Octopus component tester weirdness!

[fubar.gr]

I am trying to model an octopus componet tester circuit in LTspice. Here's the circuit with a zener diode as a device under test:



The voltage source output is a 10 Volts peak sinewave at 50Hz. The simulation runs for 0.1 seconds so there are 5 full oscillations at that time.

Here is the voltage across the resistor versus the voltage across the diode. Looks like a typical I-V curve



However, if I zoom in near the breakdown region:



It seems that each time the voltage goes up and down, the current isn't exactly the same, hence the multiple lines.

And this also happens on my oscilloscope with my real octopus tester and it is even more pronounced!

What causes this? Is it some kind of thermal noise?

[c4757p]

In SPICE it's more quantization noise than anything. Numbers are getting rounded at different points each time. If you decrease the minimum timestep, it will become smoother. SPICE doesn't simulate any noise beyond the unintentional (except for the specific noise analysis feature.)

In real life, there could be many different contributors. Thermal noise is usually small, but it could be amplitude/phase/frequency noise on the input sine wave. Could also be noise from inside the scope itself (both thermal, and capacitive coupling from other circuits) if your signals are small. If the DUT is an avalanche diode (6.2V is right around the point where it could be either true Zener or avalanche) it could also be avalanche noise.

[krivx]

The capacitance of the diode should change with applied voltage, as the depletion width changes. I don't know if it this would enough capacitance to add some hysteresis to the plot (anyone?)

If you have some similar diodes but with different current ratings this could be a fun experiment. I would guess that beefier diodes will have higher capacitances and more hysteresis.

Do you see this at both the forward and reverse breakdown? I think the effects should be different, as different kinds of capacitance are dominant for forward and reverse bias

[T3sl4co1l]

If you reduce RELTOL and maximum time step, you'll see that clear right up to a nice crisp knee.

Which is not actually physically representative, as breakdown is a noisy process that SPICE transient simulation doesn't know about.

Zener and avalanche breakdown aren't the same thing, and both are relevant to an actually 6.2V device.  Higher breakdown voltages are avalanche dominant, and the knee is fairly sharp with a consistent positive tempco.  Lower voltages are softer, and the tempco varies (but is generally negative IIRC?).  SPICE may not know about this either (the diode BV, IBV parameters are well suited to modeling avalanche breakdown at DC, but not so much zener breakdown or AC noise effects).

I don't remember what noise effects occur in zener breakdown (which is a tunneling phenomenon, so I expect it at least has full shot noise, but beyond that, I don't know).  But avalanche is a very noisy phenomenon, because it's the solid state equivalent of a spark discharge.  When the electric field across a junction is high, thermally generated electron-hole pairs are accelerated across it.  If the field is strong enough that the charge carriers attain enough speed to break more electrons / holes free after a collision, then a runaway process can occur, where one charge carrier bounces off an atom or defect, freeing more charges, and so on, until a huge cone of charge is suddenly zipping across the junction.

Avalanche is a very fast phenomenon, under a nanosecond by the time it reaches the terminals.  Avalanche breakdown of certain BJTs (planar epitaxial I think?) can result in peak currents of several amperes, drawn through a nominally ~200mA transistor, turning on in less than a nanosecond (when the transistor normally takes tens of ns to do any sort of switching or amplification).  Essentially, the entire volume of the chip turns into a short circuit due to runaway avalanche.

Avalanche diodes are usually constructed in such a way as to minimize this runaway effect, but the random surge effect is present nonetheless.  It's particularly easy to see at low currents, where the voltage rises slowly, then jumps down suddenly by a random amount, which reduces the electric field enough to prevent further avalanches, then it starts rising again, and so on.

Tim

[fubar.gr]

Thanks guys!