|
Michael Post subject: Quadrature Hybrid Design and Fabrication
Unread postPosted: Thu Aug 05, 2004 1:25 pm Good day. I am attempting
to design a branchline hybrid (Quadrature Hybrid) that will operate
between 2400 MHz and 2500 MHz. I have simulated it in ADS. I
used Rogers 4003C board material (er = 3.38, subatrate thickness = 1.542
mm (60 mils), and metal thickness = 0.0356 mm (1.4 mils or 1 ounce copper)
to fabricate the hybrid. Everything simulates perfectly, but
when I get the fabricated board and test it, it does not work. The insertion
loss is 10 dB, the return loss is less then 5 dB, and the isolation
is less then 3 dB. I solder SMA connectors (Johnson Components 147-0701-631)
on each of the four ports. I am using an Agilent 8753E calibrated
from 2350 to 2550 MHz. I use 2 - 50 ohm loads to terminate the 2 ports
that are not connected to the VNA. Both 50 ohm loads have a return loss
(S11) of at least 35 dB. I used Pozar's book - Microwave Engineering
2nd Edition Page 379 - 383 to start the initial design. Can anyone
give me some thoughts or ideas about what is happening and how I can
get the issue resolved. Thank you for your time and help.
Michael Top Itay Post subject: Unread
postPosted: Thu Aug 05, 2004 4:32 pm Sounds that you have some
basic mistake...if your ADS simulation works well, there should be no
problem in the lab. Check the following: 1) Is your board
assembled over a metal chassis? It should be assembled and fastened
with screws to a metal chassis that will act as GND plane. Are the connectors
assembled to this chassis and tightened with scrwes? As you know the
VNA ports are referenced to GND, make sure that your GND isn't floating.
2) Have you calibrated the VNA for full 2 S-parameters measurements?
3) Check that the cables you use are in good shape..check their
insertion loss and reflection. 3) Have you measured the width
of the traces and verified they are in the same width as you defined
them for the PCB vendor? Hope this help, please let me know...
Good luck, Itay Top Guest Post subject:
Quadrature Hybrid Unread postPosted: Fri Aug 06, 2004 7:42 am
I agree with Itay. Also, if you have acess to an EM simulator like
momentum , or sonnet you can performan em simulation. If the em simulation
agrees with the linear model then the problem is most likely in the
fabrication, and/or measurement. Top Guest
Post subject: Unread postPosted: Fri Aug 06, 2004 10:18 am
Yes In addition to that, EM Simulation might help if your device
is assembled inside metal enclosure. It can predict possible couplings
that can increase the insertion loss or other phenomena. Make sure that
you enter the right height of the PCB from the enclosure top to get
out the exact results. The Momentum is a powerful and fast tool for
EM simulations, you can control the simulation time by defining the
meshing of the structure. Good luck and keep us posted if the
solutions we provided helped you. Itay Top
Michael Post subject: Quadrature Hybrid Unread postPosted:
Sun Aug 08, 2004 2:29 pm Good day. Thank you very much for your
thoughts and comments. I will answer Itay's questions Check the
following: 1) Is your board assembled over a metal chassis? It
should be assembled and fastened with screws to a metal chassis that
will act as GND plane. Are the connectors assembled to this chassis
and tightened with scrwes? As you know the VNA ports are referenced
to GND, make sure that your GND isn't floating. Answer: The hybrid
is not mounted in or on anything. The center conductor of the SMAs are
soldered directly on the microstrip line and the connector grounds are
soldered directly to the ground plane. The work bench where I am doing
the testing has a wooden top with metal legs. 2) Have you calibrated
the VNA for full 2 S-parameters measurements? Answer: Yes. I
did a full 2 port calibration of the VNA from 2350 to 2550 MHz with
801 points. 3) Check that the cables you use are in good shape..check
their insertion loss and reflection. Answer: Yes. I checked and
double-checked the cables. I also used a second VNA to verify that the
first VNA was working correctly. 3) Have you measured the width
of the traces and verified they are in the same width as you defined
them for the PCB vendor? Answer: Yes. That was one of my thoughts
also, so I measured the trace widths and the substrate thickness. Everything
is as it should be. I do have access to Momentum and I did double
check my simulation results. Everything agreed very well between Momentum
and my other simulation results. Any additional comments or suggestions
are greatly appreciated. Thank you for your time and help.
Michael Top Itay Post subject: Unread
postPosted: Mon Aug 09, 2004 2:14 am Here are few more thoughts:
Well I think that you should assemble your board over a metal chassis,
and this is to ensure a good continuity of your GND plane. This is what
I used to do with designs which worked at frequencies much lower than
yours. This provides a safe way to ensure a good GND plane. Have
you checked that your termination resistor at the isolated port is OK
(50 ohm)? Can you verify that the substrate you are using in
ROGERS and not FR4? Another thought (maybe an over-kill) is to
check that your traces truly have 50-ohm characteristic impedance (controlled
impedance) with reflectometer. Hope this helps... Good
Luck, Itay Top Michael Post subject: Quadrature
Hybrid Unread postPosted: Tue Aug 10, 2004 9:55 am Good day.
Thank you for all of your assitance. I have found the problem. Lance
Lascari suggested that I should increase my frequency range to see what
was happening. The quadrature hybrid worked very well at 2.85 GHz. 400
MHz higher then the design. That pointed me in the correct direction.
The problem was the Microstrip TEEs. I used ADS's layout feature
to verify the dimensions. The dimensions were off. The dimensions of
the TEE has caused me problems in the past.
Posted 11/12/2012
|