How to prevent the instability of high gain receiver?
we need to implement a very high amplification in our RF receiver at same
frequency. We have used a single-conversion superheterodyne receiver,
but at 400 MHz IF frequency we have needed to implement more than 60 dB of gain.
We have just selected a commercial IF amplifier and the idea is to cascade several ones to obtain the required gain.
I've heard that very high amplification at same frequency may lead to have instability problems due
to Barkhausen condition.
We would like to ask you the answers to the following questions:
1) How to understand when the instability occurs?
2) Are there some guidelines to prevent any instability problems?
3) Could you tell us a book which describes this problem extensively?
Many thanks
Best regards,
Antonio L.
If the LO works as an LO and the amplifier as an amplifier, there shouldn't be any problem. Instability in amplifiers itself is an ocean of a problem. Multiple levels of feedback, negative resistance areas introduced due to wrong feedback paths and so on... Did you try non-linear HB analysis or NDF analysis on amplifiers? Perhaps, STAN tool gives some hint where you have probable cases of oscillations.
Large signal stability analysis by A Suárez is an excellent book.
Hello BMR,
thank you for reply.
Our question is more related to the system design than circuit design. We have arleady selected the commercial components of receiver.
The problem is that we have no models of components with expection of the S-parameters and data reported on their datasheet.
We would like to understand if at system level a way exist to predict any potential oscillation, above all we will implement several cascaded amplifers in order to obtain the required gain.
Anyway, we will read the book you told us to gain more confidence about this issue.
are there other informations or advices you could tell us regarding the questions?
Best regards,
Antonio L.
You can use ADS to simulate the system and check the stabilities performance.
There are different performances, but you'd better check all of it, at least some combination.
Really? And what performances do we have to check?
We have a good experience in ADS, but we don't know this particular point of view.
Regards,
Antonio L.
The potential candidate blocks are the RF amplifier and the LO. What I do is, run HB analysis for these all these blocks indvidually and combine to see the system response in terms of BER, etc... small signal response also gives you some idea, but non-linear large signal analysis is necessary. There are templates available in ADS for these kind of simulations,but ofcourse little bit of tweaking has to be made.
"wrong feedback" paths and possible instability with high gain are a matter of insufficient shielding and power supply bypassing. A receiver designed according to state-of-the-art won't suffer from it.
Analysis tools are good to set gain of individual stages, optimize noise performance etc. I doubt that the parasitic effects causing instability will show up there.
Does this approach work even for only system design? i.e. we have only S-parameters of component obtained from manufacture and non-linear behaviour data reported in datasheet. Our question is: if we insert these data into ADS's behaviour models and run HB analysis, would we have some information about stability condition?
Regards,
Antonio L.
If you have full S parameters of the amplifier then it may be worthwhile to derive it's stability factor k.
But in practice when cascading multiple amplifiers you introduce more paths for feedback which won't be characterized in the device datasheet and are very difficult to estimate. They will be functions of things like physical spacing, shielding, cable routing, supply filtering, etc.
S parameters are small signal analysis and is limited by signal amplitude levels. HB analysis works in a way by introducing signal perturbation on critical nodes. Particularly summing nodes on feedback paths. You can only model so much and predict when it comes to instabilities. I did do some modelling on these but after the prototype had some instability issues.
Ok, we have decided to try with HB analysis.
Best regard,
Antonio L.
I presume that the commercial gain block is unconditionally stable, and has a gain of 20dB. If it's not, look for another one.
The problem with high gain stages is parasitic leakage of the big output signal to the input. Simply said, if the isolation between the output of the chain and the input is worse than 60dB you have a big chance of oscillations. In fact you want more, say 70dB minimum.
Now in fact 70dB of isolation is quite a challenge. So in fact I would recommend to use screening (cans, metal housing) to shield the signals from each other. Also be careful that the signal does not leak through the power supplies of the amplifiers. But that can adequately blocked when taken into account.
The HB balance analysis might learn you something, but it also might not. In fact the onset of instability is a linear problem. My advice: look for a reliable and stable gain block. That will save you many troubles along the way.
Stability analysis using with s-parameters will be valid only and only single stage amplifiers, therefore they cannot be used for whole system that includes many stages,mixers,filters etc.
In fact, this is a general design problem and any satisfactory answer couldn't be found until now.What is the stability of a whole system ?That's the question.
HB analysis will not be a solution using with small signal s-parameters and nonlinear component models because even separate singles stages seems stable, when these blocks are connected together the system has a probability of being unstable and this cannot be easily predicted.
There are some programs to show the complex poles and zeros of a LTI system but I don't know how they are effective.Also, Normalized Determinant Function ( NDF ) technique can be another metric of the whole stability.
if you are just stringing together wideband amps on a PC board to form 60 dB of gain at 400 MHz, yes you might have an oscillation problem. For one thing, try adding a bandpass filter somewhere along that 60 dB gain chain. That way an oscillation will not happen at any frequency, but only will be possible around 400 MHz. THEN play around with shielding and phase lengths to that the output of the IF chain is out of phase with the input of the IF chain.
Sorry but I disagree. A filter will most definitely have multiple poles, so also a phase variation of several 360 degrees. In case the isolation is insufficient, the filter will increase the likelyhood of oscillation instead of reducing it. In my opinion, make sure that the isolation is okay first and then commit to the filtering.