Bipolar Avalanche Transistors - need a good list, and some ideas how to test 'em

[SMB784]

Hey everyone,

Recently I have been working on an avalanche transistor project and I have finally discovered that the avalanche transistor that I have been using has too low of an avalanche probability.  I have been browsing Mouser & Digikey for avalanche transistors, but neither of those websites has the ability to filter for avalanche type BJT's specifically.  Does anyone know of a good way to search Mouser or Digikey for specifically avalanche transistors? I know that all BJT's can avalanche if you bias them high enough, I'm specifically referring to BJT's which are specifically geared towards use in avalanche mode (e.g. the 2N2369A.  I am looking to test as many different avalanche BJT's as possible.

Also, is there a standardized way to test the avalanche probability of an avalanche BJT?  I have just been using a multimeter, a frequency counter, and an oscilloscope to compute the avalanche probability, but maybe there is an easier way or a device that is capable of doing this more readily.

I also just realized that I posted this in the wrong forum, so mods please feel free to move it to the Projects, Designs, and Technical Stuff section!  Sorry!

[edavid]

Zetex (now Diodes Inc.) made some parts that were guaranteed to avalanche, so those would be best.  The ZTX415 is still in production, although expensive: https://www.diodes.com/products/discrete/bipolar-transistors/special-function-transistors/avalanche-transistors/part/ZTX415

See the app note for test circuits.

Some other parts that people have used:

2N2501
BFG424
BFG425
BFG540
BFG541
BFR505

[mzzj]

Avalanche pulse generator thread is somewhere here on EEVBLOG...
RF transistors like BF-something seem to have lowest avalanche voltage if that's what you are after.

[T3sl4co1l]

I don't know that any were "specifically geared" for avalanche operation.

2N2369(A) is a gold-doped switching transistor -- it is to 2N3904, what 7400 series TTL logic transistors are to average analog fabs of the time (e.g., what got used on LM324s and such).  Vce and hFE are lower, but t_stg is much lower.  (Whereas a non-gold-doped '2369 might be rated for 60V, it's only rated for 15; whereas LM324 and such are typically 30-40V rated, TTL is rated 7V -- presumably it's a bit more heavily doped in comparison.)

BTW, don't fall into the trap of insisting on 2N series (JEDEC registered) parts.  You probably want PN2369 or MMBT2369 (TO-92 and SOT-23 respectively).  JEDEC registration includes package as part of the spec, so you will always get the TO-18 in this case, and you will always pay for it (whether it's more expensive new, or salvaged from something and you're getting ripped off).

Same with 2N2222(A), which is a painfully weak registered part to begin with.  Use 2N4401/3 instead, for general purposes.

Avalanche.  I suspect all BJTs will avalanche, at some combination of Ic and Rbe.  Some avalanche easily -- over a wide range of Rbe, and starting at low Ic -- and some do not.

Also, large junctions are useless.  A power transistor only avalanches in a small pinhole location, so it isn't capable of handling pulses any greater than, say, 2N3904 is, before a hole gets burned through it.

ZTX415 is the only one I'm aware of, and it's probably not so much designed for it (AFAIK, it's a completely stock FZTxxx) as tested for it (hence the greatly jacked up price -- the testing is probably slow, and at a low production rate).  Mind that it is slower than the smaller transistor types, having a typical risetime in the 4ns range, whereas 2N3904 will do 1 or 2ns and MMBT2369 will do under 1ns.

Tim

[nctnico]

Several years ago I did some experiments with trying to avalanche BJTs. I found out modern parts don't avalanche very well (at all that is). The 2N3904 however seemed to work very well IIRC. First time I needed 10V/div in 50 Ohm mode (which most scopes won't do).

[SMB784]

I don't know that any were "specifically geared" for avalanche operation.

2N2369(A) is a gold-doped switching transistor -- it is to 2N3904, what 7400 series TTL logic transistors are to average analog fabs of the time (e.g., what got used on LM324s and such).  Vce and hFE are lower, but t_stg is much lower.  (Whereas a non-gold-doped '2369 might be rated for 60V, it's only rated for 15; whereas LM324 and such are typically 30-40V rated, TTL is rated 7V -- presumably it's a bit more heavily doped in comparison.)

BTW, don't fall into the trap of insisting on 2N series (JEDEC registered) parts.  You probably want PN2369 or MMBT2369 (TO-92 and SOT-23 respectively).  JEDEC registration includes package as part of the spec, so you will always get the TO-18 in this case, and you will always pay for it (whether it's more expensive new, or salvaged from something and you're getting ripped off).

Same with 2N2222(A), which is a painfully weak registered part to begin with.  Use 2N4401/3 instead, for general purposes.

Avalanche.  I suspect all BJTs will avalanche, at some combination of Ic and Rbe.  Some avalanche easily -- over a wide range of Rbe, and starting at low Ic -- and some do not.

Also, large junctions are useless.  A power transistor only avalanches in a small pinhole location, so it isn't capable of handling pulses any greater than, say, 2N3904 is, before a hole gets burned through it.

ZTX415 is the only one I'm aware of, and it's probably not so much designed for it (AFAIK, it's a completely stock FZTxxx) as tested for it (hence the greatly jacked up price -- the testing is probably slow, and at a low production rate).  Mind that it is slower than the smaller transistor types, having a typical risetime in the 4ns range, whereas 2N3904 will do 1 or 2ns and MMBT2369 will do under 1ns.

Tim

Generally I'm not too worried about the pulse duration, nor am I too worried about the pulse amplitude.  What I am worried about is the transistor's avalanche probability (i.e. what is the chance that an electron triggers the avalanche).  It is my experience so far that this probability is obscenely low for the STMicro 2n2369A, e.g. 1 electron in every 1 billion triggers an avalanche.  I would like that number to be much, much higher, like around 10E-6 or better, but I would still accept 10E-7.  This probability, from what I have read, depends on the shape and structure of the actual NPN junction and how it distributes the electric field on the collector over the collector surface.

Any idea about which transistors are more apt to avalanche more readily on a per electron basis?

[T3sl4co1l]

Per electron?  WTF?

How would you even know?

You can get a APD I suppose, and if you insist on it being electronic in origin, put a field emission cathode nearby to trigger it.  Seems like an awfully expensive and highly roundabout way to generate pulses though.

Tim

[SMB784]

Per electron?  WTF?

How would you even know?

You can get a APD I suppose, and if you insist on it being electronic in origin, put a field emission cathode nearby to trigger it.  Seems like an awfully expensive and highly roundabout way to generate pulses though.

Tim

Sorry I wasn't clear about my intentions.  I am not trying to make a pulse generator.  I am trying to make an electron counter.  My understanding of the avalanche process is that an electron entering through the emitter passes through the emitter-base junction, is accelerated by the voltage on the collector, and causes impact ionization of more electrons, thereby generating the avalanche.  I am trying to exploit these avalanches triggered by single electrons to create an electron counter

I am doing this by grounding the transistor base, putting a small negative voltage across a large resistor to limit the number of electrons per second injected into the emitter, and putting a large positive voltage across a more modest resistor connected to the collector.  An example of such a circuit is included in the paper I have attached.  It is known as a single electron bipolar avalanche transistor (SEBAT).

From my attempts so far using a 2N2369A, the avalanche probability appears to be about 10E-9 for this particular transistor series from STMicro.

[edavid]

I am not trying to make a pulse generator.  I am trying to make an electron counter.

That's called a picoammeter.

[SMB784]

I am not trying to make a pulse generator.  I am trying to make an electron counter.

That's called a picoammeter.

A picoammeter is $1000 that I don't have, and SEBAT counters (albeit with custom transistors) can measure current in the attoamp range, far below that of any ammeter.

Also, it's a fun and interesting project.

[Wolfgang]

Hi,

I tried quite some transistors and pulser types, here:

electronicprojectsforfun.wordpress.com/pulse-generators/avalanche-transistor-pulsers/

The question is what your pulse property priorities are.

- The smaller the chip, the faster they are (BFR96, 2N2501, 2N2369A)
- The higher breakdown voltage you have, the higher the pulse amplitiude will  be
- NOT ALL transistors avalanche. You will have to screen them (See Jim Williams High Speed Amplifiers Techniques AppNote)
- Avalanching is normally no property guaranteed in datasheets. Only the ZETEX ones have this guaranteed, but they are
  comparatively slow and expensive.
- If your product of pulse energy x repetition rate is too high, device limetime could be very short
  ZETEX has graphs on this for their devices.
- Matching at the output is important for clean pulse shapes.

Much fun ! 

[mzzj]

A picoammeter is $1000 that I don't have, and SEBAT counters (albeit with custom transistors) can measure current in the attoamp range, far below that of any ammeter.

Also, it's a fun and interesting project.

I'm afraid SEBAT is totally different beast compared to normal BJT

[SMB784]

I'm afraid SEBAT is totally different beast compared to normal BJT

It is indeed different - the junction structure includes a guard ring to improve the avalanche probability - but that doesn't mean that you cant construct a SEBAT out of an off the shelf part.  I have had success in creating a SEBAT out of a 2N2369A, albeit with terrible avalanche probability (10E-9 at best), hence the question: what are the avalanche probabilities of some other off the shelf transistors?  I assume that this information isn't known about most of the off the shelf parts, so I am endeavoring to measure it.  However, in order to do that, I need a good list of transistors that reliably avalanche so that I may test their avalanche probabilities.

This brings me to Wolfgang's comment:

Hi,

I tried quite some transistors and pulser types, here:

electronicprojectsforfun.wordpress.com/pulse-generators/avalanche-transistor-pulsers/

The question is what your pulse property priorities are.

- The smaller the chip, the faster they are (BFR96, 2N2501, 2N2369A)
- The higher breakdown voltage you have, the higher the pulse amplitiude will  be
- NOT ALL transistors avalanche. You will have to screen them (See Jim Williams High Speed Amplifiers Techniques AppNote)
- Avalanching is normally no property guaranteed in datasheets. Only the ZETEX ones have this guaranteed, but they are
  comparatively slow and expensive.
- If your product of pulse energy x repetition rate is too high, device limetime could be very short
  ZETEX has graphs on this for their devices.
- Matching at the output is important for clean pulse shapes.

Much fun ! 

Thanks so much for the information!  I will order some of these parts and try them out!

[vindoline]

Sounds like a very cool project! Please keep us up to date!

[iMo]

Here is an another paper on the topic, and you may contact them..

[Wolfgang]

Your comment about larger junctions is correct; what could be tried, however, are emitter-ballasted RF transistors like the vintage 2N3375.
Any experiences with those ?

[Cerebus]

I am not trying to make a pulse generator.  I am trying to make an electron counter.

That's called a picoammeter.

No, I think it's probably a zeptoammeter. One electron/s is ~1.6 x 10^-19 amps or 160 zeptoamps. An attoamp would be 16 electrons/s, a femtoamp is 16,022 electrons/s and a picoamp is a comparative raging torrent at 16,021,766 electrons/s. As the intent is to detect individual electrons I think we have to err in the direction of zeptoammeter if we're going to stick to SI prefixes.

Of course at these levels one probably has to regard the solar wind as a source of leakage current, even after it has been attenuated the the zeptoammeter's primary guard ring, the magnetosphere of the earth. 

I have idly wondered in the past if there was any mileage in using a variant on a photomultiplier tube as an electron counter.

[Krakdustten]

So, how dit the project go?
Dit you find any transistors that where able to measure pico/femto currents?

I'm doing the same kind of project and this looks like one of the possibilities that are actually achievable.