abusing transistors

On a sunny day (Mon, 04 Jan 2010 15:38:17 -0800) it happened John Larkin wrote in :

LM317 is very temperature satbel with no load. :-)

Trying to fathom your inadequate description...

NPN transistor

Collector at ground

Base floating

Forcing a current _into_ the emitter

Is this correct?

You're measuring what would nominally be LVceo, except it's inverted, so call it LVeco ?:-)

It's leakage dominated, so nothing is firm. ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.

John Woodgate mentioned this in the LTSpice group, back on 30 Dec 2009:

"It's not like a zener diode; the breakdown voltage is much larger than the maintained voltage - the voltage across the device in the broken-down condition."

He wrote that in response to another saying this: "The base emitter diode have a reverse breakdown at between 6-7 volts in this mode. This is in series with the forward conducting base collector diode which add about 0.6 volts to the B/E zener diode."

Both were in a discussion about these pages:

Jon

At 1 mA, it's acting like a zener diode with about a 36 ohm dynamic impedance... positive, not negative. Most of that is caused by the c-b diode; the b-e zener is a lot stiffer, more like 10 ohms. That all makes sense.

The voltage seems to be increasing by about 30 millivolts per hour or so. I wonder if I'm migrating implants around. That might correspond to the known beta damage effect of zenering the b-e junction.

Looks like a b-e junction isn't a very stable zener. So it may not be a very stable noise diode.

John

The way I've always done it is tie base to collector and ground both (NPN). Then feed a (+) current source into the emitter and use the emitter voltage as a regulated voltage.

No?

Jim

Why aren't you shorting base and collector to ground, and putting current into the emitter?

--
www.wescottdesign.com

Why should I?

This way, I get to measure both the zener voltage and the forward-biased junction voltage.

Maybe I'll start with a fresh transistor tomorrow and measure both versus time.

John

P.T. Rudge published a circuit for such a voltage reference in the British Journal of Scientific Instuments in the 1968

"A plug-in transistorized shunt regulator" P T Rudge 1968 J. Phys. E: Sci. Instrum. 1 493-494

According to page 198 of my Ph.D. thesis, he used a reverse-biased

2N3638A as his reference source.

IIRR the paper suggested tweaking the current through the forward- biased collector-base diode and the reverse-biased emitter-base junction for minimum voltage shift in response to a temperature shift imposed on the device by prodding the package with a hot soldering iron.

It is built with the National Semiconductor process 63 - PNP medium poswer, non-overlay double-diffused - also used for the 2N2905 and

2N2907A.

-- Bill Sloman, Nijmegen

That happens around 6.2 volts.

Reference diodes, like the 1N827, are a zener in series with a regular PN diode, and the tc can be tweaked to zero by varying the current. GE used to make "reference transistors" which was an NPN transistor with a zener in the emitter, also tweakable to zero TC. The classic solid-state Fluke differential voltmeters mostly used reference transistors as their main voltage reference.

Maybe older diffused transistors didn't change voltage much as a function of integrated b-e zener current. The BCX70 sure seems to.

John

We had this topic here only 11 years ago. Maybe it should go into the FAQ.

From the test frequencies, you can tell I'm a ham. But I have upgraded to an Agilent 346C in the meantime. :-)

regards, Gerhard, dk4xp

The paper seems to suggest that the combined voltage drop is at around

7.0V at the zero temperature coefficient, which is what Rudge claimed to be able to set up at a current of around a few mA - less than the 7.5mA you needed for a 1N823/5/7/9 or the like (which was what you would have bought to to the job at the time).

You might take a look at the 1997 thread "ZENER DIODE OSCILLATION", where on the 5th August Winfield Hill eventually dug out the physicists papers on how avalanche discharges in higher voltage (over about 6V) "zener" diodes actually work, with most of the action happening in localised microplasmas (which get warm. as well as emitting light). He also posted a follow-up on the 6th August which clarified a few points.

A micro-plasma in a very thin base diffusion might make it a little less thin. A zener discharge at 3.25V ought to be occurring by the Zener mechanism

which ought to be better behaved, but it does have a negative temperature coefficient, which could cause the reverse current to concentrate into narrow channels which could then warm up enough to anneal the original doping profile into something a little less sharp- edged.

-- Bill Sloman, Nijmegen

Hello John,

You probably read the noise source thread also:

7bb3ce8f4c62fde?hl=3Des

I did some reverse measurements on a 1n4448 (and mentioned them in the noise thread). Though the average current can be very low (for example

100uA), the peak current was in the 40mA range with ns rise time, so not good for a Gaussian noise source. Heating the diode does stabilize the breakdown process. I know from experience that reverse biasing BE junctions, destroys low current HFE completely (BC84x series).

Maybe you can add a small resistor in series with your transistor zener, so you can observe the current with an oscilloscope. Make sure you have good (RF) decoupling from emitter to ground, so current peaks are not limited by stray inductance.

Best regards,

Wim PA3DJS

please remove abc in case of PM

If you'll forgive me for inserting a beginner's question here: Destruction of HFE by reverse biasing the b-e junction has been mentioned more than once in this thread. I never knew about this.

My usual method of quickly testing a BJT for shorts, open circuits and excessive leakages is to use an analog MM in the ohmmeter range - low range for forward bias and high range for reverse and Iceo. For identifying c and e of an unknown transistor, I set the meter to the high resistance range - driven by a 9V battery. This causes the e-b junction to break down while c-b doesn't, thus identifying which is e and which is c. The reverse emitter current is of the order of tens of microamps.

Have I been damaging transistors this way all these years?

Yes, but only very slightly. I wouldn't worry about it.

Curious, does that method work on high voltage power transistors? I've seen a lot rated for Veb = 10V or so. Not that it's a big deal, power transistors are all BCE (or GCE or GDS) anyway.

Tim

-- Deep Friar: a very philos>>

A simple one-time ID test with low current doesn't do enough damage to discern.

However, continuous breakdown currents above ~5mA will cause gradual electromigration and beta degradation. This effect is accelerated by high temperatures... as I found out the hard way under the hood of a car :-( ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.

Hello,

As what Tim and Jim says: only little. The reason is that testing the transistor takes just a few seconds and you are using tens of uA as emtiter current. As a transistor has HFE in reverse direction also, the base current will be far below tens of uA. The time*current product is of importance, not the reverse voltage itself.

When you want to do some experiment, just take a fresh general purpose small signal transistor, measure the HFE at low collector current (for example 10uA). Then do your CE determination test and measure HFE again (at same current and same transistor temperature). When you would do your determination test all day long, you will probably notice a reduction in HFE.

Best regards,

Wim PA3DJS

without abc, the address is valid

Well, for starters because the base-emitter junction is going to act differently if you're draining carriers out of the base, and that's going to change the 'zener' voltage -- possibly a lot, and possibly in a way that has a significant effect on the temperature coefficient and aging properties.

Or something like those, yes.

--
www.wescottdesign.com

It works with most power transistors IME unless they have an internal Rbe. However, since most power transistors in the same standard package have the same lead-outs, ID-ing the leads is usually not necessary. And for repairing and reverse engineering work, it's usually easy to identify the leads from their associated parts on the PCB.

It's with low-power transistors that it's sometimes necessary to employ something like my method to identify the leads, especially in the absence of a datasheet. TO-92s in particular come with all possible orientations of the leads. I even had a stock of different batches of PN2222As with different lead-outs. I thought one batch was defective until I discovered that they had the collector and emitter leads swapped.

Ah. Thanks to you, Tim and Jim for the clarification. I'm not exactly a raw newcomer to electronics, but being self-taught, I obviously neglected to delve deeply enough into semiconductor physics.