MRF101 output transformer

[paul002]

I build a rf amplifier 1.8 to 30 Mhz using 2 MRF101 mosfets. But by reading the specsheet again, I saw I missed something. The output impedance of these fets are about 12 ohm. That is if you have 2 fets I need to match 24 ohm too 50 ohm which need an 2:1 impedance transformer. Not a simple 4:1.
Does anybody know some examples of such matching?

There are some possibilities like:
1) Match 1:4 and match 9:1 (50 ohm to 200, 200 to 22) but step up 9x creates quite high voltage on the transformer connections and a lot of cores are necessary
2) 2 times a 1:4 transformer one for each fet and have outputs in parallel. I don't know if possible.
3) a 2:3 transformer, I don't know if I can build it with tubes or just use wires something to be tested.

But maybe there are more options.

[richard.cs]

There are some possibilities like:
1) Match 1:4 and match 9:1 (50 ohm to 200, 200 to 22) but step up 9x creates quite high voltage on the transformer connections and a lot of cores are necessary
2) 2 times a 1:4 transformer one for each fet and have outputs in parallel. I don't know if possible.
3) a 2:3 transformer, I don't know if I can build it with tubes or just use wires something to be tested.

Option 1 sounds clunky to me
Option 2 wouldn't match, each transformer would output at 48 ohms, but paralleling them would bring it back down to 24 Ohms. You'd need 2x 1:8 which is hard. How about two 1:9 transformers? Paralleling the outputs then gets you down to 1:4.5 so about 54 Ohms.
Option 3 sounds better, build a ~1:2 (in Z) transformer, you want about 1:1.41 in turns ratio. 2:3 in turns would give you 1;2.25 in Z so about 54 Ohms again.

So that's two ways of matching to 54 Ohms, either two lots of 1:9 (in Z) giving two 108 Ohm outputs (from 12 Ohms) you parallel for 54 Ohms, or one transformer at 1:2.25 (in Z) taking 24 Ohms in.

By the way if you are using the tube construction take care with centre taps - the cores act independently which has some implications that are often poorly understood. There's a good write-up here: https://ludens.cl/Electron/mosfetamps/amps.html

[paul002]

Option 3 sounds better, build a ~1:2 (in Z) transformer, you want about 1:1.41 in turns ratio. 2:3 in turns would give you 1;2.25 in Z so about 54 Ohms again.

So that's two ways of matching to 54 Ohms, either two lots of 1:9 (in Z) giving two 108 Ohm outputs (from 12 Ohms) you parallel for 54 Ohms, or one transformer at 1:2.25 (in Z) taking 24 Ohms in.

By the way if you are using the tube construction take care with centre taps - the cores act independently which has some implications that are often poorly understood. There's a good write-up here: https://ludens.cl/Electron/mosfetamps/amps.html

I have red Ludens comments it is an interesting piece, thanks.
I have both tested a tube construction and tested a transmissionline version. Both were (Z) 4:1 which gave problems due to the low impedance.
I found a 4th option in the book of Jerry Sevick but have to think how to build it with the cores I have, his construction was done on a toroid.
 

[richard.cs]

I found a 4th option in the book of Jerry Sevick but have to think how to build it with the cores I have, his construction was done on a toroid.
 

That looks to be a 1.5:1 (voltage) / 2.25:1 (impedance) autotransformer, followed by a common mode choke to achieve some balancing of the low-Z side, it's not really the best way to produce balance but it may work well enough. The choke is needed because the autotransformer connected that way is unbalanced input and output, it's possible a different autotransformer arrangement might be simpler than having that simple winding arrangement and then adding complexity with a second core.

[paul002]

The reservation I have to 2:3 style is that the impedance increases quite a bit with frequency compared to a tube stye 4:1.
Did some quick measurement and the 2:3 increases 57% in impedance compared to the 4:1 only 28%. (Without compensation)
I also tried 1:1.5 (just pealoff half a winding) but that is even worse. I think it still picks up inductance from the cores.

[profdc9]

The reason why tube-style is used, as I understand it, is to minimize the leakage inductance with the primary turn being brass tubes that enclose the secondary turns, and so a standard 2:3 turn transformer can be used where the primary and secondary are not connected except by magnetic flux (and capacitance).  You don't have to use the tube construction, just rather put two wire turns around the two tubes for the primary and three for the secondary to get a 1:2.25 impedance transformation.  The tubes will get hotter because there will be more leakage inductance.  A 2:3 transformer on a toroid would be an easier construction overall, perhaps using 4:6 turns depending on the AL of the core at the frequency of use.  The 4 turns could be bifilar and the 6 turns could be trifilar and interleaved with each other to mimimize inductance.  You could stack two FT82-43 cores to get over 200 ohms of impedance at 10 MHz as calculated here ( https://owenduffy.net/calc/toroid.htm ) on the primary, more than enough to get a good transformation.    Probably even one would be sufficient for 24 ohms primary impedance.

[paul002]

just rather put two wire turns around the two tubes for the primary and three for the secondary to get a 1:2.25 impedance transformation.  The tubes will get hotter because there will be more leakage inductance. 
(Attachment Link)

I think you don't refer here to tubes but to the cores. Indeed that solution generates to much leakage induction.
I understand exact your 2nd solution, but I just tested a 6 turns trifilar on a 61 core connected as described by Sevick. That gives already a better performance. The wire is maybe to thin for the rf-amp but I am out of 1mm winding wire.

[paul002]

here a pciture of the transformer

[David Hess]

I build a rf amplifier 1.8 to 30 Mhz using 2 MRF101 mosfets. But by reading the specsheet again, I saw I missed something. The output impedance of these fets are about 12 ohm. That is if you have 2 fets I need to match 24 ohm too 50 ohm which need an 2:1 impedance transformer. Not a simple 4:1.

Could you link the specifications sheet?

Does anybody know some examples of such matching?

I have never matched an output power stage based on the output impedance of the transistors.  Normally the output impedance of the transistors is much lower than the matched output impedance because of efficiency concerns.  A matched load would provide maximum output power, and also destroy the transistors.

[paul002]

The output matching for 13 Mhz is 11.3 - 6.4 J ohm that is according to the specsheet the optimum the transistor should see. A 1:4 transformer terminated with 50 ohms has an impedance of 12.5 ohm but this is balanced so every transistor sees 6.25 ohm. (in relality it is 11.9 + 3.89J for 7Mhz with the complex part increasing with frequency due to leakage inductance). With 12.5 ohm impedance you can generate 2*48V*48V / 12.5 = 368 watts. Each transitor is capable of 182 watts dissipation. So with a 1:4 transformer I would be loading the transistors maximal, and there you are correct, you don't want that a 50% margin is normaly advised (see book solid state design of Wes Hayward). I could lower the voltage, but these fets become less linear when you do that.
Finding a good 2:1 transform is also not easy, there are some designs in the Amidon book called W2FMI-HU50 which is 1:1.78
 

[paul002]

Did some experimenting and the results looks ok, a 1:2.25 is a nice candidate. With some compensation get quite good results. I ordered a FT240-52 core and some thicker wire see if I can make a compact one.

[paul002]

Made a new version of the amp with mrf101an and bn. And 2:1 (Z) transformer. Things looks reasonable now got 3rd harmonic less than 20db, but the drive is also not very linear have to fix that first to get exact measurement. This version is a bit critical if you put to much capacity between the drains it will oscillate. But with a low value I have enough compensation.