If that's the case, then how could eddy currents in the core possibly result in additional noise seen by the windings?
I mean, what would be the source of such noise except the random motion of electrons within the core?
se
Re-read my previous post.
One cannot assume that the eddy currents are pure current...they are electrons moving within a resistive medium as a result of induced voltages. The currents are creating a field which bucks the driving flux.
If the eddy currents are noisy, as current within a resistance is prone to do, the bucking flux will be noisy.
Cheers, John
any distortion discussion needs tons of qualifiers, specifics of frequency, level as pointed out
a quick sim suggests to me that high input C jfet like Mr Curl’s low noise jfet front end is going to have higher levels of intermod at high frequencies due to MC cart R interacting with the nonlinear jfet junction C – much higher than the xfmr magnetizing inductance distortion shown in the plots - more than than the high Ni xfmr lf distortion raw number too at 1 mV levels
if xfmr lf distortion's mechanism is falling ~ 1st order with frequency, and jfet nonlinear input C distortion with resistive input is 1st order rising with frequency we already have the ingredients for a more nuanced argument about importance of IMD mechanisms vs frequency with music signal spectrum
but in addition you shouldn’t ignore the RIAA pre-emphasis of the signal at the preamp input – this makes rising distortion with frequency mechanisms (jfet nonlinear input C) much worse than the xfmr low frequency magnetic hysteresis distortion which falls with increasing frequency when compounded with the RIAA pre-emphasized input signal frequency dependence
pieter may be eable to clear up the detailed distinction between the low frequency magnitizing inductance nonlinearity disortion mechanisim which I presum is of concern as core B begins to satruate and proposed the low level permabilty "flat spot" around Zero that may have hysteresis formulation adressed in Menno VAN Der Veen paper (from which I quoted earlier:
“The results of this comparison are clear and totally as
expected; higher perm cores have enough perm at
threshold levels to create almost no weakening between
20Hz and 1kHz.”
)
a quick sim suggests to me that high input C jfet like Mr Curl’s low noise jfet front end is going to have higher levels of intermod at high frequencies due to MC cart R interacting with the nonlinear jfet junction C – much higher than the xfmr magnetizing inductance distortion shown in the plots - more than than the high Ni xfmr lf distortion raw number too at 1 mV levels
if xfmr lf distortion's mechanism is falling ~ 1st order with frequency, and jfet nonlinear input C distortion with resistive input is 1st order rising with frequency we already have the ingredients for a more nuanced argument about importance of IMD mechanisms vs frequency with music signal spectrum
but in addition you shouldn’t ignore the RIAA pre-emphasis of the signal at the preamp input – this makes rising distortion with frequency mechanisms (jfet nonlinear input C) much worse than the xfmr low frequency magnetic hysteresis distortion which falls with increasing frequency when compounded with the RIAA pre-emphasized input signal frequency dependence
pieter may be eable to clear up the detailed distinction between the low frequency magnitizing inductance nonlinearity disortion mechanisim which I presum is of concern as core B begins to satruate and proposed the low level permabilty "flat spot" around Zero that may have hysteresis formulation adressed in Menno VAN Der Veen paper (from which I quoted earlier:
“The results of this comparison are clear and totally as
expected; higher perm cores have enough perm at
threshold levels to create almost no weakening between
20Hz and 1kHz.”
)
Last edited:
Actually, they are generally chosen for their poor bulk conductivity..(edited for accuracy)
Hey, do not confuse the destination for the journey...
Our discussion has been pleasant...No, no, no..
Eddy dissipation will be inherently noisy because it relies on poor conductance for the supported eddy currents. Therefore, it will produce noisy bucking flux.
A heated core does not have eddy currents which can be noisy. There is no bucking flux that will be noisy.
You are confused, I'll repeat.
The heat is being dissipated within the series effective resistance. That is the resistance component of the Ls-Rs model that my HP meter finds.
The Rs increases when a dissipating media is introduced into the time varying flux. As you may recall, I lamented the fact that the meter was unable to distinguish between the actual winding resistance as a result of proximity effects and the dissipation as a result of the eddy losses.
For small signal transformers, there will be no proximity effect worth noting at audio frequencies, but the eddy problem is lamination thickness based.
Cheers, John
Last edited:
Parasound JC-3 phono preamp is often mentioned in this thread. Parasound detailed description gives a lot of details abaut preamp topology, but with different terms like " fet helpers" which are in fact IC CCS to get them in class A operation.
But one thing confuses me: using AD 797 as a first gain block for both MM and MC cartridges. Even the best B grade of this IC has input bias current between 25o to 900 nanovolts. Accepted maximum is known to be about 100 nanovolts. Too much for using preamp without input electrolytic.
Also, that IC does not like MM level resistence since input current noise is 2.0 pa which is five times higher than humble NE 5534 IC.Result is a few db higher noise with moving magnet.
All in all , Ad 797 is not optimum solution as a universal phono front end low noise gain block.
But one thing confuses me: using AD 797 as a first gain block for both MM and MC cartridges. Even the best B grade of this IC has input bias current between 25o to 900 nanovolts. Accepted maximum is known to be about 100 nanovolts. Too much for using preamp without input electrolytic.
Also, that IC does not like MM level resistence since input current noise is 2.0 pa which is five times higher than humble NE 5534 IC.Result is a few db higher noise with moving magnet.
All in all , Ad 797 is not optimum solution as a universal phono front end low noise gain block.
JCX, think SOURCE DRIVE IMPEDANCE! (very low) THINK CASCODE! (no Miller multiplication) Think about what an input transformer does to MAGNIFY the drive impedance. (increases it 64 times, in my case) Then you will understand the tradeoffs.
For everyone else, the KEY is RC. C is somewhat nonlinear and increases with LOWER NOISE jfets What to do? First, cascode. Second, use significant voltage across the input device. Third, scale jfet part to the expected source impedance. (in my case, 1-40 ohms. Fourth, for mm cartridges, BUFFER with a low cap jfet follower, to create low Z drive for main gain stage.
NOW, do you want to parse with me, about something else, JCX? ;-)
For everyone else, the KEY is RC. C is somewhat nonlinear and increases with LOWER NOISE jfets What to do? First, cascode. Second, use significant voltage across the input device. Third, scale jfet part to the expected source impedance. (in my case, 1-40 ohms. Fourth, for mm cartridges, BUFFER with a low cap jfet follower, to create low Z drive for main gain stage.
NOW, do you want to parse with me, about something else, JCX? ;-)
agree
yup
yes again
Here's where we disagree. When we force a current through a "good" resistor that current does not cause any extra noise except that which comes from heat (sqrt(4KT/R)). Scattering from defects, etc. causes excess noise with a 1/f spectrum, and yes sidebands on a sine wave excitation. Anyone can do the Wheatstone bridge experiment I published it in 1982 EDN(?) and Ed published the same here.
To quote the PhysRev paper, "Our problem then is simply this: What conditions are imposed by Maxwell's equations on the fluctuating currents in a piece of metal if the latter is in thermal equilibrium with the radiation field? IN EQULIBRIUM THE AVERAGE FLUX ABSORBED BY THE METAL MUST BE EQUAL TO THE AVERAGE EMITTED FLUX".
So for me this points to the same mechanism as in the resistor, how can can the conductor know the difference between induced and forced current? The carriers exchange energy with acoustic phonons (heat) and in equlibrium by the central limit theorem there is emitted flux with a normal distribution and standard deviation given by Nyquist's equation.
They go on to derive a low frequency solution for the noise (same as Nyquist) and the increase with frequency due to skin effect. I would bet (maybe just a coffee) that if you took their results and the physical properties of the laminations you could derive the improvement of thiner laminations solely from this modified Nyquist equation.
So yes JC, the physics agrees with the result.
JN, I think you postulated that there would be noise sidebands on a pure sine wave excitation. I think that is the best test.
We are not disagreeing..
I am talking about scattering defects causing excess noise in the eddy currents (bucking flux generator.) And, in a lousy resistor.
You deliver??
The conductor cannot. That is why I lamented the fact that my meter cannot distinguish the two. But the core's bucking flux noise, that the coil (hence terminals) will see. The meter interprets this as increased series resistance and lowered inductance.
That is why I proposed using the Ls-Rs function to measure the combined effect.
Uh oh, you too, eh??? What is the world coming too??
edit: Let's be a tad clearer...I postulated a physics reason to explain why eddy currents would cause noise as a result of hf excitation. I believe JC is the one who postulated the noise upon excitation, I merely extended the model to one which provides a methodology of testing the effect. You expound on the measurement technique..
I leave the actual measurement to those more qualified to do such.
Dat be you.... among others, dude...
As always, should you, or any of your DIY force be caught or killed, the moderator will disavow any knowledge of your actions...
Good luck Mr. Wurcer......
Cheers, John
Last edited:
I got a wrong information reading one of article which mentioned Ad 797. But second part of your answer shows arrogant attitude. You are describing us trivial circuits as "projects" as you did with two stage mc phono preamp, which is in fact datasheet level schematic. At the same time you are keeping even the slightest design secrets of your creations . It is not easy for professional
designer to be a member of DIY ( amateur) forum.
If you could harvest the energy you would us it for something so it would eventually end up back as heat!
Kamis, perhaps it is a language difference between us, but we, designers, must make tradeoffs when we make 'cost effective' designs. My boss told me 'NO INPUT JFETS' so I did the best that I could. I don't like bias current, either, but it is not so bad.
In other words, like so many here, you make much out of so little.
In other words, like so many here, you make much out of so little.
Yes. If you shove energy into a circuit then all the energy has to go somewhere. In electrical terms this has to be a resistance (even if the energy actually ends up as sound waves, radio waves etc.). So we agree that there is a resistance in the circuit which arises from the core losses. If core losses disappear then this extra resistance goes too. It is not winding resistance, but when you measure the winding resistance you get this amount added too.
What is the temperature of this extra resistance? Do you regard that as a meaningful question? If it is meaningful, and not zero, then does this extra resistance generate thermal noise?
John,
If the material had a "square" BH loop then the clipped distortion would be the well known = Sin(x) + 1/3Sin(3x) + 1/5Sin(5x) ....
As the material is not square a better approximation might be Sin(x) + 1/6Sin(3x)... So the old standby of 15% 3rd harmonic distortion or even 16 2/3% seems to be a suitable target for defining the threshold of saturation if your only measure is distortion.
Of course the V/I curve at the input is the classic standard.
ES
ES
While correct on it's own, I didn't understand it in the context of the discussion..
Yes. When measuring the resistance at any frequency, the core losses will show up as an effective series resistance.(note, while we speak of series resistance, it could indeed be a parallel model as well, Lp-Rp. The meter I use allows this choice based on the ratio of inductive reactance to resistance, and one chooses the model based on overall accuracy).
Yup, total agreement. Although it isn't possible to distinguish winding add on from eddy loss add on, I think that distinction is unimportant..it's the overall that affects the system. If one is trying to reduce the noise, then the distinction becomes important.
I would model it as just a resistance of higher value at room temperature. But given the discussion, that resistance is frequency dependent. Higher frequency, more noise... The fact that the winding current is constricting within the conductor does not mean it is higher temperature, just higher current density in a higher resistance.
Modelling this would not be trivial..Measuring may not..
Cheers, John
actually this is Exactly the sort of "parsing" people want to see from you as recognized expert in audio design
what are the numbers you're basing your estimates on - as an example "high output" MC can be >100 Ohm Rs
Cascode can avoid Miller multiplication - but how much distortion is still there from gm nonlinearity, is it higher than the xfmr's at mid/high freq?
if cart Rs may vary from 6-150 Ohms which tech will be a better choice "on average"
or how much are you giving up by using a design optimized with one extreme for the other
the xfmr output's higher Z may not be a problem if the higher signal level allows circuit techniques that improve linearity (including common mode input impedance nonlinearity) that would be too "costly" in terms of the extra parts noise contribution at the cart output level
those of us who want to learn the engineering principle’s application want more than obscure, oracular pronouncements that look like non sequiturs based on the few terse tech comments you issue – we would like to see how you fit the intellectual/engineering pieces together
- Status
- Not open for further replies.