Transistors operating in their avalanche regions are often used to generate fast rise pulses (see avalanche pulse generator using 2N3904). Many transistors can also avalanche when the connections to collector and emitter are reversed. When operating in reverse avalanche region, these transistors are sometimes referred to as negistors.

Because the asymmetry and doping differences between the base-emitter and base-collector junctions, the avalanche voltages for reversely connected BTJs are usually magnitudes lower than their normal avalanche voltages. Here, I decided to test a few different transistors and see at what voltages the reverse avalanche occur.

The circuit I used is a simple LED flasher, similar to what was described here. As with any circuit that exhibits negative differential resistance, the principle of operation is quite simple. The capacitor is charged via a current limiting resistor. When the voltage is low, the current that flows between the emitter and collector is roughly the reverse current of a diode which is negligible. When the voltage across the capacitor reaches a certain level, the transistor enters the avalanche breakdown mode. In its avalanche breakdown region, the transistor exhibits negative resistance (i.e. the higher the current, the lower the resistance) so the capacitor discharges rapidly through the LED and the voltage across drops until the avalanche mode can no longer be sustained. At such point, the transistor enters its normal operation mode and becomes non-conductive again. Thus the cycle continues. The rapid discharging through the LED manifests itself as short blinks. The blinking frequency largely depends upon the RC constant and the transistor breakdown characteristics.


Depending on the LED you use, you may need to add a current limiting resistor to protect the LED during discharge. But for a typical 5mm white LED, no resistor is needed as the duration of the current flow is extremely short.

Using the RC values in the schematic above, I measured the minimum voltages required for the circuit to oscillate using different transistors and the results are listed below:

2N4401 ~12.5V
SS9014 ~12.5V
2N4124 ~12V
2N3904 ~12V
BD137 ~11V
BD139 ~11V
BC337 ~9V
SS9018 ~8.2V

At first glance, we can see that many transistors can be used to make this simple flasher. And for the majority the minimum voltage required is at just around 12V. SS9018, a high frequency NPN transistor, is a noticeable exception. It can oscillate at a voltage as low as 8V, which means it can be simply powered by a 9V battery. The threshold voltage for BC337 is only a tad higher at 9V.

Another interesting thing I noticed is that the same transistor from different manufacturers or even different batches can behave quite differently. For instance, while I could get one batch of 9014’s to oscillate at around 12 to 13 volts consistently, the 9014’s from another batch would not oscillate at all. So your results might be quite different than what I have here.

Also, while I could get most NPN transistors to oscillate in their reverse breakdown regions I could only get a couple of BD138 PNP transistors to oscillate using the same circuit above (power polarity is reversed). And the oscillation only occurred at a very tight voltage interval (e.g. ±0.05V).

One of the useful features of a standard avalanche pulser (like this one) is its extremely fast rise time (sub nano second), so can we use negistors to build similar pulsers?

Well, the short answer is no. After some experiments it appeared that the rise time of a negistor pulser is magnitudes higher (e.g. ~100ns) than a typical avalanche pulser.

The circuit below shows my experiment setup:

One of the key differences in this circuit between a negistor pulser a standard avalanche pulser is that the capacitance cannot be arbitrarily small. In my case, 100nF seems to be near the lower limit. Also notice the based resistor (10K) used to reverse bias the base-collector junction. Like in standard avalanche pulser the inclusion of the base resistor increases the avalanche breakdown threshold voltage which translates into higher avalanche current and shorter rise time.

The oscilloscope capture below shows the waveform measured at the collector (TP1). The pulses are repeated at roughly 500Hz when the supply voltage is around 28V.


Here is a single pulse viewed at a much smaller time base. As you can see, the rise time is no where close to the nano-second rise time we get using the standard configuration, but at under 100ns the rise time is still pretty fast.

Because these narrow pulses have extremely wide frequency spectrum, you can easily pickup the 500Hz audio signal using an AM or shortwave radio over the entire frequency range. The wide frequency spectrum serves as the carrier frequency and the pulse intervals becomes the audible signal. And the negistor pulser shown above can operate over a wide range of supply voltages, from around 20V upto more than 100 volts. As the voltage changes, the pulse interval changes as well and if you have a radio close by, you can hear the audible tone ramping up and down as you adjust the supply voltage.


The waveform at the emitter (TP2) illustrates the charge and discharge of the capacitor. Since the capacitor is only charged briefly to a fraction of the supply voltage before avalanche breakdown occurs, the linearity of the sawtooth signal is quite excellent.

Be Sociable, Share!

27 Thoughts on “BJT In Reverse Avalanche Mode”

  • Awesome work! There is a lot of info online about the Jim Williams Pulse Generator, but the transistor used can be difficult to get your hands on. There are some sources for it, but naturally these transistors are much easier to find.

    Did you test a 2222A? I’ve got a bunch lying around, I might throw this together and put up some scope shots to see if the times are comparable. I might have some other transistors as well, I’ll have to dig around. I love experimenting with avalanche and zener points of semiconductors. You can usually get some super fast stuff going on.

    Thanks again for your work!


  • Thanks for publishing your work – was fun and useful for me.
    I wonder if integrated blinking led’s are build following
    same or similar principle, and what experiments can be done.

  • LTspice’s Gummel-Poon BJT model has more parameters available to model behavior in the forward direction than the reverse. The simulation netlist below contains an “upside down” model statement that realistically reproduces a little known behavior used to build an interesting single inverted transistor relaxation oscillator.

    * UpsideDown.asc – a single transistor relaxation oscillator model for LTspice
    V1 1 0 10
    R1 1 2 1k5
    C1 2 0 1µ Rser=8m
    XQ1 0 NC_01 2 2N2222r
    .subckt 2N2222r e b c ; this subckt just turns the NPN upside down
    Q1 c b e 2N2222r
    .model 2N2222r npn Is=10f Xtb=1.5 Rb=10 ; nondirectional parameters
    + Br=200 Ikr=0.3 Var=100 tr=400p ; reverse (forward) parameters
    + Bf=7 Ikf=0.5 Vaf=10 tf=100n Itf=1 Vtf=2 Xtf=3 Ptf=180 ; fwd (rev) params
    + Re=.3 Cje=8p Ise=5p ; emitter (collector) parameters
    + Rc=.2 Cjc=25p Isc=1p BVcbo=7 ; collector (emitter) parameters
    .ends 2N2222r
    .opt plotwinsize=0
    .tran 0 10m 0 1u uic

  • Hi, thanks for your work! I need some help:

    – I made an oscillator using this technique, and tried the 2N3904 which obviously worked with 2 9V batteries.

    – Using the S9018, I managed to make it work with a single battery as you stated, but it will stop oscillating at one point as i turn my potentiometer all the way to and as it gets higher in pitch.

    I understand that it’s supposed to stop at lowest frequency when the pot is all the way to the left, so I thought i needed to reverse something on my circuit, but i can’t figure it out.
    COuld it be the transistor ?? (I tried two of them and they behave the same)

    Thanks in advance, and sorry if I wasn’t clear..

    • the only active part since your not using the base is a pn could do it with the transistor half of a optocoupler probably. it could work with any bjt on paper. the problem is ones that have no avalanche rating are not designed to be able to do it. some cant and some can even if they are from the same batch made by the same company if there is no avalanche rating. its not standard operational behavior its more like a hazard in most cases. look up the part number and check the data sheet. you have to reach a specific voltage minimum. many you find in computers will have a high threshold

  • I bought a collection of components to make a few synth oscillators based on this circuit. I used the Fairchild/ON Semiconductor 2N3904 and I couldn’t get it to oscillate at all. I tried the simplified test circuit above and the same result occurred. To people buying components for a circuit based on this principle, I suggest you avoid this make, choose another unless somebody can prove this is only a batch issue. Interestingly enough, when testing the oscillator circuit for the first time with random components I had (including a 10K potentiometer to vary the frequency), I discovered the BFY51 worked flawlessly. This transistor oscillated at 12V with improved results at 19V (these are the only supplies I have currently, I don’t have a proper variable supply).

  • Hey there, does anyone know a formular to calculate the capacitor or resistor for a specific frequency? Or is it just Try&Erro?

    • Read about RC filter. It’s basic and simple. There are always capasitors being charged via resitors. Here happens the same, it’s a charging capasitor. Look up for that.

  • Before trying it out id like to ask if it’s possible to replace the led with a signal diode, a 1n4148 for example. It can rid the need for a current limiting resistor and use the circuit as an audio oscillator instead.

    • Hi Brent, you can substitute the LED with other components for the load but you will still need a current limiting resistor unless you are using a power source with high impedance output (e.g. a 9V battery).

      • Thanks for responding, I’ll start experementing with available diodes, and see how they affect the circuit. There will be 24 of these, so I guess I’ll keep the resistor cause I’ll need to use a psu to power them all.

  • I’ve used many older,NPN-transistors’ with an 18v. power-supply and they have all worked well!I can get about ‘four-octaves’ with a 10K-potentiometer.The older,’metal-can’,NPNs’ work very well!These oscillators’,with a ‘regulated-power-supply’,could be used to build a ‘sawtooth-wave’-keyboard,or tunable-drones’,like Sam at ‘Look Mum No Computer’,on ‘youtube’!Cool,little circuit,thats’ been around for some time!The schematic I found in a book emits’ the diode! I also used a ’10ufd. cap.’.Lowering the ‘cap’ to ’20ufd.’ will drop the tone by ‘one-octave’!

    • Cool info Rob. I also followed the Look Mum No Computer project. Interesting you say about the metal can caps. The first one I tried was a pretty old BFY51. Works very well indeed.

Leave a Reply

Your email address will not be published.