Cable Length Problem

[Squirtle]

Hello guys,

I am trying to measure the gain of the LNA with a BPF using VNA. Below you see two block diagrams indicated as figure 1 and figure 2. In figure 1 band pass filter is in series with LNA and it is connected to LNA with arbitrary length of coax. In this diagram the length of the coax between BPF and LNA does not affect the gain of the LNA, we measure the gain as it is expected. However in block diagram of figure 2, the length of the coax between LNA and BPF does affect the gain of the amplifier and it is very annoying trying to find appropriate length of the coax to get the max gain that we obtained in figure 1 without any problem. I know for sure that the problem is with LNA because some other LNA types don't have this problem. Whatever I do, I could not figure out what the problem might be? Any help would be very appreciated.

 

[BradtheRad]

Is your filter made from resistors, capacitors, inductors? For best performance it needs to be designed to work with available Ampere level and voltage amplitude. This includes RC,LR, and LC ratios.

This means you cannot reverse the positions of the same BPF and the same LNA.

 

[vfone]

Usually, passive circuits as your BPF, are matched to 50 ohms at their ports.
However, active circuits as LNA's, even the designers want to match their input/output ports to 50 ohms, not always get there.
The LNA in your case may have 50 ohms input impedance (working fine with the BPF at its input), but its output impedance may be different than 50 ohms, and when connect the BPF at the output, results lower overall gain.
BPF filters also, if doesn't see close to 50 ohms at their ports, their insertion loss and ripple performances may be affected.

 

[Squirtle]

I am using combline filter and i did not get the ratios you mentioned. One interesting thing about the design, when i touch to the rf output of the LNA on PCB with my finger, LNA jumps to max gain but i cannot cut my finger and leave it on PCB

--- Updated Mar 2, 2022 ---

But the case is when i use some other LNAs which they do not have 50 ohm input and output impedance either, they work fine, not affected by cable length.

 

[ferdem]

Your LNA is able to deliver expected gain when it is terminated with an appropriate impedance (50 ohm, NA connected case).
What about input VSWR of your filter? It looks the VSWR there is not good.

The LNA starts to behave weird when the filter is connected. This may be caused by two things; the input impedance of the filter or LNA's over sensitivity to load pulling. I think the former is more likely.

You can observe that putting an 3 dB attenuator between LNA and filter will reduce the mismatch effect and the cable length will not cause an unexpected change in the gain.

 

[Squirtle]

I tried your recommendation, when i put 3 db or 6 db attenuator it works just fine and the length of the cable does not affect the gain but how can do that without attenuator and why is it working with attenuator?

 

[volker@muehlhaus]

You need to improve the filter input matching. Only if the filter is matched to 50 Ohm in the passband, you will see 50 Ohm at the line input regardless of line length.

 

[vfone]

As I mentioned, the output impedance of your LNA may not be 50 ohms, and a 3dB pad may help improving this situation. As you said, when touching the output of the LNA get higher gain, prove my theory.
I think the only think that can help you (other than a 3dB pad) is to re-match the LNA output for the required impedance of the BPF.

 

[FvM]

As vfone explained, LNA output and filter input impedance are apparently both different from 50 ohm. A VNA measurement of both components can clarify.

 

[volker@muehlhaus]

For this ripple, I think the more critical matching is at the filter input. If that is matched to the line impedance, there is no line-length-dependent ripple, even if the LNA output impedance is different from 50 Ohm.

 

[Squirtle]

I want to thank you all for your answers. I have one more question about S parameters and it is also related to matching topic. When i look at the S-parameter of amplifiers something caught my attention. As the return loss(S11) get worse outside of the amplifier's working frequency band, the gain gets better. We know that output power is equal to input power+S21(gain). It seems to me that there is a contradiction here because output power increases as the return loss get worse. So, what should i understand from this case?

 

[biff44]

have you tried to make the measurement with a signal generator and a spectrum analyzer.
the fact that the gain changes wildly if you put your finger near the amp might indicate the amp is oscillating.

 

[Squirtle]

i tried the setup using signal generator and spectrum analyzer as you said. I am using PSA4-5043+ amplifier at 10 MHz and 400 MHz where S11=-2.23 dB and S11=-11.63 dB respectively. No dramatic change at spectrum analyzer. I did not try whether it is oscillating or not.