Dave's attempt at a Null test

I know this topic has been done already.... but here I go. Originally dissatisfied by my understanding of how I think the Ethan Winer null test works,
(please accept my apologies if I have got the wrong end of the stick, I respect the effort put in to this), and intrigued by the interesting work by Max Townsend.
I hypothesize: to tell the difference between two wires the amplifier has to be part of the system. A year or two ago I did a null comparing interconnects using the left and right channels of a preamplifier one wire as a control and one EUT, and I could see a difference using a square wave and nulling on a scope.
Once upon a time I heard a clearly audible difference between a QED79 style cable and a Goertz low capacitance style cable. I imagined that it must be able to pick up this difference with a null. So apropo of nothing, but to avoid getting on with another project, I set up a circuit similar to below. I am using the 2i2 because of its balanced input to monitor with headphones.

nulltest1.JPG

My transformer is an output isolating transformer from a piece of old pro audio. I can't get the perfect null, but my initial result are interesting. Having made sure the gainclones are not clipping etc, with no speaker wires plugged in, apart from noise and buzz I get a fairly clean sounding rendition with some high and low boost. But when I plug in either wire, noticeable horrible grungy distortion. The distortion is probably different and the high capacitance wire is probably worse. I am a bit surprised, but am encouraged by this result. I have long suspected that a lot of the unpleasant sound associated with SS is its inability to drive cables without getting upset. Will try a tube amp next...

Please bring on the constructive criticism, I may have missed some one doing the same thing.
 
Last edited:
  • Like
Reactions: 1 user
AX tech editor
Joined 2002
Paid Member
I hypothesize: to tell the difference between two wires the amplifier has to be part of the system.
It is good practise in comparing something to try to isolate everthing in the setup that is not part of the something. By including an amplifier, you introduce a factor that has nothing to do with the cables you want to compare. For instance, the capacitance differences between cables can cause differences in the amplifier stability. That means you are not only comparing cables.

I understand your setup and it is quite ingenious, well thought out. But there are many factors introduced that can make a difference and with this setup it will be extremely hard to convince sceptics that the difference you hear is the cables.
As just one factor, the transformer will introduce a lot of non-linearities and will pick up stuff that can even differe with lead dress.
The use of music is also probably the worst signal you can use as it is unpredictable and extremely hard to measure difference with sufficient accuracy and resolution.
The CD player, the pot to balance the signal and the reliance on the two amps to be identical in performance is also weakening the case.

Take just a single signal generator, swap cables and measure, and be sure to equalize the signal generator load in the two cases. For instance, if one cable has lower capacitance, add the difference to the generator output to get identical loads. Etc.

Jan
 
  • Like
Reactions: 1 user
It is good practise in comparing something to try to isolate everything in the setup that is not part of the something. By including an amplifier, you introduce a factor that has nothing to do with the cables you want to compare. For instance, the capacitance differences between cables can cause differences in the amplifier stability. That means you are not only comparing cables.
Thank you for your comments. It is my intention to compare the cable and the driving device. I am completely convinced by the Ethan Winer null test results. But that he can find no audible difference, when I can hear a difference, tells me the test is missing something.

I understand your setup and it is quite ingenious, well thought out. But there are many factors introduced that can make a difference and with this setup it will be extremely hard to convince sceptics that the difference you hear is the cables.
As just one factor, the transformer will introduce a lot of non-linearities and will pick up stuff that can even different with lead dress.
The use of music is also probably the worst signal you can use as it is unpredictable and extremely hard to measure difference with sufficient accuracy and resolution.
The CD player, the pot to balance the signal and the reliance on the two amps to be identical in performance is also weakening the case.
Thankyou, and indeed there many factors, and the circuit has many problems. I just knocked it together from bits I had lying around :). I am not interested in convincing sceptics. This is for entertainment purposes at the moment, and hopefully thought provoking. To be honest I was surprised to hear a difference in the null when plugging and unplugging cables, so I thought I would share my results for anyone that is interested.

Take just a single signal generator, swap cables and measure, and be sure to equalize the signal generator load in the two cases. For instance, if one cable has lower capacitance, add the difference to the generator output to get identical loads. Etc.
It is my feeling that I will not get any useful results by using pure waves. I want to hear the hash, If I can do that with convincing repeatability I will attempt some measurements.
 
The CD player, the pot to balance the signal and the reliance on the two amps to be identical in performance is also weakening the case.

Take just a single signal generator, swap cables and measure, and be sure to equalize the signal generator load in the two cases. For instance, if one cable has lower capacitance, add the difference to the generator output to get identical loads. Etc.
That makes perfect sense for comparing two different cables, as is the topic.
When you modify the parasitics of the two different cables to be identical in order to equalize the load on the generator you are no longer comparing two different cables.
Maybe in the context of "I can hear a difference between two different cables when listening to music" it could be important to use music as test signal and have the exact amplifier in the loop that was used when the subjective assessment was made and also to use a reactive load (e.g. cheap speaker with most of the cone removed).
 
AX tech editor
Joined 2002
Paid Member
But when you go to 'I can hear a difference' you're back to subjective/personal judgement which is no longer valid as an objective judgement that is also valid for other people.
You need to make a choice what you want.
It is my intention to compare the cable and the driving device.
But how do you know then, that what you hear is the cable, and not the amp?

Jan
 
But how do you know then, that what you hear is the cable, and not the amp?
That makes perfect sense. But in real life you can't listen to a cable. Or an amplifier. So the question if one cable sounds different than another maybe also doesn't make sense and looking at the sum of various non-perfect devices and them interacting with different parasitics and their behaviour being influenced by these parasitics is what needs to be looked at? The question if you hear a difference between two cables or if it is phantasie can only be answered by a null test if as much of the gear that is used listening is included in the loop of the null test.
Otherwise you would just measure resistance, inductance and capacitance of various cables and compare these.
 
Jan & HBC

There are several valid points hat Jan brings forward, but some of those points can be easily neutralized.
Here are a few tips.

1) What signal to be used.
IMO white noise would be the most neutral and repeatable signal.
This has to be checked with an FFT over at least a 200msec recording, giving frequency bins of 5Hz, somehow corresponding to the the auditory sensitivity of humans.
This FFT, averaged over multiple spectra will show the used spectrum after the two gain clones and should be absolutely identical.

2) Transformer.
It would be easy to feed the transformer on both sided from one gain clone, and again measure the spectrum on the secondary side.
This will show the CMRR against frequency of the setup with transformer

3) Noise
The secondary noise in test 2 will be a measure how deep we will be able to discriminate distortion from the input signal.
This should be well below -100dB.

4) Load
I don't know in how far the circuit diagram in #1 is representative, but for testing interlinks as in the video, 8R is way too low and should be at least 10K or higher.
Suppose you make a short circuit with interlinks, then this setup will tell that the cables don't add anything.
Anyhow, the load should be as neutral as possible and not become part of the test. So the should be non inductive, having a very low voltage coefficient and absolutely no wire wound resistor should be used.

5) Gain Clone neutrality
It will be easy to test what happens when one gain clone is loaded with resp 1nF, 10nF and 100nF.
This will show in howfar the Gain clones can be regarded as being insensitive to a capacitive load.

Just my two cents,
Hans
 
AX tech editor
Joined 2002
Paid Member
Agreed Hans.
And the simplest way to exclude the gainclones from the equation is to use one and the same with both cables. Just switch the cables.
My biggest issue with the proposed system is that there's so much stuff in there that is not under control that can easily sway any results.
Whatever the results are, anybody that sees the setup will never accept the result because it cannot be trusted to be only dependent on the cables.

Jan
 
It is my feeling that I will not get any useful results by using pure waves.
I read this often on forums and I am gonna use this one as an example.

The reason why using sine waves first, is to set a baseline.

Because with deductive reasoning, we know that if a certain difference is impossible (or barely) to hear with just pure sine waves, it's most certainly impossible with music.

From a experimental point of view it also works a lot better for eventually troubleshooting the setup as well as giving a much clearer path to understand what's going on.
Or in other words, linking the practical results with the hypothesis.

I totally agree with Jan, introducing more stuff will only lead to more variables and potential problems.
Which in the end might even influence the results.
In fact, that will be extremely likely!

It's most definitely not the way to go when you actually want get proper and objective results.
 
Jan & HBC

There are several valid points hat Jan brings forward, but some of those points can be easily neutralized.
Here are a few tips.

1) What signal to be used.
IMO white noise would be the most neutral and repeatable signal.
This has to be checked with an FFT over at least a 200msec recording, giving frequency bins of 5Hz, somehow corresponding to the the auditory sensitivity of humans.
This FFT, averaged over multiple spectra will show the used spectrum after the two gain clones and should be absolutely identical.

2) Transformer.
It would be easy to feed the transformer on both sided from one gain clone, and again measure the spectrum on the secondary side.
This will show the CMRR against frequency of the setup with transformer

3) Noise
The secondary noise in test 2 will be a measure how deep we will be able to discriminate distortion from the input signal.
This should be well below -100dB.

4) Load
I don't know in how far the circuit diagram in #1 is representative, but for testing interlinks as in the video, 8R is way too low and should be at least 10K or higher.
Suppose you make a short circuit with interlinks, then this setup will tell that the cables don't add anything.
Anyhow, the load should be as neutral as possible and not become part of the test. So the should be non inductive, having a very low voltage coefficient and absolutely no wire wound resistor should be used.

5) Gain Clone neutrality
It will be easy to test what happens when one gain clone is loaded with resp 1nF, 10nF and 100nF.
This will show in how far the Gain clones can be regarded as being insensitive to a capacitive load.

Just my two cents,
Hans
Thanks with regards to 1, 2 and 4. In this experiment I am testing speaker cables, hence 8r, but actually I am testing the amplifier and cable as a system. I believe the decent quality speaker to be fairly transparent so the change I am looking for must be between the source and the speaker terminal. I would like to investigate a difference in sound that was clearly audible to me and others, I have tried it in 3 different systems, all SE tube, with the same audible results. I don't think I would hear the effect with white noise. Yes I will check the CMRR of the transformer as suggested, good call. I am using gain clone initially as it was to hand.


I read this often on forums and I am gonna use this one as an example.

The reason why using sine waves first, is to set a baseline.

Because with deductive reasoning, we know that if a certain difference is impossible (or barely) to hear with just pure sine waves, it's most certainly impossible with music.

From a experimental point of view it also works a lot better for eventually troubleshooting the setup as well as giving a much clearer path to understand what's going on.
Or in other words, linking the practical results with the hypothesis.

I totally agree with Jan, introducing more stuff will only lead to more variables and potential problems.
Which in the end might even influence the results.
In fact, that will be extremely likely!

It's most definitely not the way to go when you actually want get proper and objective results.
Yes I agree, and I have indeed set up the experiment using a signal generator, and using square wave I can see my frequency response anomaly. But I suspect I am looking for some complex inter modulation by-products, and I know what it sounds like, So I use music.
 
But I suspect I am looking for some complex inter modulation by-products, and I know what it sounds like, So I use music.
Introducing (unintentional) bias, is the least thing you want to do.

So measurements first where you know you can rely on.
Intermodulation is extremely easy to see and value when measuring pure sine waves.

Our own listening isn't reliable tool at all.
Mostly because we don't just only objectively listen, but our brain and other senses WILL mess with it.

Like I said before, get a objective baseline first, move from there.
Any other approach makes the whole test very quickly invalid.

The only exception is doing double blind tests. These tests can be very troublesome by themselves (also practically speaking), they also don't give proof of the hypothesis. They are only there to exclude variables and/or show us potential differences and hooks to work with.
From a statistical point of view they can be a big can of worms sometimes as well.

Also, I would advise doing some literature study first.
Knowing the orders of magnitude we are talking about, your average sine wave generator isn't gonna cut it.
A YT rip with already compression in it, is also not the way to go.