Why different terminologies for FET vs Transistor

[IanB]

Oh no they don't!

(You might not get the British cultural reference here.)

[IanB]

Reference:

In an integrated circuit there is no difference between drain and source of a mosfet. The fet is symmetrical. There is also not this reverse diode.

(The arrow on the circuit symbol does not represent the body diode, it represents the source of charge carriers. The body diode is formed by a connection between the bulk substrate--the fourth terminal--and the source. If this connection between the bulk and the source is not made inside the device there is no body diode. What you have in fact is two diodes back to back so that one of them is always reverse biased and forward conduction never occurs.)

[c4757p]

There is also not this reverse diode. this diode is formed between the bulk and the channel. we connect those bulks in such a way they don't bug us.

I think we might be arguing over a difference in terminology. I'm not saying there's always a diode from source to drain, but unless I've been horribly mistaken for a long time, there is a diode from the substrate to the channel (or reverse, in the case of a P-channel FET). As f_e said, the substrate in an IC is usually connected to whichever power rail is most convenient to keep the diode always reverse-biased.

I'm fairly certain, though, that the substrate-channel diode is always present.

[A Hellene]

All the MOSFETs are constructed as four-terminal devices...

You can read more here or, even better, here.


-George

[IanB]

I think we might be arguing over a difference in terminology. I'm not saying there's always a diode from source to drain, but unless I've been horribly mistaken for a long time, there is a diode from the substrate to the channel (or reverse, in the case of a P-channel FET). As f_e said, the substrate in an IC is usually connected to whichever power rail is most convenient to keep the diode always reverse-biased.

I'm fairly certain, though, that the substrate-channel diode is always present.

There are two diodes, not one. Take a look at this diagram from the Wikipedia article:



There are two islands of N-type material sitting on a substrate of P-type material. Each of the N-P interfaces forms a diode junction, so between source and drain you have an N-P-N sandwich making two diodes back to back. If you connect the body terminal to the source terminal as is commonly done, then one of the diodes is shorted out and the other diode remains as the body diode, as there is now a single P-N junction between source and drain.

On an IC, or in some other applications, the body terminal might be connected to some other equipotential point apart from the source and then there is no parasitic diode. The FET becomes symmetrical and source and drain become equivalent.

[c4757p]

Right. I was mistakenly picturing the channel as one conductive region, as if we were talking about depletion-mode FETs.. Still, two diodes - but my point stands, that there is always a diode there. The substrate can't be left floating, so it has to be connected somewhere, and then you have one diode each from the substrate to source and drain. It's not right to say that you can avoid the diode entirely by building it inside an IC, or that small MOSFETs like BF998 don't have one, even if it is technically incorrect to call it one diode.

In other words, regardless of how you explain it or how the MOSFET is constructed, a diode tester will always be triggered when it's connected from the substrate to either end of the channel. Always, no matter what kind of MOSFET it is.

[IanB]

There are certain circumstances where apparently you can avoid the diode entirely by building the transistor inside an IC, when using silicon on sapphire technology. In this case the semiconductor substrate is replaced by an insulating sapphire substrate and then no diode junctions exist.

[c4757p]

That is pretty cool, I'll admit.

Still, I don't think most of us will ever have to worry about what happens when our substrates are made of sapphire.

OK, replace "all MOSFETs" by "all silicon-only MOSFETs".

[SeanB]

SOI still has the diode, just it is connected to the one terminal so it will still have an effect when voltage is high enough to cause leakage current to flow. The substrate connection will just be a small isolated area on the bottom of the device, but it still will have the diode as part of the intrinsic construction. The only advantage of SOI is a high tolerance to radiation SEU and a low leakage current.

[wbeaty]

Heh.  What if we're looking at 3-terminal active components in general?

If it's got a BASE, then BJT

If it's got a GRID, then vacuum tube.

If the input is the GATE, then obviously it's ...an SCR or Triac!   Or FET.


Once I saw a brief mention of a very cool 3-terminal electronic device.   It was a vacuum tube with a wide, fan-shaped electron beam.  (More of a cathode-ray tube, I guess.)   A beam of microwaves was aimed at the electrons.  Free electrons are a conductor, so a small part of the microwave beam was reflected.   A third electrode caused the electron beam to sweep back and forth.

Not very efficient, since electrons in vacuum are so sparse.   But any time you need to scan a microwave beam back and forth at tens of megahertz, they've got you covered.

[madires]

If the input is the GATE, then obviously it's ...an SCR or Triac!   Or FET.

Could be also an IGBT :-)

[free_electron]

There are two diodes, not one.

It's worse than that. That shit is actually a transistor in many cases , and it has some gain !
For large geometry fets you can consider it two diodes. The smaller we become the narrower the gate and the thinner that base layer becomes. There is a point where that stuff begins to work as a bipolar transistor !

Somebody mentioned silicon on sapphire. You can also construct on depleted silicon.( that is a perfect isolator)

[nukie]

Please, just watch this guy's videos then all will be clear. 
https://www.youtube.com/user/booksofscience/videos

[megajocke]

It's worse than that. That shit is actually a transistor in many cases , and it has some gain !
For large geometry fets you can consider it two diodes. The smaller we become the narrower the gate and the thinner that base layer becomes. There is a point where that stuff begins to work as a bipolar transistor !

I think I had this bite me in a circuit where a 16-to-1 MUX selected the input to an ADC. One of the sources had a problem, sending 5 V through a 10 kohm resistor to one input of the MUX which was on a 3.3 V supply. Many sources were high impedance, and the current being injected into the MUX output through the parasitic bipolar transistor on the faulty channel gave rise to an offset of about 1/3 full scale on all channels...

[free_electron]

Could be. The analog mux chips frequently have no protection structures on the analog signals... So if you pull the input higher or lower than the rails by 0.7 volts you may trigger the transistor... Since you had 10k there you did not fry it but were 'steering' it ...

That is another case of the old rule : never, ever apply a signal beyond the rails hoping the internal protection structure will clip it... You simply dont know how it is built and what it can handle.

Only if it is explicitly stated by the manufacturer , and is listed as a marketing feature, can you do that...

[SeanB]

Even if it is do not rely on it though, add the diodes. you never know just how that junction will degrade with time and if it will still behave the same with a change in the die during a redesign or process change. Unless it is a part of the design and is given specs and typical and min/max values in the datasheet and is tested in production it is only a feature and not a designed in spec, subject to change without notice*.

* often undocumented or buried deep in small print in a small errata note in a datasheet for another device that coincidentally uses the part as a unit in a reference design.

[A Hellene]

... and this problem exists, especially for the Silicon On Insulator (SOI) technology devices, because the electrical potential of the pseudo-floating (see: 'insulated') bulk material (the fourth terminal of those four-terminal MOS devices) can easily cause local channel latchups under the specific condition of an I/O exceeding the absolute maximum Vdd/Vss thresholds.

In the modern semiconductors (devices of a few tens of atoms thick for each transistor(!) --see the modern PC RAM modules) even cosmic radiation can cause random reversals of an I/O state, oftenly resulting in computer systems uncalled crashes...

Life can be a real bitch; and it seems that this is not only for us, the humans!


-George