Using transistors in reverse

[nsled]

What is the negative to flipping a transistor to change its properties?

The only place I've seen it is here

 

at the 4:50 mark where collector and emitter are flipped but it's more a case I've "I've got this transistor and I'm damn well going to use it" in an expedient sense.

Has anyone felt a need, desire or mistakenly flipped a transistor? what generally is the result?

[TheUnnamedNewbie]

When you flip a 4-terminal MOSFET or 3-terminal JFET, you usually get the exact same device (in the case of a planar mosfet the source and drain are identical). BJTs usually have a 'reverse beta' in datasheets to illustrate the 'reverse operation' performance. It just becomes a shittier transistor, I don't know of any good reason why you would do it? But old-timers probably have loads of good stories where it just happened to work out they needed that transistor performance!

[Keith956]

The negative is that bipolar devices in general are not designed for reverse operation.

It's a bit like fitting the seat in your car facing backwards and driving in reverse.  Sure you can do it but it makes no sense.

[Kleinstein]

With a BJT the reverse gain is usually quite a bit lower lower, though the number is not very common in the datasheets.
The leakage goes up and the maximum voltage before breakdown it also lower (often only some 5-10 V).
There is mainly one positive effect: the CE saturation votlage is usually lower in reverse. This was used when lower saturation is needed. However more available MOSFETs are even better in that respect - so the trick is no longer needed.

In the circuit is the lower gain that in this circuit may be a slight advantage - though higher values resistors are also available. Some RF transistors also have relatively low gain to start with.

[Whales]

It's possible to use mosfets backwards too; however you need to keep Vds very low (so the body diode doesn't conduct).  They can still amplify

[kosine]

Reverse mode on BJTs is actually a whole lot more common than people realize - All the inputs for TTL chips are emitters. https://en.wikipedia.org/wiki/Transistor%E2%80%93transistor_logic

The gain is usually less than 10, and you can only apply 5-6V before reverse EB breakdown occurs, but it's a perfectly valid use for BJTs. Has a few advantages including a reduction in Miller effect.

I suspect the voltage limit is why 5V supplies became standard, You couldn't use much more for TTL circuits, and we're still "TTL compatible" to this day.

[magic]

Reverse mode is not used in any meaningful way in TTL. It only causes some current sinking from inputs that are driven high.

One example of reverse BJT actually being used on an IC is LM318 and Q5, which biases the "drivers" of its darlington input stage.

[T3sl4co1l]

Yeah, TTL is all NPN so there's not a lot of funny business going on.  (Maybe in the more analoggy types like timers?)

Actually, how do inputs pull up to the ~3V they do?  Or am I misremembering the OCV?  It should be 1.2V from the diagram, with additional current sunk at higher voltages due to inverted operation.  (Mainly, that input arrangement works more like a pair of diodes; I suppose coupling them together performs a bit better than ye olde DTL, but it's basically the same function.)

Reverse mode was used quite extensively in the olden days, lateral PNP for example.   Maybe that's arguable; can you really invert a symmetrical transistor?

Personally, I've used the mode for low leakage or saturation, or higher voltage range, like in these examples:



The middle line comes from a current-limited signal, and its positive value is clamped close to the top voltage.  It should work alright the other way around, but I found E-B leakage was objectionable (compared to a 1-10k ohm source).




This discrete synchronous rectifier works, a little slowly (100s ns), but does what it says on the tin.  Max rating is Vceo of the leftmost transistor (which should be matched to the other inverted type, of course; I suppose mismatched Vbc's might be interesting to play with, however).  Response time manifests as forward and reverse recovery (with forward recovery being limited by the FET's body diode, which, I've never seen noticeable forward recovery in MOSFETs, curiously enough; thus, the apparent forward recovery transient is limited to the body drop, until the channel begins conduction), and I think the feed-forward path makes reverse recovery depend on dV/dt (and via Ls, dI/dt in turn), analogous to the case for real diodes.

Tim

[David Hess]

As hfe increases the Early voltage decreases toward zero (mu is inversely proportional to beta), which is another way to say that the collector resistance decreases, so there might be some application for a reversed transistor where the highest possible collector resistance is required like for maximum voltage gain or lowest leakage in the off state.

I occasionally run across precision analog integrated circuits where this is done but the schematics rarely show the reversed emitter and collector.

Transistors with high hfe have higher voltage errors, and transistors with low hfe have higher current errors.

In the past they made special "chopper" bipolar transistors with a symmetrical emitter and collector.  So they had equal hfe with the collector and emitter swapped and equal base-collector and emitter-collector breakdown voltages.  Still available parts not long ago were the 2SC2878, 2SC3326, MMBT404, and CMPT404A.

[T3sl4co1l]

Interestingly, low-Vce(sat) transistors generally have quite high ihFE -- this is no accident, but a consequence of the saturation region being symmetrical around zero.  Curiously, these still have low Veb (5-7V), but ihFE is close to I think 50% of forward hFE under typical conditions.  Pretty neat.

Also curious, the SPICE models for them do not usually report ihFE correctly (BR), probably because they were tuned for forward parameters only.  And either the Gummel-Poon model misses something, or the parameter is more sensitive to other forward characteristics (the exact shape of the curves?) so it doesn't end up quite correct for that mode.

Tim