John Curl's Blowtorch preamplifier part II

Status
Not open for further replies.
Emphasis was on the "Losses" to make clear that the linked document deals with Eddy Current Losses only, with no hint of inference to Eddy Current Noise.

Ok. I just didn't understand why you made mention of it again since there was never any disagreement regarding eddy current losses.

But reading your:
I was (and still I’m) under the impression that you were pulling a leg toward the skin effect and proximity effect (which are actually describing the action of eddy currents within the circuit’s current conductor own material).
If I failed to read correctly your message, my apologies, but please explain your message’s reasoning.

All I was saying was that the thermal noise of a simple length of wire should also exhibit a frequency dependent level due to eddy current losses, i.e. skin effect.

A simple length of wire such as the phono cable connecting the turntable to the phono stage.

se
 
Sooner or later, eddy current losses and their effects will be better understood by my 'critics' here. One only has to measure the Q of the coils with frequency to understand it almost completely, and of course, that is what I was specifically looking for in my quest in the JAES for a graph that was clearly defined and explicit.

More meaningful would be to actually quantify it in a moving coil stepup transformer instead of making blanket statements that any such transformer that's using 13 mil laminations is crap and not worth considering.

se
 
With regard to eddy current losses, one way to think of these is to add an extra winding to the transformer then put a resistor across the winding. This represents both a loss mechanism, and a source of thermal noise. The thermal noise is present whether there is a signal or not.

However, this might not be a perfect analogy as eddy current losses increase with frequency but a loaded secondary is flat. Perhaps a better analogy would be a loosely coupled secondary, so that the induced voltage rises with frequency but there is only a weak back reaction onto the primary.

Eddy current losses are losses generated inside the ferrous core due to induced circulating currents. They are in a direction transverse to the magnetic field. This is why cores are laminated and insulated from each other electrically. The thinner the laminations, the less the losses for a given alloy.

Hysteresis losses are losses in the core due to the inefficiency of the coercive field to produce a corresponding induced field. At first application the curve is extremely non linear much like one of the family of characteristic curves of a triode. It curves slowly up from zero and then rises steeply and linearly. Once the magnetic domains are in motion, the retrace is usually linear but the curves in opposite directions don't coincide. The area between them is a measure of hysteresis loss. Short fat curves that are widely separated are characteristic of material that is poor for most transformers and inductors (saturable core reactors used in some circuits may be an exception.) Tall nearly vertical curves that are closely spaced such as for supermalloy are best.

Hysteresis and eddy current losses can be eliminated completely by building air core transformers. However, the lack of a core to concentrate magnetic flux may reduce the overall efficiency of coupling the primary and secondary windings. This might still be the best option...if you have to use a signal transformer. It seems to me that with low noise transistors, you have a far better choice.
 

Attachments

  • connie.jpg
    connie.jpg
    38.1 KB · Views: 193
Member
Joined 2002
Paid Member
Originally posted by jneutron
Eddy currents do not couple back in this direct fashion, so I would not use this method. (good thinking however, I also considered this test but after thinking a while, understood the limitations.)

What is needed to evaluate this is:
1. Wind 4 C cores with pickup windings. One core solid iron, one 1 mm laminations, one 1 mil laminations, and one comprised of ferrite.
2. Excite the core using a rotating neodymium magnet. The drive must be low noise.
3. Examine the output waveforms for indications of noise amplitude consistent with frequency.

John, may I propose an easier to implement set-up for testing the eddy current induced signal noise idea?
Take 3 small 1-2 Watt cheap power E-I transformers, all of the same type, from the same manufacturer.
Dismantle them, one at a time.
Leave the laminations of one as is(actually, no need to dismantle this one).
Remove the electrical insulation from half the laminations of the second.
Remove the electrical insulation from all the laminations of the third.
Clean the laminations.
Reinstall the coil bobbins and reassemble the transformers.

Apply a signal to the primary (10KHz-20KHz ?) and record (the properly terminated) secondary voltage from each transformer. Better still to simultaneously record the primary voltage as well.

Examine the output waveforms for indications of noise amplitude consistent with frequency.

This test can be accomplished with only one transformer. Three measurements again, but the first with all laminations as is, the second with half the laminations stripped from insulation and the third with all laminations stripped.

(Edit:Regardless if one or three transformers are used,) on the three measurements, core eddy current losses will be different. Core hysteresis losses will be the same. Wire losses of any kind will be the same.

Regards
George
 
Last edited:
It is not noise caused by the presence of eddy currents, but noise caused by the possibility of eddy currents. The simple fact that there is an energy dissipation mechanism present is sufficient to create thermal noise from that mechanism. The mechanism does not have to be dissipating energy from a signal. The mechanism just has to exist. That is all. Just like a resistor; its thermal noise does not depend on the voltage across it, just its temperature.

The "possibility"? Shirley you jest...next you'll be tossing a cat in a box..:p

Let's extend your argument, shall we?

Take the 4 cores I mentioned earlier. Put all 4 in an isothermal, zero field box, with only the wires exiting the box. Your statement ""The simple fact that there is an energy dissipation mechanism present is sufficient to create thermal noise from that mechanism means that I should be able to distinguish the core design by examination of the terminal voltage despite the fact that there is no external excitation.

Therefore, the one with the higher terminal voltage is the one that has the possibility of eddy currents.

While not exactly Maxwell's demon, you come close. You postulate that despite the total lack of a magnetic field to couple the coil to the core, it knows nonetheless..

I'm not buying it..

Current does not produce shot noise. It is uncorrelated discrete charge carriers which create shot noise. Almost all currents use highly correlated charge carriers, so no shot noise.

Hmm...for some reasons, my textbooks do not agree with your statements.

Please explain the difference between correlated discrete charge carriers and uncorrelated discrete charge carriers...in a metal.

The exception is when charge carriers have to cross a potential barrier, which shields them from each other and reduces correlation.

Are you talking conduction and valence bands?

As Q is a measure of dissipative loss, it will correlate with thermal noise.
Yup...I did say that earlier as well.
If John Curl measures lower Q and hears more noise, then that suggests that the noise is thermal in origin. If he also finds that thinner laminations give higher Q and less noise, then it is quite likely that the mechanism is eddy current losses.
Said that as well...so I guess we agree?

Carver Mead argues to the equivalence of the two (just to add confusion). :)
This is a good read.

http://www.rle.mit.edu/avbs/publications/journal_papers/journal_16.pdf

Nice read. Thanks for the link.
Funny, yesterday I was thinking of the old slide the supermagnet down the copper pipe experiment but you could never get the stiction noise low enough. I would try dropping the magnet in a vaccum.

You are very close!!! (ran for lunch, had time to think)..

Take a C shaped laminated core which has two neo magnets and a 1/2 inch wide gap such that the flux crosses that gap.

Wind a pickup coil around the back leg of the C such that any variation in the permeability of the gap will produce a voltage on the coil.

In steady state, the coil voltage will be only thermal noise..

Now, build a long arm with a copper blade much like a sythe, pivoted such that only the copper blade passes through the flux gap.

Any additional noise pickup by the coil is a direct result of the eddy current loss mechanism of the copper.

Easy as pie..where's my invite to Sweden...

Cheers, John
 
John, may I propose an easier to implement set-up for testing the eddy current induced signal noise idea?George

Not bad, I like the way you think.

Practical problems:

Cleaning the glue off the laminations.

Packing density.

Noise from the current source.

Proximity effect based noise in the wires (the increase of Rs with frequency.)

I was thinking along the v drive line as well, but decided I could remove the current induced noise of the primary by using a rotating magnet. Yes, I would then have to contend with the bearing noise of a pair of ABEC 9 unsealed bearings in a vacuum connected to a flywheel. But what the heck, this is all thought experiments anyway...

Cheers, John
 
Take the 4 cores I mentioned earlier. Put all 4 in an isothermal, zero field box, with only the wires exiting the box. Your statement ""The simple fact that there is an energy dissipation mechanism present is sufficient to create thermal noise from that mechanism means that I should be able to distinguish the core design by examination of the terminal voltage despite the fact that there is no external excitation.

Well, don't the coils have Johnson noise? So that, except at 0K, the core is getting excited.

edit: To be clearer, the fact that the box is isothermal means that there has to be energy removed from it by external means.
 
John, may I propose an easier to implement set-up for testing the eddy current induced signal noise idea?

I was thinking of something similar.

If what jcx says is true, i.e. that that noise would be independent of signal level and would be manifest even in the absence of a signal, then to quantify noise due to eddy current losses in the core it should be a simple matter of measuring the high frequency noise in an existing transformer and compare that to the calculated noise based on simple winding resistance.

se
 
Well, don't the coils have Johnson noise? So that, except at 0K, the core is getting excited.
his premise is that there will be coupling to the coil despite zerofield, and that it's possible to distinguish cores based on the possibility...
edit: To be clearer, the fact that the box is isothermal means that there has to be energy removed from it by external means.
Hence my goin ta sweden. Always wanted to meet King Gustav..:D

I was thinking of something similar.

If what jcx says is true, i.e. that that noise would be independent of signal level and would be manifest even in the absence of a signal, then to quantify noise due to eddy current losses in the core it should be a simple matter of measuring the high frequency noise in an existing transformer and compare that to the calculated noise based on simple winding resistance.

se

Exactly...and it would mean that swapping core types will show a difference.

Yet, there is no coupling mechanism between the two if there is no current..

Hence the conundrum..

Cheers, John
 
the wire "skin effect noise" example is a good use of "reductio ad absurdum" - with the twist that the physical principle is "sound" but the absurd conclusion is that it is large enough to have an effect on sound

I suspect the same is true in MC xfmr


- just a thought but maybe cable lifters work by keeping the wires at more nearly the same gravitational potential - after all electrons do have mass

anyone want to trademark "frame acceleration compensated" cables? - apparently it is always good to reference Einstein in your marketing
 
Last edited:
Member
Joined 2002
Paid Member
Originally posted by jneutron

Practical problems:

Cleaning the glue off the laminations.

Packing density.

Noise from the current source.

Proximity effect based noise in the wires (the increase of Rs with frequency.)


Cleaning the glue off the laminations: It can be managed.

Packing density: Correct. The stripped-off insulation has a thickness that can be made-up with paper or plastic inserted between bobbin and core(I don’t think it will be more than 5-6 mills).

Proximity effect based noise in the wires (the increase of Rs with frequency.) and Noise from the current source. : Both will be the same on the 3 measurements.

They are meant to be comparative measurements, not absolute ones.
And it was proposed as a doable, not as a thought experiment:)

Regards
George
 
his premise is that there will be coupling to the coil despite zerofield, and that it's possible to distinguish cores based on the possibility...

Hence my goin ta sweden. Always wanted to meet King Gustav..:D



Exactly...and it would mean that swapping core types will show a difference.

Yet, there is no coupling mechanism between the two if there is no current..

Hence the conundrum..

Cheers, John

but we're pretty sure there is thermal noise current in the core material - some components of it will have "the right" physical/spatial structure to couple to the winding

again the simplest version of the question is can you distinguish by noise measurement at the primary different resistances connected to a xfmr secondary?
 
the wire "skin effect noise" example is a good use of "reductio ad absurdum" - with the twist that the physical principle is "sound" but the absurd conclusion is that it is large enough to have an effect on sound
I have coils on my desk which are made using 700 or so turns of 15 guage solid copper, and some made using equivalent guage litz. Measurement of both from 20 hz to 50 Khz clearly shows the equivalent series resistance of the solid goes nuts, exceeding 300 to 400 ohms at 20Khz.

Am I being told that both will have the exact same noise should I push 20 Khz 1 ampere into them?

- just a thought but maybe cable lifters work by keeping the wires at more nearly the same gravitational potential - after all electrons do have mass

Ah, but when they flow downhill they pick up speed therefore mass, and uphill, the opposite, so net neutral..:eek:

Cleaning the glue off the laminations: It can be managed.
Packing density: Correct. The stripped-off insulation has a thickness that can be made-up with paper or plastic inserted between bobbin and core(I don’t think it will be more than 5-6 mills).

The lams have to be electrically connected here to change the eddies..we'll gain packing factor if they touch, but they must..
Proximity effect based noise in the wires (the increase of Rs with frequency.) and Noise from the current source. : Both will be the same on the 3 measurements.

True if reluctance is exact. Eddies fight flux penetration, so one would expect a variation in penetration therefore proximity effect...given the level one wishes to measure, I'm not sure how close the cores must be to eliminate their influence.
They are meant to be comparative measurements, not absolute ones.
And it was proposed as a doable, not as a thought experiment:)

Regards
George
Doing it requires someone with equipment, time, and something to gain by the measurements. Maybe a xfmr manu somewhere? I point out that I will most certainly provide whatever assistance I can should anybody care to setup these experiments.

There are noise currents in the windings.

se
Yup..

but we're pretty sure there is thermal noise current in the core material - some components of it will have "the right" physical/spatial structure to couple to the winding

Actually, the entire core structure will indeed be setup such that it will couple to the windings. Buuuuuut....

Thermal molecular movement will be totally random in direction. There will be NO net carrier movement, therefore there will be no net fields generated at the macro level.


again the simplest version of the question is can you distinguish by noise measurement at the primary different resistances connected to a xfmr secondary?
That was not the question I was discussing. I was discussing the distinguishing of the core via the primary (with no secondary load), but simply because the eddy losses will generate noise.

Cheers, John
 
Last edited:
To complete my previous thought (the coil inducing current in a plate) the only added noise is due to the increase in the plates temperature.This mechanism is present at equilibrium with or without signal. Take my sub-sqrt(2qI) noise maker, a noiseless 1000V battery connected to a 1M resistor as an example. The only increase in noise is due to heating of the resistor (sqrt(4KT/R)).

Remember excess noise would be present in a plate full of defects, etc. and is different.

I rest my case awaiting the experiments. :D
 
I have coils on my desk which are made using 700 or so turns of 15 guage solid copper, and some made using equivalent guage litz. Measurement of both from 20 hz to 50 Khz clearly shows the equivalent series resistance of the solid goes nuts, exceeding 300 to 400 ohms at 20Khz.

Am I being told that both will have the exact same noise should I push 20 Khz 1 ampere into them?

Shouldn't you have the same absolute amount of noise whether at 20 Hz or 20 kHz?

I mean, the only source of noise in the wire is the thermal energy in the wire, yes?

So if we leave out joule heating for the moment, then you should have the same amount of noise at 20 Hz as 20 kHz, yes?

Now, if the resistance of the wire goes up with frequency due to skin effect, then you'll have some attenuation of the signal because of it and therefore some reduction in signal to noise ratio, but the noise itself should remain the same, correct?

se
 
Shouldn't you have the same absolute amount of noise whether at 20 Hz or 20 kHz?

I mean, the only source of noise in the wire is the thermal energy in the wire, yes?

se

My first thought would be no. The shorted coil would exhibit a current noise that would have to obey the same physics. The open circuit voltage noise would be due to the bulk R (there is no current i. e. no skin). Experiment No. 3 :0
 
Last edited:
To complete my previous thought (the coil inducing current in a plate) the only added noise is due to the increase in the plates temperature.This mechanism is present at equilibrium with or without signal. Take my sub-sqrt(2qI) noise maker, a noiseless 1000V battery connected to a 1M resistor as an example. The only increase in noise is due to heating of the resistor (sqrt(4KT/R)).

Remember excess noise would be present in a plate full of defects, etc. and is different.

I rest my case awaiting the experiments. :D

Makes sense to me.

And I've been thinking about John's Q measurements.

John is assuming that Q dropping with frequency is due to some increase in resistance and therefore there must be a commensurate increase in noise.

But eddy currents in the core cause the core's effective permeability to drop.

All else being equal, as permeability goes down, so does inductance.

Would this not also result in a lower Q with increasing frequency?

se
 
Status
Not open for further replies.