I output 100W(@100MHz) under the following conditions, but I do not know the reason why the temperature of the device becomes high and the solution.

Source matching: Function generator → Cp(82pF) → Ls(22nH) → Cp(18pF) → MRF151

Load matching: MRF151 → Directional Coupler → Cs(82*2pF) → Ls(13.5nH) → Cp(68+82pF) → Ls(43nH) → Cp(18pF) → Load(121.130 + i17.425 Ohm)*Since the assumed load is around 100Ω, adjust with a matching circuit.

The constant of the matching circuit is the result of changing the element value and performing source and load.

I think it is necessary to know the impedance of Tr in order to find the Q of the matching circuit. How do you find it?

Please give me some advice.




First, the output load was 50 Ohm for load pull. After that, the output impedance of Tr was calculated back from the matching circuit that outputs 100W. Utilizing the output impedance, I tried to operate with a matching circuit that lowers the Q on the Smith chart to 100 Ohm. However, the output was very low, so the result of load pulling again with a load of 100 Ohm is the matching circuit described when I asked

I forgot to mention the parts that generate heat. One is a dual directional coupler with a pattern between the transistor and the matching circuit. The other is not a matching circuit, but a coil placed in the power line.

2020-09-02 enter image description here fig1 The planned circuit.

2009-09-08 An image of the impedance measurement of each element additionally performed by the spectrum analyzer. The element between the transistors was removed from the measurement point for measurement.

enter image description here fig2 Circuit when 100W output.

enter image description here fig3 circuit diagram

enter image description here pic1 pcb

enter image description here pic2 DDC

enter image description here pic3 C309

enter image description here pic4 L303

enter image description here pic5 C311

enter image description here pic6 L304

enter image description here pic7 C313

enter image description here pic8 Simulated impedance in each element.

  • 2
    \$\begingroup\$ Draw a proper circuit please. \$\endgroup\$
    – Andy aka
    Commented Sep 1, 2020 at 10:07
  • \$\begingroup\$ Andy-san Thank you for your support. I attached a circuit diagram, and when I measured the board with a network analyzer, it showed different impedance from the circuit diagram. These are also difficult to understand. \$\endgroup\$
    – kimura
    Commented Sep 2, 2020 at 0:33

1 Answer 1


Ok, so firstly a ferrite core inductor in the drain supply is almost certainly NOT what you want, go with an air core coil instead the losses will be much more reasonable. Usually I do an air core here, with a deliberately lossy ferrite the other side of the bypass caps (That will carry significant current and can usefully be ATC or similar), sometimes I kill the Q of the ferrite inductor with a few ohms of parallel resistor to reduce the tendency to low frequency oscillation.

You probably want to shunt C302 with a few uF in series with 10R or so again to damp the low frequency tendency to honk.

I am not sure how you are getting that much bias, given that R313 is making a voltage divider with R307, might want to look at that. Might also want to re do the bias generation such that an open circuit wiper contact on the trimmer will not be an expensive disaster. On the subject of bias, where is the thermal compensation?

Why do you have a directional coupler there?

The drain impedance will be both nothing like 50R and the signal will have a mess of harmonic content, I would not expect to get anything meaningful out of a reflectometer bridge in that position, it should go after the output matching and LPF networks. The heating is probably dielectric losses in the main, the harmonics are more then high enough in frequency to heat the dielectric on cheap FR4 significantly.

Incidentally, why is your output a tuned network? The usual approach in a band II amplifier is to go with a transmission line transformer (And usually a push/pull stage to get the useful reduction in second harmonic, makes the filter MUCH nicer to design).

Pro tip for RF devices using that package, bend the outside few mm of the leads upwards before soldering the device down, makes removing it much less sweary.

  • \$\begingroup\$ Dan-san Thank you for all the suggestions. What should we focus on for the drain power supply inductor? When choosing the L301, we only focused on the dimensions and having an inductor at 100MHz. bourns.com/docs/Product-Datasheets/SRP7028A.pdf Please let me know if there is a document regarding the power line filter or bias circuit. \$\endgroup\$
    – kimura
    Commented Sep 9, 2020 at 1:54
  • \$\begingroup\$ I haven't implemented R313, so I was thinking that it might be necessary to set the gate voltage to 0 at startup. There was a disaster in the explanation of the trimmer resistance, but will the contacts burn? I do not understand the need for thermal compensation and am currently investigating. Please let me know if you have any documents. \$\endgroup\$
    – kimura
    Commented Sep 9, 2020 at 1:54
  • \$\begingroup\$ I thought that accurate information could be obtained if the position was closer to the transistor. I thought that harmonics could be solved by the filter on the secondary side of the coupler. Does the coupler have to be 50Ω except on the secondary side (forward direction, reverse direction)? I understand dielectric loss, I shouldn't place the coupler pattern near the output terminals. \$\endgroup\$
    – kimura
    Commented Sep 9, 2020 at 1:55
  • \$\begingroup\$ The reason for using the Tuned network is because we want to reduce the size. 50Ω coaxial cable (length λ/4) is very large at 100MHz. Can a push-pull transformer be used with a single transistor like the MRF151A? Should I use it again? Please tell me the reason why the lead is deformed. Is it possible to reduce the stress on the transistor and solder due to heat? I'm using a translator, so I'm sorry if there is something wrong. \$\endgroup\$
    – kimura
    Commented Sep 9, 2020 at 1:55
  • \$\begingroup\$ @kimura Carbonyl core material at 100Mhz? Not a good idea, go air core you will be much happier. You don't need a 1/4 wave line for a transmission line transformer, that is a matching stub, and in fact a TLT is always electrically short. Not having to tune the thing is a real advantage. You deform the lead to make it much easier to remove the device when you blow it up. RF development always blows up transistors, may as well make your life easy. Have a look at some broadband 2M amp designs for ideas. \$\endgroup\$
    – Dan Mills
    Commented Sep 9, 2020 at 15:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.