High Frequency Oscillation in Amplifier circuit

[vorashrenik]

I am trying to use a dual MMIC matched amplifier to amplify a differential signal. Te schematic for the same is attached. The components used in the circuit are listed below:

Amplifier: MERA-556+ **broken link removed**

Transformers: TC1-1-13M **broken link removed**

RF Choke: TCCH-80+ **broken link removed**

The board seems to have a constant oscillation at about 2.6GHz with a strength of about 350mV. In the presence of an input, this strong noise signal rides on the input. The input frequency is between 200MHz-1GHz.

The layout of the board is also attached, I have used FR4 board(1.6mm) and the main trace width is 97mils.

Any ideas on how I can get rid of the high frequency oscillation?

 

[vorashrenik]

Does anyone have any ideas on this?

 

[FvM]

I see a problem, that except for the said "main trace" all connections are considerably above 50 ohm impedance. Also the ground layout is possibly unsuitable for 1.6 mm substrate. If you are not able to reduce the parasitic layout inductances, you should try to add series resistors and/or RC "snubbers" effective at 2-3 GHz.

Also reducing all interconnections and transmission lines to a minimum length possibly can help. Using 0.5 mm substrate (not necessarily Rogers) is most likely the best way.

 

[E-design]

I will start by trying to determine where the feedback problem occur. Lift TX2 and check if the signals out of the amps are clean. You can also check each amp without the transformers to see what happens at the output when you inject a single ended signal into each in turn. There are various ways to take it apart step by step to check for instability.

 

[FvM]

I think it's difficult to identify an instability cause by arbitrarily disconnecting circuit parts. The matching is completely changed in any case.

The specified amplifier S-parameters suggest, that the device should be stable with most source/load impedances. Thus I assume, that an additional ground inductance caused by the thick 1.6 mm substrate reduces the achieved isolation.

 

[vorashrenik]

@FvM,
What do you suggest I do? Redesign the board using the suggested rogers board? Also, I do now have small enough vias. The smallest vias I can do is 0.6mm. Is there a way to get around that?

 

[E-design]

It will not be that difficult to drive the input of each amplifier in turn with a controlled 50Ohm source and checking the output on a SA. It may be an issue with the transformer/connections that is causing the problem. At least he can see if it is stable at the lower or any frequencies. The FR4 (1.6mm) should not be that much of a problem at 1GHz @ 18dBm power level. I have designed quite a few boards using similar MMIC's (not Minicircuits) at 20dBm, 26dBm output running at 980MHZ using standard 1.6mm-FR4. I would do some more tests before just redesigning and condemning the board outright. My experience is that these type of amps are very forgiving and you must really be sloppy with layout to get them to go unstable. We use them for many applications, including oscillators (low flicker noise).

 

[keith1200rs]

Where are the decoupling capacitors?

Keith

 

[ucah]

Vorashrenik,
if your input signal is differential, you do not need the baluns at the inputs and outputs of your amplifier. You just need to feed/extract the signal pairs individually to each device port and that's that. You would use the TCP-1-13M baluns if you wanted to convert from balanced (differential) to unbalanced (single-ended) and vice-versa.
With regards to stability, you can evaluate the Rollet stability factor from the S-parameters of the device, and if the device is found potentially unstable or unstable, you can take measures to stabilize it at simulation level. This is effected by adding a resistive element at either (i/o) port of the device. If all is fine with the simulation and the prototype is unstable, the problem is usually down to poor grounding and/or decoupling, or other parasitics not taken into account at simulation time. Have you used s-parameters for those chokes at the output of the devices? I usually use simple self-resonant inductors of the 0805 or 1008 variety.
10nF of decoupling capacitor is certainly not effective at 2.5GHz. You need to go down to the tens of pF range.

 

[keith1200rs]

10nF of decoupling capacitor is certainly not effective at 2.5GHz. You need to go down to the tens of pF range.

and put them right next to the chip - they are so far away I didn't notice them.

Keith.

 

[FvM]

1.6 mm FR4 isn't a problem as such, neither the vias. The problem is in the parasitic ground inductance and TL impedances. I'm rather sure, that the amplifier can be stable operated on 1.6 mm FR4. E.g. additional ground pour around the signal traces can be used to adjust the impdedances.

 

[BigBoss]

Vias should be underneath of the amplifier, not far from.These amplifiers are sensitive to grounding connections, so grounding vias must be placed just underneath of the amplifier.
But in generally, your PCB layout is bad.The circuit should have "guided" transmission lines and ground connections must be everywhere possible.

 

[FvM]

Well, the mini-circuits reference design uses a similar via pattern as shown above and microstrip TL, too. But it uses 0.5 mm substrate, which makes the difference.

I guess, the manufacturer avoids to suggest vias in the enhanced pad, because they can cause problems in soldering. But they are better in any case and possibly sufficient to compensate for the 1.6 mm substrate. For a hand soldered prototype, they should always work.