Transistors as diodes

This is more of a self explained answer than a question, but it might help someone else

Please correct me if I am wrong though.

A NPN BJT transistor in its most fundamental form is two N doped regions with a P doped region that separates them.Thus two PN junctions (diodes) pointing away from each other. Now if you would short one of these PN junctions then you would effectively have a single PN junction that can behave as a diode.

Why would we do this?
1) It is handy if you do not have diodes on hand
2) Transistors are better documented and characterized in most cases than diodes. Especially at RF frequencies.
3) I ran out of more reasons

So now which part of the transistor should we short? base-collector or base-emitter?

well if you dig deeper into the working of the transistor you would find that one of these two PN junctions has to allow current to flow opposite than what the diode wants it to flow. This is interesting and it would follow that these two diodes need to be different to allow this behavior. Therefor when doping the transistor regions the collector and emitter regions, although they are both doped as N regions, are doped to different extent and have physically different geometries.

This is the reason why using two diodes configured as a transistor possibly wont work as you would expect it to.

Now back to the transistor, if you short the wrong PN junction you would still get diode behavior but it will probably not work as you expect it to. In fact you could use the collector as an emitter and the emitter as a collector in an amplifier circuit and it will work; but only to some extent and probably not half as good as when it was connected correctly(be careful as you could cause a breakdown in one of the junctions and blow the transistor)

I cant remember exactly how they are doped though but I know that for a NPN transistor it would be in favor of the diode application to short the base-collector junction.

Hope this helps someone who needs a bit more in depth explanation. This explanation can work wit the PNP as well as FETs, but I am not sure how the FET will behave if the Drain and source is swapped.

You are correct. As transistors are in most cases more expensive than an equivalently rated diode they are not normally used as replacements. However, there are some circumstances where they are used that you didn't mention:

1. where the junction voltage drop has to match that of the same device being used as a transistor, especially where good thermal tracking is involved.
2. in current mirror applicatons.
3. occasionally, with the base and collector linked together to make a diode with sharp V/I characteristics. As the B-E junction starts to conduct it then also passes current through the C-E path.

JFETS are basically symetrical but the shape of the junction may be skewed in favor of current flowing one way. Power MOSFETs are in most cases different because they have a PN junction across their source and drain pins so they always conduct in one direction, normally it is used in reverse biased mode.

Brian.

The emitter-base junction of a transistor has a reverse-biased avalanche breakdown at 5V to 7V and is not robust to handle the breakdown current so the transistor will be damaged. Then maybe the base-emitter should be shorted to make a diode.

I've only used small signal transistors as diodes to provide temperature compensation. But I've always shorted the base-emitter terminals together.

Audioguru said:

The emitter-base junction of a transistor has a reverse-biased avalanche breakdown at 5V to 7V and is not robust to handle the breakdown current so the transistor will be damaged. Then maybe the base-emitter should be shorted to make a diode.

Click to expand...

These e-b diodes in avalanche mode can be used as nice noise generators up to 13 GHz. A series resistor must be adjusted for an optimum noise output. Lower current, under 1 mA, allows to generate a maximum noise output under 1 GHz, at 10-12 GHz I use 5...10 mA for ENR of 30 dB.
SMD RF transistors are best for the higher frequency noise.

I have heard that the noise produced by an emitter-base having avalanche breakdown is caused by its fragile junction having small holes burned in it. Then the hFE of the transistor is reduced depending on the breakdown current and its duration. I was told to never use a transistor that has had its emitter-base junction have avalanche breakdown. I never tried one as a noise source but I wonder if the noise level slowly decreases as the junction is slowly destroyed? A zener diode also makes avalanche breakdown noise but it is built to survive the punishment.

Audioguru said:

I have heard that the noise produced by an emitter-base having avalanche breakdown is caused by its fragile junction having small holes burned in it. Then the hFE of the transistor is reduced depending on the breakdown current and its duration. I was told to never use a transistor that has had its emitter-base junction have avalanche breakdown. I never tried one as a noise source but I wonder if the noise level slowly decreases as the junction is slowly destroyed? A zener diode also makes avalanche breakdown noise but it is built to survive the punishment.

Click to expand...

I have never heard about a fragile junction with holes. My noise dipole with eb diode of a SMD transistor I made 20 years ago was adjusted for 8 mA to make a strong noise at 11 GHz, and I use it to calibrate and check my radio telescopes. No change was observed except I have to replace the 9V alkaline battery once in 2 years.
I made more than 20 such noise generators for experiments, and I only burned one when the avalanche current exceeded 10 mA.
The mechanism of Zener-diode noise generators I think is not based on avalanche breakdown. Zeners keep the junction voltage when the current is changed. Some Zeners even have a negative resistance which can generate noise, and the junction is large, so its capacitance would limit the generated noise to audio range.

I think it depend greatly on your use of the diode then, but I think everyone agrees that the use of a transistor is only usable with small signal inputs. The moment the reverse bias current becomes to great it will destroy the device. But the way that I am planning on using it is for a nonlinear terminating load to phase shift a given frequency as a function of its amplitude. Thus the input signal voltage can be controlled, and I have two transistors in parallel with opposite polarities. So they will never have a big enough reverse biased voltage over them to damage the transistors.

Yet the above uses is quite interesting.

I have never had the need to create noise though, and it would make sense to create it to calibrate some systems.

Audioguru said:

The emitter-base junction of a transistor has a reverse-biased avalanche breakdown at 5V to 7V and is not robust to handle the breakdown current so the transistor will be damaged. Then maybe the base-emitter should be shorted to make a diode.

Click to expand...

If BE junction or BC junction are shorted, where is the problem with the reverse biased PN junction if it is shorted (0 V across it). Where would be avalanche if there is no reverse biased PN junction ?

I have worked in technologies where the design support
had models only for BJTs and none for a plain PN diode
of any kind. Mostly these were medium voltage analog
meant for op amp products, but which we needed to
use for other applications. In general it was easier to
"just use what's in the kit" than try to get someone in
another functional area to give you PCells, models and
rules decks within your project timeline. So E-B or C-B
transdiode were the options. When I got to run my own
process development, I insisted to have PN diodes in
the kit.

Now, transdiodes have some plusses and some minuses.
Plusses include a low effective series resistance and a
better range of log-linear behavior (nice for bandgaps
etc.). The active BJT behind the E-B transdiode helps
local-feedback-style. But one minus is the low BV and
another is a long storage time in the C-B junction which
although it appears outside saturation, becomes not-so
when you start to run significant current - your real
C-B junction is not the external 0V (B=C) but If*Rc
and this can put you into significant charge storage
and high reverse recovery times.

C-B transdiodes have some of this but are pretty benign.
The shorted E, B may improve recovery time some, it
enhances intra-base charge gettering. But you'd have to
put similar counter-plugs in the collector as well to get
a true fast recovery style, and goodbye to any model
fidelity (if the BJT-as-diode works right to begin with).

Now, funny story. My entry to high voltage and analog
was a power MOSFET driver design in one aforementioned
technology. Back before the days of computer LVS, a
hand drawn schematic & pencil checks was it. I drew
a C-B diode but it got laid out E-B (because that's what
was the norm among the "real analog designers" and
their parts). Diode terminals, P and N, checked out. Part
got fabbed. Standing on the probe floor at first wafer
test debug, I looked into the probe station microscope
and saw little flashes of orange light and felt something
sting my face. Turns out the E-B diode which should
have been C-B, was not up to the 30V reverse bias
and was instantaneously heating to orange-hot and
launching its silicon body out of its dielectric tub.
Not cool. A minor mask revision fixed it but not the
desired "first pass success"....

Mr freebird, could you maybe give more information as to when a BJT or even HBT model fidelity will go out of the window? Or why the BJT will not behave as a diode at certain frequencies, if say the models fidelity was tested way above the frequency that you are interested in?

I would like to gather as much as possible information on this possible downfall on my application.

Just a thought, would the lower base resistance of the HBT increase the maximum workable frequency of the diode?

And when you moved from the transdiode to the diodes after you got to run your own development, were there any pros or cons against the diode than the BJT? (what I am asking is: is there a situation where you would want to revert to the transdiode? Better models for higher frequencies mayhap?)

I would expect a HBT foundry to be pretty good about
frequency domain modeling, but then again they may
have an ingrained expectation that (say) the transistors
are only going to be used as forward-active amplifiers
and neglect the C-B diode application altogether - so
ignore fine fitting of inverse hFE, saturation recovery
in the case of the E-B diode and do on. This was the
case where I was at, the process was supposed to be
for linear applications and no effort made to characterize
or model saturation entry / exit at all. And swinging in /
out of saturation (as a diode would, as a switch /
commutation element) is large signal while frequency
response is often thought of as small signal, so what
the diode's small signal frequency behavior is may not
have much to tell you about a different application.

Not really about the device or the model per se, but
the priorities of the modeling people. But transistors
do have these other "dimensions" that a diode does
not, which makes fitting simple stuff not-simple.