BJT E-B zener

Danke,

A 2n3904 b-e should be a stable zener, good for the transistor's normal dissipation. Long-term zenering will wreck the transistor-mode beta but is otherwise pretty stable.

The b-e junction is small compared to a regular zener, so has less capacitance. Keep the current low.

Some npn transistors can be used backwards (c-e, collector negative) as "reference diodes." The forward biased c-b junction is in series with the zenering b-e junction and the temperature coefficients cancel, around 6.8 volts as I recall. 1N821 does this dual-junction tempco trick. These sorts of things typically have some current where the tempco is zero.

But cheap bandgap references have made the transistor tricks mostly history.

That's reversed. Lower voltage breakdown diodes, "real zeners", use a tunneling effect and have soft knees. Higher voltage parts, technically avalanche diodes, have sharp knees and low current just below breakdown. They have opposite tempcos. The fuzzy boundary between types is roughly 6 volts.

A 10 or so volt avalanche diode can have nA current just below breakdown voltage. They do fun stuff at low currents too.

Nowadays people call both kinds "zeners"

Exactly right. Clamping the output of a 33V TVS with as little leakage as possible. The E-B Zener has very low leakage until it goes into breakdown but lowers the overall Vin at that point so I can clamp it more tightly just below 5 for my ADC input. In my case I'm looking for a overall error of less than 10 to 20nA leakage from my protection circuit because this device will be able to see -50mV to 5.000V of valid signals though not all at once. The leakage of a regular zener is simply too high for this to work.

whatever are people thinking when communicating with unexplained web links?

I ignore them all.

You might use two official diodes to clamp to some power supply and ground. Transistor c-b junctions make excellent low-leakage diodes, pA or for some parts fA reverse leakage. Official diodes, like 1N914 types, tend to be leaky and photosensitive.

The PAD series of "low leakage" diodes are terrible.

The sadly departed BFT25 made a great diode. I measured about 10 fA leakage, but that was about my measurement resolution.

I sometimes use depletion fets as the "resistor" in a clamp circuit. See AoE3 fig 5.80.

Are TVS diodes leaky? I never tested for that.

There are USB protectors that work pretty well for that, e.g. PRTR5V0U2. You can get it in SOT-143 (a four-pin SOT23), which is convenient.

It has a zener plus two pairs of Schottkys.

Cheers

Phil Hobbs

Interesting part. The diodes are < 1 nA leakage typ.

Actually, it will be siting in the overlap region between the Zener mechanism and the avalanche region.

Rubbish. Back in the 1960's the British Journal of Scientific Instruments - now Measurement Science and Technology - published a paper on using a gold-doped switching transistor as a cheap alternative to the 1N821 - 1N829 series. You had to tune the forward current in the 1ma to 5mA range to get a precisely zero temperature coefficient, but it gave you a very stable reference voltage close to 6.2V. The 1N821 series have a nominally zero temperature coefficient at 7.5mA, but the selection process that picks are the premium 1N829 part just picks the parts where the zero happens at close to 7.5mA. The stable breakdown voltage range is 5.9V to 6.5V - +/-5% - even for the 1N829, and you have to calibrate that out.

The 'much better leakage' relates to test protocols, more than reality. The recommendation to use a 2N3904 as a reference goes back to National Semiconductor's (now TI) process for making those; the Motorola (now ONsemi) process might be different (and breakdown is only specified as 'over 6.0V').

E-B breakdown also applies the opposite to 'normal' field on the surface of the transistor, where the B and E contacts are made; such surfaces are passivated, but field reversal can move dirt on the surface to age the component abnormally.

So, no one knows 'how stable it is' for this usage, unless they've time/temperature tested a few batches from a known process and the manufacturer doesn't change anything. That '2N3904' part number was defined many decades ago, there have been lots of variant parts holding to that loose specification, sold under that part number.

It'll probably work. That doesn't mean it's good design practice to depend on it.

You can get a 0.5% bandgap voltage reference for 4 cents.

That is my gut feeling too. I was looking for better corroboration of if it's a good idea to design into something that can go months before anyone comes by to service the flaky analog input section.

Don

Another non-risky option is uA723 or LM723 or MC1723; the innards is basically a few general purpose transistors, plus a temperature compensated zener. That was the old-school part, with LOTS of second sources.

A legacy part, used by legacy designers, in much the same way that the NE555 still has it's fans. You can do better, but often you don't have to bother.