.

[spdif-usb]

Flat response DC-26khz, distortion at or below .1%, and 20 bit levels of SNR with good digital filtering. That pretty well covers it.

The linear frequency response has to be in air – and the in air qualification matters. Moreover, the -60 dB hearing threshold for .1% 2nd-order harmonic distortion applies only to mono presentation, i.e. not stereo sound.
Add to that the fact we do not only hear sound, but also build a map of sound objects in our cortex while we listen. And the fact Bob Stuart told Robert Harley measurements are not the be-all and end-all in audio.

[esldude]

The linear frequency response has to be in air – and the in air qualification matters. Moreover, the -60 dB hearing threshold for .1% 2nd-order harmonic distortion applies only to mono presentation, i.e. not stereo sound.
Add to that the fact we do not only hear sound, but also build a map of sound objects in our cortex while we listen. And the fact Bob Stuart told Robert Harley measurements are not the be-all and end-all in audio.

I find it odd you seem attached to Stuart's less specific comments about audio to a generalist magazine versus a bit more rigorous treatment he gives in AES papers. His statement that measurements are not the be all and end all in audio can be looked at in any number of ways. Given his other writing, my guess is Mr. Stuart didn't mean to imply there something highly mysterious going on with listening. I imagine it was more about preferences, the differences in listening your in living room when trying to reproduce a huge recorded space etc. etc. And yes, the .1 % at -60 db and all that is for a particular test. I find it odd how people continue to want to believe that much more complex sound like music or other live events will let them be even more discerning than when using the simple test signals. I don't get that.

[spdif-usb]

I find it odd you seem attached to Stuart's less specific comments about audio to a generalist magazine versus a bit more rigorous treatment he gives in AES papers. His statement that measurements are not the be all and end all in audio can be looked at in any number of ways. Given his other writing, my guess is Mr. Stuart didn't mean to imply there something highly mysterious going on with listening. I imagine it was more about preferences, the differences in listening your in living room when trying to reproduce a huge recorded space etc. etc. And yes, the .1 % at -60 db and all that is for a particular test. I find it odd how people continue to want to believe that much more complex sound like music or other live events will let them be even more discerning than when using the simple test signals. I don't get that.

Bob Stuart said he doesn't believe in Black Magic, and neither do I. However, that doesn't mean the human brain isn't highly mysterious, even to world's best neuroscientists. I find it odd how electrical engineers continue to want to believe simple test signals can teach us everything there is to know about the human auditory system, let alone about how the human cortex interacts with it. What could be more biased than the assumption the human hearing and its ability to spatially localize sound objects as well as to identify them against a noise background will work the exact same way for natural, broadband sounds inside a room as it works for narrowband, artificial test signals though a pair of headphones? I am talking about the natural process that occurs in the cortex when we listen to real music at home and enjoy ourselves doing so, not a bunch of sine tones being used to measure the contents of our heads in an awful test lab. Do you know what endorfins are and how they define alot of what we hear?
Here's from the second paragraph in the interview article that you loathe so much.

What I think is outrageous is to say we understand everything about how the human hearing system works, because what we do know is that it’s incredibly sensitive to certain kinds of differences and very tolerant of others.

[orgel]

Ever hear of the Audio Mythbusters?

Thanks for the video, mitchco. Super interesting and educational. The part about masking was really enlightening.

And thanks to esldude for all his hard work and his patience in explicating it. This thread's given me a lot to chew on.

--David

[prufrock]

The thing is, we don't need to rely just on the EEs to tell us what is or isn't happening here. Other tools are available. It is true that the physiology involved is very complex and you would need to be very careful in extrapolating from test signals to actual music.
What many are falling victim to here is a type of intellectual arrogance (Nothing personal here btw - just talking about the argument, not the person). Most of the people who hear differences cite some related expertise or years spent listening as a reason to put them outside the circle. No need/point in testing us they say. We are experts. Its hard to imagine a more basic failure in proper thinking than this. This is pre age of reason stuff. Their arrogance enables them to totally discard all the evidence that shows that when a cable/interconnect/most amps are unsighted and no other cues are available, they cant tell one from the other. Instead of concluding their brain has deceived them (if they cared to look they would find it does this on a regular basis), they conclude a testing protocol that has proven itself time and time again is not applicable or faulty.

[Paul.Raulerson]

Most of the people who hear differences cite some related expertise or years spent listening as a reason to put them outside the circle. No need/point in testing us they say. We are experts. Its hard to imagine a more basic failure in proper thinking than this. This is pre age of reason stuff. Their arrogance enables them to totally discard all the evidence that shows that when a cable/interconnect/most amps are unsighted and no other cues are available, they cant tell one from the other. Instead of concluding their brain has deceived them (if they cared to look they would find it does this on a regular basis), they conclude a testing protocol that has proven itself time and time again is not applicable or faulty.

This is so south of the mark... everyone, including me, that can hear differences has pretty clearly said that (1) We don't understand why, (2) That it applies only to the individual, (3) That training and experience does make a difference and most of all (4) that people should listen for themselves and decide.

If you think that is arrogance, well, then we disagree. I think it is quite humble, especially given the qualifications of the people it comes from (excluding myself, in this instance).

The argument you seem to be making is that training and experience do not under any circumstances make a difference in what one can hear. That is clearly is clearly a false assumption - training and experience always make a difference.

You also assert a testing protocol that has been proven "time and time" again. Please provide details of that protocol. As far as I know, the testing protocol for each test has been unique, and usually takes the subject out of his or her well known environment and puts them in a stressful "testing" environment. Stress easily affects a user, unless the user has been trained to overcome it, and has enough experience to be able to do so when desired.

I think some see it differently, but - the overweening arrogance I sense is the implicit assumption that people have to prove what they clearly hear to satisfy the objections of some "group" or another. Obviously, that "group" isn't the group spending money on cables, why would they? If they are doing so, they should run their own tests to ascertain for themselves what they can and cannot hear, and what matters to them.

Let me put that another way, in a search of the system, I cannot find one instance of a person being berated for not being able to hear a difference. There are numerous instances where people who can hear a difference are berated by others. It's silly, it's rude, it isn't scientific, it *is* bullying, and it never ever has changed the opinion of a single person. (shrug) So why am I attempting to shed some reason here? Obviously, because I am a fool who believes people can have reasoned debate and that the truth is out there to be found. And because whenever there is a perceived paradox, it is always because some critical information is missing.

-Paul

[esldude]

I think I am going to start a new thread discussing whether or not null testing is flawed and why it might or might not be. Nothing wrong with what is going on here. But the thread has gotten very long, which is good shows there is some interest. Also lots have questioned how good a null is good enough. The answer to that isn't exactly obvious. But a new thread related to just the ins and outs of null testing and how it relates to what we hear seems like a good idea at this point. Bothering to consider this in light of the testing I have done has given me some ideas I wouldn't have considered if not stimulated by those questions. So I can put out how it looks (sounds) to me and see what other ideas that prove useful we can bounce off each other.

[prufrock]

Paul
I actually appreciate your coming out and admitting you hear differences. That is an honest thing to do and it enables us to have a decent debate on the issue. I am just taking the other side on this particular issue and pushing my points to the max. Sorry if this comes across as a little aggressive at times. Rest assured, it is just debate and nothing more than that.

[Paul.Raulerson]

Paul
I actually appreciate your coming out and admitting you hear differences. That is an honest thing to do and it enables us to have a decent debate on the issue. I am just taking the other side on this particular issue and pushing my points to the max. Sorry if this comes across as a little aggressive at times. Rest assured, it is just debate and nothing more than that.

No sweat- we are all pretty passionate about audio, or we would not be here. Same from my side. I admit though, I think there is some middle ground that will make sense to everyone. Wish we could find it...

Check out the new discussion Dennis started- I think you might have some great thoughts to share in that discussion.

-Paul

[esldude]

I mentioned earlier that I was bothered that a recorded signal compared to the digital original only nulled out to around 44 db in Diffmaker. Comparing the difference in Audacity didn't look much different. The difference would show between 40-50 db difference for most frequencies.

I exported the spectrum plots from the two into a spreadsheet and noticed something odd. To compare in Audacity I normalized the two files as you pretty much have to in order for it to mean anything. Something is throwing off or otherwise interfering with the normalization. As what I noticed was the recorded file was a pretty consistent amount louder. I am pretty sure Diffmaker does normalization so likely it was being thrown off the same way. The level shift was the same out to 2 decimal places with slight variation in the third position. The only areas is wasn't this minor amount was below 20 hz and above 19 khz (this latter likely from anti-alias filtering being done twice for playback and record).

So I averaged this amount over the 20hz-19khz range and reduced the level of the recorded file this much. Differencing against the original in Audacity the remaining signal dropped by over 20 db. As that probably puts it in the mid 60 to nearly 70 db null range that seems better to me. Putting this corrected file in Diffmaker did improve the result from 44 db to 48 db. So I still think it is normalizing and gets thrown off by something.

As there must be some distortion both in playback and recording a difference of -60 db or a bit more versus the digital original doesn't look too bad.

[goldsdad]

I am pretty sure Diffmaker does normalization

Yes, if I recall correctly about the one occasion on which I used Diffmaker, that was confirmed by differencing a file with a significantly attenuated version of the same file, the reported result being an extremely small difference. My memory may be tricking me, so try that for yourself.

[esldude]

Thought it might be nice to include the later results of this difference testing. As I have gotten various issues figured out and gotten a quieter background to work with results are improved over the initial posting of sound files I did. This is a Wilson recording excerpt of 30 seconds. Audioquest Diamond was used for one recording. A generic set of interconnects was used for another recording. Then the two recorded files were subtracted from one another leaving only the difference between them.

Not sure there much more to discuss, but this has been a good thread with lots of discussion politely and respectfully done by all involved. Thanks to all who took part.