The mysteries of RF amplifier stability...

An unconditionally stable transistor would be of no use other than as a door stop, (at least in any definition of unconditionally stable I know of). Perhaps you can define what you mean by "unconditionally stable transistor".

You can always take a transistor and turn it into an oscillator if you try to.

An unconditionally stable amplifier is a different beast all together. That is a collect of parts. In theory it will not oscillate no matter what the source and loads are, even if they are out of band. Even a short or open at DC. In practice, with microwave devices it is very hard to realise in practice.

A chance for a shameless plug to mention the yahoo rfamplifiers group.

Come on over and ask your questions...

cheers, skipp

: In sci.electronics.basics billcalley wrote: : Hi All,

: I'm having a very hard time understanding the full picture on : amplifier stability, even if I am an engineer. For instance, if we : select a transistor that is unconditionally stable at our frequency : band of interest -- let's say from 900MHz to 1000MHz -- but the : transistor is only conditionally stable at all other frequencies: So, : we bias, match, and resistively stabilize the transistor so that we : finally see, in the linear simulator, that K is greater than 1, and B1 : is greater than 0. This should indicate that we now know that the : amplifier will not oscillate under any input/output impedance : conditions. But what happens when you place a filter at the input or : output port of our newly stabilized amplifier? Since the stopbands of : the filter are anything *but* 50 ohms when the amplifier is looking : out-of-band, wouldn't there still be a chance that this : "unconditionally stable" amplifier could oscillate, since it is only : based on a conditionally stable transistor?

: Thanks!

: -Bill

The definition of "unconditionally stable amplifier" is that it will stay stable with any passive load attached -- so unless your filter presents an impedance with some negative resistance you're OK.

You can see this when you plot stability boundaries on a Smith chart -- the chart doesn't really end at the gamma = 1 circle, it continues on outward for reflection coefficients higher than one, which imply negative resistances. So if the stability boundary is entirely outside the usual chart it doesn't mean that _nothing_ will make the amplifier unstable, it just means that nothing _passive_ will make the amplifier unstable.

Thanks Tim and Dave! Great stuff! I'm a bit confused by Dave's answer, since I had strongly thought that there *was* such a thing as unconditionally stable transistors (stability as calculated by the Rollett Stability Factor 'K', and the Stability Measure 'B1'), and not just unstable amplifiers -- and that in-phase feedback and gain made the amplifiers into oscillators. I'm probably missing something here though...

-Bill

Yes, if the K & B1 conditions hold for _all_ frequencies, not only those you are interested in. Then it doesn't even matter if the transistor initially was unstable on the intended operation frequency.

It would not oscillate, no matter what the source or load impedance were. (Well, unless source or load exhibit negative return loss, but that would be unfair.) It would oscillate only if there was _additional_ external feedback around your amplifier core. That feedback might be invisible in your circuit diagram, like waveguide modes of your box or long vias thru thick epoxy..

regards, Gerhard

Hello Bill,

There aren't. This would require ratification of the UST act in congress.

SCNR, Joerg

Transistors have enough capacitance from collector to base (or drain to gate) that with the right inductances on the collector (drain) and base (gate) they will oscillate. Many overtone crystal oscillators are made this way, with the crystal providing inductive reactance to the base and an LC tuned circuit providing inductive reactance to the collector.

There are many, many circuits for which you just don't have to worry about a transistor bursting into oscillation -- perhaps this is why you thought they were unconditionally stable?

Thanks Gerhard and Tim for the clarifications. Much appreciated!!

-Bill

The capital K refers to the overall (device embedded in the circuit) stability, k is the Rollet stability factor. Even if k 1 at all frquencies by lossy embedding and or feedback. Having done this the overall circuit will be unconditionally stable for all non-feedback terminations with reflection coefficients no greater than1. Dick

Thanks Dick! I hadn't known that there was a K and a k. Makes sense though, and the books never seem to fully clarify this either.

-Bill