Something like a diplexer

[Joel Kolstad]

I'm trying to build a circuit to split two signal paths apart, process them
differently, and them put them back together. In particular, I have
incoming RF signals from ~25MHz-3GHz (they're going to a wideband receiver)
and I have some notch filters that I want to be able to selectively engage
between 25-500MHz.Due to the topology and parasitics of the filters, they
tend to roll off above 500MHz and have significant loss by the time you get
to 3GHz.

Hence, I'd like to run 25-500MHz though one signal path and 500MHz-3GHz
through another (I want to leave the 500MHz-3GHz part untouched). The
'cross over' region doesn't have to be particular 'clean' (i.e., it can vary
+/-5dB easily).

I tried desgining a 5th order diplexer, and it cleanly splits the two
frequency ranges into separate path going to their own terminators.
However, if instead of terminating the low pass output and high pass outputs
to their own loads I connect the two outputs together, I get something of a
mess -- not at all a 'flat line' like I was hoping for.

I next tried hooking up two of these diplexers 'end to end,' and while the
response is almost flat, it has a very sharp null right at the 500MHz corner
frequency, and another very sharp set of nulls a couple hundred MHz above
and below this (they're mirror images). Hmm...

So... any hints how to do this properly? I thought for certain the end to
end diplexers would have done the trick.

Thanks,
---Joel Kolstad

[B.Viel]

To split a signal you can use a direction coupler.
I have seen "hybrid" ring models made of transmission line with specific
length in UKW-unterlage A4 typed edition 1980's.
To inject a signal you can use a direction coupler too....
>and them put them back together.
Maybe Google for Dubus artikels.
I'am just giving a hint.
A piece of thick coax, braid a wire for about 1/4 wavelength under the
outer layer, remove plastic cover first.
Connect one end to a resistor 50 Ohm (same as the transmission line
impedance) other end of the resistor to earth.
The other end of the wire to get signal from or put signal in.
Idea from the book antennenbibel Karl Rothammel.


"Joel Kolstad" > schreef in bericht
...
> I'm trying to build a circuit to split two signal paths apart, process
them
> differently, and them put them back together. In particular, I have
> incoming RF signals from ~25MHz-3GHz (they're going to a wideband
receiver)
> and I have some notch filters that I want to be able to selectively engage
> between 25-500MHz.Due to the topology and parasitics of the filters, they
> tend to roll off above 500MHz and have significant loss by the time you
get
> to 3GHz.
>
> Hence, I'd like to run 25-500MHz though one signal path and 500MHz-3GHz
> through another (I want to leave the 500MHz-3GHz part untouched). The
> 'cross over' region doesn't have to be particular 'clean' (i.e., it can
vary
> +/-5dB easily).
>
> I tried desgining a 5th order diplexer, and it cleanly splits the two
> frequency ranges into separate path going to their own terminators.
> However, if instead of terminating the low pass output and high pass
outputs
> to their own loads I connect the two outputs together, I get something of
a
> mess -- not at all a 'flat line' like I was hoping for.
>
> I next tried hooking up two of these diplexers 'end to end,' and while the
> response is almost flat, it has a very sharp null right at the 500MHz
corner
> frequency, and another very sharp set of nulls a couple hundred MHz above
> and below this (they're mirror images). Hmm...
>
> So... any hints how to do this properly? I thought for certain the end to
> end diplexers would have done the trick.
>
> Thanks,
> ---Joel Kolstad
>
>

[Asimov]

"Joel Kolstad" bravely wrote to "All" (14 Feb 05 16:51:45)
--- on the heady topic of "Something like a diplexer"

JK> From: "Joel Kolstad" >
JK> Subject: Something like a diplexer
JK> Xref: aeinews rec.radio.amateur.homebrew:2136

JK> I'm trying to build a circuit to split two signal paths apart, process
JK> them differently, and them put them back together. In particular, I
JK> have incoming RF signals from ~25MHz-3GHz (they're going to a wideband
JK> receiver) and I have some notch filters that I want to be able to
JK> selectively engage between 25-500MHz.Due to the topology and
JK> parasitics of the filters, they tend to roll off above 500MHz and have
JK> significant loss by the time you get to 3GHz.

JK> Hence, I'd like to run 25-500MHz though one signal path and
JK> 500MHz-3GHz through another (I want to leave the 500MHz-3GHz part
JK> untouched). The 'cross over' region doesn't have to be particular
JK> 'clean' (i.e., it can vary +/-5dB easily).

JK> I tried desgining a 5th order diplexer, and it cleanly splits the two
JK> frequency ranges into separate path going to their own terminators.
JK> However, if instead of terminating the low pass output and high pass
JK> outputs to their own loads I connect the two outputs together, I get
JK> something of a mess -- not at all a 'flat line' like I was hoping for.

JK> I next tried hooking up two of these diplexers 'end to end,' and while
JK> the response is almost flat, it has a very sharp null right at the
JK> 500MHz corner frequency, and another very sharp set of nulls a couple
JK> hundred MHz above and below this (they're mirror images). Hmm...

JK> So... any hints how to do this properly? I thought for certain the
JK> end to end diplexers would have done the trick.

JK> Thanks,
JK> ---Joel Kolstad

You missed a step called a directional coupler. It's a sort of
transformer with 2 input ports and 1 output port. The 2 inputs don't
see one another but their power is combined at the output.

A*s*i*m*o*v

.... We're young, rich, and full of sugar, what do we do?

[budgie]

On Monday, 14 Feb 2005 23:53:26 -500, "Asimov"
> wrote:

>"Joel Kolstad" bravely wrote to "All" (14 Feb 05 16:51:45)
> --- on the heady topic of "Something like a diplexer"
>
> JK> From: "Joel Kolstad" >
> JK> Subject: Something like a diplexer
> JK> Xref: aeinews rec.radio.amateur.homebrew:2136
>
> JK> I'm trying to build a circuit to split two signal paths apart, process
> JK> them differently, and them put them back together. In particular, I
> JK> have incoming RF signals from ~25MHz-3GHz (they're going to a wideband
> JK> receiver) and I have some notch filters that I want to be able to
> JK> selectively engage between 25-500MHz.Due to the topology and
> JK> parasitics of the filters, they tend to roll off above 500MHz and have
> JK> significant loss by the time you get to 3GHz.
>
> JK> Hence, I'd like to run 25-500MHz though one signal path and
> JK> 500MHz-3GHz through another (I want to leave the 500MHz-3GHz part
> JK> untouched). The 'cross over' region doesn't have to be particular
> JK> 'clean' (i.e., it can vary +/-5dB easily).
>
> JK> I tried desgining a 5th order diplexer, and it cleanly splits the two
> JK> frequency ranges into separate path going to their own terminators.
> JK> However, if instead of terminating the low pass output and high pass
> JK> outputs to their own loads I connect the two outputs together, I get
> JK> something of a mess -- not at all a 'flat line' like I was hoping for.
>
> JK> I next tried hooking up two of these diplexers 'end to end,' and while
> JK> the response is almost flat, it has a very sharp null right at the
> JK> 500MHz corner frequency, and another very sharp set of nulls a couple
> JK> hundred MHz above and below this (they're mirror images). Hmm...
>
> JK> So... any hints how to do this properly? I thought for certain the
> JK> end to end diplexers would have done the trick.
>
> JK> Thanks,
> JK> ---Joel Kolstad
>
>You missed a step called a directional coupler. It's a sort of
>transformer with 2 input ports and 1 output port. The 2 inputs don't
>see one another but their power is combined at the output.

You don't actually *need* a directional coupler. I have seen window
preselectors with five separate sections for sub-bands between 403 and 520 MHz.
The configuration is symmetrical (in/out) with simple bandpass filter segments
and coaxial split/combine harnesses. Not a DC or hybrid in sight. I am
presuming - not having swept one - that at "off" frequencies each parallelled
leg presents a high enough impedance to the split junction that the effect is
negligible.

[Asimov]

"budgie" bravely wrote to "All" (16 Feb 05 12:52:14)
--- on the heady topic of "Re: Something like a diplexer"

bu> From: budgie >
bu> Xref: aeinews rec.radio.amateur.homebrew:2181

>You missed a step called a directional coupler. It's a sort of
>transformer with 2 input ports and 1 output port. The 2 inputs don't
>see one another but their power is combined at the output.

bu> You don't actually *need* a directional coupler. I have seen window
bu> preselectors with five separate sections for sub-bands between 403 and
bu> 520 MHz. The configuration is symmetrical (in/out) with simple bandpass
bu> filter segments and coaxial split/combine harnesses. Not a DC or
bu> hybrid in sight. I am presuming - not having swept one - that at "off"
bu> frequencies each parallelled leg presents a high enough impedance to
bu> the split junction that the effect is negligible.

Yes but it requires some effort and cost to build the 2nd BP filters so
I wonder if it's any more difficult to do either? What do you figure?

A*s*i*m*o*v

.... Always look on the bright side of life...[Monty Python]

[budgie]

On Wednesday, 16 Feb 2005 09:42:58 -500, "Asimov"
> wrote:

>"budgie" bravely wrote to "All" (16 Feb 05 12:52:14)
> --- on the heady topic of "Re: Something like a diplexer"
>
> bu> From: budgie >
> bu> Xref: aeinews rec.radio.amateur.homebrew:2181
>
> >You missed a step called a directional coupler. It's a sort of
> >transformer with 2 input ports and 1 output port. The 2 inputs don't
> >see one another but their power is combined at the output.
>
> bu> You don't actually *need* a directional coupler. I have seen window
> bu> preselectors with five separate sections for sub-bands between 403 and
> bu> 520 MHz. The configuration is symmetrical (in/out) with simple bandpass
> bu> filter segments and coaxial split/combine harnesses. Not a DC or
> bu> hybrid in sight. I am presuming - not having swept one - that at "off"
> bu> frequencies each parallelled leg presents a high enough impedance to
> bu> the split junction that the effect is negligible.
>
>Yes but it requires some effort and cost to build the 2nd BP filters so
>I wonder if it's any more difficult to do either? What do you figure?

I'm not sure we are on the same wavelength here (no pun intended). For
*non-overlapping* filter sections, the bandsplit through two bandpass sections
25-500 and say 550-3000 should be able to be achieved with T-pieces at the input
and output ends (although it appears the O/P didn't achieve this).

[J M Noeding]

On Thu, 17 Feb 2005 17:35:11 +0800, budgie > wrote:

>On Wednesday, 16 Feb 2005 09:42:58 -500, "Asimov"
> wrote:
>
>>"budgie" bravely wrote to "All" (16 Feb 05 12:52:14)
>> --- on the heady topic of "Re: Something like a diplexer"
>>
>> bu> From: budgie >
>> bu> Xref: aeinews rec.radio.amateur.homebrew:2181
>>
>> >You missed a step called a directional coupler. It's a sort of
>> >transformer with 2 input ports and 1 output port. The 2 inputs don't
>> >see one another but their power is combined at the output.
>>
>> bu> You don't actually *need* a directional coupler. I have seen window
>> bu> preselectors with five separate sections for sub-bands between 403 and
>> bu> 520 MHz. The configuration is symmetrical (in/out) with simple bandpass
>> bu> filter segments and coaxial split/combine harnesses. Not a DC or
>> bu> hybrid in sight. I am presuming - not having swept one - that at "off"
>> bu> frequencies each parallelled leg presents a high enough impedance to
>> bu> the split junction that the effect is negligible.
>>
>>Yes but it requires some effort and cost to build the 2nd BP filters so
>>I wonder if it's any more difficult to do either? What do you figure?
>
>I'm not sure we are on the same wavelength here (no pun intended). For
>*non-overlapping* filter sections, the bandsplit through two bandpass sections
>25-500 and say 550-3000 should be able to be achieved with T-pieces at the input
>and output ends (although it appears the O/P didn't achieve this).

it looks like different wavelength to me, but I believe it is
described in ARRL handbook how to build such combiners

73, LA8AK
---
J. M. Noeding, LA8AK, N-4623 Kristiansand
http://home.online.no/~la8ak/c.htm

[J M Noeding]

If it was supposed to be for the same wavelength is another
interesting construction shown at
http://my.athenet.net/~multiplx/cgi-bin/wilk.main.cgi


---
J. M. Noeding, LA8AK, N-4623 Kristiansand
http://home.online.no/~la8ak/c.htm

[Gary Schafer]

On Thu, 17 Feb 2005 17:35:11 +0800, budgie > wrote:

>On Wednesday, 16 Feb 2005 09:42:58 -500, "Asimov"
> wrote:
>
>>"budgie" bravely wrote to "All" (16 Feb 05 12:52:14)
>> --- on the heady topic of "Re: Something like a diplexer"
>>
>> bu> From: budgie >
>> bu> Xref: aeinews rec.radio.amateur.homebrew:2181
>>
>> >You missed a step called a directional coupler. It's a sort of
>> >transformer with 2 input ports and 1 output port. The 2 inputs don't
>> >see one another but their power is combined at the output.
>>
>> bu> You don't actually *need* a directional coupler. I have seen window
>> bu> preselectors with five separate sections for sub-bands between 403 and
>> bu> 520 MHz. The configuration is symmetrical (in/out) with simple bandpass
>> bu> filter segments and coaxial split/combine harnesses. Not a DC or
>> bu> hybrid in sight. I am presuming - not having swept one - that at "off"
>> bu> frequencies each parallelled leg presents a high enough impedance to
>> bu> the split junction that the effect is negligible.
>>
>>Yes but it requires some effort and cost to build the 2nd BP filters so
>>I wonder if it's any more difficult to do either? What do you figure?
>
>I'm not sure we are on the same wavelength here (no pun intended). For
>*non-overlapping* filter sections, the bandsplit through two bandpass sections
>25-500 and say 550-3000 should be able to be achieved with T-pieces at the input
>and output ends (although it appears the O/P didn't achieve this).

I am not sure exactly what is trying to be accomplished here but if it
is to combine two different bandpass filters then yes a T. The cable
from the low pass filter to the T should be a quarter wave length at
the frequency of the high pass filter.

The cable from the high pass filter to the T should then be a quarter
wave length of the low pass filter.

The low pass filter will look like a short circuit at the high pass
frequency. With the quarter wave length cable that short circuit will
then be transformed to a very high impedance to the hi pass circuit.
That effectively isolates one from the other.

Do the another T and cable setup at the input and output.

73
Gary K4FMX

[budgie]

On Thu, 17 Feb 2005 21:36:05 -0500, Gary Schafer > wrote:

>On Thu, 17 Feb 2005 17:35:11 +0800, budgie > wrote:
>
>>On Wednesday, 16 Feb 2005 09:42:58 -500, "Asimov"
> wrote:
>>
>>>"budgie" bravely wrote to "All" (16 Feb 05 12:52:14)
>>> --- on the heady topic of "Re: Something like a diplexer"
>>>
>>> bu> From: budgie >
>>> bu> Xref: aeinews rec.radio.amateur.homebrew:2181
>>>
>>> >You missed a step called a directional coupler. It's a sort of
>>> >transformer with 2 input ports and 1 output port. The 2 inputs don't
>>> >see one another but their power is combined at the output.
>>>
>>> bu> You don't actually *need* a directional coupler. I have seen window
>>> bu> preselectors with five separate sections for sub-bands between 403 and
>>> bu> 520 MHz. The configuration is symmetrical (in/out) with simple bandpass
>>> bu> filter segments and coaxial split/combine harnesses. Not a DC or
>>> bu> hybrid in sight. I am presuming - not having swept one - that at "off"
>>> bu> frequencies each parallelled leg presents a high enough impedance to
>>> bu> the split junction that the effect is negligible.
>>>
>>>Yes but it requires some effort and cost to build the 2nd BP filters so
>>>I wonder if it's any more difficult to do either? What do you figure?
>>
>>I'm not sure we are on the same wavelength here (no pun intended). For
>>*non-overlapping* filter sections, the bandsplit through two bandpass sections
>>25-500 and say 550-3000 should be able to be achieved with T-pieces at the input
>>and output ends (although it appears the O/P didn't achieve this).
>
>I am not sure exactly what is trying to be accomplished here but if it
>is to combine two different bandpass filters then yes a T. The cable
>from the low pass filter to the T should be a quarter wave length at
>the frequency of the high pass filter.
>
>The cable from the high pass filter to the T should then be a quarter
>wave length of the low pass filter.
>
>The low pass filter will look like a short circuit at the high pass
>frequency. With the quarter wave length cable that short circuit will
>then be transformed to a very high impedance to the hi pass circuit.
>That effectively isolates one from the other.
>
>Do the another T and cable setup at the input and output.

If I understood the O/P he was trying to split and then recombine the 25-3000
band with a split ("crossover") at 500 MHz. Assuming a 50R in/out impedance
across the passband, the fiter sections presumably present a substantially
different (usually much higher) port impedance out of band. A simple T combiner
therefore presents a minimal impedance mismatch except near the split frequency.

Also I'm still thinking about how to make a quarter-wavelength line at 25-500
MHz or at 500-3000 MHz..

[Gary Schafer]

On Fri, 18 Feb 2005 11:42:56 +0800, budgie > wrote:

>On Thu, 17 Feb 2005 21:36:05 -0500, Gary Schafer > wrote:
>
>>On Thu, 17 Feb 2005 17:35:11 +0800, budgie > wrote:
>>
>>>On Wednesday, 16 Feb 2005 09:42:58 -500, "Asimov"
> wrote:
>>>
>>>>"budgie" bravely wrote to "All" (16 Feb 05 12:52:14)
>>>> --- on the heady topic of "Re: Something like a diplexer"
>>>>
>>>> bu> From: budgie >
>>>> bu> Xref: aeinews rec.radio.amateur.homebrew:2181
>>>>
>>>> >You missed a step called a directional coupler. It's a sort of
>>>> >transformer with 2 input ports and 1 output port. The 2 inputs don't
>>>> >see one another but their power is combined at the output.
>>>>
>>>> bu> You don't actually *need* a directional coupler. I have seen window
>>>> bu> preselectors with five separate sections for sub-bands between 403 and
>>>> bu> 520 MHz. The configuration is symmetrical (in/out) with simple bandpass
>>>> bu> filter segments and coaxial split/combine harnesses. Not a DC or
>>>> bu> hybrid in sight. I am presuming - not having swept one - that at "off"
>>>> bu> frequencies each parallelled leg presents a high enough impedance to
>>>> bu> the split junction that the effect is negligible.
>>>>
>>>>Yes but it requires some effort and cost to build the 2nd BP filters so
>>>>I wonder if it's any more difficult to do either? What do you figure?
>>>
>>>I'm not sure we are on the same wavelength here (no pun intended). For
>>>*non-overlapping* filter sections, the bandsplit through two bandpass sections
>>>25-500 and say 550-3000 should be able to be achieved with T-pieces at the input
>>>and output ends (although it appears the O/P didn't achieve this).
>>
>>I am not sure exactly what is trying to be accomplished here but if it
>>is to combine two different bandpass filters then yes a T. The cable
>>from the low pass filter to the T should be a quarter wave length at
>>the frequency of the high pass filter.
>>
>>The cable from the high pass filter to the T should then be a quarter
>>wave length of the low pass filter.
>>
>>The low pass filter will look like a short circuit at the high pass
>>frequency. With the quarter wave length cable that short circuit will
>>then be transformed to a very high impedance to the hi pass circuit.
>>That effectively isolates one from the other.
>>
>>Do the another T and cable setup at the input and output.
>
>If I understood the O/P he was trying to split and then recombine the 25-3000
>band with a split ("crossover") at 500 MHz. Assuming a 50R in/out impedance
>across the passband, the fiter sections presumably present a substantially
>different (usually much higher) port impedance out of band. A simple T combiner
>therefore presents a minimal impedance mismatch except near the split frequency.
>
>Also I'm still thinking about how to make a quarter-wavelength line at 25-500
>MHz or at 500-3000 MHz..


Usually a bandpass filter will present a low impedance out of band.
That is where I was with the quarter wave cables. But you are right,
it may be somewhat of a problem with that wide of a span. Guess I
didn't read close enough.

I read the original post again and maybe understand better what is
trying to be accomplished.
With the back to back diplexers there may be a termination problem
with combining them. ?? Maybe try a load resistor at the junction
where they recombine.

The simplest solution seems to be two hybrid power splitters back to
back. No band pass filters needed. Like two tv cable splitters. Of
course that means 6 db of loss.

73
Gary K4FMX

[Joel Kolstad]

Hi guys,

"budgie" > wrote in message
...
> If I understood the O/P he was trying to split and then recombine the
> 25-3000
> band with a split ("crossover") at 500 MHz. Assuming a 50R in/out
> impedance
> across the passband, the fiter sections presumably present a substantially
> different (usually much higher) port impedance out of band.

Yes, this is correct; the problem is that close to the 500MHz split, the
filters combine together so as to create several resonances. I'll try the T
combiner approach; this should tend to 'de-Q' the resonances.

An isolator (or circulator) would work here, but I'm not aware of any that
cover such a wide bandwidth.

The suggestions for 'quarter wavelength' anythings aren't appropriate given
the wide bandwidth.

Thanks for all the help,
---Joel Kolstad