avalanche transistors

I tried that, albeit stacking only two 2n3904 transistors and self-triggered. Exceedingly noisy, in that no two pulses were the same amplitude. I got 4ns, 70V pulses into 50 Ohms. I did not pursue it any further.

Jeroen Belleman

Am 21.07.20 um 19:02 schrieb snipped-for-privacy@highlandsniptechnology.com:

Not a string, only 1 transistor fed from 350V. The collector capacitor was a piece of coax, makes a nice flat top. Cable Z should be low against the

50R in the emitter or one loses amplitude. Parallel RG-188 is OK. 2N3904 is said to avalanche nicely. We used something different, don't remember exactly what for generating ultrasonic pulses with BaTitanate transducers.

Zetex/Diodes Inc. makes a transistor specified for avalanching. Costs an arm & a leg.

We went later to an array of 32 300W amplifiers for a beam steering phased array. Building the power stages would have been a dream job for me as a ham, but I had to do the phase & amplitude control. The transmitter transistors were 4 each of the newest & best from Motorola in Toulouse, nearly in Class A for a short time for each shot.

Gerhard

Sounds like the triggering was insufficiently powerful to ensure that both 3904 units were fully on, simultaneously.

In power electronics, long serial strings of hocky-puck SCRs are used to switch very large voltages (meaning a large factor larger than the breakover voltage of any one SCR). While this can be done with trigger transformers, it's far more common to use optically triggered SCRs fed via optical fiber by an optical power splitter driven by a single very healthy laser pulse generator.

I think this is also done with IGBTs. It ought to work with phototransistors.

Joe Gwinn

The Zetex parts are about $5 to 10 each, which isn't bad to predictably make a sub-nanosecond, 350 volt pulse. They have some appnotes about series connections for more voltage. I just wanted to see if anyone here has done it.

Sure. Works fine, but you get N * Rds(on) as it were, and if your load is

50 ohms or whatever regardless of stack height, well...

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website:

tirsdag den 21. juli 2020 kl. 19.02.23 UTC+2 skrev snipped-for-privacy@highlandsniptechnology.com:

?

What if the series avalanche transistors are spaced along a transmission line? Gotta Spice that.

The Zetex SOT-23 parts can dump 60 amps!

Am 22.07.20 um 02:06 schrieb Tim Williams:

I don't think that in avalanche breakdown the transistor has much Rce. It's beyond ON.

cheers, Gerhard

Sounds like you want a Percival distributed amplifier.

Bill Percival proposed the idea back in 1936. He was still working at EMI Central Research when I was there (1976-79). It's a brilliant conception.

is a bit older - 1924 - and a slightly different idea, but again a series string.

--
Bill Sloman, Sydney

"in most junctions breakdown is due to secondary ionization which results when minority carriers enter the reverse biased junction. Because of the similarity to breakdown in gases this has been named avalanche breakdown."

Here is a sprawling paper with lots of info about avalanche transistors:

On page 8 an equation is given that shows the collector current above the CE breakdown voltage, where Hfe is negative, and is in terms of an emitter efficiency y, a transport factor B, and the avalanche multiplication factor for electrons, M_n.

"If an avalanche transistor is to be used as a bistable switch the collector voltage during conduction should be made as low as possible when compared to the collector voltage during the time the transistor is turned off. cE(off) is limited by the breakdown voltage of the collector junction, that collector voltage at which M_n = (1/y*B)/(1 + Ib/Ic).

In order for this to occur at as low a value of M_n as possible the y*B product should be as large as possible. This product cannot have a value greater than 1, and it is at least 0.95 in any practical transistor, so very little can be done to further control this quantity. The alternative approach is to design the collector so that the required value of M_n occurs at a lower collector voltage."

On page 46 and 81 they give some circuits for testing avalanche transistors and a fast pulse generator, respectively

At Ferranti in the 70's we used a stack of Ferranti/Zetex e-line transistors ZTX 3xx or 4xx series as a Laser Q switch modulator. The transistors had to be specially selected for breakdown

Similar to this ,but with a lot more transistors as the ZTX had a lower breakdown. We abandoned it when we switched from KD*P TO lithium niobate cells which needed a higher voltage.

Brian

--
Brian

Cool, 1964, obviously done on a fairly worn-out typewriter.

The Zetex parts are, to my knowledge, the only transistors specifically designed to avalanche. They are apparently made on an ancient Russian fab line, probably all diffused. Epitaxial transistors seem to not avalanche much.

But given the Zetex monolopy, I'm more interested in circuits and physical PCB topologies, and especially actual experience.

Interesting, but I'm not designing semiconductors, especially not with germanium.

The old HP and Tek sampling scopes seemed to use selected 2N-type parts. Biased to maybe 60 volts or so, they could be triggered to make maybe a 30 volt sampling spike under 100 ps wide. The 7S14 uses an avalanche transistor to sample two channels and gets 2 GHz typical bandwidth.

The old transistors roughly half-discharged the available voltage. HP and Tek no doubt got selected parts from the makers; a random 2Nxxxx would be unpredictable. The Zetex parts fire like an SCR, all the way down to zero volts/shorted, and the SOT23s can dump 60 amps in around a nanosecond. That's worth $6.

--

John Larkin         Highland Technology, Inc 

Science teaches us to doubt. 

  Claude Bernard

Cool.

I did one HV pulser that used a drift step-recovery diode. Got about 7 ns FWHM at 2KV, into the tip of an atom probe. More recently, I did a

5 MHz, 1200 volt Pockels Cell driver with SiC fets. Both were water cooled and both were a PITA to develop.

Science projects! Fun but rarely profitable.

--

John Larkin         Highland Technology, Inc 

Science teaches us to doubt. 

  Claude Bernard

I don't think the math for silicon diffused transistors would be that different vs. germanium, just different fudge-factors

The ol' 2D21 thyratrons, gas tube relay triggers for like jukeboxes and stuff I guess, had very fast (ionization? de-ionization?) times for their era. I dunno what the equivalent slew rate is exactly but the rising edge on a relaxation sawtooth is fast enough that divided down the usual jellybean opamps as unity-buffers with GBWs of 10s of MHz and

10 or 20 volt/usec slew rates can't keep up.

I saw a contract job posting where the client wanted circuit to charge a

400n capacitance to 60V in 40 nanoseconds. Maybe you want that job. I don't want that job. Particularly if they're from the Middle East

600 amps for 40 ns isn't bad. A few EPC GaN fets could do that. Or even some mosfets. It would be mostly a packaging problem.

2N3904s are about 10 ohms. It seems Rce isn't enhanced much -- avalanche does not, a UJT, make. (Related note: JFETs don't seem to enhance much when the gate is forward-biased, at least not the common types I tried. A shame; I recall reading that UJTs are basically failed JFETs (unable to pinch off), would be fun if it actually worked.)

The better types have lower resistance, but still a nonzero amount. The point of diminishing returns is easily calculated, in any case, once this is known.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Design 
Website: https://www.seventransistorlabs.com/

I have seen mesfets and phemts turn on radically as the gate starts to conduct. Almost a bipolar effect.

JFETS and depletion phemts do enhance a little. A depletion phemt Ron can be reduced about in half from the zero gate voltage value.

The Supertex depletion mosfets enhance too.

Fractional microsecond. Very similar to SCRs actually, and for very similar physical reasons, I suppose.

Deionization is slow as balls, milliseconds. Xenon (2D21 and such) and mercury (bigger power thyratrons) are very slow ions. Hydrogen, or deuterium for that matter, is the fast one, with nanoseconds risetime and, I forget what exactly, fractional microsecond deionization?

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Design 
Website: https://www.seventransistorlabs.com/