[SOLVED] 60dB multi-stage LNA

[trozio]

Hi all,
I have managed to obtain a stable 60dB gain, multistage LNA at 1.5GHz (S11=-23dB, S22=-12dB, NF=1.3dB).
The problem comes when I put the whole circuit inside a through hole, PCB mount shielding.
In-band oscillations arise as S12 gets higher. Even by removing the covering lid, the amplifier appears not to be stable anymore.
I am starting to wonder if it is physically possible to obtain such a high gain in one single enclosure.
Any way to get over this problem?
Thanks

 

[BigBoss]

The Power Gain is pretty high and interstage couplings can create feedback and obviously be cause oscillations.
Have you used shielding walls between the stages ?? How about GND paths and VIA configuration ??
Are you able to put your PCB and schematic here ??

 

[trozio]

The Power Gain is pretty high and interstage couplings can create feedback and obviously be cause oscillations.
Have you used shielding walls between the stages ?? How about GND paths and VIA configuration ??
Are you able to put your PCB and schematic here ??

Thanks for the feedback.
Take into account that the circuit works without shielding, so I suppose that GND paths and VIAs have been well designed.
However, I am curious about the shielding walls point. This means that I should separate each stage and link them with through a top-bottom-top path, right?
I am sorry but I cannot upload any schematic or PCB layout.

 

[BigBoss]

Thanks for the feedback.
Take into account that the circuit works without shielding, so I suppose that GND paths and VIAs have been well designed.
However, I am curious about the shielding walls point. This means that I should separate each stage and link them with through a top-bottom-top path, right?
I am sorry but I cannot upload any schematic or PCB layout.

Shileds sometimes can cretate some EM feedbacks from output to input since they have not been well designed.60dB Power Gain is a huge gain for an amplifier chain and a small amount of EM diffraction can reach to input with a same phase and the circuit may oscillate.Therefore stages should be seperated by shield walls.

 

[ge]

I'm with BigBoss on this. Put some absorber walls as a proto. Metal will have less loss but this will tell you if it's feedback. Trozio is making a good point of having the shield very well attached to your ground plane. 60dB is a lot of gain.

 

[Terminator3]

There was a topic recently, where author cutted LNA PCB to separate pieces, so each stage was on smaller PCB, separated stages by walls, and connected only by small area. Then his problem was solved, there was some leakage through PCB itself.

 

[biff44]

my rule is to never put more than 30 db of gain under one metal shield. You need to split it up into two, or maybe 3 shield covers. Also u need lowpass filtering in the bias lines

 

[E-design]

We once did a design for a 9kHz - 2.9GHz tracking gen module for a spectrum analyzer. We had seven stages each in its own cavity giving a total of around 75dB gain. Everything worked like a charm with the agc keeping it flat over this range. The problems started when we tried fitted the shielding covers. It went unstable, and it took a lot of work trying absorbers and other tricks to get it stable. When dealing with high gain as mentioned can be a nightmare.

 

[trozio]

Thanks to all.

I have managed to obtain a stable 64dB gain LNA with 4 discrete transistor amplifiers plus two bandpass ceramic filters @ 1.5GHz.
Following BigBoss' wise advice, I have inserted an internal wall divider in order to split the first two stages from the other ones.
The principle has been validated with this configuration but for the next version I am going to separate each amplifying stage.