How to reduce S21 of an amplifier at undesired frequency to improve the stability?

I often find that at some frequency(usually below the working band) S21 is still large(say above 15dB) while both S11 and S22 are poor(sometimes closed to 0dB). This obviously deteriorates the stability factor and makes the amplifier conditionally stable.

How can I reduce S21 at those undesired frequencies, or make the amplifier unconditionally stable?

Thank you.

i read about adding resistor at in or output may help, but also you get lower gain and noise. In real designs i never saw such resistors. Playing with smith chart and see stability circles, adding stubs. But practically i learned that more negative biasing for FET with common source can give a good result. I just used negative voltage generator IC and tuned resistor between GND and negative voltage until oscillation stopped

Apply feedback ( series or series+parallel) to reduce the Gain and improve the reflections.It will also improve the stability...

Its true that,although most text books talks of putting series or shunt resistors to achieve broadband stability,but I haven't come across one such practical circuit with such a stability circuit.
Terminator, are you tuning a resistor in series with gate connected to -ve bias voltage? Or you are tuning a shunt resistor to -ve supply voltage for stabilizing.



Big boss, whether this feedback technique will work at 12Ghz?
Can you please share,some practically implemented feedback circuit
at higher frequencies like 10-12Ghz?

I'm designing a cascode amplifier for 2.45GHz right now. At 2.45GHz, All S-parameters are fine: S11=-13.7dB, S22=-14.7dB, S21=17.7dB, S12=-26.8dB.

But at undesired frequency, like 1.2GHz, S21 is still so large(S21=21.2dB) and both S11 and S22 are poor(-1dB and -3dB). Why is S21 still so large? I don't get it.

For input/output matching network try to use a high-pass topology at the input, and low-pass topology at the output.
This might solve the problem of high (and unstable) gain at lower frequencies.

Five Common Methods for LNA Stabilization

Also, see "Stability" on page 8 in this appnote

I designed a frequency tripler :
I/p. 4Ghz
O/p 12Ghz
I checked the small signal stability in simulation,it was fine.
The practical circuit oscillates if terminated by any load other
than 50ohms,even a isolator at input won't help.

What can be done here? Practically how to improve upon the stability
Of this circuit?

Feedback is independent from the frequency, it's always applicable but there might also be some practical constraints.For instance I applied series feedback to a narrowband amplifier which was potentially unstable @5.8GHz, this feedback has obviously decreased the gain but improved reflections and stability.Even a small amount of resistor at Source/Emitter may play an important role on the performance of an amplifier.
I have never involved into design of feedback circuit @10-12GHz but I'm pretty sure it would work by applying a proper value.

Well, according to my limited knowledge, cascode is to improve stability by reducing the Miller capacitor, right? Should I have to use a RLC feedback in a cascode amplifier?

I know adding a resistor at the gate of the common-gate stage could improve the stability, I've already done that. That resistor is already 750Ohm now. Should I increase it?

afz23, i use voltage divider for gate biasing. In final design it can be two resistors. While tuning i use potentiometer. Using voltage divider i tune gate bias between 0v and -3v.

Also i read somewhere on internet about making hole in PCB under active device in power amplifier, maybe some page related to DB6NT stuff. Not sure whats it about.

How about adding out of phase feedback? I mean delay signal by 180deg. make a little tee from output and couple it to input. Not sure if it is correct.

From the second appnote:

"Stability compensation is usually implemented with RF feedback to filter excessive out−of−band gain. Resistive damping also can be applied. It makes sense to use lossy stabilization techniques only after filtering and feedback techniques are applied.
Resistance at the output is always preferable to resistance on the input. It also might be possible to use source (emitter) inductive loading to improve stability. This option, however, presents additional trade−offs in gain and noise figure that can be undesirable."

afz23, maybe it was fine only in simulation because your real grounding have problems? As written in appnote above.
Because actually adding inductance (small length of short-circuited microstrip) to base of BJT transistor or to common gate of FET is a recommendation for building an oscillator. If your substrate is thick then vias can give inductance that does not exists in simulation. Or if you have shielding enclosure it can reflect some signal back and it re-amplified. Also i have doubts about common-gate resistor torbai used. It can actually work as inductance too, maybe need to use a tiny one, 0603 or smaller.

Also some time ago someone mentioned on this forum, that using lossy substrate like FR4 can help to absorb unwanted feedback of amplifier.

And if nothing works maybe changing transistor applicable for lower frequency (lower gain) or with other s-parameters would be good.

It's a 0402 resistor.

what should we do if 0402 resistor pad width is much smaller than micrpstrip line width it need to be attached to?