There are millions of wilkinson combiner design programs out there but when it comes to creating a practical design for high power usage there seems to be a marked scarcity. So here goes a tool for taking into account the power combiner practicalities. This is an on-line version of an Excell sheet version.
When combining power amplifiers (a typical Band 11 amplifier would be 1Kw to 1.5Kw), efficiency, power dissipation and fault behaviour are important. Power amplifiers are never exactly equal, the optimum load may not be 50 ohms. If auto gain control or power trimming is used the amplifiers will not track together. So the power loss in the balance resistor needs to be known.
Rather than suffering a complete broadcast loss on failure of a power amplifier it is often desirable that the broadcast continue at reduced power, this means the balance resistor needs to loose considerable power to a substantial heatsink. We also need to ensure the back leakage to the failed amplifier is restricted. The line distance between the combiner and the amplifier may need to be set for the complex impedance present by a failed amplifier.
If there is a combiner there will be a splitter. After all how are you going to ensure the phase of the signals to be combined is correct.
Since a combiner and splitter are identical in format, here we are going to deal with Compact Wilkinson Splitters (Webb, Scardelletti), Lumped element multiport, Bagley (odd, even variants).
Even number combiners usually just cascade Wilkinsons. This gives a good bandwidth/isolation compromise. Single stage combination of odd numbers is possible but because of the increase differential between the drive resistance and summing point resistance the bandwidth becomes increasingly restricted. Also minor differences in the amplifier phasing make for sensitivity problems
Combining three amplifiers requires just that, a three way combiner. However in broadcast applications the combiner will be followed by a harmonic filter, so what better than combine the two. Instead of a combiner 50 in to 50 out + 50 to 50 filter we combine the two. A 50 ohm in (or what ever is optimum for the amplifiers) to a low R summing point then an even order Chebychev type 11 harmonics filter back to 50 ohm out.
Above three amplifiers it is probably better to use an even number of stages, eg: 6 x amplifier of 1Kw for a 5Kw Tx, the amplifiers can be down rated ( 6 x 900W instead of 5 * 1.15Kw), run cooler have better excess rho protection. This is because above 3 the isolation between ports degrades (Theoretical isolation for n = 2, 3 is infinite, n = 4 drops to 21.6db, n = 5 to 19.5db and so on to an asymtopic limit of 14db only), thus any differences beteen amplifiers is accentuated.
If a more complicated life is desired, then any odd order combiner can be made from 2 and 3 way combiners, with the proviso that there an un-equal split is used. Eg: One possible design for a a 5 way combiner is: two off 3 way + 2 off 2 way. The two 3 ways have two equal split arms and the unequal split off the third arm. The two unequal arms are summed wilkinson fashion to give a total of 5 equal power arms. Finally the two 2 off 3 ways are wilkinson summed.
This is a space saving technique that can be applied to any type of combiner. The classical method in Broadcasting is to use an amplifer combiner plus a low pass filter. The whole assembly works on the 50 Ohm chain. Amplifiers to 50Ω loads, combiner 50Ω inputs to a 50Ω output. The filter 50Ω to 50Ω as well.
However there is an alternative: In many cases due to constraints of the matching the amplifier will be more efficient with a non 50Ω load. Likewise why not take advantage of the source-load resistance transformation possible and couple the combiner sum point directly into the filter. This is what we do here.
Further more we use a lumped element approach to condense the size, a chebychev type II to fix traps at the harmonic frequencies, again to reduce sizing parameters.
Here we look at two possible 4 way combiners. Using the Wilkinson variants we come up with several pro/con solutions. The straight forward way is a a Wilkinson cascade 2 x 2:1 + final 2:1. This gives a rather long design but provides good isolation, distribution of power dumping in the case of a PA failure. The multiple 1/4 lambda length can give rise to increased insertion loss, so for a HPA combiner we need to maximise the exposed copper cross section by use of low Z lines, Low Z combining junctions. So we allow selection of the three combiner node R's.'
Another approach is a direct sum of 4 inputs transformed to a single node. This will not be a true Wilkinson since planar connection of all the ports via dump resistors is not possible. This means isolation is low and minor differences in the amplifier phasing make for sensitivity problems.
Then there is the Gysel style approach, a bit like the a branch line coupler. Big advantage the dump load can be a water cooled off board resistor.
Capacitors come in many forms but power capacitors are a special breed. Here we are looking at capacitors that pass many amps of reactive current. Now you may say well so what a reactive current cannot dissipate, true, but it does "flow" and it is impossible to make a capacitor that has no real part. So there must be dissipation (otherwise of what use is tan delta that manufacturers supply) and that dissipation is what power capacitors are very concerned with. Not just because any power loss in the capacitor results in less power to the load but, also, because heat is generated in a capacitor.
If you think that a few watts in a capacitor is nothing, then put an excess of amps through a ceramic capacitor and wow, an exploding capacitor is a sight to see, debris everywhere. Ok, just make the capacitor larger (more parallel paths) but there is a trade off, size, both size v frequency and size v economics.
Here we look at some capacitors suitable for power filters, especially the sum-filter and harmonic filters
Inductors also come in many forms and like power capacitors dissipation is a major concern but, with a different emphasis. Inductors are made of metal and metals conduct heat as well as current allowing an inductor to run at a much higher temperature than a power capacitor. Power inductors tend to be large, run hot, have to be kept away from other components, eminate a high magnetic field.
Thus we see the design problems are different, orientation is important to minimise interaction between the magnetic fields of an inductor chain. Special mounting techniques are required to dissipate the heat generated. For example an inductor in a MW, AM duplexor may be 50cm dia, 100cm long, use 25mm copper pipe and have fins soldered on to the pipe to enhance air cooling!
Here we look at some inductors suitable for power filters, especially the sum-filter and harmonic filters
The hybrid combiner/splitter differs from an inline combiner like the Wilkinson in that the split is 90 degrees. This confirs some special qualities when used for combining amplifiers (Normally called the balanced amplifier approach). eg: load source protection when an amplifier fails.
One big advantage when used as a combiner is that the "balance" load is not integrated into the combiner like the Wilkinson, so this load can be mounted far away from the actual combiner ( eg: a water bath) this easing considerably the problem of fault power dissipation.
Hybrid couplers come in many forms, microstrip, lumped, waveguide. My first use was a 1MHz lumped design as a Uni project, then as a twit (TWT's for non microwave visitors) combiner with K band waveguide. However since this a Broadcast RF site I will deal with the multi-box lumped format.
The branch line combiner/splitter differs from an inline combiner like the Wilkinson in that the split is 90 degrees. This confirs some special qualities when used for combining amplifiers (Normally called the balanced amplifier approach). eg: load source protection when an amplifier fails.
One big advantage when used as a combiner is that the "balance" load is not integrated into the combiner like the Wilkinson, so this load can be mounted far away from the actual combiner ( eg: a water bath) this easing considerably the problem of fault power dissipation.
Branch line couplers come in many forms, microstrip, stripline, lumped, waveguide. Here we will deal with multi-section hybrid coupler in microstrip, stripline and slabline..
