Placement of coaxial stubs and bandpass filters in line
Ken_Silverman_at_Wiget_Lane at atlas.ccmail.PacTel.COM
Ken_Silverman_at_Wiget_Lane at atlas.ccmail.PacTel.COM
Wed Oct 27 20:07:43 EDT 1993
Well, many of you probably have been following my trail of questions. Thanks
for all the help. Attached is a message which I sent to some of my "advisors"
on the issue of coaxial stubs. I think you will find it intesting.
The tests below were done trying to simulate the overall band rejection of
combining a bandpass filter on the back of the radio, and a coaxial stub on the
back of the amplifier for additional harmonic rejection. These "simulations"
were done using a network analyzer - and the real question is how to adapt them
to a "real" system. That's where my call to you comes from.
The position of the coaxial stub in line with respect to the bandpass filter on
the back of the tranceiver greatly impacted the overall system band rejection.
The responses from the original addressee's was mixed+. This topic seems like
unexplored ground. I hope some of the more quiet people on the reflector will
have something to add on this issue.
In a nutshell: the question raised by these tests is if you have a system hook
up as follows: RIG--bandpassfilter--amplifier--coaxialstubb--antenna; How
does the insertion of the amplifier effect the interaction between the bandpass
filter and the coaxial stub? You will see in my tests below, that the line
length between the bandpass filter and the coaxial stub had a GREAT impact on
the overall band rejection in my "simulation".
Mark's thoughts were that you had to treat the 2 filters independent from each
other. The bandpass filter is for the rig, and the stub is for the amplifier.
Sounds logical, but I still get this funny feeling that you have to look at all
of these resonant devices as a system. ((Mark, I greatly value your opinion,
but something inside says there might be something more to it))
Do any of you have any thoughts on this? Lets hear from all of you technical
guys who wrote the book on this stuff. Much obliged. I hope you find this
interesting, if nothing else, amusing.
Hi Mark AG9A (cc: K3LR, K2MM, KE7X, K9ZO),
I was playing around again this past weekend with some coaxial stubs, and ran
into some problem (maybe) and I thought you might have had some experience with
what I am seeing on the spectrum analyzer.
In one of your notes on coaxial stubs, you said that additional AND identical
stubs can be added in line to obtain additional rejection. If you do that,
space them an 1/8 th wave down the line.
Well, I started playing first with the spacing between 1 coaxial stub and the
I.C.E. filter. (I had built all of the stubs and shipped them down to Caracas
already - except for the 160m stub.) I just made the 160m stub from some
Belden 9913, for the ICE 160m adjacent band performance is not what I expected.
I also made another coaxial stub for 160 using RG8-X just to test the dual in
line stub response.
I made a series of mesurements on the network analyzer, which I will put all in
one chart. I will then talk about some of my concerns etc afterwards. In the
chart, I will tell you how things were hooked up ex: ICE-barrel-9913. This
means the ICE filter was only separated by a barrell connector with the coaxial
stub made with Belden 9913 coax. Another ex: ICE-tuned 160m 1/4 wave-9913.
This means that the ICE filter was separated from the Belden 9913 coaxial stub
by a tuned 1/4 wave section between the two. I hope this is understandable.
This was all done to model how the path length between the stubs and the filter
impacted the ultimate rejection. I only tested a few cases at the operating
frequency of 1.8 mHZ just to benchmark insertion loss. I also only tested up
to 14 mHZ since I am assuming that 160m will not interfere with the other bands
- since it is very unlikely that 160m and 10/15 will be open at the same time.
The 160m stubs, (like others I have built) are balanced between the responses
of the different bands. That is, for some reason, the 40m notch is not twice
the frequency of the 80m notch. The same goes for the difference between the
40m notch and 20m notch. I have only given you the attenuation on exactly 1.8,
3.5, 7.0, 14.0. Primerily because these coaxial stubs are built for CW only.
The depth of the notch may be slightly higher or lower, thus the ultimate
rejection frequency may be another 2-4 dB better than what I have given you. I
figured this way was the best for comparison.
Attenuation in dB @ 1.8 3.5 7.0 14.0
ICE 160m bandpass Filter 1.0 14 24.5 34
9913 coaxial stub for 160m .25 22 19 15
RG8X coaxial stub for 160m .25 18 14 8
ICE-barrel-9913 1.0 28 30.5 34.5
ICE-barrell-9913-barrell-RG8X 33 31 32
9913-barrell-RG8X 26 21 16
9913-20'coax-RG8X 42 37 22
* Note: 20' is what I had handy - it is near 1/4 tuned on 40m. For some
of the next measurments, I cut a tuned 1/4 wave stub for 40m @ 23
ICE-20'length coax-9913 33 48 48
ICE-tuned 1/4 wave on 40m-9913 35 48 37
ICE-1/4 wave on 160m-9913 29 36 47
ICE-1/4 wave on 160m + 20'coax-9913 33 48 52
ICE-7'coax-9913 18 41 52
*Note: this 7' length was used to simulate the connection between the
filter output, through the linear, and then to the stub.
What is causing me some heartburn is the last measurement. I ran it 3 times to
be sure. But when the ICE filter was separated from the tuned stub by what I
was assuming to be a standard length, the adjacent band (3.5) was only
marginally better than the ICE filter alone. This configuration is what I
would have considered "typical" in that you would set your equipment up like
Xcvr - ICE filter bolted to back of radio - jumper to linear - linear - coaxial
stub inserted at direct output of amp. The 7' allowed for a possible coax
switch in there too.
Thus the insertion of the coaxial stub at a random length from the bandpass
filter does not seem like a way to acheive maximum benefit of the two devices.
It seems that the coaxial stub needs to be separated from the ICE filter by a
tuned PATH length. This path length should be tuned for what benefits you want
to see. I still haven't come to grips on how to calculate that path length,
but will give it some thought.
One problem comes from the path length through the amp. Since these
measurements were done on the network analyzer, the fact that the amp will be
in line may change these results. I do not know how to model the path length
thru an active amp that is in line. When the amp is in standby, you bypass the
tuned circuits. Thus there is no way for me to test a live system. If you
transmit on a frequency where there is a null, the SWR is very high (what you
would expect), and the radio shuts down immediately.
So, any thoughts on what I am doing? Have you given any consideration to this
idea? Does all of this get nullified with the amp in line? If so how? If you
have to account for the path through the amp, is there any way of doing so? I
am rather suprised at the overall results and findings if they are true. This
is new ground for me.
Another observation is the maximum rejection of 22 dB of the Belden 9913 stub
on 160m. For most of the other stubs I built, I was seeing typically 25 dB
notches (for the second harmonic) using new RG213. For those of you that are
interested, the following is the response curves of the coaxial stubs I built:
1.8 3.5 7.0 14.0 21.0 28.0
160 .25 22 19 15
80 1.0 >.25 26 24 22 20
40 11 5.0 >.25 32 36 34
20 9.0 2.0 34 >.25 27 28
15 4.0 2.0 0 27 >.25 24
10 Not built
One important note - each "band" filter may have been comprised of 2 stubs to
achieve the proper nulls on each band. Example, the 20m band needs one stub
for 40m and 15m nulls, and another one for the 10m null. If tested separately,
each null was a few dB less than the combined system. I have only given you
the combined system results.
Well, that's it for now. Any thoughts would be appreciated. This may be a
topic for discussion on the reflector, but I first wanted to keep it limited to
see if I have missed something along the way.
Many thanks, Ken WM2C
>From Alpha Personal Systems -- DTN 223-5747 -- MLO3-6/A8 27-Oct-1993 2221 <reisert at wrksys.enet.dec.com> Thu Oct 28 03:17:18 1993
From: Alpha Personal Systems -- DTN 223-5747 -- MLO3-6/A8 27-Oct-1993 2221 <reisert at wrksys.enet.dec.com> (Alpha Personal Systems -- DTN 223-5747 -- MLO3-6/A8 27-Oct-1993 2221)
Date: Wed, 27 Oct 93 22:17:18 EDT
Subject: 286 and CT
Message-ID: <9310280217.AA26309 at us1rmc.bb.dec.com>
Jay, WS7I wrote:
>386 that they had was working fine. But the 286 got some kind of DOS error
>message and wouldn't load.
You must have the file ZPM.EXE in the same directory as CT286.EXE. This is
the Zortech memory manager that CT286 requires to run.
73 - Jim AD1C
>From n7stu at thetech.com (Robert Brown) Thu Oct 28 03:30:25 1993
From: n7stu at thetech.com (Robert Brown) (Robert Brown)
Date: Wed, 27 Oct 93 19:30:25 PDT
Subject: Instant License .....what's a VE to do next
Message-ID: <eHZ4Bc1w165w at thetech.com>
What a joke! The FCC ran this same scam back in the mid-70's-80's with
CB licenses. Each rig came with a temporary license form and your call
sign was K + initials + ZIP code or something like that. I'm not a CB
basher (a lot of junk goes on there) but this idea is not the way I want
to see ham radio progress.
73, Robert N7STU/YB2ARO
n7stu at thetech.com (Robert Brown)
The Tech BBS (408) 279-7199 San Jose, CA
>From Tony Brock-Fisher <fisher at hp-and.an.hp.com> Thu Oct 28 13:54:45 1993
From: Tony Brock-Fisher <fisher at hp-and.an.hp.com> (Tony Brock-Fisher)
Date: Thu, 28 Oct 93 08:54:45 -0400
Subject: Coax Stubs and Filters
Message-ID: <9310281254.AA01327 at hp-and.an.hp.com>
Here are my thoughts on the coax stub / ICE filter problem posed by Ken.
There are two issues I will commnet on: the 'tuning' of the length of the
cable, etc between the filter and the stub, and the effect of adding
a linear amplifier to the fray.
On tuning the length:
The tuned stub idea uses the principles of transmission lines to effectively
place a short across the line at a certain frequency. It is also a happy
coincidence of transmission line behavior and our FCC frequency allocations
that results in additional shorts appearing for other useful frequencies.
Due to dielectric losses in the cable used for the stub, we don't get a perfect
short but some other very low complex impedance.
The ICE filter (or other bandpass filter) is engineered to have a input and
output impedance that matches the Zo of our system (nominally 50 ohms)
AT THE DESIGN FREQUENCY! At other frequency, it probably has very different
input and output impedances. Add to this the fact that at frequencies other
than the design frequency, the antenna at the far end of the line has a
impedance which is almost guaranteed NOT to be 50 ohms! In addition, the
radio's output impedance is also NOT 50 ohms at other than the operating
Now lets examine what the system looks like. At the operating frequency,
hopefully, everything in the system is matched to 50 ohms. The SWR is
nearly flat. The impedance seen anywhere along the line is 50 ohms. However,
at the frequencies that we wish to attenuate, nothing is matched! We have the
rig output (NOT 50 ohms) driving the filter (NOT 50 ohms) with some length of
cable, driving the antenna (probably not 50 ohms). Therefore, at this frequency
the SWR is not flat - it's probably very high. Therefore, the impedance VARIES
along the transmission line! Now, if you apply your stub, which is a small but
finite impedance across the varying impedance along the line, of course the
attenuation is going to be a function of position! The only way to predict
where to put the stub is to know all the impedances in the system and spend
three weeks figuring out Smith charts! (Or you could just cut and try,
which is what most of use do anyway)!
On the addition of the linear amplifier:
The linear amplifier in some ways actually simplifies things. While it's
output impedance is NOT 50 ohms at other than the transmit frequency, it
acts as a buffer amplifier and eliminates the radio and ICE filter on transmit.
On receive, of course, the amplifier just looks like a length of transmission
line and the rig and filter are back in. But I think we only really care
about what is happening on transmit.
Hope this helps ...
73 de Tony, K1KP
More information about the CQ-Contest