Ethan Winer Builds a Wire Null Tester

[pkane]

well impedance and voltage are both complex numbers with imaginary numbers, which can sometimes be zero.

thing is, why are you obsessed with how people interpret circuit like it was 100 years ago.

one thing here is the VU meter. It’s designed for 600 ohms source output impedance, and that’s a fairly old standard. Modern solid state audio devices have a typical output impedance of 100 ohms, with actual value ranging from 50-300 across manufacturer. Jumping from 100 ohms to 600 ohms, with modern recording interfaces and long cables, you will get -7dB signal loss in 10 khz compared with nothing before the jump.

You keep changing the argument without ever answering a direct question. And you keep bringing more and more nonsense into the discussion that has nothing to do with the topic being discussed. I suggest some reading and further research before you try to argue things you don't understand.

 

[pkane]

Well you need to read about what noise correlation coefficient means in conplimentary noise reduction.

are you a bot? It’s a serious question.

 

[solderdude]

One simply cannot be this thick, pig headed and non comprehensive.
Only a troll can sprout the nonsense posted.
Even Paul McClown could not come up with these theories.
There is either a complete lack of understanding nulling or a deliberate attempt to discredit the nulling process by using false (bogus and non related) arguments and appeal to authority which he does not understand.

period.

The A & B inputs have 115dB SN ratio. The signal supplied to it is just 1 channel (left or right). The 'B' input gets its signal from the A input (via the DUT) and adds its own noise on top of that of channel A.
The null circuit thus cancels the noise on input A while showing the difference between A and B (the noise of B + its own noise).
That's the bottom noise floor of the null tester (noise of B + noise of null circuit + noise of the amplification after it)

It is abundantly clear the mono signal applied to A will also be the same (within the audible range and well outside) on input B in amplitude and time (phase). So regardless of what is connected to A is also present in the exact same manner on B and thus can be cancelled out completely even if there are small amplitude differences between A and B input and of the nulling circuit. That's where the null pot comes in.

The Sansa and its performance have NOTHING to do with it as the voltage coming out of the left or right, whatever it may be.. noise, an 8 bit signal, music, test tones is exactly the same on A + B (with the added noise of the input/nulling) so cancels except the noise.
Even if the source has a S/N ratio and distortion level of just -30dB it does not matter because the analog voltage is not stepped. it is analog and thus also contains 'values' well below 115dB BUT they will be the same for A and B and thus can null completely.

The voltage meter simply measures the (amplified) output voltage of the entire device. It is completely irrelevant what sensitivity it has and what impedance. The meter is 'dumb' and 'agnostic' it just displays the peak input voltage and it only needs to have a -30dB S/N ratio to function properly.

I have built and used my own null tester about 30 years ago and know what can be achieved by it and what not. (Mine was noisier but can fit better opamps if I wanted to).

 

[Thomas savage]

This is why I had quantum stickers on my old amp's transformers.

They mitigated issues in non observable dimensions.

 

[SIY]

I simple have no idea why "random" noise shooting into both channels of the splitter would have the same pattern so you can cancel them out later, as if they are not random after all.

And none of you ever managed to explained why in error calculation, addition and subtraction treat error exactly the same way, by summing them up. What's statistical meaning of that +- sign in front of error?

I saw that Star Trek TNG as well.

 

[Wombat]

http://ipl.physics.harvard.edu/wp-uploads/2013/03/PS3_Error_Propagation_sp13.pdf

http://lectureonline.cl.msu.edu/~mmp/labs/error/e2.htm

In quadrature calculation only produces '+' result.

 

[solderdude]

I simple have no idea why "random" noise shooting into both channels of the splitter would have the same pattern so you can cancel them out later, as if they are not random after all.

Noise is random. A signal containing random noise is just a signal that aside from the intended signal (music) also contains some random noise.
That signal (including its own noise) is equal for both A and B input so can be nulled.
This has been explained by quite a few people.
ONLY the noise (at -115dB of input B) and the noise of the summing opamp can NOT be nulled as only that differs. That noise is around -110dB.
So even if the null of the test signal would null down to -200dB one would only see the noise floor of the circuit (around -110dB).
BUT when the music is nulled at say -115dB you will not measure that signal (as music is also random) but can still hear something related to the music even in the -110dB noise floor.
The noise from a source (most of which is random) and its 'equivalent resolution' is completely irrelevant.

I believe such is not hard to understand.

Ethan's null circuit thus works as advertised (so does mine) and your assumptions are clearly incorrect.
On top of that also in software it is possible to get excellent nulls providing there are extremely small or well documented and compensated for errors between both recordings.

And yes... you can null noise just as well as music and test tones.

 

[Wombat]

View attachment 47248

Once signal splitted, random noise are considered statistically independent in error estimation. They don't cancel out in phase reversal.

After 4 opa1612 (BJT input stage), random noise in the 2 channels from the sansa clip are totally independent.

You can only record it to cancel out random noise as it has a "moving average" over a given period of time.

PS. High school psychics are knowledge 200 yrs old.

Psychics we are not. Please provide the physics or math, high school level or otherwise.

(There is a .pdf copy of the book on the WWW from a remote Russian university - 36.4Mb, 2588 pages. I couldn't be bothered wading through it to find one sentence in context).

 

[Blumlein 88]

View attachment 47248

Once signal splitted, random noise are considered statistically independent in error estimation. They don't cancel out in phase reversal.

After 4 opa1612 (BJT input stage), random noise in the 2 channels from the sansa clip are totally independent.

You can only record it to cancel out random noise as it has a "moving average" over a given period of time.

PS. High school psychics are knowledge 200 yrs old.

You still have the little problem of the fact this works as advertised. Which indicates with 100% reliability you are in error somewhere in your chain of thinking. We all see where it is, and have pointed to it, but you refuse to look there.

Perhaps if you wrote out a simple circuit description of this tester in use you'd see where you are making a mistake. Perhaps not. There however is a mistake because the device functions as described.

 

[Wombat]

I'm still waiting for the maths/physics proof. You have said it is at high school level so you must have accessed it.

 

[Wombat]

The isolated attachments/references need a coherent context. Are you able to explain your point in any coherent and comprehensible form?

If not you are wasting your time and ours.

 

[DonH56]

A careful look at the schematics will reveal that OA3 is non-inverting and OA4 inverting in the first schematic, and in the second signals are applied to the inverting and non-inverting inputs of OA5. That is what does the subtraction.

 

[Blumlein 88]

This is the actual circuit. You'll notice the nulling occurs prior to active electronics. Then you have gain on the residual.
https://audioxpress.com/article/you-can-diy-building-a-null-tester-device

 

[DonH56]

DonH56, if you assign logical meanings, I think you are right.

Please check image below. I updated another 2 probing. The 0.001V different is introduced between OA2 and OA4. This seems to be a systematic error of this circuit itself. Because of this, I think it is inappropriate to assume this circuit is capable of doing -110dB measurement accuracy of the input signal.

In fact, if I change input of upper channel to 1.0013V and center the trim, result is ever smaller.

I design analog circuits for a living and have for a few decades now. So I don't mean logical complementing like digital, I mean signal polarity inversion. With less than infinite op-amp gain-bandwidth and real input and output impedances you'll find you will not get perfect cancellation even in simulation (let alone the real world). There is also the limitation of the simulation -- you may have to tweak convergence parameters and limit the step size in transient simulations to improve simulation accuracy and precision (not the same thing, alas). You can tweak the DC point but will still have AC (transient) errors.

You'll note in Ethan's circuit that he places a ten-turn pot across a small resistor to make it easier to fine-tune the null and as @Blumlein 88 noted the active circuitry is after the null. All that has to be carefully tweaked as well, of course.

 

[pkane]

Like I said many times, his experiment is fundamentally flawed.

In below example, I change input of upper channel to 1.005V. And I can trim down to -100dB on the output.

View attachment 47272

I see you are coming up with yet another way to attack the null tester. What happened to all the other attack vectors? Let's recap, this is just from the last 24 hours:

1. FFT of white noise is white noise, therefore white noise has no phase
2. Noise doesn't have phase and therefore can't cancel out
3. Noise will never cancel out because errors only stack up
4. Only in-phase noise from source will cancel, out-of-phase noise from the source will not cancel
5. Quantum noise cannot be cancelled, uncertainty principle, etc.
6. Noise cancellation can only happen if the signal is captured, otherwise it can't cancel (probability wave collapse?)
7. Splitter introduces its own noise so the signal will not cancel
8. The VU meter has the wrong impedance for the circuit and so can't measure accurately
9. Your simulation shows that the circuit can never get to -110dB, even though Ethan measured the actual, real thing and it does

I'm sure I missed a few. What's next?

 

[Thomas savage]

I see you are coming up with yet another way to attack the null tester. What happened to all the other attack vectors? Let's recap, this is just from the last 24 hours:

1. FFT of white noise is white noise, therefore white noise has no phase
2. Noise doesn't have phase and therefore can't cancel out
3. Noise will never cancel out because errors only stack up
4. Only in-phase noise from source will cancel, out-of-phase noise from the source will not cancel
5. Quantum noise cannot be cancelled, uncertainty principle, etc.
6. Noise cancellation can only happen if the signal is captured, otherwise it can't cancel (probability wave collapse?)
7. Splitter introduces its own noise so the signal will not cancel
8. The VU meter has the wrong impedance for the circuit and so can't measure accurately
9. Your simulation shows that the circuit can never get to -110dB, even though Ethan measured the actual, real thing and it does

I'm sure I missed a few. What's next?

Ethan's null tester dose not cook fried chicken..

It's incomplete at best.

 

[Blumlein 88]

Ethan's null tester dose not cook fried chicken..

It's incomplete at best.

How are you going to null a chicken? It will have to involve a quantum singularity.

 

[Hugo9000]

How are you going to null a chicken? It will have to involve a quantum singularity.

Wouldn't eating the chicken nullify it? Whatever inedible/nasty bits are left over would be equivalent to noise. Maybe?

 

[solderdude]

Ah ... didn't see his latest ramblings and only reacted to his brilliant poll.

b.t.w. the circuit posted by @Blumlein 88 is only a small part of the actual circuit. It does not show the input circuit (which our good friend objects to) which is a buffer and phase rotation.

He can't have chicken and eat it too.

 

[solderdude]

Here comes a Spice simulation of Null tester with 2x 1V input. Output was a 50k ohms resistor connected to ground as load.

A: The output resistance of the 1V input is NOT 50k at all. You even missed the most basic part of electronics here.
The output resistance seen from the opamp input is 1.1k.
The load the 0 Ohm output resistance of the 1V source sees is 50k

Back to electronics 101 for you here.

Probing shows after the opamps, inversed channel introduced a 0.001V (or -60dB difference against 1V input) voltage difference in absolute terms with input. This voltage bumped out between OA2 and OA4.

Assuming perfect circuit, this is 0.1% difference in voltage against input in absolute terms.

So you are saying there is merely 60dB signal noise ratio left in the inversed channel ?
Question: If that were the case and this would be random noise opposite the non inverting channel and that would be amplified 80dB then the noise level shown by the meters would be +20dB ?
Why is this not happening ?
Is it possible you got this wrong somehow, or did you mean the amplitude is 0.1% off compared to the non inverting opamp.
Did you simulate for possible tolerance difference extremes as well ?
In such case would the 22 Ohm nulling pot not be able to compensate for a level difference of the source signal ?

Which is it ?

If you centered the trimmer, you will get a -64dBv signal against 1V input in this circuit, which is "error" due to Ethan's design.

As Spice points out, the Volt difference across load is V10+V12, not V10-V12. Physically, there is nothing being "subtracted" by this circuit. But this tester is capable of hiding a 0.001V signal difference.

Why would anyone center the trimmer. It cannot possibly work this way. He could have left the trimmer out if that is the point of the circuit as a trimmer pot most likely would have a physical error in the mechanical centering anyway.
NO you do not understand this well.
The whole object of the game is to set the pot so that the input signal in level is exactly the same in amplitude, but reverse phase, to that of the non inverting output when looked at from the center tap of the pot.
You do that by ear or meter. THAT determines the null.
The only difference remaining is the added noise + distortion of the input circuits + distortion of the DUT (tested cable)
When nulling cable A and B and the result is the same there is no difference between cable A and B (in the audible range).

True proponents of the nulling process would moan over the sound quality would degrade so much using opamps (as they all sound bad compared to the heaven sent discrete amps or tubes) and because of this all cables null the same because the sound of the cable is changed due to the opamps used.

Again, based on NwAvGuy's claim, 0.1% difference is distinguishable. This shows under perfect condition, null tester would not show a minor audible difference.

0.1% = 0.00868dB level difference. Where did NwAvGuy claim you could hear such differences ?
When you speak of 0.1% distortion = -60dB then in certain cases you can actually hear that depending on many factors.
You seem to say the pot cannot null perfectly despite the small adjustment range (fine control of the diff signal)

If you claim you can detect -110dB difference in both channels, you are simply wrong.

Nope because you can even hear below the noise floor so even if it were your -100dB (where did the -60dB go to ?) you can still null below your calculated -100dB.

And where did the Sansa go that seemed to be the biggest issue in your gripe ?