Use of Diodes in Filters??

[Magnus Danielson]

>>>>> "m" == media  <media at mail1.nai.net> writes:

 m> First, I'd like to again thank everyone who has mailed me sources of VCF
 m> schematics.

 m> I apologize if this seems an uneducated inquiry, but I am curious as to the
 m> use of diodes in several filter schematics I've looked at.  In all cases
 m> they seem to be common silicon signal diodes.

Things like the all popular 1N4148 or possible 1N914 ;)

 m> In several designs there seem to be three or more diodes in series.  What
 m> different function does that serve than that of a single diode??

 m> On the MS-50 schematic there seem to be six, three in series in each
 m> direction, running parallel to the resistor in the negative feedback path
 m> of an op-amp.  Even more curious, there is what appears to be a bridge
 m> rectifier in the middle of the circuit!!

 m> On the EMS filter also shows several diodes in series.  There are also two
 m> rows of diodes with capacitors running across them.  The Mini-Moog VCF
 m> schematic shows a similar structure using transistors in place of diodes.
 m> I presume these are the famous "diode ladders" -- how do these function??

 m> I hope these questions aren't too tiresome.  I tried looking into several
 m> books, finding numerous non-resonant low-pass filter designs which make
 m> perfect sense, but I am yet to find a description of these uses for diodes.

OK, back to basic... better do a crash-course just in case..
that a diode is rectifying is kind of well known
these days... but they fortunatly do more!

In a PN-junction (such as a diode) will the behaviour be something
like

            qV
            --
            kT
 I = Is * (e   - 1)

(doesn't ASCII formulas look great/ugly??)

Where
 I is the current througth the diode
 Is is the backdirection leakage (usually at about 10 fA or so)
 q is the unity charge, the charge of an electron
 V is the voltage over the diode
 k is the Boltzmann constant (about 1.38 * 10^-23)
 T is the PN junction temperature in Kelvin

So, q and k is physical constants (you can't fiddle with them).
e is a standard mathematical constant (you can't fiddle with that
either... it's about 2,718281828... and is really more basic than the
more known pi).
The temperature T will usually be something like a warm and health 300 K
(or at least I would like it to be more than the 276 K we have here in
Stockholm rigth now... brr...).
The Is value is a property of the diode you happend to find in the
back of your closet, so you can't fiddle with that either, but it is
almost of no importance since it is so small!
So, you apply a voltage V and you get a current I and there is an
exponential behaviour between them.

So, if you look at the small-signal properties around a certain base
voltage you will get a certain voltage to current conversion factor
gm, this is called transconductance.
For small signals you will get this conversion factor, but for larger
signals will the excess voltage cause so large changes such that the
transconductance will really be diffrent, it will modulate itself.
This is a non-linear behaviour, but the diode is a non-linear
component and the transconductance factor was just a linear
approximation at a certain point.
The smart thing is that you can move the large offset to change the
transconductance factor and all this is similar as changing the
resistance of the diode. So, the diode is used as a electrically
tunable resistor. But it is non-linear for large signals.

Also, you usually don't get more than 0.75 V over most diodes, it will
just be a lot of current going thougth it... this put a upper limit on
the voltage applied and this results in a need to keep signals down
before you clip.

One trick one may play with the diode is to stack multiple of thems
such that you migth get better signal to noise ratio (less noise that
is). An extreme case of this is found in the EMS Hi-Fli where the
phaser section contains a stack of 20 diodes per AP filter, there is a
total of 12 in the Hi-Fli which results in just 240 diodes for the
core phase block itself.

Have you got any wiser or was it just a waste of your time and
bandwidth?

Cheers,
Magnus