Robin van Spaandonk wrote:
A diode is not of course a very good switch and has a gently
changing V/I slope (ie impedance) near zero bias.
Which is precisely why you put the transformer in between. That
shifts the voltage up the curve of the diode away from the zero
bias point.
Bear in mind that we are talking of AC (noise) voltages so one
cannot expect to work with any DC bias. Also a diode has an
exponential V/I relationship (Shockley equation) and so with
appropriate scaling I believe it can be considered to work
just as well in a high impedance circuit with zero bias as it
does in a low impedance circuit when biased to the 0.65 volt
so-called "knee" (which value is entirely dependent on the
scale on which you choose to plot the exponential - scale the
axes in microamps and millivolts instead of volts and milliamps
and the "knee" moves down to zero).
However you would need an incredible transformer ratio,
and the resulting minute current on the diode side may be
"lost" in the noise of the diode. This depends somewhat
on whether or not these purported signals ...
There is no "purported"-ness about these signals. It is a
standard experiment performed by 3rd year physics students
to measure this noise voltage and from it determine absolute
temperatures (to ~4 digits with ~hours of integration), or
knowing a single temperature, to determine Boltzmann's
constant from the noise voltage.
... from the resistor can be "ganged"
together. Since they would have random phase relative to one
another, they would likely at least on occasion enhance one
another leading to a "spike" that might be transformed and
rectified.
My point was that a transformer provides nothing that
simply choosing a different valued resistor would provide.
A high value resistor gives high voltage with low current
but still only 4kT watts per Hz of bandwidth.
Similarly ganging resistors together provides nothing
different from what a single resistor would with the same
value as the ganged set. (There is a small difference -
and that is how well the resistor is heat sunk or connected
to the heat bath - but for the power flows under discussion
better connection to the heat source/sink is hardly an issue!)
Thus it must also generate Johnson noise by the same
mechanism (whenever there is a path for electrical power to
be dissipated as heat, then there is the reverse path in
which the heat bath can generate electrical power - this is
called the "fluctuation dissipation theorem" in physics).
Presumably this noise power source/sink will vary slightly
in impedance with the voltage/current fluctuations
The transformer "transforms" the impedances, so that there
is a deliberate mismatch between resistor and diode.
I think you missed my meaning - the exponential V/I
relationship (Shockley equation) of the diode means that
it will behave just like a resistor who's resistance (or
impedance) varies (only minutely with thermal level I&V)
as the voltage or current in it varies. This is after
all what provides the rectification effect - current in the
forward direction sees the diode as a much lower valued
resistor than current in the reverse direction. It is just
possible that this effect could produce some net cohering
of the statistical fluctuations. But like I said, I doubt
if nature would make it that easy to beat its second law!
- but I am sure nature will have organised
it such that no configuration you can dream up will
allow net power to be generated from thermal energy!
A solar cell already does this, it just operates at a
higher "ambient" temperature.
A steam engine also works well when you have a significant
temperature difference - such as that between the surface
of the sun and the ambient on earth. But beating the 2nd
law requires that it work without a temperature difference
- ie turn random thermal energy into ordered electrical
energy which can then be used to say heat an isolated
resistor above ambient while slightly cooling the ambient
heat bath in the process.
Its built in diode, acts like a 0 K heat sink.
More like a 300 K heat sink!
