MAR-6 MMIC amplifier chain saturation?

[Ravaner]

Hello. I have devolopped a line amplifier with 3 x MAR-6 MMIC. Normal daisy chain is : antenna followed by a LNA which is followed by my line-amp. If nothing is connected at the input ofLNA, all 3 Mar-6 have normal operating voltage ie 3.5 V and resulting noise is very low (seen with a 100 MHz scope). Since horn antenna is connected, and even if just a 1/4 wavelenth wire (51 mm) is connected at the input of LNA, the 3rd Mar-6 has is DC voltage drop to 1.8 V ! As it works far from 1 dB compression point, how this can be explained ?

 

[Baluncore]

My guess would be that you are getting coupling from the #3 MAR-6 output back to the 2” stub you connect as the input. It could be coupling partly through the power supplies or the nearby chassis. The input may need to be terminated in 50 ohm across the band in order to remain stable. How do you screen the stages from each other, and how do you isolate the power supply to each MAR-6 from the others?
Have you checked the frequency of the output? Is it stable or does it change when you capacitively load the circuit with a finger to find if there is a sensitive point in the feedback loop.

Do you have a link to the circuit and layout used?

 

[Ravaner]

Thanks for reply. Schematics is very simple 4 x 10 pF separated by 3 x Mar-6. The "line" supporting all these components is a 50 ohms stripline. Output signal looks like noise. No significant frequency detected with my scope. If circuit is oscillating at frequency over 100 MHz I can't display it. There is no screen between each stage.

 

[Baluncore]

You have the LNA and then more than 60dB gain from the MAR-6s in that line. You need at least 60dB of isolation and attenuation in the power supply line that feeds current to the individual MAR-6 outputs. There will usually be more components in the power supply than in the RF signal chain. Do you have a complete ground plane on the bottom of the PCB? Is it stitched with many vias or links to a ground plane for the MAR-6 grounds and power supply bypass?
A picture is worth 1000 words.

 

[Ravaner]

Total ground plane under RF circuit. Vias for grounding each MAR-6

and also very close to input and output SMA connectors. By-Pass for power supply is 2.2 nF.

I hope image properly attached ...

 

[Baluncore]

You need many large diameter vias to bond the ground plane to the signal side peripheral ground track. The same goes for the MAR-6 ground pins, where a local ground plane either side of the MAR-6 on the signal side of PCB should have 6 or more spaced vias to reduce the ground inductance. Remember that many separated parallel paths make low inductance. Large diameter vias or through pins also make low inductance. Sometimes narrow slots are cut in the PCB and a shim is inserted and soldered on both sides.

Are the coaxial connector ground connections through hole pins that bond to ground plane below or only surface mount to outer track?

I would wrap shim round the edge and solder, or place many more vias around that peripheral ground track, it looks more like a 20 ohm transmission line than a ground.

Introduce the external power at the output end of the strip and put the big cap there also. The +ve supply distribution line looks like a 50 ohm track. Cut it between MMICs and put in 10 ohm series resistors.

Those RFCs might couple. Can you turn the input end RFC sideways to reduce coupling?

Can you test the board for oscillation with a 50 ohm termination at each end connector?

 

[Ravaner]

Thanks for this really complete answer. May be just one more precision : Output voltage of each Mar-6 is Ok if nothing is connected at the input. And also no oscillation ... Problem occurs only if 'something', even a short wire, is connected to input. May be this works as a trigger for oscillation.

 

[Baluncore]

May be this works as a trigger for oscillation.

It is almost impossible to tame a poorly laid out wideband MMIC chain.

You should be able to chain together three stages, terminate the input with a 50 ohm resistor, then see the broadband Johnson noise from the resistor raised by 60 dB on an analyser, all without in-band oscillation. From that you should be able to calculate the temperature of the resistor.

Two inches of wire may be all it takes to highlight inductance between the surface and the ground plane.
Two inches of wire may be all it takes to pick up your mobile phone or a blue tooth device.