Issues with SWR/Return Loss of Splitter in FDM System

Discussion in 'Wireless & RF Design' started by Management, Jul 25, 2009.

  1. Management

    Thread Starter Active Member

    Sep 18, 2007
    So let me describe by setup:

    Coax Line carrying 8 signals in the range of 0.8 GHz to 2.2 GHz. Each spaced by 200 MHz.

    This coax goes into a 1-to-8 Splitter. On the output (8 outputs) of the splitter there are 8 BPFs. Each filter has a different center frequency of it's particular channel.

    Then there are 8 amplifiers.

    Then each goes into its own mixer with corresponding LOs for down conversion to IF, etc.

    My issue is that when I take the 1-to-8 splitter and terminate all 8 outputs with 50 terminations then I get really good return loss and corresponding swr at the input (S11). BUT when I put on the BPFs and terminate those outputs at 50 Ohms then my S11 (return loss) looking into the splitter again goes into the toliet.

    Why is that and how can I rectify this situation if I want to use a splitter?

    Secondly, if I can't use a splitter is a mux the only way? I would rather try to make the splitter work though.

    I can post plots if need be.

    Thanks for the help.
  2. Management

    Thread Starter Active Member

    Sep 18, 2007
    any takers?
  3. Tesla23

    Senior Member

    May 10, 2009
    The problem is that the way that most passive filters work is that they either pass the signal to the load or they reflect it to the source. So what is happening is that when you send a signal at F1 into the splitter, 1/8th of it goes to the F1 filter and to its load, but the other 7/8th is sent to the other filters, which don't pass it, simply reflecting it, back to the splitter and (depending on the phasing and splitter losses) mostly back to the source.

    Your choices:
    1. put up with it
    2. add some attenuation before the filters - each 3dB you add should improve your return loss by 6dB
    3. design a diplexer (well multiplexer actually)

    A diplexer will send (ideally) all the power at F1 to the F1 load etc. To get the idea of a diplexer imagine that you designed your filters so that they are open-circuit in the stop-bands, then you can simply parallel them and not use a splitter. It's a bit more complex than this in practice, but that's the idea.
  4. Management

    Thread Starter Active Member

    Sep 18, 2007
    Thank you so much for you incite on this issue.

    I understand now. So I guess a mux is gonna be the only way to get some serious performance increases.

    Where do you get that 6 dB improvement in return loss number from? Is that just what one nominally sees when you add 3 dB of attenuation?

  5. Management

    Thread Starter Active Member

    Sep 18, 2007
    Any advice on the design of a mux for an application such as this?
  6. rjenkins

    AAC Fanatic!

    Nov 6, 2005
    As each filter input is (nominally) 50 Ohms at it's pass frequency and apparently way off at the other spot frequencies, you may actually get better results with a direct 'star' connection rather than the 50 Ohm splitter.

    Use the absolute shortest possible connections, ideally a 'tree' of T-connectors with no cable or at least if the connection length is near 1/4 or 1/2 wave on any channel use that on it's matched filter.

    Whatever you use for splitting/muxing, if there is any common earth connection - even capacitive - after that point, use ferrite sleeves on the link cables in & out of the filters or it will *not* work as expected - that is a possibility for problems with your initial tests..

    Coax only works 'as described' in isolation. When you get multiple paths like this & grounded equipment, the RF current will be everywhere unless you use ferrite to force the current balance within the coax.
    (Speaking from experience running a Ham radio BBS with seven radios inc. some full duplex).