Power splitter - unbalanced outputs

[SG61-addict]

Hello,

I am using a minicircuits SCN-3-13+ power splitter/combiner (3 ways, 0°).

Before the splitter, I have a SMA connector, some LF filtering and a LNA. This input block is connected the the SUM port of the splitter.

Each output port of the splitter, is connected to a SAW filter and LNA, giving 3 amplified and filtered signals in 3 different frequency bands.

Simulating the whole design would give me approx. +23dB of gain for all 3 channels.

In the real life, I have a huge unbalance :
- 847MHz channel : +20dB
- 868MHz channel : +13dB
- 897MHz channel : +30dB

If I remove the splitter, measuring S-params of all pre and post block give me the expected values.

My conclusion is that the pre/post blocks composed of filtering and LNA are ok but the splitter don't works as planned. I am not a RF expert and not sure this device is compatible for this usage (split 3 low-power signals from an antenna).

I tried to removed the 3 required isolation resistors. Things are a little bit different but still wrong:
- 847MHz channel : +23dB
- 868MHz channel : +16dB
- 897MHz channel : +31dB

Which type of resistance am I supposed to use for the RF application? Current design is using thick-film resistor.

Any suggestion will help
Thank you

 

[FvM]

Simulating the whole design would give me approx. +23dB of gain for all 3 channels.

Does the simulation include SAW filter S-parameters over the full frequency range?

 

[SG61-addict]

Does the simulation include SAW filter S-parameters over the full frequency range?

Yes. I previously made a characterization board and measured each device with a VNA, 1MHz step from 1M to 6G. For all components... Excepted the splitter, for which I use the manufacturer s4p file.

 

[BigBoss]

Have you applied the recommended circuit for the Splitter ??

 

[SG61-addict]

Have you applied the recommended circuit for the Splitter ??

Yes !
But things are better without theses recommended resistors (my second set of measurements, first post).
R values checked... ok.
PCB layout is very similar to the recommended layout.