Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.
Navigate to the Microwave Encyclopedia. There you will find 11 varieties of Wilkinson power splitter, including lumped element, stripline, multi-port, etc, and even a Excel spreadsheet.
Use these for design purposes. Once you know the electrical lengths of the parts, you get to the physical lengths by going after the phase velocity of the transmission lines you intend to use for its construction. The physicals are reduced by the fraction of the phase velocity compared to the speed of light. The lengths are multiples of a quarter-wavelength at 3.6GHz
For simulation (if you still need it), there is AWR Microwave Office, or Ansoft, or Agilents ADS. These costs will make you cry! If you need a nice Spice simulator with transmission line models, or similar, LT-Spice IV from Linear Technology website is free, as are the various public domain simulators for Linux systems.
If you load up a Debian-based Linux system, there is ngSpice, which can use the schematic entry from the gEDA suite, or qucs (Quite Universal Circuit Simulator), just a few clicks away from the (huge) repository. The last one is, I think, a full harmonic balance simulator.
Thanks Darktrax for the suggestion ... only with regard to the simulation, should I use, for reasons to university, Ansoft Designer SV. Is there someone I can please describe the steps to follow with that software. Thanks for your kind attention and availibilty.
- - - Updated - - -
Oh I forgot, how do I calculate the physical length? For the simulation,is the same thing, use a divider with a circular section or divider straight section?Thanks
OK - we try again.
First is to visit the links and understand how this works. Wikipedia shows the math.
A Wilkinson divider is built using two transmission lines of impedance Zo*SQRT(2), and are quarter-wavelength (lambda/4) long. You know the frequency. You know the speed of light. So you can figure the wavelength.
Now you also know that the speed of the signal down a transmission line is slower than light. If you choose to use microstrip, then the tools in Ansoft will tell you the phase velocity, and how to make a resonant length. A Wilkinson divider is a resonant device. I leave you to figure how much shorter to make the lines.
Straight or curved, or any shape you like, the important property is that they are a quarter wave long. These are called "quarter-wave transformers" because they transform the impedance at one end into another impedance at the other.
The feature that makes the device a Wilkinson Divider is the presence of a resistor 2*Zo between the Port_2 and Port_3.
Normally, a lossless 3-Port device cannot have all ports simultaneously matched. Wilkinson added the resistor to cheat, and then discovered that the resistor does more than allow all ports to be matched, it also isolates Port_2 from Port_3 at the centre frequency. The resistor does not dissipate power because the voltages at its ends are in phase
Very commonly, a quarter-wave section is added at Port_1 to match to 50 ohms.
All these bits you do with a calculator. It takes only a few minutes. Then, use Ansoft tools to simulate. It will tell you the dimensions of the transmission lines. A divider with a quarter-wave section 42-ohms, split into two quarter-wave sections 59.4 ohms and a 100R resistance between the outputs is a wider-band better performing build.
Very nice thank you ... and if I wanted to design a Wilkinson coupler with circular section, how should I do to calculate the radius, so as to represent him in Ansoft?
When I tried it, I was doing a "Rat-Race" combiner, which is circular, with 3 of the ports separated by lambda/4 and the rest of the circle was lambda/2. I simply made the length of the arc in the centre of the track equal to (Lambda/4)*(Vp/c) where Vp is the phase velocity of the transmission lime, and c is the speed of light. I was using stripline (like microstrip, but with dielectric and groundplane on both sides).
It does not matter the shape of the curves, many microstrip lines are made as a series of wiggles, usually to make it fit in a smaller area. You just take the length along the centre. Possibly, if its a circle, then the "real" electrical length of the line might be slightly short, making the resonant point slightly higher in frequency.
In Ansoft, and other simulators, its easy to scale the device to make that right. You can make the line length one of the parameters for adjustment for the optimizer, and then have it calculate the revised design. In never saw it as necessary. Making a successful Wilkinson divider does not need a big simulation. There is a (free) design tool from Agilent, designed to promote their products, which has sections where you can figure out any transmission line impedance you desire.
If you Google for AppCAD you will find it availble from many places, but I include one of them here --> http://www.hp.woodshot.com/
But by chance, could you bring me to an example that you used with any CAD, to solve my problem of Wilkinson combiner? I would be immensely grateful ... I'm sorry but my first experience
But by chance, could you bring me to an example that you used with any CAD, to solve my problem of Wilkinson combiner? I would be immensely grateful ... I'm sorry but my first experience
In fact, thanks to Darktrax, I got the physical parameters:electrical length=90°, Z0 = 48 ohm, 2.45 mm width microstrip, length (lambda / 4)= 44.51mm. Zc= 67.88 ohms, width= 1.404 mm, length (lambda / 4)= 15.10mm. I hope to be corrected. Now how should I proceed to make "Planar EM Design" in Ansoft? How do I figure out where to put the resistor 2Zo? Thanks
In fact, thanks to Darktrax, I got the physical parameters:electrical length=90°, Z0 = 48 ohm, 2.45 mm width microstrip, length (lambda / 4)= 44.51mm. Zc= 67.88 ohms, width= 1.404 mm, length (lambda / 4)= 15.10mm. I hope to be corrected. Now how should I proceed to make "Planar EM Design" in Ansoft? How do I figure out where to put the resistor 2Zo? Thanks
Ok, but what you use CAD? Could you post some photos with the design and simulation, inserting my parameters? Me with Ansoft, I do not know where to put their own, too bad that I'm forced to use it. Thanks for your kind collaboration.
Could you post some photos with the design and simulation, inserting my parameters? Me with Ansoft, I do not know where to put their own, too bad that I'm forced to use it.
With Ansoft Designer, from back in Version2, and very likey most later versions, the Help -->Contentents--> Getting Started with Ansoft supplies a tutorial on how to choose the planar properties, and enter your parameters to implement various things like couplers and filters. Especially in File-->Open-->Examples-->Planar EM you will probably still find Wilk_c.adsn and Wilk_x.adsn.
They include the pads for the resistor, and the resistor itself in the simulation circuit.
These may be tutorial examples, so if they are partly implemented, then follow through the tutorial to finish them. Tutorials can also be found using internet search, or free from the Ansoft site.
Do not consider it bad that the available tool is Ansoft Designer. You will find it to be as good as any.
In Ansoft Designer, you can edit the properties of a port, to set the impedance, and the amplitude and phase for simulation.
That you ask this suggests you should now follow through the help index, and try out tutorial examples enough to become familiar with how you can make a circuit equivalent, add and alter ports, and create and connect the parts in between, and set up simulation. You will then see how to get results from the Wilkinson examples. You may find the ports and stimulus properties are already there.
Note also that for 3.6GHz, the Wilk_c example may have dimensions closer to the one you want, but you would need to alter the sizes anyway to get the lengths to be resonant at 3.6GHz.
It being a combiner, it combines. Most folk choose 1V amplitude, for at least one of the ports.
You edit the properties of the port. One property you probably do not want to change is the impedance. You make that 50 ohms. The PHASE of the ports is also specified. If you make both 0 degrees (or 0 radians) then they are in phase. The full amount of both will appear at the output, less the insertion loss, and maybe some phase shift from unbalance.
If you make the phase of one to be + or - 180 degrees compared to the other, they are out of phase, and the combination will cancel. You get nearly zero output.
If you put 1V into the Port_1, you get -3dB at Port_2 and at Port_3. Operating it backwards like this, it is a splitter.
Since the Port_2 is identical to Port_3, you often need not bother to simulate both. You just put a 50 ohms termination on two of the ports, and a 50 ohms signal source port on the last.
You would use it to combine the signal from (say) two antennas, where you want the incoming signal from Port_2 not to have any large portion simply re-transmitted out the other antenna on Port_3. This exploits the isolation property between those ports.
You can use two Wilkinson devices, to split and then re-combine a signal, putting amplifiers in the split signal routes. This way you can use two low power amplifiers to be a high power amplifier, and know that if one fails, the signal continues to be delivered at half-power (3dB down) via the remaining route
In Ansoft, you would set up the simulation conditions to sweep the frequency from (say 1GHz to 6GHz) or maybe 3GHz to 5GHz.
You would want to know the S11 (reflection coefficient) plot for each port. You would want a plot of the transmission coefficient between ports, to see the loss with frequency (S21). The tutorials will show you how to do this on other example designs. Do these before you then apply the knowledge to the Wilkinson example - or do them together if your computer is fast enough.
No calculation required. Isolation between the Port_2 and Port_3 is a basic property of the Wilkinson.
Please read the page --> **broken link removed**
You will find there the most complete descriptions of Wilkinson splitter/combiners.
There is also the Excel spreadsheet for designing them.
Can someone plz tell me how to implement the following corporate feed in Ansoft designer.? I have calculated the lengths and widths of the 50 and 70 ohm lines.
The combiner shown is of the sort that might be used to with (say) 4 x 50 Ohms impedance antennas, all pointing the same direction..
To implement in Ansoft Designer (or any) simulation software is much the same.
You select the physical technology, in this case probably microstrip.
Program needs to know the substrate thickness, Er, Tan(Delta) characteristics, etc.
You enter the circuit using MSTRL for those physical lengths you want to force.
Use MSTRLE for those lengths in the quarter-wave transformers you want to have controlled to be optimum for the frequency.
For the T-junctions, use either MSTEEE, MSTEEC, or MSCTEE if you want to have a "compensated" shape.
View the help for these to decide which best suits your need.
For the mitred bends, use MSBENDC, or maybe MSBENDO for optimally mitred bends.
Before you go further down this road, give consideration to using stripline, ie. another substrate with groundplane on the topside. Depending on the frequency, this forces a more coaxial-like TEM nature with fewer radiation losses.
Note also that you do not need Ansoft or other simulation at all for this.
Much of the design is just folded up bits of 50-Ohms line done to place the connectors to save space.
You can separately calculate mitres. Here is a link --> **broken link removed**
The lengths of the lines with bends in them are clearly arbitrary, so long as they are equal to each other. Since 70.7 is the geometric mean of 50 and 100 ie. sqrt(50*100), the 70.7 sections transform 50 up to 100 ohms, where two 100 Ohms in parallel at the TEE bring it back to 50 Ohms.
That means the lengths of all the 50 Ohm lines is arbitrary, so long as those doing similar jobs are equal to others of their kind.
You could make this one in a "H" shape, and lose all the mitre bends altogether!
Your posting is in a well-expanded "Wilkinson" thread. You could use Wilkinson combiners in this application, but you may have chosen not to.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
