Winfield said:
A complete and accurate analysis will have to include not only
the transformer's leakage inductance but the diode's reverse-
recovery time model. You have some more learning ahead of you.
Back to your snubber, this is to damp the inductive pulse and
high-frequency ringing that occurs from the rectifier diode's
fast snap-off after the reverse-recovery time is finished. In
this case it's the transformer's leakage inductance that was
charged during the reverse-recovery time and forms a resonant
circuit with the winding and diode capacitances. You add more
capacitance in parallel to lower the ringing frequency and take
control of the resonant energy, plus a series resistor to lower
the Q and damp the ringing to a single cycle. (We've written
up a nice story about this scene, complete with scope photos
and spice models for the 3rd edition of AoE.)
Anyway, instead of a very sharp high-voltage spike (too short
to see with a scope unless you trigger properly) followed by
RF ringing, you get a single small long smeared-out innocuous
pulse. If left unattended to, the magnetic radiation from the
inductive spikes can be picked up and create an audible buzzing
signal (harmonics of 120Hz) in low-level phono or mic inputs.
Well, I put my measured values in:
for my small 50VA toroid with 17% loss 220/18V I got from primary to
secondary 25uH leakage inductance and 240pF interwinding capacity with both
secondaries in parallel. From one secondary to the other it was 6uH and
3500pF, seem to be wound bifilar. I have not corrected the reading for the
DC-resistance(0.7R each secondary, 36R primary), so the actual values should
be smaller.
IMHO a high leakage induction is bad. The pulse is not spread out, but the
peak is delayed and the pulse has a much longer risetime, whereas the
falltime gets shortened. This *creates* the problem with the spike. The peak
current and rms value stay almost identical.
Since the charge stored in the diodes is carrier_lifetime * current it will
be better to take a faster diode. There is *no spike at all*, when the
carriers recombine before the voltage gets reversed.
One big mistake would be to put a capacitor across the secondary, or
capacities to gnd from there. now the current which is stored in the leakage
induction will oscillate nicely through this cap instead of the diffusion
capacity of 2*2 diodes in series. People think it is good to put 10nF across
the diodes, but this is almost as bad. Sure the frequency gets lower, but
the damping not, unless you also use a resistor in series with C.
Anyway, to spread out the pulse you will need a resistor in series. To
dampen the ringing, a resistor across the secondary which is big enough not
to dissipate too much will serve better against the "high voltage spike".
Also the current of this ringing is really low, so I doubt it will have too
much effect. The higher the frequency, the more damping because of skin
effect etc.
