Diode reverse recovery time and the speed (snap back) in which recovery takes place

Question

I'm looking at generating a circa 1 nano second current pulse by utilizing the snap-back time (usually measured in the tens of pico seconds) when a diode is "forcibly" reverse polarized after being forward biased for a short time. This answer from Neil gives me some confidence in what I believed to be true i.e. a common or garden sloppy old diode like the 1N400x series (or the 1N540x as per Neil's suggestion in the link above) can achieve this.

For instance, in simulation, I apply a 1 MHz sinewave biased up at 10 volts (green waveform) to a diode whose anode is at 3 volts. The sinewave voltage is fed via a few ohms for current limiting: -

The blue waveform is the diode voltage and, you can see (highlighted in orange), it takes a while to recover but, when it does recover, it "snaps back" pretty darned quickly. And if I used a pulse (with moderate rise and fall times instead of a sinewave) I can get improvements in the current delivered into a 1 Ω load via a 100 pF capacitor: -

The above simulation uses ideal wires to connect things hence, I'm only showing its "results" as a way of demonstrating the potential fast snap-back time of sloppy old diodes.

Here I can get a 1 amp pulse of about 1.5 ns duration and about 100 ps rise time. But, do I really believe it? There isn't anything in the data sheets for these types of diode that indicate what the snap-back time is.

So, how can I calculate it for say the 1N540x or 1N400x series of diodes (question)?

Maybe I can't calculate it; maybe it's just a bit of pot-luck and trial and error. Maybe there is a formula somewhere that can reveal what I need to know? As you might expect I have searched for this on google and I'm aware that this sort of technique is used but, mainly with step-recovery diodes (hard to get and actually not what I think I want). Step recovery diodes have a very short recovery time and therefore operate differently to when using a long recovery time diode.

_{A few words about why I choose a diode with long reverse recovery time; the longer the reverse recovery time, the more time I have to reverse the applied pulse to a large reverse bias voltage and therefore the easier life is to produce a reverse current that is larger. A larger reverse current (prior to snap-back) means a higher value snap-back pulse.}

Answer 1

My experience with my own recovery test jig, is that regular rectifier diodes exhibit specified behavior (\$t_{rr}\$ and softness) for \$t_{on} \gg t_{rr}\$. This is a hidden assumption that they don't specify. They can be pushed into drift step recovery* by using short pulses. This has to do with injecting a layer of charge with a brief forward bias, then reversing and waiting for it to snap. Consider the function of \$t_b\$ vs. \$t_{on}\$: at most pulse lengths (over \$t_{rr}\$ or so) it's as specified and fairly constant, but as you reduce \$t_{on}\$, it gets suddenly MUCH shorter; it's pretty cool to watch.

*I'm not real clear on the precise definitions of SRDs and DSRDs, so, something like this, but check the literature to be sure (Grehkov, etc.).

Note that \$V_f\$ may be quite large during such a pulse -- some types have worse forward recovery than others, but the key is that you're still injecting charge and the junction isn't in quasi-equilibrium yet. So things can get quite strange looking, like putting 40V across a diode that you're certain isn't entirely due to lead inductance.

Answer 2

I believe--but am not 100% certain--that this "snap-back time" is what we in the semiconductor industry call \$t_b\$ (with \$t_a\$ being the rest of the recovery time--\$t_{rr}=t_a+t_b\$). Generally diodes are designed to have a slow snap-back to reduce high-frequency noise. The ratio \$\frac{t_b}{t_a}\$, called the "softness", can be found on datasheets for FREDs, and may be helpful for your purposes.

As I recall from using them to test some recovery measurement equipment I built, Vishay's E4PH and EPU diodes have particularly poor (low) softness, which may be what you're looking for here.

Answer 3

The topology of this circuit your are testing is unusual for a SRD, as it involves such low impedances, and ignores some key parasitics.

Usually, when we come to build a circuit physically, the L/C ratio of the unavoidable parasitics will tend to push us up into the 10s to 100 ohm range, and 1 ohm loads will be out of the question, especially in the face of a few nH series L in the switching diode.

This is the sort of basic circuit that is more typically used and analysed.

^{simulate this circuit – Schematic created using CircuitLab}

During the charge stage, current flows from the source through L1, D1, L2, storing charge in D1.

During the discharge phase, current flows back out of the diode, through L1 and L2, storing energy in L2.

When D1 snaps off, the interrupted current in L2 generates a voltage that goes through C2 to the load. The now low self capacitance of D1 means that L3's energy does not couple significantly to the output. Additional biassing may be used to time the snap off to occur at the peak discharge current.

What to look for in diode data sheets?

Most rectifier diodes, if they mention it at all, will boast of soft recovery, which is not what you want. A snappy diode will not be advertised as such, at least not by the manufacturer.

One of the things to look for is evidence of high doping adjacent to the junction. PIN diodes for instance are exactly the wrong thing. We have had best performance in the 130 MHz in 1.5 GHz out ballpark in commercial equipment using BA482/682 switching diodes. These are heavily doped for low forward on resistance at modest currents.