Continue to Site

Eng-Tips is the largest engineering community on the Internet

Intelligent Work Forums for Engineering Professionals

  • Congratulations GregLocock on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!

Duplex filter for 0-100 mhz and 100mhz-6Ghz.

Status
Not open for further replies.

watchjohn

Electrical
Mar 28, 2007
27
.

This question is related to the previous
( "Wideband antenna in wristwatch 10-6000 mhz" )

I´m looking for a duplex filter circuit (RX only)

I have 2 antennas each for its own frequency-band

a) 1 to 100 Mhz
b) 100 to 6000 Mhz

Ofcourse they overlap...can´t avoid that

I want to connect (couple) both antennas to 1 receiver.

To avoid "nulling out" problems when a certain signal is received at both antennas...(in fact they are all)
i understand that the best would be to have filters in
both antenna lines so a certain frequency will only be received 1 time (and not cause nulling out)
Such filters will also match the impedances to 50 Ohms.


I understand that basically i have to combine 2 filters

1) a low-pass with a cutoff of at 100 mhz
2) a highpass-filter with a cut off at 100 mhz

One side of the filters are connected together at the receiver input.
The other sides go each to there own antenna.

See an example here...for 144 and 430 mhz
the crosspoint (cutoff) is probably somewhere in the
middle...at 280 Mhz or so:



Now i want something like that for 1-100 and 100-6000 Mhz
It´s for receive only...so the C´s and L´s can be much
smaller.

So i downloaded a piece of free software called "diplexer"
by "tonne-software" it´s free and can be downloaded here :


This software lets you design exactly what i want
and allows to make all kinds of adjustments too.
it will also plot a graph.

So i input my values...and i get a nice drawing whit all the values..perfect

So i put in a crosspoint of 100 Mhz...and voila

Problem is it doesn´t show what happens at the end of
both the lower and higher range...actually you only see the crosspoint...the lower and upper ends go on forever !!
wich can´t be correct.

In other words...i get a design for a crosspoint at 100 mhz
but i cant see what happens at 6000 mhz (high end)
and at 1 Mhz (low-end)

It could very well that i have i nice crosspoint at 100 Mhz
but that at for example 5.8 Ghz there is a loss of 30 dB
caused by the circuit

Ofcourse i will use components that have specs up to 6 Ghz

Anyone can tell me what i´m doing wrong...or how the plot
really looks like at the ends of both bands.

I´m afraid that other online filter-calculators will have
the same problem ?

Thanks again !


WatchJohn
 
Replies continue below

Recommended for you

>It could very well that i have i nice crosspoint at 100 Mhz but that at for example 5.8 Ghz there is a loss of 30 dB caused by the circuit

That "diplexer" is a very nice piece of software. You are modelling ideal components and so the response of the low pass arm will go to DC and the high pass end will go to daylight. Notice that the ideal circuit model uses very small lumped elements. These will go self-resonant at some point, limiting the performance. So you need to select some real components, model the Q and self resonances and re-simulate in SPICE.
 
.

Thanks...that was a ery clear answer !

So now i know that...what would be the next step ?

Here´s what i think i should do :

1) Design the filter using the "diplexer" software
and neglect the fact that the software doesn´t show
the lower and upper-end responses (it can´t).

2) Find the components
(SMD coils and SMD cap´s for Ghz applications)
since i don´t want to work with stripline stuff.

3) Make shure that the choosen components have good specs
up to 6 Ghz...
meaning that they should have low loss below 6 Ghz and
self resonating freqs should be above 6 Ghz,

if that is not possible then the selfresonating freqs
should be at some point in the band that is of less or
no importance (there will be a dip at those freqs)

Is that it ?


Thanks again


Radiotechscan




 
Yes that is correct.

On point 3, if you run below the SRF of any of the components you have to simulate to say what the response will be. It has to be said that any imperfection will make the input VSWR worse and the transfer response less. You simulate to quantify the effect.
 
I agree, that is a nice program.

For the side that is the 100 MHz and lower output, you will not have any problem. Since the lowpass filter is comprised of series inductors, you will have unity gain at 0 Hz, so you know what is going on at the low end!

At the high end, like others said, you have to worry about component resonances and board (transmission line) effects. So getting it to work up at 6 GHz may be tough.

I would layout the entire filter so that the 100 MHz parts were close to each other, and that the 6 GHz path parts were close to each other, and that at the common junction there two filters were close to the junction.

I would also make all of the inductors in the 6 GHz path to be made up of two values. For instance, lets say the 6 Ghz path has one inductor of 40 nH value. I would put a 5 nH (small non-resonant chip)inductor in series with a 35 nH bigger (but still the best non-resonant frequency chip inductor that I could find). That way, when the 35 nH inductor resonates, at say 3 Ghz, it is somewhat dampened bythe 5 nH which may not resonate until 7 GHz.
 
.

Thanks you "logbook" and "bif44",

I´m new to this forum...and i´m happy
that i get good reply´s
wich is not always the case on other forums.


"inductors in the 6 GHz path to be made up of two values"

Yes i understand that.



Question 1
----------

A related inductor question:

I´m using some 6 ghz amplifiers (MMIC) like this one


If you scroll down that pdf-file
you will see a small table called
"Recommended Component Values for Key
Application Frequencies"

There they recommend some (bias) coil values for different
freqeuncies. (every manufacturer of MMIC´s does that)

The purpose of those coils there are to block the RF
from leaking to the power-supply.

Now do i have a problem there ?
since i want to receive the whole range 100-6000 Mhz
wich value should i take ?

For the the whole range from 100 to 6000 Mhz
I have these options i guess:

A) use the same trick ? (more then 1 coil in series).
B) use the value for the highest freq.
C) use the value for the lowest freq.
D) use 2 values in parallel (wich values?)

Wich one is correct ?


Question 2
----------

I also ran into these new "conical" coils
the manufacturer states that they have no SRF
up to 40 Ghz !


They look very funny
...obvious that conical shape makes them wideband.
i suppose i can look at them as a whole lot of coils
of different values...in series

The point is that they come only in 4 values in the
nH range...wich are wrong values for above mentioned
amplifier (or not ?)...so cant use those i suppose...wich amazes me since they advertise them exactly for that purpose


Greetz

WatchJohn
 
Please use "which" instead of "wich". This forum is in English not text speak, and the mis-spelling is annoying.

Don't use C1 in the application information for the MMIC. Your diplexer handles the input coupling.

A conical inductor is ideal for your application, although they are expensive. Something around 150nH would seem right according to the application note data.
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor