23cm Low Noise Amplifiers

I was quietly collecting parts to build a 23cm transceiver for local terrestrial use. I had a NE32684 and loading the S parameters into RFSim99 and using similar matching as other designs produced the following:

Circuit, Gain and Input return loss.

But I wasn't entirely happy with it. While existing designs had enjoyed great success, I was concerned about the stability. It's hard to convey in words but sweeping the frequency slider in RFSim99 showed that while the input was unconditionally stable, the potential existed for sub 1GHz oscillations depending on the load.

RFSim99 - about as easy as it gets to test. Here showing potential for instability with the output load.

Since I don't have a lot of 23cm test gear I decided to see how negative feedback would help. I was surprised to see that even with just a little resistive feedback the potential for oscillation disappeared and input matching with LC networks could be used. That gave me the following circuit:

This would not be optimum from a noise figure perspective, since we are both matching to 50 ohms instead of using a deliberate mis-match for lower noise figure, and we have used resistive feedback which generates noise.

However, on balance I was happy that my front end filter would see a reasonable termination at the image rejection frequency and this would prove to be suitable.

Then I chanced upon two cellular base stations: one at 900MHz and one at 2100MHz. Pulling htem apart and trawling the internet showed they had low noise front ends using ATF54143 and ATF58143 transistors. Both had comparable noise figures to the NE32684 at 23cm, but not as good at 10GHz.

So I decided to spin the wheel again (and again) to see what these transistors could deliver. With a bit of fiddling in RFSim99 I found that the ATF54143 gave me a useful front end RF amplifier:

Yes, it looks wrong.  With feedback between the gate and the capacitor you get quite a different response. What I have here is something with a good return loss and gain from LF to 23cm.  The ATF58143 was not vastly superior at first blush so I put that to one side.

Now I have something that works like a mmic. I know there are a couple of low noise mmics that have good return loss and better noise figure than this "theoretical" circuit, but I didn't have one to hand. So I fired up another software package to see what the noise figure might be. The simulation results were near identical, always a comfort, and the overall predicted noise figure is 1.09dB. I have yet to try to see if this could be improved.

It's time to build a board and add this part of the radio to the front end filter that is now in place. Nothing like building something to prove if the maths was correct.


Update:
I decided not to pursue this. Firstly, it looked wrong and when I checked it still had stability issues. Secondly, I found something better. A GPS LNA which I will write up soon. Board layout for that device is progressing as time permits. It looks like the best device yet.


73's