pcf@galadriel.bt.co.uk (Pete French) (11/01/89)
A simple problem - I need to make (or buy) tranformers to operate at 999kHz
to split powqer R.F. from a main feeder into several sub-feeders in the ration
I want - this is normally 1 in with 4 out equally, but some points in the
system need 1 in and 2 out split 2:1 between them.
Both the inputs and outputs must be matched to 50 ohms.
A simple enough requirement, I even have a design that has been in use since
the early 70's on the ssystem - eexcept that nobody makes thee cores
anymore.
So I need a new design ... and there seems to be nowhere that can help me do
it. We have contacted firms all over the U.K. that do custom design. None of
them can do it. We have asked electronics lecturers - none of them can do it.
We have looked in text books - they all say how to wind the things, but
tactfully avoid mentioning how to match input-output to 50 ohms. This seems
to be the major sticking point.
Help ! (please)
-Pete.
--
-Pete French. | "It was cold,
British Telecom Research Labs. | and the person beside him
Martlesham Heath, East Anglia. | had faded badly..."
All my own thoughts (of course) | -BAUHAUSelliott@optilink.UUCP (Paul Elliott x225) (11/08/89)
In article <498@galadriel.bt.co.uk>, pcf@galadriel.bt.co.uk (Pete French) writes: > > A simple problem - I need to make (or buy) tranformers to operate at 999kHz > to split powqer R.F. from a main feeder into several sub-feeders in the > ration > I want - this is normally 1 in with 4 out equally, but some points in the > system need 1 in and 2 out split 2:1 between them. > > Both the inputs and outputs must be matched to 50 ohms. There are several ways you can go here... Ground Rules: Power level was not specified. Moderate power is assumed, range from milliwatts to tens-of-Watts. Frequency range / bandwidth was not specified. Since the frequency "999 KHz" was mentioned, narrow-band as well as wideband solutions are presented. I ramble on about semi- esoteric RF circuits because I like to hear myself talk, and I am a semi-frustrated RF designer now doing gate-array design (but I like that too). I Take No Responsibility (tm) for ANYTHING I say here. For that, I like to get paid. 1) Duplicate the existing design. The transformer core material is probably not critical in a 1 MHz splitter (assuming a broadband splitter of good design). Any moderately high permability RF ferrite should do. A uL of 125 to 1000 should work, if the Xl of the windings is at least 5x the impedance level (5x 50 Ohms = 250 Ohms). If the current design uses tuned circuits however, the core material becomes very critical. Note that a splitter is _not_ made by merely making a multiple-winding transformer. As was mentioned in the original question, matching impedances is a goal, and so the circuit is necessarily more complex. 2) Buy ready-made splitters from Mini-Circuits Labs or Merrimac. These and other manufacturers make broadband N-way splitter/combiners at a reasonable price. The performance is specified, which greatly simplifes things. This is probably the best solution. 3) Build a broadband splitter. I recommend the Motorola RF Transistor databook app notes for design info here. Try to find AN749 (Broadband Transformers and Power Combining Techniques for RF). 4) Use resistive splitters and distribution amps. The practicality of this depends on the power level and other things. 5) (My favorite, because I think it's neat): Build a Wilkinson power divider. This circuit is a narrow-band solution. It provides good port-to-port isolation and can be configured for any number of ports. The component tolerance sensitivity is moderate, and it will operate well over several percent bandwidth. The termination resistors only dissipate power in the case of a mis-match, so for reasonable splitting applications, the power rating can be low. There is no power lost in the splitter (assuming ideal components and impedances). Wilkinson described this in "An N-WAY Hybrid Power Divider", IRE _Transactions_on_Microwave_Theory_and_Techniques_, January 1960. The circuit works as an N-to-one power combiner, as well. The general prinicple is that for an N-way splitter (or combiner), you take N quarter-wave impedance-matching sections, each matching the source/load impedance Z to (N * Z) then parallel them at the source. The required characteristic impedance of the quarter-wave section is sqrt(Z1 * Z2), where Z1 = the source Z and Z2 = N * Z. This of course reduces to sqrt(N * Z**2). Thus, for the 50 Ohm 2-way splitter, the line Z = 70.7 Ohms, and for the 4-way, the line Z = 100 Ohms. Termination resistors are placed at the outputs to cancel any load-to-load coupling (the load1 - source - load2 path has a 180 degree phase shift, and the resistor path cancels this). The quarter-wave sections can be replaced with equivalent lumped-impedance PI sections, where the reactance of the components is equal to the line impedance. The source-end capacitors can be combined into a single capacitor (XCs = Zline / N). For a 1 MHz 2-way splitter: Zline = 70.7 Ohms, Cs = 4.5 nF, Ls = 11.75 uH, Cl = 2.25 nF. For a 1 MHz 4-way splitter: Zline = 100 Ohms, Cs = 6.37 nF, Ls = 15.9 uH, Cl = 1.59 nF. A two-way Wilkinson power divider: (XLs = XCl = 70.7 Ohms, XCs = 70.7 / 2 Ohms) 50 Ohm in o--+----[ Ls ]---+----+----o 50 Ohm out source | | | load | | | Cl | Rt | | 100 Ohm | gnd | | | | +----[ Ls ]---+----+----o 50 Ohm out | | load Cs Cl | | gnd gnd A four-way Wilkinson power divider: (XLs = XCl = 100 Ohms, XCs = 25 Ohms) +---+---+---+ | | | | | | | | Rt Rt Rt Rt (Rt = 50 Ohms) | | | | | | | | | | | | 50 Ohm in o--+----[ Ls ]---+----|---|---|---+---o 50 Ohm out source | | | | | load | | | | | Cl | | | | | | | | | | | | | gnd | | | | | | | | | | | +----[ Ls ]---+----|---|---+-------o 50 Ohm out | | | | load | | | | Cl | | | | | | | | | | gnd | | | | | | | | +----[ Ls ]---+----|---+-----------o 50 Ohm out | | | load | | | Cl | | | | | | | gnd | | | | | +----[ Ls ]---+----+---------------o 50 Ohm out | | load Cs Cl | | gnd gnd I hope this has been at least amusing, if not helpful. -- Paul M. Elliott Optilink Corporation (707) 795-9444 {pyramid,pixar,tekbspa}!optilink!elliott "I used to think I was indecisive, but now I'm not so sure."