Вы находитесь на странице: 1из 122

Passive devices versus active devices A passive device contains no source that could add energy to your signal,

with one exception. The first law thermodynamics, conservation of energy, implies that a passive device can't oscillate. An active device is one in which an external energy source is somehow contributing to the magnitude of one or more responses. The important properties of a passive network are:

whether it is reciprocal or non-reciprocal whether it is lossy or lossless whether it is impedance matched or unmatched.

What is the exception to the "passive rule" about not adding energy? Mixers! Here the local oscillator adds energy, but because of the way that a mixer works, no signal gain is possible. Unilateral versus non-unilateral devices When we are discussing unilateral devices, we are talking about ideal amplifiers or other active devices. Two-port S-parameters almost always use the convention that S21 is the forward direction (the direction of gain) and S12 is the reverse-isolation direction. A unilateral device has the property that S12=0. In practice, this is impossible, but it sure makes analysis easier. One problem with non-unilateral amplifiers is that they have poor directivity. Properties of reciprocal and non-reciprocal networks A reciprocal network is one in which the power losses are the same between any two ports regardless of direction of propagation (scattering parameter S21=S12, S13=S31, etc.) A network is known to be reciprocal if it is passive and contains only isotropic materials. Examples of reciprocal networks include cables, attenuators, and all passive power splitters and couplers. Anisotropic materials have different electrical properties (such as relative dielectric constant) depending on which direction a signal propagates through them. One example of an anisotropic material is the class of materials known as ferrites, from which circulators and isolators are made. Two classic examples of non-reciprocal networks are RF amplifiers and isolators. In both cases, scattering parameter S21 is much different from S12. A reciprocal network always has a symmetric S-parameter matrix. That means that S21=S12, S13=S31, etc. All values along the lower-left to upper right diagonal must be equal. A two-port S-parameter matrix (at a single frequency) is represented by:

If you are measuring a network that is known to be reciprocal, checking for symmetry across the diagonal of the S-parameter matrix is one simple check to see if the data is valid. Here is an example of S-parameters of a network that is either a non-reciprocal network, or your technician has a drinking problem. (Need to add figure)

Although the data shows the part is well matched (S11 and S22 magnitudes are low), and low loss (S21 and S12 magnitudes are high). The magnitudes of S12 and S21 are equal, so what is the problem? The phase angles of S12 and S21 are significantly different. That can't be right. Properties of lossless networks For a network to be lossless, all of the power (or energy) that is incident at any one port has to be accounted for by summing the power output at the other ports with the power reflected at the incident port. None of the power is converted to heat or radiated within a lossless network. Note that an active device is not in the same category as a lossless part, since power is added to the network through its bias connections. Within the S-parameter matrix of a lossless network, the sum of the squares of the magnitudes of any row must total unity (unity is a fancy way of saying "one"). If any of the rows' sum-of-thesquares is less than one, there is a lossy element within the network, or something is radiating. Why are we looking at sum of the squares instead of sum of the elements themselves? Because the S matrix is express in terms of voltage, and as we said, we are accounting for power. Power is proportional to voltage squared, get it? Guess what? You can never make a lossless network. But you can come extremely close, especially with waveguide structures. How does the sum-of-the-squares-equals-unity property of lossless networks make your life better? You can use it as a check to see if data is bullcrap. If your Grandmother hands you S-parameters of a two-port test fixture and it looks like this, you can tell her for Christsakes to re-measure it after she checks the calibration:

The "test" that it fails is that 0.8^2 + 0.7^2 =1.13, which is different from unity (1.00). Apparently this two-port device has gain. If it really did, you could file a patent and become fabulously wealthy. But it doesn't, so quit dreaming and get to the bottom of your measurement problem. A good place to start is examining your test cables for flakiness. For some stable VNA cables at a good price, check out Storm Products Company. Properties of matched networks

A matched network is one in which all of the ports are matched to the same impedance (Z0). A desirable quality, you must agree. Looking at the scattering matrix, this means that the diagonal elements from top left to bottom right are all zero. Need to add a figure! A little more explanation of this property (still waiting for that figure...) If a network is matched to fifty ohms, its reflection coefficients have magnitude zero. This means we are at the center of the Smith chart, right at Z0. If you look at the expression for reflection coefficient: gamma=(Z-Z0)/(Z+Z0), when gamma (which in this case could be S11, S22, etc.) equals zero, Z=Z0. S-parameters are in units of volts/volt, not ohms. A special property of three-port networks The math behind the theory of three-port circuits such certain as couplers and splitters is not all that complicated and has a certain elegance, like a proof of the Pythagorean Theorem. For those of you who enjoy some good matrix algebra derivations, we refer you to Pozar's book Microwave Engineering, which you can find on our book recommendation page. We are going to skip the math and tell you the conclusion: it is impossible for a three-port network to be reciprocal, lossless and matched all at the same time. You can only have two of these properties. Reciprocal three-port junctions are characterized by the fact that a change in the terminal conditions at one port affects the conditions at the other ports. This effect is particularly pronounced when the junction is dissipationless (loss-less) . The reason is that the insertion of a matching network at one port changes the impedance characteristics at the other two ports. This lack of isolation between ports can limit the usefulness of three-port junctions, particularly in power monitoring, combining and divider applications. If a three-port is lossless, and contains no anisotropic materials, then it will have to be reciprocal. But it cannot have all three ports matched to 50 ohms at the same time. Is this a bummer or what? Not really. This is the reason for the isolation resistor in a Wilkinson power splitter. A Wilkinson is an example of a reciprocal matched three port network. It is only lossy between ports 2 and 3, which has no effect on its efficiency as a combiner or splitter. The resistor isolates the output arms as well. Even and odd mode impedances Updated December 14, 2008 Click here to go to our main page on characteristic impedance

Click here to go to our page on baluns Click here to go to our page on coupled line couplers This material came from Trev45, who frequents our message boards. To which we say, Thank you sir, may we have another! In preparing this text, I am acutely aware of being a dwarf climbing onto the shoulders of comparative giants. Much of the credit must go to Polar Instruments Ltd., who produced a very informative Application Note AP157 related to high-speed circuit PCB testing, from which I have shamelessly plagiarized the diagram. You might ask why bother to explain a topic when the answers can be computed by any of a thousand electromagnetic field simulation software packages? Well, the short answer is "to understand the answer, you first need to understand the question." Impedance is a vector quantity that expresses the ratio of electromagnetic voltage (or electric field strength) to current (or magnetic flux density). Imagine the familiar coaxial transmission line, a center circular conductor within a circular ground plane. There is a straightforward analytic formula for the characteristic impedance: Z0 = (138/ (
R

)^0.5)* Log(D/d)
R

where d is the inner diameter, D is the outer diameter, and

the relative permittivity.

Now imagine a simple balanced two-wire line. Again, there is a straightforward analytic formula for the characteristic impedance: Z0 = (276/ (
R

)^0.5 )* Log(2*s/ D)

where s is the spacing and D is the conductor diameter, is the relative permittivity Now imagine a transmission line comprising a balanced twin conductor pair inside a circular conducting tube, in other words a shielded twisted pair transmission cable. Is there a straightforward expression for Z0, and anyway, what is meant by Z0 in this situation? Well, a) there isn't and b) it's complicated. There are three impedances that characterize the structure, characteristic impedance, even mode impedance and odd mode impedance. Whilst in this case there is no precise analytic formula that applies to any geometry, such as circular-cylindrical or strip-line of any ratio of dimensions, there are approximate formulas that apply to say strip-line or coaxial cylindrical forms over a restricted range of parameters. A reasonably accurate formula for screened twisted pair is to apply a small correction to the balanced line formula to take into account the proximity of the overall screened ground cylinder. The formula is

Z0 = (276/ (

)^0.5) * Log(2*s / D) * ((1 - G2)/ (1 + G2))

G being the ratio of center to center inner conductor spacing to outer diameter (thanks for the correction, Bert!) There are two modes of current flow in an electromagnetic situation such as this: the first is one flowing down one conductor with a contra-flow current back up the other conductor caused by displacement current coupling between the two conductors. This is termed the odd mode or differential mode current, and has an associated odd mode characteristic impedance, styled Z0o. Imagine it conceptually like this. Energy couples from one line into the other, flowing away from the source to the matched load. No energy returns from the far end matched loads, but some flows back out towards the source, so there is a plus arrow going away down one conductor and a minus arrow coming back out of the other conductor. The other component of current flows by displacement current between each center conductor carrying the same polarity, and the ground that is common between them. Hence this is called the common or even mode current, and has an associated even mode characteristic impedance, styled Z0e.

Figure 1 Even and Odd Mode electric field lines From the electrostatic field patterns, it is clear that there are 3 values of capacitance involved for the odd mode, and each must figure in the Z0 formula somewhere. The first is the direct capacitance to ground of each trace, the second is the mutual capacitance between the traces, and the third represents the distortion of each caused by the presence of the coupled element (and a small contribution due the thickness of each). The same argument applies for the inductance when the magnetic fields are considered, and the combined system impedance at very high frequencies in a non-dissipative system is then

Z0 = (L/C)^0.5 where L is the inductance per meter and C is the capacitance per meter. Whatever the geometry, the following holds true Z0 = ( Z0o* Z0e)^0.5 which is the characteristic impedance of the system. Also, intuitively, from an examination of the field pattern, the even mode impedance Z0e is likely to be high because the inductance increases and the capacitance decreases as the coupling becomes tighter. This because the field is more concentrated between the center conductors, and less field exists directed to the common ground The coupling between the two conductors is put to good use as the well-known transmission component, the directional coupler. The amount of coupled energy resulting from any geometry depends on Z0e and Z0o, or perhaps those two impedances depend on the coupling. Whichever way you view it, they are interrelated and the expressions for wavelength matched lines are: Z0e =Z0 * ((1 + k)/(1 - k))^0.5 where k = 10^(-0.05*C) and C = coupling in decibels, and Z0 is the line characteristic impedance Z0o=Z0 * ((1 - k)/(1 + k))^0.5 From these, it is clear that Z0 = ( Z0o* Z0e )^0.5 In a 50 ohm system, for weak coupling, say 30 dB, k = 0.032, and so Z0e should be around 51.6 ohms and Z0o around 48.42 ohms. Intuitively, the coupling is weak, so the two impedances are almost identical to the system impedance, 50 ohms. With weak coupling there is minimal field distortion relative to the case where only one conductor exists. For a 3 dB broadside coupler in a 70.7 ohm section (Z0'), k = 0.707 so Z0e is approximately 170.7 ohms and Z0o is 29.3 ohms. Because the coupling is tight, these impedances are considerably different from the system impedance 70.7 ohms. Trace-widths and spacing in micro-strip or strip-line can be found in the standard reference works, or even some day on this site 8-). To be honest, there are almost as many modeling formulas as there are microwave engineers working in this field.

Gysel power splitter Updated March 31, 2011 Click here to go to our main page on couplers and splitters New for April 2011! We continue the analysis of Gysel splitters onto the following two pages: Rat-race versus Gysel splitters N-way Gysel splitters Ulrich Gysel published a IEEE paper titled "A New N-way Power Divider/Combiner Suitable for High-Power Applications" in 1975. His solution looks like a cross between a branchline and a Wilkinson, but it is also closely related to the rat-race coupler. It provides in-phase outputs, is configured most commonly as a five-port (for a two-way split with two terminations) but can serve as an N-way splitter as well. We will describe a two-way Gysel on this page. The Gysel divider is often used in kilowatt-level power combining, for example, if you want some redundancy in a 50,000 watt television transmitter you could use a five-way Gysel combiner with 15,000 watt tubes and be able to remove one of the tubes for service or replacement without taking the transmitter off-line. The terminations will need to be oil cooled to dump all of the wasted power, but that is doable. Gysel dividers are finding their way into the millimeter-wave spectrum, when gallium nitride amplifiers are combined in solid-state power amps (SSPAs) that will compete with vacuum tubes such as TWTs. New for April 2011! How do you pronounce "Gysel"? We didn't know, until we contacted Ulrich! It's simple, it rhymes with "diesel" and the "g" is the same sound in "good". Got that? The beauty of the Gysel, compared to the Wilkinson, is that the isolation resistors become oneports, which makes it much easier to realize good thermal performance, and they are much more forgiving in terms of parasitic phase response (they don't need to be "zero length").

Equal-split Gysel Gysel didn't provide closed-form equations for his splitter, he used a CAD program to optimize the line impedances. That's what we did when we designed one. We didn't use the transformer Z1 (let it be 50 ohms). The other three impedance work out to be: Z2=70.35 ohms Z3=50 ohms Z4=25 ohms (other values are possible, it affects the bandwidth). The big advantage of the Gysel power splitter is its power handling. In a Wilkinson splitter, the resistor is embedded into the network, and must provide a short phase length for the scheme to work. The terminations in a Gysel are equal to Z0, and can be high-power loads if power handling is a requirement (such as in a transmitter). The loads can be external to the power splitter, any length of Z0 transmission line can be added between the loads and the splitter. Etched onto a thin-film, there is no way to measure the resistor in a Wilkinson, because it is shorted out by the transmission lines around it. The Gysel allows the two resistors to be measured in parallel, even if they are grounded to the substrate. Below is a plot of the insertion loss from input to the two output ports. The one dB bandwidth is a remarkable 61.8%, which is more than double the bandwidth of a simple branchline power splitter.

Power split of an ideal -way Gysel power splitter The final plot shows the return loss at port 1, as well as the isolation between ports 2 and 3. The return loss bandwidth for 1.5:1 VSWR (-14 dB return loss) is 44.4%. This again beats the simple branchline which provides 20.8% bandwidth at 1.5:1 VSWR.

Return loss (blue) and isolation (red) of ideal two-port Gysel power splitter

Updated November 6, 2011 Search for directional couplers on EverythingRF.com

Click here to go to our main page on couplers and splitters Click here to go to our page on hybrid (3-dB) couplers Click here to go to our page on basic network theory Click here to go to our page on coupled-line couplers

Directional couplers are four-port circuits where one port is isolated from the input port. Directional couplers are passive reciprocal networks, which you can read more about on our page on basic network theory. All four ports are (ideally) matched, and the circuit is (ideally) lossless. Directional couplers can be realized in microstrip, stripline, coax and waveguide. They are used for sampling a signal, sometimes both the incident and reflected waves (this application is called a reflectometer, which is an important part of a network analyzer). Directional couplers generally use distributed properties of microwave circuits, the coupling feature is generally a quarter (or multiple) quarter-wavelengths. Lumped element couplers can be constructed as well. What do we mean by "directional"? A directional coupler has four ports, where one is regarded as the input, one is regarded as the "through" port (where most of the incident signal exits), one is regarded as the "coupled" port (where a fixed fraction of the input signal appears, usually expressed in dB), and one is regarded as the "isolated" port, which is usually terminated. If the signal is reversed so that it enter the "though" port, most of it exits the "input" port, but the coupled port is now the port that was previously regarded as the "isolated port". The coupled port is a function of which port is the incident port. Looking at the generic directional coupler schematic below, if port 1 is the incident port, port 2 is the through port (because it is connected with a straight line). Port 3 is the coupled port, and port 4 is the isolated port. For a signal incident on port 2, port 1 is the through port, port 4 is the coupled port and port 3 is the isolated port. Just follow the lines! Note: this paragraph was corrected in November 2011 thanks to Jim!

If you ever have any schematic questions, there is an IEEE standard that you can probably find with a google search: Graphic Symbols for Electrical and Electronics Diagrams, IEEE Standard 315-1975. Definitions Let's first look at some definitions using S-parameters. Let port 1 be the input port, port 2 be the "through" port. For a backward wave coupler, port 4 is the coupled port and port 3 is the isolated port. Ideally, power into port 1 will only appear at ports 2 and 4, with no power at port 3, but in real couplers some power leaks to port 3. For an incident signal at port 1 of power P1 (and output powers P2, P3 and P4 at ports 2, 3 and 4), then: Insertion Loss (IL) = 10*log(P1/P2)=-20*log(S21)

Coupling Factor (CF) = 10*log(P1/P4)=-20*log(S41) Isolation (I) = 10*log(P1/P3) = -20*log(S31) Directivity (D) = 10*log(P4/P3)=-20*log(S31/S41) Note that these numbers are positive in dB. Quite often, microwave engineers present these quantities as negative numbers, it is not a great faux pas, just look at the magnitude, Dude! Note that directivity requires two, two-port S-parameter measurements, the other quantities require only one. Directivity is the ratio of isolation to coupling factor. In decibels, isolation is equal to coupling factor plus directivity. Please send us any comments on the preceding statements, we are operating under a state of partial dyslexia and there is a possibility that we slipped up on a minus sign! Forward versus backward wave couplers Waveguide couplers couple in the forward direction (forward-wave couplers); a signal incident on port 1 will couple to port 3 (port 4 is isolated). Microstrip or stripline coupler are "backward wave" couplers. In the schematic above, that means for a signal incident on port 1, port 4 is the coupled port (port 3 is isolated). Coupler rule of thumb

The coupled port on a microstrip or stripline directional coupler is closest to the input port because it is a backward wave coupler. On a waveguide broadwall directional coupler, the coupled port is closest to the output port because it is a forward wave coupler. The Narda coupler below is made in stripline (you have to cut it apart to know that, but just trust us), which means it is a backward wave coupler. The input port is on the right, and the port facing up is the coupled port(the opposite port is terminated with that weird cone-shaped thingy which voids the warrantee if you remove it. Luckily Narda usually prints an arrow on the coupler to show how to use it, but the arrow is on the side that is hidden in the photo.

On the waveguide coupler below, the input is on the left, while the coupled port is on the right, pointing toward your left ear. There is a termination built into the guide opposite the coupled port, although you can't see it.

Bethe-hole coupler This is a waveguide directional coupler, using a single hole, and it works over a narrow band. The two guides are configured to (sorry we need to finish this section!!!) In waveguide, a two-hole coupler, two waveguides share a broad wall. Holes are 1/4 wave apart. In the foreword case the coupled signals add, in the reverse they subtract (180 apart) and disappear. Coupling factor is controlled by hole size. The "holes" are often x-shaped, and... Bi-directional coupler A directional coupler where the isolated port is not internally terminated. You can use such a coupler to form a reflectometer, but it is not recommended (use the dual-directional coupler you cheapskate!)

Dual-directional coupler Here we have two couplers in series, in opposing directions, with the isolated ports internally terminated. This component is the basis for the reflectometer.

Hybrid couplers A hybrid coupler is a special case, where a 3 dB split is desired between the through path and the coupled path. There are two types of hybrid couplers, 90 degree couplers (such as Langes or branchlines) and 180 degree hybrids (such as rat-races and magic tees). We have a separate page on this topic, click here! Reflectometer This is the component that allows you to measure S-parameter magnitudes using a network analyzer. A directional coupler only does what it is supposed to if it sees a matched impedance at all four ports. Errors due to finite directivity

Directivity can cause errors if load is not matched. 40 dB directivity will have a very small error, 20 dB may be unacceptable accuracy.

d (3 dB) couplers

Updated October 15, 2

h for hybrid couplers on EverythingRF.com

here to go to our main page on couplers and power splitters

here to go to our page on directional couplers

here to go to our page on coupled line couplers

here to go to our page on Lange couplers

d couplers are the special case of a four-port directional coupler that is designed for a 3-dB (equal) power split. Hybrids come in 90 degree or quadrature hybrids, and 180 degree hybrids. Why isn't there a "45 degree hybrid" you ask? Maybe it wouldn't isola urth port! Anyone that can submit a proof of this statement will win a gift!

d couplers are often used in creating reflection phase shifters.

the couplers discussed on this page have separate pages that go into detail on their operation. This page will help tie the entire m er.

egree hybrid couplers

include rat-race couplers and waveguide magic tees. Here we will look at the rat-race and introduce the vector and shorthand on that is often used when referring to 180 degree hybrid couplers.

a plot that shows the ideal, "classic" rat-race response (equal split at center frequency).

t-race gives about 32% bandwidth for a phase error of +/-10 degrees from the ideal 180 degree split.

gree hybrid couplers

are often called quadrature couplers, and include Lange couplers, the branchline coupler, overlay couplers, edge couplers, and slot hybrid couplers. Here we will just look at a branchline, and show you some of the "short hand" notation that is often used wh ng to hybrids.

the branchline is used as a combiner. The input signals are vectors of magnitude A and B, then the outputs are as shown. Note t se we are dealing with voltages, the outputs have a square-root-of-two factor. Power is split exactly in half (-3 dB), equal to the of the voltages.

et's look at it as a divider. Here only an input signal is present at port A. It splits by 3 dB at the two outputs, and is isolated from ally zero energy comes out this port).

erting a branchline coupler to a 180 degree hybrid

me applications like a monopulse comparator, available 180 degree hybrids complicate the layout because the "sum" port is betw lit ports. But it is as easy as adding a 90 degree section to one of the ports of a branchline. Below we've lengthened "Input A" by r-wavelength (impedance is Z0). Now we've got a sum and a difference output, just like a rat-race.

mage was corrected on March 28, 2008, previously the short-hand notation was wrong (our apologies for the error). Thanks to L ernhard who both pointed this out!

et's look at the response of this component, and compare it to the "classic" rat-race:

andwidth is less. If we just look at the frequency where the 180 degree split is within +/- 10 degrees, it is about 20% (0.9 to 1.1 Referring to the classic rat-race above, it has 32 percent bandwidth for the same phase error. back-to-back hybrids to achieve an RF-RF crossover

scading two hybrid couplers, you can create a four-port network that provides isolation between two RF paths that cross each oth posed to an airbridge or wirebond crossover). This should only be attempted where there is plenty of room for it, and it is probab over only 10% bandwidth. a schematic:

ere's the response. S41 is the path that is coupled, S21 and S31 are isolated, at least at the center frequency (in this case 1 GHz).

tructure has a severe bandwidth limitation. In most cases you should consider a 3-dimensional RF crossover first, like an airbridg irebond, before you resort to this!

Quadrature couplers Revised October 23, 2010 Search for quadrature couplers on EverythingRF.com

Click here to go to our main page on couplers and splitters Click here to go to our page on balanced amplifiers Click here to go to our page on power combining Click here to go to our page on reflection phase shifters See below for our discussion of the disappearing reflection coefficient in a balanced amplifier.

A quadrature coupler is one in which the input is split into two signals (usually with a goal of equal magnitudes) that are 90 degrees apart in phase. Types of quadrature couplers include branchline couplers (also known as quadrature hybrid couplers), Lange couplers and overlay couplers. Here's a clickable index to out material on quadrature couplers and splitters: Branchline coupler (separate page) Single-box branchline coupler Double-box branchline coupler Lumped-element branchline coupler Unequal-split branchline coupler Coupled-line coupler Overlay coupler Lange coupler Short-slot waveguide coupler Traveling wave splitter (new for February 2010!) Quadrature combining advantages on load pull effects (tres important!) One excellent use of quadrature couplers is to impedance match pairs of devices. The devices are arranged so that reflections from them are terminated in a load that is isolated from the quadrature coupler's input. This trick is possible only because of the 90 degree (quadrature) phase difference of the coupled and through arms. (This needs a figure, coming soon!). When FETs are combined using quadrature couplers, this is called a balanced amplifier. Quadrature couplers are also used to make reflective attenuator devices (such as shunt PIN diodes) absorptive. Another use for Lange or overlay couplers is to form a diplexer, where one port passes DC while the other port passes RF. This structure can be used as a bias tee. Overlay couplers Overlay couplers were very popular back in the 1970s within soft substrate stripline boards. Recently they are being used again, in some of the new multilayer media such as LTCC. The problem with overlay couplers is that none of the CAD software packages provides an easy way

to design them. So you end up spending a week with electromagnetic simulations in what would take five seconds if it was a planar quadrature coupler. Our page on coupled line couplers continues this discussion. Disappearing reflection coefficient

Was Smitty really marooned on the island all week long? Quadrature couplers are often used to power combine amplifiers, such a structure is called a balanced amplifier. This allows an important degree of freedom, so long as the amplifiers have the matched reflection coefficients on inputs and outputs, the network will have good impedance input and output impedance matches. What's going on? The figure below will help you visualize it. A signal of amplitude "V" at phase angle zero enters the network, and splits. The upper amplifier is excited with half of the signal (0.707 x voltage) at phase angle zero, while the lower amp see the same signal but phase shifted -90 degrees.

Now, what happens in the ideal case where your two amplifiers have the same reflection coefficient? The signal returned to the input port cancels out, and all of the reflected power goes to the load.

Branchline couplers Revised May 11, 2011 Click here to go to our main page on couplers and splitters Click here to go to our page on quadrature couplers Click here to go to our page on Lange couplers New for July 2009! Reflection attenuators are one application of branchline couplers. Here's a clickable index to out material on branchline couplers: Single-box branchline couplers Double-box branchline couplers Lumped-element branchline couplers Unequal-split branchline couplers Three-way branchline couplers (New for April 2011!) Patch coupler (New for May 2011!) Single-box branchline couplers The branchline the simplest type of quadrature coupler, since the circuitry is entirely planar. A ideal single-box branchline coupler is shown below. Each transmission line is a quarter wavelength. However, 3/4, 5/4 or 7/4 wavelengths (etc.) could also be used on each arm if the circuit layout requires it, the penalty is paid in decreasing bandwidth. A signal entering the top left port (port 1 in the figure) is split into two quadrature signals on the right (ports 2 and 3), with the remaining port 4 fully isolated from the input port at the center frequency. Remember that the lower output port (port 3) has the most negative transmission phase since it has the farthest path to travel.

Ideal branchline coupler The next figure shows the response of an ideal branchline coupler where the each side is a quarter wavelength at 10,000 MHz (10 GHz). The first graph shows the losses from the input to the two output arms. S21 is the transmission loss from the top port to the upper right port, S31 is from the input to the lower right port. Using the ideal transmission line impedances shown above provides a equal 3 dB split at the center frequency. The markers have been aligned to show the 1-dB bandwidth of the coupler, which is 2580 MHz or 25.8%.

Power split of ideal branchline coupler The second graph shows that the bandwidth where the device has better than 14 dB return loss (1.5:1 VSWR) is 2080 MHz, or 20.8%. The isolation (power coupled to the terminated port) is also plotted here and is very nearly equal to the return loss.

Return loss (blue) and isolation (red) of ideal branchline coupler The next plot shows the phase difference between the two outputs (ideally 90 degrees, remember?) For +/-10 degrees the bandwidth is about 4300 MHz, or 43%.

Phase response of ideal branchline coupler

Double-box branchline coupler

New for February 2006: we've added a separate page on the double-box coupler, which explores this topic further! As with the Wilkinson power splitter, the bandwidth of a branchline coupler can be improved by adding sections. The next figure shows a "double-box" branchline coupler with its ideal impedances. We've never seen this in a text book, have you? Microwaves101 rules! In a 50-ohm system, the line impedances of the end vertical segments work out to be 120.71 ohms, and the center vertical segment 70.71 ohms. In practice, it may be hard to accurately achieve the 120.71 ohm impedance lines accurately.

Ideal double-box branchline coupler The next three figures show the frequency response of the ideal double-box branchline coupler, centered at 10,000 MHz (10 GHz). In this case, the 1-dB response of the coupled arm is 35%, the 14 dB return loss band (1.5:1 VSWR) is 41%, and the +/-10 degrees phase difference is 50%. However, the tradeoff for the extra bandwidth in real life will be added loss of the second box section, not to mention the added size.

Power split of ideal double-box branchline coupler

Return loss (blue) and isolation (red) of ideal double-box branchline coupler

Phase response of ideal double-box branchline coupler

As a final illustration of improving the bandwidth of the branchline coupler, a double-box structure was tuned to increase the frequency bands for 1-dB coupling and 14 dB return loss. Note that the equal 3 dB power split at the center frequency must be somewhat corrupted as a tradeoff. The arm impedances of this coupler are now 38 ohms for the series arms and 100 ohms and 65 ohms as shown for the three shunt arms, as shown below. The figures speak to the results. The 1-dB band is now 55.6%, while the 14 dB return loss band is 58.4%.

Double-box hybrid coupler tuned for more bandwidth

Power split of tuned double-box hybrid coupler

Return loss (blue) and isolation (red) of tuned double-box branchline coupler Lumped element branchline coupler New for March 2007: here's a page on lumped element Wilkinsons!

Lumped elements can be used to approximate transmission lines in a branchline coupler. The structure for a quarter wave transmission line can be realized with a pair of shunt capacitors of equal value, separated by a series inductor (a "pi" network). By optimizing the inductor and capacitor values, different line impedances can be "faked". Why would you want to use shunt elements? So you make branchline couplers at lower microwave frequencies (UHF through Sband) and not have to deal with huge transmission line lengths. Below we present a lumped-element quadrature coupler which was optimized to work at 100 MHz. The lumped elements are ideal; if this was a real design that we were getting paid for we would have included all of the parasitic elements into the capacitor and inductor models, which is necessary if you want a true prediction of how the circuit will perform on the bench. Note that if you want to scale the design to a different frequency, all you have to do is scale the element values with inversely with frequency (use half the values for a 200 MHz design).

Schematic representation of 80 to 100 MHz lumped-element quadrature coupler Below are the power splits of port 2 and port 3. Note that we managed to get them within maybe 1.5 dB of each other at the center frequency. Maybe you could do better, Mr. Smart Guy.

Below is the phase difference between the paths 1-2 and 1-3. Note that 90 degrees is achieved with less than 1 degree of error from 80 to 100 MHz. In real life it won't be as good because you will have to allow for tolerances on the parts.

Finally, we show the port matching (S11) and the isolation (S23). Note they are similar, this is true of most quadrature coupler designs.

Here's a branchline coupler in MMIC representation, with the quarter wavelength arms changed into lumped inductive and capacitive elements. It's a double-box structure on four-mil GaAs. It worked great at S-band, with less than 1 dB resistive loss in spite of all those spiral inductors! Such a structure can make an efficient power combiner for the IF output of a higher-frequency image-rejection mixer.

Lumped-element branchline coupler (MMIC representation) Unequal-split branchline couplers By varying the impedances of the opposite arms in a branchline coupler, unequal power splits can be obtained, as shown in the figure below.

Unequal branchline power splitter The equations for the line impedances Z0A and Z0B are given below, as functions of the power split PA/PB and the system impedance Z0.

The plot below shows the characteristic impedances Z0A and Z0B, for a fifty ohm system, as a function of the coupling ratio PA/PB expressed in dB. Note that two very different topologies result when PA is greater than PB (Z0A and Z0B are higher impedance than in an equal-split branchline) and when PA is less than PB (Z0A and Z0B are lower impedance than in an equal-split branchline).

Line impedances of unequal-split branchline coupler

The chart above does not completely tell the story of the tradeoffs made when you select which port provides the most power. Check out the power split responses for PA/PB=0.25 and PA/PB=4.0 below. The bandwidth for PA/PB=4.0 is far superior.

Unequal-split branchline frequency response, PA/PB=0.25

Unequal-split branchline frequency response, PA/PB=4.0 Check out our unequal-split power divider calculator, it handles Wilkinsons, rat-races and branch-line couplers!

Coupled-line couplers Updated September 16, 2008 Click here to go to our even and odd mode analysis page Click here to go to our main page on couplers and splitters Click here to go to our page on directional couplers New for October 2008! Here's a new page on a microstrip "3 dB" coupler! It's about time we dealt with couplers that are truly coupled lines, as opposed to "direct coupled" couplers such as the Wilkinson and the branchline. The Lange coupler is one form of coupled line coupler. Coupled lines occur when two transmission lines are close enough in proximity so that energy from one line passes to the other. Coupled lines are used in couplers (usually quadrature couplers) as well as transmission line filters. Line can be coupled in three ways:

edge coupled end coupled broadside coupled

In order to make a quadrature coupled-line coupler you need to couple a quarter-wave section; end-coupled structures are not useful in this case. That leaves two broad categories of coupled line couplers, edge coupled, and broadside coupled (and perhaps some gray territory in between!) Both can be realized in microstrip or stripline, but stripline is best. Theory of coupled line couplers For two coupled lines forming a four-port network, there are two things that have to occur with coupled lines to become a useable coupler with directivity and quadrature phase: 1. The coupled section must be a quarterwave 2. The product of even and odd mode impedance must be equal to Z0^2 It's time to define some port numbers. Let port 1 be the input port. The port that is directly coupled to port 2, which is one of the two output ports. The other output port is directly across from the input port, we'll call it port 3. Under ideal conditions, a signal incident on port 1 will transfer zero power to port 4; this is called the isolated port.

Advantages of coupled line couplers Bandwidth is better than direct coupled couplers like the branchline. Why does the coupled-line have a natural 90 degree phase split? This is a great question, and if you've a better answer then the one provided, please send it in! The coupling occurs via two mechanisms, voltage, and mutual inductance (current). The mutual inductance coupling has a minus sign associated with it, the voltage coupling does not. The combined effect not only reverses the signal flow in the coupled line (backward coupling) but it puts the two signals 90 degrees out of phase. Ideal coupler ADS model In ADS you can use an ideal coupled line, which is described by its even and odd mode impedances and center frequency. The product of the two impedances being Z0^2, you can easily create an equation to solve one from the other.

Now you can use ADS's tune feature to vary Ze until a 3 dB split is achieved. It turns out that to get a 3 dB split (equal power) in fifty ohm system impedance the even mode needs to be ~121.5 ohms and the odd mode impedance must be ~20.6 ohms (in our ADS network Zo is calculated automatically from Ze and Z0).

For the ideal coupler, you don't have to plot the phase between the two output arms (port 2 and port 3 in this case). It's automatically 90 degrees! Analyzing coupled line couplers using Excel As you know by now, one motto of Microwaves101 is that "anything that can be analyzed in Excel, should be analyzed in Excel". Coupled line couplers are no exception. Here's the same coupler we analyzed in ADS, now in Excel. You can get a free copy of the spreadsheet that did this remarkable piece of work in our download area!

Here's a slightly undercoupled coupler.

Here's the coupler, slightly overcoupled.

Lange couplers Updated September 16, 2008 Click here to go to our main page on couplers and splitters Click here to go to our page on quadrature couplers Click here to go to our page on microstrip Click here to go to our page on even and odd mode impedances Click here to go to our page on balanced amplifiers New for October 2008! Here's a different way to create a microstrip "3 dB" coupler!

We are off to a good start on this ubiquitous microwave component, thanks to Microwaves101 contributor and Hall-of-Famer Julius Lange! Lange couplers are a modern miracle of microwaves. Here is the quadrature coupler at its best: low loss, wide bandwidth, compact layout, and CAD elements good to go. This coupler spawned an entire industry of wideband hybrid-style amplifiers that is still here today in spite of a lot of good competition from MMICs. Here's a clickable index to this page: History of the Lange coupler Theory of the Lange coupler Modeling a Lange coupler Limitations of Lange couplers Layout examples History of the Lange coupler Here we'll let Dr. Lange tell you in his own words how the famous coupler was developed: "In 1969 we at Texas Instruments were building microwave amplifiers on thin film ceramic substrates. We were using the scheme invented by Engelbrecht at Bell Labs, which required 3dB quadrature couplers. The challenge was to get tight coupling on single layer microstrip. On the other hand our transistors had too much coupling between the interdigitated base and emitter fingers. So why not an interdigitated coupler? I built it; and it did not work well. Then I remembered that geometric symmetry guarantees quadrature, a 90 split between the outputs. So I moved some of the crossovers from the ends to the middle; and it worked! We had a microstrip interdigitated quadrature coupler with low loss and wide, one octave, bandwidth. " Theory of the Lange coupler For a given input on a Lange coupler (or other types of quadrature couplers for that matter), the three output ports can be denoted

isolated port through port coupled port (-90 degree transmission angle compared to through port)

Referring to the six finger Lange below, if the bottom left port is the input, the top left is the "coupled" port, the top right is the "through" port and the bottom right is the "isolated" port. You can find the "through" port easily in a Lange because it has a DC connection to the input. The

isolated port is on the same side of the coupler as the input for a normal Lange. More about abnormal Langes later!

Time for a Lange rule of thumb:

The physical length of a Lange coupler is approximately equal to one quarter-wavelength at the center frequency on the host substrate. The combined width of the strips is comparable to the width of a Z0 (fifty-ohm) line on the host substrate. Modeling a Lange coupler All modern CAD programs for linear simulation of microwave circuits have the capability to model a Lange coupler. Here we will show you an example using Agilent's ADS. The model is shown below: we have chosen to build our Lange on 15 mil alumina. Because the length of the Lange is 100 mils, which works out to a quarter wavelength at 12 GHz, that is the center frequency. We played around with the strip width and gaps until we achieved exactly 3 dB coupling at the center. Widths and gaps of 1.25 mils are considered doable in a good thin-film shop.

Below is the response predicted by ADS. For a signal input at port 1, we see a three-dB split at ports 2 and 3. The isolated port gets a signal that is down by more than 25 dB.

One more important plot is the phase difference between the output ports. Here is one of the major attractions to the Lange coupler, you won't see such a beautiful quadrature response on a branchline coupler!

Now let's play around with the gap dimension. Below are two response, the first one the gap has been increased to 1.5 mils. Notice the coupled port receives less power than the through port. This coupler would be called "under-coupled".

The next figure shows what happens when the gap dimension is reduced to 1.0 mils. Now we see an "over-coupled" response. This is often the most desirable case, especially when your application is wide band. The "coupling error", defined as the difference in magnitude between the two output ports, is less than 1.0 dB from 8 to 16 GHz, an octave of bandwidth. Referring to the first case where exactly 3 dB was achieved at 12 GHz, the coupling error of 1 dB is only maintained from 9 to 15 GHz.

Limitations of Lange couplers Lange couplers have been used from UHF to Q-band, perhaps higher. But as you go up in frequency, you'll need to reduce your substrate height to get microstrip to behave (see microstrip height rule of thumb). Reduced height means reduced strip width, which is the ultimate limitation. At some point the strips get so narrow that even if they don't fail your design rules, they will start to become lossy because there just isn't much metal to provide a conductor. Langes on alumina are usually restricted to applications where the substrate is 15 mils or thicker; this means you'll see alumina Langes operate no higher than 25 GHz. If you attempted to make a Lange on 10-mil alumina, the strip widths would need to be less than 1 mil (25 microns). In MMIC applications, Langes can be made on 4-mil substrates, but it is a fools errand to try to make them on 2-mil substrates. According to our rule of thumb, that means you'll never see a Lange above 80 GHz (four mil GaAs craps out there). If you attempted to make a Lange on 2mil GaAs, the strip widths would need to be about five microns. Forgetaboutit! Layout examples Lange couplers have many different layout permutations, some of which have been patented! Four and six-finger Lange couplers are two implementations. Below is a photo of a four-finger Lange coupler on 25 mil alumina. If you look closely you can see four bond wires, which are needed to connect the strips (one on the left side, two in the center, and one on the right side). Some thin-film vendors can make airbridges that will eliminate these tiny bondwires, but you will add some serious money to the fabrication cost.

Microstrip "3 dB" coupler Updated November 27, 2009 Click here to go to our page on coupled-line couplers Click here to go to our Lange coupler page Click here to go to our main couplers and splitters page Click here to go to our main page on microstrip New for October 2008! How can you create a 3 dB (equal split) edge coupler on a microstrip board? If you downloaded our even/odd impedance analysis spreadsheet (look in the download area) , you'd compute that for equal spit in a fifty ohm system, even mode impedance is 120.7 and the odd mode impedance is 20.7 ohms. Now think about how difficult it is to realize a 120.7 ohm line in microstrip, it would be a very skinny... then think about whether edge coupling could

ever accomplish 20.7 ohms impedance. Mathematically it's possible, but you might need the gap between the lines to be 10 micro-inches (250 nanometers!), which is not going to happen in real life. Plus, the structure would be very sensitive to strip thickness. Sorry if we just crushed your dream! Before we go on, let's point out that a Lange coupler is a 3 dB microstrip coupler that accomplishes the required even/odd impedances for equal split. Julius Lange divided the lines into multiple smaller lines to provide more than one edge to couple to. The lines and gaps on an alumina Lange coupler are on the order of 1 mil (25 microns), which is manufacturable but you will pay a premium price. You'll never see a Lange coupler on a copper printed wiring board, because tolerances on the order of one mil. But there's another way! This technique was suggested by Kevin a while ago. It turns out that you can cascade edge couplers and the coupling factor will increase. Kevin predicted that two 8.2 dB couplers could make a 3 dB coupler. We made our own version of this in ADS without thinking about it too much (a dangerous practice!) We went right to microstrip stimulation (nit ideal T-lines), in the example below we chose 10 mil alumina. We played around with the coupling of a single coupler, and arrived at 9 dB as a good value to cascade (we need to post an image of the coupler by itself, check back in a few days). Note that the network requires one termination (Term2 on the schematic) and it will require at least one jumper when you lay it out. The important thing is that a 9 dB edge coupler on microstrip is easy to realize because the even mode impedance is 72.5 and the odd mode is 35.5 ohms. The line width on 10 mil alumina is 7.7 mils while the gap is 2.5 mils. No worries!

TL1 and TL2 are necessary to lay out the coupler. We found that the response is very sensitive to their lengths, which have to be equal. Here's the frequency response, including return loss and isolation. Note that by symmetry, all ports will have the same insertion loss if the rest of the ports are terminated in matched loads.

Let's talk about another problem with microstrip couplers in general. The velocity that signals travel (speed of light) is different for even and odd modes, which degrades performance. In the example the best isolation at 9 GHz is not achieved at the coupler's center frequency (10 GHz). This is a disadvantage compared to stripline. If you tried this trick on stripline you'd get a much better response! Here's a close-up of the direct and coupled port transmission coefficients. The coupler is overcoupled, but we look at that as a good thing because it increases the one-dB bandwidth.

If anyone want to analyze the signal flow, be our guest!

Short-slot waveguide hybrid Updated April 9, 2008 Click here to go to our main page on waveguide Click here to our main page on couplers and splitters Revised for April 2008! We now have an example posted, thanks to Paul Klasmann! Scroll down a piece... The short slot hybrid is a four-port device, a form of quadrature coupler in waveguide. It provides an equal split, with outputs 90 degrees apart (quadrature). We are waiting for permission from IEEE to post some pictures of this component that we found in a 1963 article titled Semiconductor Switching and Limiting Using 3-dB Short-Slot (Hybrid) Couplers by V. J. Higgins. The original paper on the topic is by Henry J. Riblett, entitled The Short-Slot Hybrid Junction, published by the IRE in February 1952. The basic construction of the short-slot coupler uses two rectangular waveguides that share a wall. Between then a single slot couples three dB (half) of the incident power from one guide into the second guide. The two resulting waves are 90 degrees out of phase (in quadrature). As is the case with most waveguide structures, it takes some tuning to get this puppy to work. It's a narrow-band device, with isolation of 20 dB over maybe 10% bandwidth.

94 GHz short-slot coupler example This example came to us from Dr. Paul Klasmann, who works at Q-Par Angus Ltd. 94 GHz is smack in the middle of W-band. If microwaves were audible, only a dog could hear this frequency! The plot below was taken from an electromagnetic simulator tool, showing E-field intensity within the coupler. The input port is on the lower right (where the E-field is seen to be strongest). The "through" port is on the upper right and the quadrature (90 degree) output is on the upper left. The two output ports are of equal intensity indicating that a 3 dB split has been achieved. Note the absence of a discernible E-field at the isolated port (lower left). We regret that prior to April 9, 2008 the input/output ports were designated incorrectly in this paragraph, but they are correct now.

The figure below shows the bottom half of the waveguide block with the ports numbered 1 to 4 anticlockwise starting at the bottom right. Note to readers: you have to be very careful when you split a waveguide assembly this way, the braze joints have to be perfect or the E-field won't see the ideal short circuit it needs at the short wall. In this case the coupler block is split at the top of the short wall so that there's no discontinuity in the region of peak current (at the middle of the short wall). The two posts you see inside the slot are tuning slugs.

S-parameter prediction is shown below. From this you can get an idea of the useful bandwidth of the short-sot coupler, it should be considered for relatively narrow band applications only. However, at 94 GHz, 1% bandwidth is about one GHz wide!

Here's a photo of the completed assembly. Click on the image for a higher resolution image!

Initial measurements, further adjustment of stubs are required to improve the isolation and amplitude balance. Ideally both outputs should be at -3.01 dB but there is around 0.5 dB "real" insertion loss (S21 and S31 are ~-3.5 dB).

Phase between outputs should ideally be 90 degrees. The measurement below shows that this has been achieved at the center frequency.

If you get in touch with Q-Par Angus about this coupler or other millimeterwave products, tell them they should cough up some dinero and sponsor this page! Saleh power divider Updated November 11, 2005 Click here to go to our main page on power dividers and couplers New for November 2005! Here's a nifty "quadrature-like" power divider that makes a four-way split where the adjacent outputs are 90 degrees apart in phase. It was invented by Adel Saleh and first published by the IEEE in 1981 in an article entitled Planar Multiport Quadrature-Like Power Dividers/Combiners. We'll add more to this write-up soon, but meanwhile, here's a model of a four-way Saleh divider and some predicted performance.

Traveling wave splitter Updated January 30, 2010 Click here to go to our main page on couplers and splitters

Click here to go to our page on quadrature splitters New for February 2010! The traveling wave splitter is worth considering if you are combining four amplifiers, that are on the order of a quarter-wavelength in width, as the four outputs are located at intervals of a quarter-wavelength. The traveling wave power divider/combiner was first described in the reference below: Bert and Kaminsky, The Traveling Wave Power Divider/Combiner, IEEE MTT-S, May 1980. Note that on this web site, we try to use the convention that the word "splitter" is for passive networks that can be used as dividers and combiners alike, with the stipulation that they don't have an isolated port (we call that a "coupler). Let's also point out that we used the American spelling of "traveling", not the British or Aussie version which has two L's. Don't get your panties in a bunch over the two spellings, there are plenty of other problems that better deserve your attention... Properties of the traveling wave splitter include the quadrature phasing between the four split ports, i.e. they are 90 degrees apart in phase. This provides the "magical property" that mismatched amplifiers that have identical reflection coefficients will combine to a an ideally perfect input match (the reflections are dissipated in the loads). The schematic below come right from the reference, which of course includes all manner of equations for determining the element values. For now we will let it stand alone, if you want to design a traveling wave combiner, you can scale this one, you can use an optimizer, or you can look up the reference. Surely there are many degrees of freedom, starting with the input transformer values.

Transmission magnitude

Transmission phase

Port match

Isolations

Let's take a look at the claim that mismatches at the split ports don't affect the input match. In the schematic below, we have rmodified the port impedances on the four outputs to be 100 ohms, which represents a mismatch of 2:1 VSWR, all at the same phase angle (0 degrees).

Now look at the response. The transmision coefficient is down by about 0.5 dB, this is a direct result of mismatch loss. The transmission phase characteristic is nearly unaffected. The return loss at port 1 is preserved, while the other ports are reduced to 2:1 VSWR (9.54 dB return loss). The isolation is also preserved at center frequency.

Transmission magnitude

Transmission phase

Port match

Isolations

Rat-race couplers Revised March 20, 2011 Click here to go to our main page on couplers and splitters Click here to go to our page on magic tees (a waveguide network with similar properties to the rat-race) Click here to go to our page on cascaded ratrace couplers (new for April 2011!) Click here to go to our page on rat-race versus Gysel couplers (new for April 2011!) Applications of rat-race couplers are numerous, and include mixers and phase shifters. The ratrace gets its name from its circular shape, shown below. The circumference is 1.5 wavelengths. For an equal-split rat-race coupler, the impedance of the entire ring is fixed at 1.41xZ0, or 70.7

ohms for a 50 ohm system. For an input signal Vin, the outputs at ports 2 and 4 (thanks, Tom!) are equal in magnitude, but 180 degrees out of phase.

Rat-race coupler (equal power split) Note: Richard from Australia reminded us recently that the rat-race can also be used as an inphase splitter. If you feed port 4 in the above figure, port 1 and port 3 will receive a 3 dB split, and port 2 will be isolated. This is why port 4 is referred to as the "sum port" and port 1 is referred to as the "delta port", an important consideration in designing monopulse comparators. We'll expand on this idea later. Thanks! Also, coming soon... a two-stage rat-race design! If you are interested, please email us and we'll try to move this up in the queue. The coupling of the two arms is shown in the figure below, for an ideal rat-race coupler centered at 10 GHz (10,000 MHz). An equal power split of 3 dB occurs at only the center frequency. The 1-dB bandwidth of the coupled port (S41) is shown by the markers to be 3760 MHz, or 37.6 percent.

Power split of ideal rat-race coupler The graph below illustrates the impedance match of the same ideal rat-race coupler, at ports 1 and 4. By symmetry, the impedance match at port 3 is the same as at port 1 (S11=S33). For better than 2.0:1 VSWR (14 dB return loss), a bandwidth of 4280 MHz (42.8%) is obtained.

Impedance match of ideal rat-race coupler The next graph shows the isolation between port 1 and port 3 (S31). In the ideal case, it is infinite at the center frequency. The bandwidth over which greater than 20 dB isolation is obtained is 3140 MHz, or 31.4%.

Isolation of ideal rat-race coupler Below the phase difference between arms 2 and 4 is plotted. At the center frequency. a perfect 180 degree difference is observed. The bandwidth that better than +/- 10 degrees is maintained is 3200 MHz, or 32%.

Unequal-split rat-race couplers In order to provide an unequal split, the impedances of the four arms are varied in pairs, as shown below.

Unequal-split rat-race power divider Equations for the Z0A and Z0B line impedances, as a function of the power split PA/PB, are given below:

Z0A and Z0B are graphed below versus the power split express in dB (coupling ratio) for a 50-0hm system. Click here for info on how to think in dB.

The graph below shows the frequency response for a rat-race coupler where PA/PB=0.25. This corresponds to a 50-ohm power divider where the power out of port 2 (PA) is six dB below the power out of port 4 (PB). Solving the above equations for the line impedances yields Z0A=111.6 ohms, and Z0B=55.9 ohms. Note that in many real-life cases, this coupler may prove impractical because a line impedance as high as 111.6 ohms may be difficult to accurately achieve in a 50ohm system.

Unequal-split rat-race frequency response, PA/PB=0.25

The graph below shows the frequency response for a rat-race coupler where PA/PB=4.0. This corresponds to a power divider where the power out of port 2 (PA)is six dB higher than the power out of port 4 (PB). The line impedances are opposite to the case where PA/PB=0.25; here Z0A=55.9 ohms, and Z0B= 111.6 ohms.

Unequal-split rat-race frequency response, PA/PB=4.0 Check out our all-new unequal-split power divider calculator, it handles Wilkinsons, rat-races and branchline couplers! Magic tees Updated March 17, 2008 Click here to go to our main page on couplers and splitters Click here to go to our main page on waveguide Click here to go to our page on rat-race couplers Oh ho ho, it's Magic by Pilot, a Scottish group, 1975 A magic tee is a four-port, 180 degree hybrid splitter, realized in waveguide. Originally developed in World-War II, and first published by W. A. Tyrell in a 1947 IRE paper, it has very similar properties to the rat-race coupler, which is usually realized in microstrip or stripline. Like all of the coupler and splitter structures, the magic tee can be used as a power combiner, or a divider. It is ideally lossless, so that all power into one port can be assumed to exit the remaining ports.

The convention used in Pozar's book "Microwave Engineering" is shown on the following figure, though not all waveguide vendors adhere to it. Port 1 is the (sum) port, and is sometimes called the H-plane port, and sometimes called the P-port for "parallel". A signal incident on port 1 equally splits between ports 2 and 3, and the resulting signals are in phase. Ports 2 and 3 are sometimes called the co-linear ports (thanks Bill!), because they are the only two that are in line with each other. Port 4 is the (difference or delta) port, and is sometimes called the E-plane port, or the S-port for "series". A signal incident on the difference port splits equally between ports 2 and 3, but the resulting signals are 180 degrees out of phase (thanks Harald!)

The math behind the magic tee is too much for us to present here for now. Maybe it's just better to leave it as "magic" and not try to analyze it. We used HFSS v10 to model a magic tee, using an example right out of the Ansoft HFSS book. This exercise will help you visualize how the E-field of a signal entering the sum port remains in the same up-and-down direction and polarity as it splits to ports 2 and 3, while the E-field of a signal entering the delta port wraps around into two opposing polarities as it splits between ports 2 and 3. The interior dimensions of the waveguide are 50 mm by 20 mm. This is not a standard waveguide size, the broad wall is approximately two inches, which puts it close to WR187. You can tell that Ansoft is run by mathematicians, not microwave engineers, or they would have picked a "real" waveguide band. Below is the model:

The next picture shows how it was meshed:

The next two pictures show the E-field vectors for signals entering the sum port, then the delta port. Now you can see how the delta port excites opposing phases in the CO-linear arms.

Cool stuff! The next plot shows the phase of the transmission coefficients out the CO-linear ports, when driven by the delta port. Note the 180 degree difference.

Last, here are some of the S-parameters of the fourport network, including the transmission coefficient between sum and delta ports (red trace), which is better than -50 dB. The input match S11 (blue trace) could be better, which would require some tuning. Guess that's why you'd never buy a magic tee from Ansoft! If you go back to Tyrell's paper, he suggests adding tuning to the and arms. This can be done with tuning screws, rods, or plates. We're not going to get into this right now.

The magic tee in the photo below is WR-62. If you cut it open you could see how it was tuned. We like it just fine in one piece.

Resistive power splitters Revised July 4, 2011 Click here to go to our main page on couplers and splitters

Click here to go our page on resistive taps. Click here to check out the Chris Owen's resistive unequal splitter Click here to check out Greg Adams' resistive unequal splitter Wye and delta equal-split resistive splitters If you need an unequal split, check out the Owen and the Adams splitters (links above). Use the Owen splitter for maximum isolation, or the Adams splitter for highest efficiency. The wye and tee provide equal split. Resistive power dividers are easy to understand, can be made very compact, and are naturally wideband, working down to zero frequency (DC). Their down side is that a two-way resistive splitter suffers 10xlog(1/2) or 3.0103 dB of real resistive loss , as opposed to a lossless splitter like a hybrid. Accounting for 3.0103 dB real loss and 3.0103 dB power split, the net power transfer loss you will observe from input to one of two outputs is 6.0206 dB for a two-way resistive splitter, so they are often called 6 dB splitters. Dig? For applications where loss is critical such as power amplifier combiners, the extra loss of a resistive splitter is an unacceptable compromise. But in others, especially in test equipment where power is just an outlet strip away, resistive splitters have their place. For the 2-way resistive dividers shown below, one half of the power that flows through it is wasted in the resistors. For example, a one watt signal at port 1 will result in two quarter-watt signals at ports 2 and 3 (down by -6 dB). Because of the loss within the network, you have to carefully consider the power dissipation and resistor power ratings. You can put many watts through a "lossless" divider such as a rat-race or branchline coupler. But a watt might burn out a resistive divider. Another disadvantage it the none of the ports are truly isolated from each other. It's time for a Microwaves101 rule of thumb!

The isolation of a resistive splitter is equal to its insertion loss. The 6 dB three-port splitter has (ideally) 6.02 dB loss from any one port to any other port (S21=S31=S23). The advantages of resistive dividers are size (it can be very small since it contains only lumped elements and not distributed elements), and they can be extremely broadband. Indeed, a resistive power divider is the only splitter that works down to zero frequency (DC). It's so broadband, that we didn't even bother to make a frequency response plot for you! Below are schematics of the two options for three-port resistive splitters, the "delta" and the "wye". Resistor values as shown will ensure that each port is impedance matched to Z0. These schematics and many others are available in a Word file that you can get in our download area,

it's called Electronic_Symbols.doc. You'll find it comes in handy for creating simple block diagrams in Word, PowerPoint or Excel.

Delta 6 dB resistive splitter

Wye 6 dB resistive splitter N-way resistive splitters You can make N-way resistive splitters easily from the Wye splitter. The Delta resistor becomes a nightmare network for more than a 2-way split, it can't be constructed in two dimensions. The appropriate resistors for an N-port Wye splitter are found by the equation: R=Z0x(N-1)/(N+1) For examples, a three-way splitter needs resistors of Z0/2, while a four-way splitter needs resistors of 3xZ0/5, and so on. The efficiency of a resistive splitter gets worse and worse the more arms you split to. The transferred power ratio is (1/N)^2, as opposed to a lossless splitter that varies as 1/N. So for a four-way splitter, only 1/16 of the power makes it out to one of the four matched loads. It's time for another Microwaves101 rule of thumb!

To put it simply, the resistive splitter has double the dB loss compared to a lossless splitter's insertion loss. Thus a two-way resistive splitter transfers -6.02 dB power to each arm, a

three-way splitter transfers -9.54 dB, a four-way transfers -12.04 dB, etc. An infinite-way resistive splitter would lose 100% of the incident power and transfer nothing to the loads! Fractional dissipation in the Wye resistive splitter New for February 2007! We've solved this algebra problem for the Wye splitter, here it is. If the resistor closest to the common port is designated "resistor A" and the other two resistors are designated "resistor B" and "resistor C" then: Power dissipated in resistor A: PDissA=Pin x (N-1)/(N+1) Power dissipated in B and C are equal by symmetry. PDissB=Pin x (N-1)/[(N+1)xN^2)] Let's tie it all together. The table below shows the fractional dissipation and the output powers. The "power factor" is a measure of the efficiency of the network, it is the sum of all the output powers divided by the input power. N 2 3 4 5 6 PDissA 33.3% (-4.77 dB) 50% (-3.01 dB) 60% (-2.22 dB) 66.7% (-1.76 dB) 71.4% (-1.46 dB) PDissB 8.33% (-10.79 dB) 5.56% (-12.55 dB) 3.75% (-14.26 dB) 2.67% (-15.74 dB) 1.98% (-17.02 dB) Pout 25% (-6.02 dB) 11.1% (-9.54 dB) 6.25% (-12.04 dB) 4% (-13.98 dB) 2.78% (-15.56 dB) Power factor 50% (-3.01 dB) 33.3% (-4.77 dB) 25% (-6.02 dB) 20% (-6.99 dB) 16.7% (-7.78 dB)

Fractional dissipation in N=2 Delta resistive splitter We only use the delta splitter for N=2, it doesn't make sense for higher order splitters because it becomes a 3D nightmare. The fractional dissipation in an N=2 delta splitter is easy to calculate. The resistors that are in series with the split ports dissipate as much power as the two outputs. If the series resistors are labeled "resistor A" and the resistor that shunts the output ports is "resistor B", then the dissipation is given by: N PDissA PDissB Pout Power factor

25% (-6.02 dB)

25% (-6.02 dB)

50% (-3.01 dB)

Resistive power taps Updated February 19, 2007 Click here to go to our main page on resistive splitters

New for March 2007! Oh, no! This particular power tap is unmatched, resistive, and has zero directivity, why would you want to use that? Fear not, Mr. Bill, we haven't seen Sluggo around in ages. But on this page we'll show you that the resistive tap is an excellent technique for coupling power, and often has advantages over all other techniques. The resistive power tap is the answer to the question "how can I create an extremely compact -20 dB coupling network with less than 1 dB loss over infinite bandwidth?"

Advantages of the resistive tap include ultra-wide bandwidth, low insertion loss for low coupling values. For a 20 dB coupler, your circuit will have a hard time noticing the loss in the primary path due to the missing power that is coupled off! The resistive power tap schematic is shown below. In this case, we use R1 and R2 to steal power from the primary path, with the stipulation that port 3 sees an impedance matched to Z0. Using Ohm's law and some manipulation, R1 can be expressed in terms of R2 and Z0 as: R1=(Z02+Z0R2)/(2R2-2Z0) The expression for R2 in terms of the voltage coupling factor "CF" from port 1 to port 3 is simply: R2=Z0/(1-2xCF) Thus for a -20 dB, 50 ohm coupler, R1 is 225 ohms and R2 is 62.5 ohms. We'll post expressions for insertion loss and return loss soon...

A common mistake that engineers make is to create a power tap using just R1. Sure, this brings power to port three, but port 3 will also see a very poor impedance match. Using a combination of R1 and R2, we can match port 3 to Z0 in all cases. This puts a lower limit of 25 ohms on R1 (for R2=infinite), at a maximum coupling of -6.02 dB from port 1 to port 3. At the other extreme, R1 becomes infinite and R2 is 50 ohms when the coupler provides perfect isolation of port 3.

Here's a plot of the insertion loss versus coupling factors. By symmetry, S13=S23, and S11=S22. Not plotted is S33, because it's a perfect match to Z0.

More to come! Adams' resistive splitter Updated July 30, 2011 Click here to go to our main page on resistive splitters Click here to check out the Owen resistive splitter, it can also provides unequal splits Click here to go to our main page on unequal splitters New for August 2011! Here's a link to a cool calculator that will provide you with Adams' splitter resistor values. Thanks to Brian!

This page describes an unequal-split resistive splitter conceived by Greg Adams. We first came across it on the web site that no longer exists. They didn't give closed form equations for any of the elements, just a few example resistor values for three different splits.

The splitter was later fully described in an article by Greg Adams in the March 2007 issue of High Frequency Electronics. (Thanks to David for correcting the link!) In the article there are equations for solving for the resistors. Greg describes a clever simplification that we overlooked! We asked Greg what is the significance of the nomenclature Rs, Rt and Ru. Rs was named because it is a series resistor, Rt and Ru are merely the next letters in the alphabet! For "fun" we tried to solve the Adams splitter resistor values in terms of split values. Which was a lot harder than it looks. Here is what we came up with for the resistor values: First choose Rs. Then Rt can be solved as a function of Z0 and Rs, but it's an ugly quadratic equation. Ru is easily solved from RS and Rt. Greg's article explains some clever observations that make this splitter much easier to model than the brute force method we employed! (one of these days we'll post some more info on this!!!) For the example splitter, there is only one unique solution for the three resistor values, RS, RT and RU (we'll use the same notation that was used by RF Cascade). Below they are plotted against the coupling ratio.

Owen resistive splitter Updated July 30, 2011 Click here to go to our main page on resistive splitters Click here to go to our page on L-pads Click here to go to our main page on unequal splitters Click here to check out Adams' unequal resistive splitter, it offers less loss that the Owen splitter but lower isolation between arms (and it doesn't offer 3-way or higher splits) This idea for this content was contributed by Chris Owen, who designed this cool splitter. No where else in the universe but Microwaves101 will learn about this, thanks to Chris who chose to "publish" it here!

Update October 2010: Looks like MiniCircuits now markets an Owen splitter. Someone there must have read about it on this Microwaves101 web page! Look for part number ZFRSC-123 on this page. Here's an example of an Owen splitter. The Owen splitter provides multiport capability (2-way to N-way), with unequal splits possible to any of the ports. Like all resistive splitters it offers wideband performance. What sets this splitter apart from others (wye, delta, and the Adams splitter) is that the Owen splitter achieves maximum output to output isolation because none of the resistors are in the common path. In this respect there is no better resistive splitter! Why would you want to consider this power splitter? Everything in engineering is a tradeoff. Here's the pros and cons: 1. Like all resistive splitters, the only limitation on bandwidth is due to parasitic elements, there is nothing intrinsic to the design that would limit bandwidth. 2. It is simple to expand the design from 2-way to 3-way to 4-way to N-way. But with increasing N, the efficiency is reduced. 3. Like all resistive splitters, you will incur "real" loss when you use it. It is actually lossier than classic wye and delta resistive splitters described here. 4. It provides better isolation between split ports than classic wye and delta splitters. 5. It provides a means to vary the voltages of the signals of split ports, so it can be used as an unequal splitter. We envision this design as a wideband equal splitter on a thin-film or thick-film chip, with 2way, 3-way and 4-way versions. Anyone interested, give us a shout! Here's the description Chris sent us in January 1007: In a semiserious fashion I wish to lay claim to the Owen Resistive Splitter (similar to the Wilkinson splitter!) if it hasn't already been claimed, I have never seen it elsewhere. This particular resistive splitter design provides 10 dB power loss from the input to both outputs, maintains 50 ohms match on all ports, AND provides 19 dB of isolation between each of the splitter outputs. This splitter design may be useful for wideband applications (dc to 18 GHz!) where attenuation is not an issue and where a little isolation is useful and where simplicity and low cost and low space are also important.

Chris's splitter is based on L-pad attenuators. Unequal split, two-way Owen splitter Below is a schematic for a 2-way Owen splitter with the ports defined. Port 1 is the common port, as a divider this is where the incident signal enters. We'll use this resistor nomenclature (R1, R2, R3 and R4) in the analysis below.

Let's review what is meant by coupling factor and isolation. We (Microwaves101) define coupling factor as the ratio of the output voltage at a coupled port to the voltage incident on the common port, which we will call (port 1):

Most often the coupling factor is expressed in dB, which is 20xlog(CF). The power factor is the square of the coupling factor:

PF in dB is 10xlog(PF), at which point the coupling factor and power factor are identical. Isolation is the ratio of voltage incident to a coupled port to another coupled port. For a two-way splitter the isolation from port 2 to port 3 is:

The isolation from port 2 to port 3 is merely the sum of the coupling factors of these two ports (you can prove that for a homework!) This is why the Owen splitter offers the maximum isolation of any resistive splitter. The efficiency factor is a measure of the efficiency of the device. It is the power output at all of the coupled ports divided by the power input at the common port. Expressed in terms of the coupling factors the efficiency is:

Now it's time to solve some equation equations for resistor values. The two-way, unequal Owen splitter resistor values can be solved by choosing one resistor and solving for the rest, noting that all ports need to be matched to 50 ohms. We'll skip the painful algebraic steps that get you to the solution, for this you'll need a large pad of paper, two cups of coffee and a #2 Pokemon pencil if you're anything like the rest of us... R2 can be expressed in terms of the coupling factor of its arm:

We solved the equations for R1, R2, R3 in terms of R2:

And last, we present equations for the two coupling factors in terms of resistors that make up their arms:

All of this stuff is great material for a spreadsheet, we've been there and done that. If anyone is interested, ask us nicely and we'll share it with you. Here's a plot of the four resistor values for unequal splits and Z0=50 ohms, versus the ratio of the two coupling factors (in dB):

Below is a table of coupling factors, isolation and resistor values for 2-way Owen splitter (Z0=50 ohms). Looks like you can never get less than 19 dB isolation with the Owen splitter! You can't say the same thing about the Wilkinson splitter. The efficiency ( ) is theoretically equal to 100% when all of the power is coupled to one of the arms (which would not be a useful coupler.) Efficiency bottoms out at about 22% for intermediate coupling values. Update August 2007: previously the data in R1 and R2 columns was inadvertently reversed, it's fixed now. Thanks to Robert! CF2 (dB) -1 -2 -3 -4 -5 CF3 (dB) -30.57 -24.32 -20.58 -17.89 -15.76 CF2/CF3 (dB) 29.57 22.32 17.58 13.89 10.76 (%) 80% 63% 51% 41% 34% Isolation (dB) -31.57 -26.32 -23.58 -21.89 -20.76 R1 Ohms 5.77 11.61 17.61 23.85 30.40 R2 Ohms R3 Ohms R4 Ohms 53.05 56.47 60.31 64.62 69.46

869.55 843.81 436.21 409.69 292.40 265.06 220.97 192.78 178.49 149.42

-6 -7 -8 -9 -9.54 -10 -11 -12 -12 -14 -15 -16 -17 -18 -19 -20

-14.01 -12.53 -11.24 -10.10 -9.54 -9.10 -8.20 -7.39 -6.66 -6.01 -5.41 -4.88 -4.39 -3.95 -3.55 -3.19

8.01 5.53 3.24 1.10 0.00 -0.90 -2.80 -4.61 -6.34 -7.99 -9.59 -11.12 -12.61 -14.05 -15.45 -16.81

29% 26% 23% 22% 22% 22% 23% 25% 27% 29% 32% 35% 38% 42% 45% 49%

-20.01 -19.53 -19.24 -19.10 -19.08 -19.10 -19.20 -19.39 -19.66 -20.01 -20.41 -20.88 -21.39 -21.95 -22.55 -23.19

37.35 44.80 52.84 61.59 66.67 71.15 81.66 93.25 106.07 120.31 136.14 153.78 173.46 195.43 220.01 247.50

150.48 120.49 130.73 99.82 116.14 52.84 104.99 72.18 100.00 66.67 96.25 89.24 83.54 78.84 74.93 71.63 68.83 66.45 64.40 62.64 61.11 62.48 54.52 47.87 42.24 37.41 33.23 29.58 26.37 23.54 21.03 18.80

74.88 80.97 87.80 95.46 100.00 104.06 113.70 124.53 136.67 150.30 182.74 182.74 201.99 223.58 247.81 275.00

Two-way Owen splitter layout Here's an excellent layout for a "wideband" two-way Owen splitter, this image was contributed by Chris Owen himself. Note that the four resistors are all located as close to each other as possible, and as close to the common node as possible. Any significant track length between the resistors acts as distributed transmission line which will cause a phase shift; at higher frequencies the increasing phase shift will eventually skew the response (coupling factors will shift and port VSWRs will increase). At some frequency any transmission line will become 1/4 wave and then all bets are off! The grey/black circles represent vias that tie the backside ground to the topside grounds.

N-way, equal coupling factor Owen splitters Here the arms would all have the same resistor values, R1 and R2. After some math, the formulas for R1 and R2 (R1 is series resistor, R2 is shunt resistor) such that all ports are matched to Z0 are: R2=NxZ0/(N-1) (solve this resistor first!) R1=Z0x(NxZ0+R2(N-1))/(Z0+R2) Here's the values for R1 and R2 for N-way splitters in 50 ohm system. We also show the "coupling factor", which is the ratio of arm output to splitter input power (we'll post that equation later). For N=2, R1=66.7, R2=100, CF=-9.54 dB For N=3, R1=120, R2=75, CF=-13.98 dB For N=4, R1=171.4, R2=66.6, CF=-16.08 dB For N=5, R1=222.2, R2=62.5, CF=-19.08 dB For N=6, R1=272.7, R2=60, CF=-20.82 dB Yes, it is possible to create an N-way Owen splitter with output arms set to different coupling values. Rather than trying to solve for the resistor values algebraically, we recommend you use an optimizer to do this job for you.

Wilkinson power splitters Revised October 15, 2010 Search for Wilkinson power dividers on EverythingRF.com

Click here to go to our main page on couplers and splitters Here's a clickable index to our treasure-trove of material on Wilkinson power splitters: Two-port Wilkinson (this page) Designing Wilkinsons using Excel New for August 2008! Wilkinson isolation Rules of Thumb Lumped-element Wilkinson Multistage Wilkinson (separate page, multiple examples!) Unequal-split Wilkinson (separate page) Compact Wilkinsons (separate page, multiple examples!) N-way Wilkinsons (separate page) N-way Wilkinsons with unequal split (separate page, new for September 2009!) The Wilkinson power splitter was invented around 1960 by an engineer named Ernest Wilkinson. It splits an input signal into two equal phase output signals, or combines two equalphase signal into one in the opposite direction. Wilkinson relied on quarter-wave transformers to match the split ports to the common port. Because a lossless reciprocal three-port network cannot have all ports simultaneously matched, Wilkinson knew he had to cheat so he added one resistor and the rest is history. The resistor does a lot more than allow all three ports to be matched, it fully isolates port 2 from port 3 at the center frequency. The resistor adds no resistive loss to the power split, so an ideal Wilkinson splitter is 100% efficient. Two-port Wilkinsons In its simplest form, an equal-amplitude, two-way split, single-stage Wilkinson is shown the figure below. The arms are quarter-wave transformers of impedance 1.414xZ0 (thanks for the

correction, Rod!) Here we show a three-port circuit (the most common in practice by far, but Wilkinson described an N-way divider).

Ideal two-port Wilkinson splitter

S-parameters of ideal 2-way Wilkinson power splitter

Here is how the Wilkinson splitter works as a power divider: when a signal enters port 1, it splits into equal-amplitude, equal-phase output signals at ports 2 and 3. Since each end of the isolation resistor between ports 2 and 3 is at the same potential, no current flows through it and therefore the resistor is decoupled from the input. The two output port terminations will add in parallel at the input, so they must be transformed to 2xZ0 each at the input port to combine to Z0. The quarter-wave transformers in each leg accomplish this; without the quarter-wave transformers, the combined impedance of the two outputs at port 1 would be Z0/2. The characteristic impedance of the quarter-wave lines must be equal to 1.414xZ0 so that the input is matched when ports 2 and 3 are terminated in Z0. Okay, what about as a power combiner? Consider a signal input at port 2. In this case, it splits equally between port 1 and the resistor R with none appearing at port 3. The resistor thus serves

the important function of decoupling ports 2 and 3. Note that for a signal input at either port 2 or 3, half the power is dissipated in the resistor and half is delivered to port 1. Why is port 2 isolated from port 3 and vice-versa? Consider that the signal splits when it enters port 2. Part of it goes clockwise through the resistor and part goes counterclockwise through the upper arm, then splits at the input port, then continues counterclockwise through the lower arm toward port 3. The recombining signals at port 3 end up equal in amplitude (half power or the CW signal is lost in resistor R1, while half of the CCW signal is output port 1. And they are 180 degrees out of phase due to the half-wavelength that the CCW signal travels that the CW signal doesn't. The two signal voltages subtract to zero at port 3 and the signal disappears, at least under ideal circumstances. In real couplers, there is a finite phase through the resistor that will limit the isolation of the output ports. Below we show an example of extending the bandwidth of a Wilkinson splitter by placing a quarter-wave transformer on the common-node and optimizing its impedance along with the impedances of the quarter-wave legs.

Example of Wilkinson with input transformer

S-parameters for above Wilkinson with input transformer

Lumped-element Wilkinson splitters Updated March 4, 2007 Click here to go to our main page on Wilkinson power dividers Click here to go to our main page on lumped elements New for March 2007! Content was donated by Dr. Antonije Djordjevic, a Professor at University of Belgrade. The real Belgrade, not Belgrade, Maine, by the way. Thanks, Tony! This design below provides lumped-element equivalent of simple one-stage Wilkinson power divider with nominal port impedance 50 ohms. The center frequency was picked to be 1.592 GHz, so that angular frequency =1010 radians/second. You can apply scale factors to the capacitors and inductors to change center frequency, both are inversely proportional to frequency. Note that the isolation resistor value is the same as in a "normal" Wilkinson, at 2xZ0. The normalized inductor reactance at center frequency: XL1=sqrt(2) The normalized capacitor susceptances at center frequency: BC1=sqrt(2), BC2=sqrt(2)/2 Let's review the formulas for capacitive and inductive reactance while we're on the subject:

Here's how it performs over frequency:

Now for the derivation... One way to derive the lumped-element equivalent of the Wilkinson power divider is to begin by deriving a lumped-element equivalent of a quarter-wave transmission line. Suppose that the characteristic impedance of the line is ZC=Z0, where Z0 is the port nominal impedance. If the line length is /4, the matrix of scattering parameters of the line is:

We consider the pi-network of lumped elements, which consists of capacitors in shunt branches and an inductor in the series branch. We want the network to have the same scattering parameters as the transmission line. Due to the symmetry, the two capacitors must be identical (C1=C2=C). We place a generator of emf E=2V, angular frequency , and internal resistance R1=Z0 at one port. We terminate the other port in R2=Z0.

Let V1 and V2 be the input and output voltages (the lower nodes are assumed grounded). To obtain the required scattering matrix, we should have complex voltages V1=1 Volt and V2=-j Volt. The nodal equations for this circuit read:

and:

Substituting V1= 1V, V2=-jV, R1=R2=Z0, and C1=C2=C into the second equation, we readily obtain j C+1/(j L)=0 and L=Z0, so that C=1/Z0. Hence, the reactance of the coil is XL=Z0 and the susceptance of the capacitor is BC=1/Z0. In the transmission-line Wilkinson power divider (for 50 Ohm nominal port impedances), we have two quarter-wavelength transmission lines whose characteristic impedance are 50xSQRT(2)=70.7 ohms. Each line can be replaced by a pi-network with XL=70.7 Ohms and BC=1/70.7 Siemens, which are the same values as in our scheme. (Two capacitors at port 1 are merged into an equivalent capacitor with two times larger capacitance.) References for this work: Djordjevic, A. R., Badar, M. B., Vitoevic, G. M., Sarkar, T. K., Harrington, R. F., Scattering Parameters of Microwave Networks with Multiconductor Transmission Lines (software and user's manual), Artech House, Boston, 1989. Djordjevic, A. R., Badar, M. B., Harrington, R. F., Sarkar, T. K., LINPAR for Windows: Matrix Parameters for Multiconductor Transmission Lines, Version 2.0 (software and user's manual), Artech House, Boston, 1999. Multistage Wilkinsons Updated July 26, 2008 Click here to go to our main page on Wilkinson power splitters Click here to go to our page on designing Wilkinson splitters with Excel Multistage (or multi-section) Wilkinson power splitters can perform remarkably well over more than a decade of bandwidth. On this page we will show you a procedure that will help you design a multi-section Wilkinson quickly and efficiently, instead of using guesswork and optimization. Then you can cheaply build your own splitters instead of buying them from Narda for thousands of dollars! Example 2 presents info on the exact design of Chebyshev and maximally flat multistage Wilkinson splitters, using our impedance transformer spreadsheet that you can download! Does Agilent offer this in ADS? Heck no, Dude!

Calculating isolation resistors For a multistage Wilkinson, we haven't (yet) developed a detailed method for determining isolation resistors along the chain. Harlan Howe's book on stripline design has some nice tables that can help, he also shows some equations. Seymour Cohn wrote an excellent article on the subject way back in 1968, you can download it from IEEE Explore if you are a member. Sorry, we try to respect restrictions on copywrited material so we won't post the article. Here's the exact reference: A Class of Broadband Three-Port TEM-Mode Hybrids Cohn, S.B., Microwave Theory and Techniques, IEEE Transactions on Volume 19, Issue 2, Feb 1968 Page(s): 110 - 116 Thanks to Tim for this IEEE reference! Table 1 gives normalized quarter-wave line impedances and resistor values for various bandwidth designs. Cohn does give closed-form equations for the two resistors in a two-stage Wilkinson for equal-ripple response, then shows some approximations for cases higher than N=2. In the end, you will probably have to use an optimizer such as you'll find in Agilent's ADS to squeeze out the maximum isolation opver your design bandwidth. Isolation resistors are sometimes referred to as "dump" resistors, an example of microwave slang. Also thanks to Tim! New for August 2008! We created a spreadsheet that will allow you to analyze multi-section Wilkinsons in Excel, including isolation between the split ports. It's explained on this page. Example 1: designing the hard way... We put together the two-stage Wilkinson combiner below to illustrate the extra bandwidth that you have achieve with a second set of quarter-wave sections. The values for the line impedances and isolation resistors were obtained by shamelessly using the optimizer in our ancient old copy of Eagleware Superstar... sorry to disappoint anyone that thought we would derive equations for the optimum values!

Example of two-stage Wilkinson with input transformer

S-parameters of above 2-stage Wilkinson with input transformer Example 2A, two-stage, 75 ohms, the easy way! This two stage divider was calculated using our transformer spreadsheet located in the download area. The arm impedance has to be twice that of the transformer solution, because two arms are in parallel. The circuit is a two stage, Chebyshev Wilkinson, with characteristic impedance 75 ohms (instead of the usual 50 ohms, just for fun). The required frequency band was 950 to 2025 MHz, because that's what a Microwaves101 reader requested. The center frequency is 1.488 GHz (=(F1+F2)/2). The load and source resistance that the transformer is matching to is 75 ohms (the common port) and 37.5 ohms (the split ports, two 75 ohm loads in parallel). The Chebyshev transformer spreadsheet calculates 61.248 and 45.920 ohms for the two sections; you need to double these values to 122.5 and 91.8 ohms, as we stated before, because the two arms are in parallel. Note that the impedances are a function of the percentage bandwidth as well as the termination impedances. You can view the VSWR of the input port from within the spreadsheet, this is very handy and will help you decide how many sections to use for your application. The isolation resistors values were found by referring to a graph in Harlan Howe's book "Stripline Circuit Design". You can find this book in our page on microwave engineering books. Note that more than one solution exists for the resistors, depending on what you want to optimize (isolation at center frequency, bandwidth of S22, etc.)

Example 2B, two-stage, 75 ohms, with transformers In the next examle, we added a pair of transformers to reduce the 75 ohm terminations to 50 ohms at each port (61.2 ohm quarter-wave three transformers did the job). This reduces the maximum impedance of the arms, they are now Z1=61.2 and Z2=81.7 ohms (exactly 2/3 what they were before, because we reduced 75 ohms to 50 ohms). Z2 at 122.5 ohms might have been unrealizable previously. The isolation resistors were also scaled by 2/3 from the previous example. The resulting input impedance (S11) stays close to equal-ripple in the passband.

Example 2C: 950 to 2025 MHz, three stages Below we designed a three stage Wilkinson, this time with a "physical" model. The transmission lines are microstrip on Rogers 4003 (ER=3.38), 32 mils thick with half-ounce copper (0.7 mils). Again, we are looking for 950 to 2025 MHz bandwidth, but we want more isolation (hence three stages). We used the starting values for line impedances from our transformer and microstrip calculators, took the isolation resistor values from Harlan Howe's book, then did a little

optimization using Agilent ADS. The next two plots show the model and predicted performance. We rounded off the resistors to 1% RETMA values. The isolation that you get will probably be six dB worse than the prediction (which was greater than 30 dB), because the resistors won't act ideal.

Example 3 Yet another request from the Microwaves101 message board. This is for 800 MHz to 6 GHz. We arbitrarily chose four sections because we don't have a specification to go by.

If you download the transformer calculator to use as a multisection Wilkinson 2-way design, we recommend you start with the equal ripple page, enter input impedance = 50 ohms, output=25 ohms (Z0/N, where N=number of outputs...) for a three-way, output=16.7 ohms etc. Then enter the frequency range, 0.8 to 6 GHz for this example. The number of sections can be traded for VSWR (more sections=lower VSWR ripples). The calculator will give you the section impedances, for example if you have a two-way divider (50:25 ohms) with four sections, the transformer spreadsheet will give you these impedances: 41.243 37.299 33.513 30.308 These you will use as the arms for the Wilkinson. But first you have to double them because the two arms will be in parallel. The low values go toward the 2-way side, the high values toward the 50 ohms side. The "magic" you will have to perform on your own is to calculate isolation resistors, Next we simulated the Wilkinson on ADS, and used 200 ohm isolation resistors. Why 200 ohms? It's a pidooma! Here's the schematic and the response:

The isolation is pretty good at 17 dB, but it could be better. We used the ADS optimizer and came up with 20 dB isolation across the band:

That took all of 15 minutes! Here's the group delay of the splitter:

Designing Wilkinsons with Excel Updated July 26, 2008 Click here to go to our main page on Wilkinson power splitters

Click here to go to our page on multisection Wilkinsons New for August 2008! This page came about from some email correspondence with Tim who lamented about the lack of a straightforward way of selecting isolation resistors in a multisection Wilkinson. The selection of "optimum" resistor values for a design bandwidth beyond the N=2 Wilkinson is messy and almost always forces engineers to use an optimizer which is offered in expensive software such as Agilent's ADS, which not everyone has access to. Using transformer analysis is an easy method to select line impedances of a multi-section Wilkinson splitter. We showed you how to do that here (look at Example 3). Now we have an Excel spreadsheet that analyzes the Wilkinson divider, including the isolation resistors, to give you the full three-port response. You can get a copy in our download area. We still haven't perfected the math, but our new spreadsheet will allow you to fiddle with the isolation resistors and watch their affects on isolation and reflection coefficients in (almost) real time. Even and odd mode analysis The trick to making a mathematical model of Wilkinson's power splitter is to apply even and odd mode analysis. The full technique is described in a 1968 paper by Seymour Cohn in an IEEE paper titled A Class of Broadband Three-Port TEM-Mode Hybrids. We merely put his math it into a spreadsheet (OK, we also had to put in some Chebyshev polynomials, and we're working on adding the isolation resistor calculations that Harlon Howe described in his book Stripline Circuit Design but Rome wasn't built in a day, was it?) Some day the even/odd mode technique will be better explained on this page. The analysis below is also explained in Cohn's paper. First, look at the generic multi-section Wilkinson schematic. We adopted Cohn's designations of the R's and Z's starting with subscript "1" closest to the split ports:

The even mode analysis is done by assuming that the excitations on ports 2 and 3 are in-phase. Then the voltages across the isolation resistors are zero and they can be ignored:

In the odd mode, the excitations are 180 degrees out of phase. In this case a virtual ground lies across the axis of symmetry. The resistor values all get chopped in half:

From the even and odd mode schematics, the reflection coefficients can be calculated from the impedance looking in (Zin) using the transformer equation:

Next, following Cohn's paper, we can easily arrive at the reflection coefficient at port 1, because it is the same as the even-mode reflection coefficient:

The transmission coefficient (essentially S21 or S31 in S-parameters) can be calculated from the reflection coefficient at port 1 assuming that the network is lossless:

The refection coefficients at ports 2 and 3 are the average value of the even and odd-mode coefficients:

And last, the isolation between ports 2 and 3 (the transmission coefficient) is equal to half of the difference between even and odd mode reflection coefficients:

Here's a few things worth pointing out: The input (port 1) reflection coefficient is not a function of the resistors, it is merely a function of the line impedances (Z1, Z2, etc.). In other words, it behaves like a multi-section transformer, which can be solved in different ways. We chose to use Chebyshev in our spreadsheet, but it just as well could have been maximally flat. The choice of resistors

As Mr. Cohn pointed out, this model does not take into account the effects of the sections coupling to each other, which changes even and odd mode impedances and will throw off the response a bit. If you really need first pass success, use an EM simulator for the final design. Examples of using the spreadsheet The plots below used the values that Mr. Cohn developed in his paper, for seven sections. For convenience we put Cohn's computed values for his examples into our spreadsheet on a worksheet named "Cohn", you can copy them into the "Cintorl Panel" worksheet to examine the various responses he achieved. . Here's the VSWRs of the ports for a 10 GHz, seven section splitter with 164% bandwidth (F2 is 10xF1). Note the equal ripple on port 1, and the non-equal ripple on port 2 (which is because Cohn didn't have an exact way to calculate the resistors for equal ripple and neither do we!)

Also, note that the impedance matches at ports 2 and 3 are better than the match at port 1, this is typical. Here's a plot where we converted SWRs to retun losses, if you prefer units in dB:

Fnally, here's the isolation, which isn't exactly equal ripple either:

N-way Wilkinson splitters Updated June 6, 2010 Click here to go to our general discussion of Wilkinson power splitters Click here to go to our main page on couplers and splitters Click here to go to our page on the Lim-Eom three-way splitter Click here to go to a page on the Kouzoujian N-way splitter New for June 2010! Here's an analysis of planar three-way Wilkinsons. New for August 2009! Here's a page contributed by Paul Hubbard and Greg Ordy, on N-way, unequal split Wilkinsons. Here's an example contributed by Kjer of a broadband eight-way splitter.

Here's an example contributed by Limey Mark of a unique four-way splitter. Ernest Wilkinson's original paper was on an N-way combiner, and it is only fitting that we should deal with the subject of higher-order Wilkinson splitters here. For N-way combiners, the number of arms is equal to N, while the number of ports is equal to N+1, at least that is the convention we will maintain here. Radial N-way Wilkinson combiners Radial combiners are called that because they have radial symmetry. Above N=2, the splitter cannot be laid out in two dimensions. For N=3 split, there are two ways to realize the isolation resistors, the "star" and "delta" configurations. A three-way Wilkinson is shown below, with "star-resistor" configuration. The arms have impedance SQRT(3)xZ0, and the resistors have impedance Z0. By the way, the figures below, and many more, are available in our download area in a Word file, ElectronicSymbols.doc, for you to use in presentations and papers.

The next figure shows a three-way Wilkinson with the "delta" resistor configuration. The arms again have impedance SQRT(3)xZ0, but now the resistors have impedance 3xZ0.

Now that we've looked at the three-way Wilkinson, it is easy to guess what higher-order Wilkinson combiner resistor networks look like. We'll give you a hint... Every port has to be connected to every other port symmetrically. The "delta" resistor pattern gets really ugly, so stick with the star pattern for N=4 or higher. Let's state the rules for arm impedance and isolation resistors of N-way Wilkinson combiners as a Microwaves101 rule of thumb:

For a basic N-way Wilkinson combiner, the arm impedance is SQRT(N)xZ0. If you use star resistors, they are equal to Z0. If you use delta resistors, they are equal to NxZ0. Planar N-way Wilkinson combiners In many instances, it is more convenient to use a two-dimensional approximation of the radial Wilkinson shown above. In this case, one of the resistors is deleted from the layout, and a "fork" arrangement is the result. Go here for an analysis of three types of three-way splitters, in 50 ohms (very basic). In the figure below, contributed by Lou from Honeywell, a two-stage, three-way planar Wilkinson is shown, along with its predicted responses (thanks, Lou!) Note to readers: this is a very specialized splitter, it provides 12.5 ohm impedance for the split ports and 50 ohms at the combined port.

What does this do to performance? The primary thing that you give up is isolation between the arms. Instead of greater than 20 dB, you might get 15 dB isolation. Also, there will be measurable differences between the inside and outside arms. An example from Limey Mark Limey Mark is a good frend of the Microwaves101 message board and cotributed this novel design technique during October 2006. Thank you sir! Now in his own words... It is possible to make a N-way splitter on microstrip with no cross over. This technique has been done at 100MHz. I would be interested to see if any one can do this at a higher frequency. Below you will see the image of a four-way splitter. This representation was knocked up in my garage with some old copper foil and a clapped out Weller soldering iron so my apologies for the unprofessional approach.

Lim-Eom 3-way power splitters Updated December 22, 2008 Click here to go to our main page on couplers and splitters We now have two examples of using Lim and Eom's unique splitter to combiner power amplifiers: Click here to learn how to use the Lim Eom splitter as a three-way combiner

Click here to learn how to use the Lim Eom splitter as a two-way combiner This is a truly new power splitter (a rare event in microwaves!), first detailed by Jong-Sik Lim and Soon-Young EOM in a 1996 IEEE MTT-S paper titled A New 3-way Power Divider with Various Output Ratios. As a three-way splitter, it fits only niche applications, because of the phase relationships of the three outputs, as well as the asymmetric layout, but still, it's a cool splitter that you should consider. One application that comes to mind in the LO splitter for the classic three-channel monopulse receiver. By the way, Lim and Eom were at the Seoul National University's Applied Electromagnetics Laboratory when they wrote the origianl paper. The AEL posts many of their publications on their web site, definitely worth checking out! We have a page on SNU go there if you want to check out the link! We made a model of Lim and Eom's six-port neteowrk using Eagleware Genisys. We also entered the equations into Excel, just to play with the split ratios to see the effect on line impedances. The splitter resembles a double-box hybrid. Below is the schematic from our Genisys project. Port 1 (top left port) is the common port, while ports 2, 4 and 6 are the split ports. Note that you have to terminate ports 3 and 5 to operate it. The line impedances shown will provide an equal three-way split in a fifty ohm system.

Lim and EOM's paper gives equations for line impedances that you can use to obtain different power split ratios. The integers m, n and k are what sets up the power division. For an equal split,

m=n=k=1. The impedances given in the above schematic are for equal split. Here is how we entered them in Genisys: m=?1 n=?1 k=?1 delta1=m+n+k delta2=n+k Z0=50 Z1=(delta1/delta2)^.5*Z0 Z2=(delta1/m)^.5*Z0 Z3=Z0 Z4=(delta2/N)^.5*Z0 Z5=(delta2/K)^.5*Z0 These are the variables that Lim and EOM defined in their paper. Note that m, n, k don't have to be integers (and in Genisys, the question mark is what allows you to tune them while viewing the response). It is easy to convert the equations so that you input the ratios in dB (that's the first thing we did, in an Excel file we'll give you if you ask nicely!) Note that the impedance Z3 is has no effect on the power split. We left it at 50 ohms like they did in their paper. The plot below shows the frequency response for the equal split case (-4.77 dB each arm). Notice that the bandwidths of the three split arms are all different. The bandwidth isn't all that good relative to a Wilkinson splitter.

Below is the isolation of ports 3, and 5, along with the input match S11. From this point of view you could use the splitter over a 20% bandwidth and get 20 dB isolation.

Let's look at the phases of each path. The longest path is S14 (one wavelength), followed by S16 (half-wavelength), then S12 (quarter-wavelength). We see that at center frequency that S14 and S16 are +/- 90 degrees out of phase with S12, and the phase relationship is not very constant over frequency. This might be a limitation in certain applications, for equal phases you should consider a three-way Wilkinson. You might think of the Lim-Eom splitter as a combination of a 90 degree hybrid and a 180 degree hybrid.

If you wanted to achieve equal phase, equal split, here's a layout for you. The input port is in the center of the north side, and the west, south and east sides have transmission lines that give nearly equal phase outputs. Note the "fan stub" terminations, whoever did this cool design deserves a fat raise!

Now let's try out an unequal split, in this case 1:5:5. You might do this if you wanted most of your power to feed two devices, and the third port going to a detector to monitor power. We followed the equations and they do what Lim and EOM claimed! The two facored ports are 7 dB above the starved port.

More to come! We still need to look at line impedances versus coupling ratios.

Вам также может понравиться