Ciao,
vi posto una mail che e' comparsa per brevissimo tempo sul forum di MEG-Builders di Yahoo.Non so per quale motivo e' stata rimossaer corretezza non rivelero il nome della persona che l'ha postata.

Detto questo vi conviene leggerla poiche finalmente da' una risposta alle domande sul MEG:


>Ever since I first ran across J. L. Naudin's Web site, I've been
>looking at his data as a round of the fun "find the other guy's stupid
>mistake" game.
>
>It didn't take me long to recognize that his output voltage and
>current figures disagree with the stated load resistance by an order
>of magnitude or more, or that pumping 28W into a 1W carbon comp
>resistor should produce smoke and flames which are notable by their
>absence.
>
>The data provided don't make it possible to determine whether these
>discrepancies are the result of dumbnitude or deception. End of story
>there, but from an electronics geekery POV there were still some
>apparent anomalies:
>
>The drive circuit is a garden-variety switching inverter, which should
>produce a square-wave output. The output voltage and current, and the
>primary current, are sinusoidal. WTF?
>
>The output voltage is nearly twice as high as would be predicted from
>the voltage swings available in a center-tapped push-pull drive
>circuit and the transformer turns ratio. WTF?
>
>Naudin reports that removing the magnet reduces the output
>voltage. Anyone who has ever worked with transformers knows that DC
>flux in the core contributes nothing to the output and is generally
>considered a Bad Thing.
>
>WTF?
>
>I think I've got it sussed. The key to understanding it is to replace
>the transformer in the circuit diagram with an equivalent circuit of a
>real-world transformer, you know, the one with an "ideal transformer"
>and the parasitic reactances and resistances reflected to one side of
>the transformer.
>
>Basically, the load reflected to the primary consists mainly of the
>transformer's leakage inductance in series with the parallel
>combination of the secondary's distributed capacitance and the load
>resistor.
>
>This circuit will exhibit series resonance, that is, the impedance
>magnitude will pass through a minimum at a frequency determined by the
>LC product (and detuned slightly by the R in parallel with the
>C). Below resonance, the impedance will be capacitive, at resonance
>resistive and above resonance inductive.
>
>That explains the sinusoidal primary cirrent. That's exactly how a
>series tank will behave if you whang a step voltage change across it.
>
>From the point of view of the load, we can replace the ideal
>transformer of the equivalent circuit with a voltage source reflected
>from the primary drive circuit. This generator drives the load
>resistor via an L-section consisting of the leakage inductance in
>series and the distributed capacitance in shunt- in other words, a
>second-order LCR low-pass filter.
>
>Now, the voltage transfer function of such a filter being wn^2/(s^2 +
>d*wn*s + wn^2), where wn= filter natural frequency, d= filter damping
>factor and s= complex frequency, when s=jwn (the condition for drive
>at the filter's natural frequency) the transfer function magnitude =
>1/d. If d<1 then the transfer function magnitude is >1; as d is
>reduced the filter becomes more and more "peaky". d is inversely
>proportional to the load resistance; the lighter the load the peakier
>the response.
>
>That explains the voltage boost. The output circuit of the MEG is a
>seriously underdamped low-pass filter driven at its natural frequency,
>that is, right at the response peak. The attenuation of harmonics by
>the filtering action explains the fairly clean-looking sine wave seen
>at the load.
>
>This also explains why the thing has to be wound for such a high
>output voltage. Increasing the number of turns in the secondary
>increases the leakage inductance seen looking into the secondary and
>also increases the secondary's distributed capacitance. Since the
>natural frequency of the network formed by thee parasitics is
>inversely proportional to sqrt(LC), adding turns reduces the
>transformer's self-resonant frequency, until it eventually coincides
>with the drive circuit switching frequency.
>
>A quick estimate based on a 25 kHz operating frequency, a 100k
>resistive load and a d of 0.5 (about right to account for the output
>voltage boost) gives me Lleak=318 mH and Cstray=127 pF. (roughly- I'm
>not accounting for the detuning effect).
>
>Are these realistic values? A bit over 100pF doesn't seem unreasonable
>for a 1500 turn secondary, which must include several feet of wire. As
>for the leakage inductance, a secondary leakage of 318 mH seen looking
>into a winding where the magnetizing inductance is around 5.7H is
>rather high- that would be a pretty s****y transformer by normal
>design standards.
>
>Naudin's transformer design is a veritable checklist of Things Not To
>Do if you want to minimize leakage inductance:
>
>1. Wind the coils short and fat so that as many turns as possible are
>spaced as far as possible from the core.
>
>2. Wind the coils on separate bobbins and mount them separately on the
>core (About the only time you see this in a practical transformer is
>when primary-to-secondary isolation is a paramount concern; usually
>one winding will be wound over the other or the two will be
>interleaved or wound bifilar to minimize leakage).
>
>3. Stick a magnetic shunt in the core to divert some of the flux
>induced by the primary and prevent it from linking the secondary
>winding.
>
>That brings us to the role of the magnet. Entirely apart from
>producing a DC flux in the core, it's acting as a chunk of
>magnetically permeable material which "short-circuits" some of the
>flux induced by the primary. That will act to increase the
>transformer's leakage inductance. The DC flux might also play a role-
>if the core is biased near saturation its incremental permeability
>will be reduced, which will reduce the magnetizing inductance and
>increase the leakage inductance.
>
>That explains why removing the magnet reduces the output
>voltage. Without the magnetic shunt in place, the leakage inductance
>decreases, increasing the natural frequency of the network formed by
>the parasitic reactances, so that the magnitude peak is no longer at
>the drive frequency. In other words, it throws the filter out of
>alignment.
>
>The experiment of removing the magnet is a great example of failing to
>isolate variables. Taking the magnet away or installing it
>simultaneously changes two variables: it alters the magnetic circuit
>of the core and it adds or removes the DC flux. Assuming that a change
>in circuit operation is entirely due to the latter is unjustified; the
>correct way to test the effect of the DC flux would be to remove the
>magnet and replace it with a nonmagnetized but otherwise identical
>piece of identical material, or to replace the permanent magnet with
>an electromagnet and compare circuit operation with the electromagnet
>energized or not energized.
>
>Basically, it looks to me like Naudin has stumbled on a variant of the
>series resonant converter and doesn't recognize it because the
>components forming the resonant circuit are the invisible parasitic
>reactances of the transformer, and because he hasn't enough experience
>with transformers to remember that those invisible components are
>there.
>
>A few years working in pro audio, contending with output transformers
>that show ultrasonic peaks and ringing without a terminating
>resistance connected to the secondary and with passive filter sets
>designed back when everything had tubes and 600 ohm source and input
>impedances which show wildly peaked responses when operated under the
>very low source z and high input z conditions found in modern audio
>gear should cure that aspect of Naudin's ignorance. Perhaps.
>
>Anyway, if you want, give me a few days and I can whomp up a couple of
>equivalent sche's (easy) and try to work as much of the math as I can
>(yuck). I hope that this basic qualitative description might be
>enough, and that you have time to let me know what you think.


Purtroppo il mio inglese non e' buono e quindi non sono riuscito a comprendere tutto ,ma la buona parte di quello che ho capito mi ha fatto cadere gli ultimi dubbi che avevo sul MEG di Naudin.
Ciao