Hi. I've read some threads about stability of various amplifiers. But I am still looking for a good answer that is more in line with my problem.

I am designing a radiometer RF receiver using commercially available surface mount components.

To reach sufficient gain, I use three low-noise amps as the noise figure must be kept low. Now, an earlier experiment showed that if you cascade all three LNA's in close succession, the system will begin to oscillate within the passband. There was not done any stability calculations to verify if they would be stable.

So, my task has been to re-design for stability. Without losing gain, and without increasing noise factor. I've read for ages about stability, but I still do not understand all of it. I have no experience in designing amplifiers. In any case, since we use pre-fabricated amps, I do not know the topology. So to make any stability calculations I can only use the provided S-parameters as far as I know.

The only software I have available is CST Microwave Studio. If I simply do a S-parameter analysis, there are no signs of instability. I assume I can use the S-parameters for this device to calculate stability of a single LNA. This tells me (by both rollets and mu-stability factor) that the LNA is unconditionally stable.

From what I gather, I cannot compute the same parameters for all LNA's in cascade. Is this correct? I HAVE done this, just to check, and for a cascade of three such LNAs, they are still unconditionally stable. However, I know from experience, this is not true. What could influence the cascade to become unstable?

Without doing too much thinking, I ended up using an isolator to de-couple the first two LNA's and the last. I tried implementing this, and from all experiments, the system is now stable. So it worked. But I do not really know how to theoretically support it.

If anyone could share any throughts on my questions it would be highly appreciated. As of know, the only thing I can think of in my naivity, is that the isolator does two things:

1) It prevents the highly amplified signal to be reflected back and forth between the last two LNA's. This may prevent oscillation.

2) In addition, the distance between the devices is increased such that they do not influence each other through coupling of their electromagnetic fields.

Thank you. The only problem is, I do not think have the necessary means to do what they talk about the S-probe article.

But just to calrify:

1) Is it so, that rollets stability condition and its equivalents, are only valid for considering the stability of ONE active element, meaning only one amp ?

2) I have CST Microwave Studio at my disposal. In which I can simulate the total S-parameters of for example all three LNAs. Can I use this for more advanced stability analysis ?

Currently, I have tried, just for the sake of it. To compute stability parameters of the entire cascade, based on simulated S-parameters. This does not give me an indication that the cascade will oscillate.

3) Repeating from the first post. The problem is now solved, I inserted an isolator that operates in the passband of the LNAs. As two LNAs in cascade are stable, I thought, by de-coupling the last LNA, this could work. And it did. But the problem of theoretically supporting this idea still stands, as I am unable to present any simulated of calculated results to show the system will be stable.

Hehe.. I just don't know. Amplifiers and stability is not really my field. This is why I have some problems understanding everything about stability, consequently I also have some problems explaining my difficulties.

Is it possible that the only thing that was needed, was to increase the electrical distance between the amplifiers to change the phase output?

This new design also resulted in using separate power supply lines, and I guess this would have had a positive effect also?  

EDIT:

I see you edited you post while I was typing this. So sorry for any repeated questions.

Tjuby wrote on Oct 10^th, 2010, 4:40am:
See http://www.designers-guide.org/Forum/YaBB.pl?num=1240334878/18#18

Tjuby wrote on Oct 10^th, 2010, 4:40am:
Tjuby wrote on Oct 10^th, 2010, 4:40am:
If a cause of unstability is due to amplifier's coupling via power supply line or ground,
you can not take such effects into your simulation,
since you treat each amplifier as two-port S-parameter block.

Can you characterize each amplifier as three-port S-parameter block with power supply as 3rd port.

pancho_hideboo wrote on Oct 10^th, 2010, 4:09am:
Sorry, I was wrong.
If you increase the electrical distance between the amplifiers, phase rotation at output will large.
This could make cascaded system more unstable.

I mean reducing feedback from output to input by increasing the electrical distance between input and output of cascaded system.
For example, don't place three amplifiers in "U" figure or "L" figure.

Tjuby wrote on Oct 10^th, 2010, 4:40am:
Do you mean a physical size of "Isolator" ?

Tjuby wrote on Oct 10^th, 2010, 4:40am:
No.