news:595eda2d-1281-4bf2...@d20g2000vbh.googlegroups.com...
> Thanks Tim, What's different about the above LED (and also low power
> ~20V zeners) is that it looks like the whole device discharges at
> once. You get a big spike of charge, then the whole thing recharges
> for another blast.
Not unusual for some regimes of avalanche -- 2N3904s reliably go from
reasonable open circuits (<mA leakage) to about 10 ohms in the span of <2ns
(probably under 1ns, but I only have a 200MHz scope to see it).
Though the 2N3904 is rated to 60V or so, actual breakdown is in the 80-110V
range. To achieve reliable avalanche, connect a 2.2k resistor B-E, and
apply voltage C-E through a low current source (usually 100VDC in series
with 10k ohms). Parallel capacitance (C-E) of a few pF helps generate a
visibly sized pulse.
Zener diodes, and intentional avalanche structures in general, are most
likely designed with load balancing in mind. This means either minimizing
impurities to an unusual degree, or ensuring their homogeneity throughout
the junction. Consider a garden variety zener diode: at low currents,
discrete avalanche is visible as short falling edges amid waves of rising
sawteeth (bias charging junction capacitance), but the important
characteristic is the falling edges are very short, fractional volts, and
the amplitude drops with increasing bias (basically mean value theorem in
action, noise is proportional to 1/sqrt(current)).
Surprisingly, to date one of the most accurate, low noise voltage references
remains the buried zener structure. By burying a junction inside the
silicon, surface states and dislocations are avoided. I forget if epitaxy
is used, or if it's formed entirely by diffusion; I seem to recall a few
etch and grow steps are required. This is important because epitaxy
deposits silicon randomly, leading to more dislocations, and can't be
annealed at the same temperature as the whole wafer to fix them (obviously,
that would ruin the doping underneath!). (It's my understanding, the
development of high speed power diodes was made possible by epitaxy: whereas
1N4001 is a simple diffused junction diode, the dislocations of an epitaxial
layer provide more recombination sites, reducing reverse recovery time -- at
the expense of greater voltage drop and junction resistance, I suppose due
to differences in doping profile (longer junction, higher built-in
potential?), necessitated by the lower critical field strength. Ultrafast
diodes even have additional dislocations induced by electron bombardment or
transition metal doping to achieve the best non-schottky speeds.)
> "(they only spec them to 5V!), " Yeah that's just strange. Most of
> the LED's I tried I couldn't see any leakage (at the ~10nA level) for
> voltages up to 50V.
How big are the actual spikes, and what fall time? You said you measured
through an amplifier, so perhaps the amplitude isn't much to begin with?
The risetime on the scope shots wasn't very impressive, likely due to slew
rate limiting. The actual performance of these phenomena can be very fast
indeed! (The next fastest phenomena are step recovery in specially doped
diodes, ranging from 300 down to 20ps or so, and all-out monolithic "shock
line" generators, usually implemented in long InP schottky junctions I think
($$$!). Anything less than that is optical domain -- femtosecond lasers and
such.)