January 2019 LDMOS Amplifier

Here’s my latest dual 1K80H amplifier, mostly in pictures. Actually, this is not “my” design. I built this in an afternoon and it most resembles the Elecraft amplifier from the drains back (output section).

The Elecraft output section design looked pretty elegant and simple. So I wanted to try and copy it and see how it performed.

Of course, this is not an exact copy and there are many minor differences which all add up.

That said, none of these amps with the CMFC scheme you see here, and other similar ones, can ever be linear.

Getting cleaner, smarter, more efficient all the time. Already have ideas for Rev 6. Here we have back to back 4:1 Guanella TLTs.
Much better 50V DC choke, design and placement. Needs to be as close to the Drains as possible.
About 6 months of designing, experimenting and simulating went into this new Bias circuit. Perfectly compensates the Gate Bias Voltage by -2.5mV/deg C and maintains the initial Ids no matter what the temperature.
Where did all the capacitors (little bombs) go?
The Guanella balun is BACK! Provides excellent isolation, no reason not to use it.
Where’s the feedback? It may come back. Looking at that right now. Feedback used to be very important with older devices with a lot of internal parasitics. But now?

Comments

38 responses to “January 2019 LDMOS Amplifier”

  1. Joe Avatar
    Joe

    Test it for maximum power ?

    1. N4GA - Rob Avatar
      N4GA – Rob

      Hi Joe,

      I don’t like to test these for MAX power because they will typically keep going until something blows up. I test up to 1600-1700W CW key down. Then I usually run it up to 2kW just to see what kind of headroom above 1500W that it has.

      This pallet will easily do 2kW across all bands except 6m.

  2. Donald R Solberg Avatar
    Donald R Solberg

    Rob,

    Your amplifier is really good looking, very professional design. We can all learn from your experience. I would love to see a schematic if you have one.

    73,

    Don
    K9AQ

    1. Dennis Pennington Avatar
      Dennis Pennington

      Do you sell them?

      1. N4GA - Rob Avatar
        N4GA – Rob

        Not right now. Some day perhaps. Right now its more of a R&D thing for me.

  3. Tim Avatar
    Tim

    Yes, Please post the schematic to your Bias circuit.

    Tim
    WB0WUQ

    1. N4GA - Rob Avatar
      N4GA – Rob

      Hi Tim,

      Good morning. I post what I am comfortable posting. Too much time and effort went into getting the circuit perfect. More importantly, the values are only for a very specific set of criteria and LDMOS devices used.

      I.e., for a dual 188XR with 51.4V Vds with a target Ibias adjustable from 3-5A and a resulting -2.5mV/degC compensation, I have values calculated.

      For my dual 1k80h pallet, the values are different for the same outcome. If I change Vds, the values all change for either amp.

      I don’t want to put info out there that are invalid because someone is not using this exact setup, like maybe a 48V supply, or a 65V supply, or some other LDMOS, or maybe just a single, etc.

  4. Sebastian Avatar
    Sebastian

    Hello, this design seems to be very similar to the KPA-1500. I would suggest to add negative feedback and a ground resistor at the gates to avoid common mode oscillations. Do you supply 65 volts to it ?

    1. N4GA - Rob Avatar
      N4GA – Rob

      Sebastian – yes, it’s very similar to the KPA. Good eye. If you design a 16:1 using two 4:1 Guanella TLTs, also use a 50V common mode choke and a 50V input inductor, it’s going to look pretty much like this. This configuration has it’s limitations however, and requires quite a bit of compensation (caps) to work well from 1-50 Mhz.

      These are 1k80h 65V devices, run at 54 volts.

      I’m familiar with parasitic oscillations when using parallel MOSFETS, this is well known. Talk to us about “common mode oscillations.”

      1. Sebastian Avatar
        Sebastian

        Rob, Thank you. Looking more closely, I see that you already have a path to a 10 ohm ground at the biasing network through the transformer secondary. I would have placed the ground path closer to the gates and not in series with the transformer inductance.

        I am building a similar amplifier, but I am evaluating a different approach.

        1. Sebastian Avatar
          Sebastian

          Ohhh and I forgot the compensation: For the first 1:4 transformer you used TC-12 coax, which has around 10-11 ohms Zo instead of the ideal 6.25 ohm Zo. This along with the PCB traces to the drains represent some leakage inductance or mismatch that you will have to compensate to reach 6m band. However, I think the output capcitance of the LDMOS devices will probably be enough… or not. What are your findings ?

  5. Bill Avatar

    61 or 67 material?
    do you sell the PCB or a kit to build?

    de K4KYB
    Bill

  6. Don Beaver Avatar

    I wanted to know about the heat dissipation capabilities when not bolting down the LDMOS or Soldering the bottom to the copper Heat sink.

    1. N4GA - Rob Avatar
      N4GA – Rob

      Unfortunately you won’t find that here. I’m a firm believer in soldering down the device. All the years of building audio amps with mica insulators and thermal grease. Seemed too good to be true that I could solder down the LDMOS! Never looked back..

  7. STEVE Avatar
    STEVE

    WOULD you suggest to solder the base if the package is TO247. I am trying to find best possible way to tackle heat in a TO247 package of around 600watt dissipatation

    1. N4GA - Rob Avatar
      N4GA – Rob

      My only experience with that is the MRF300 devices. With those I used that liquid metal compound between the To247 and the copper spreader. I think you can solder them if you want. But I was making like tear down prototype amps that live like 3 hours before being taken back apart, so I did not even consider it. The MRF300s run very cool just bolted down.

  8. Mike KF8OD Avatar

    Great work Rob! Thanks for sharing!

    1. N4GA - Rob Avatar
      N4GA – Rob

      My pleasure, Amigo.

  9. Oscar Luciani from Argentine Avatar
    Oscar Luciani from Argentine

    Excelent work , very nice congratulations

  10. Tony Avatar
    Tony

    Hi Rob you do good work, I wish I could imitate your work, I’m a noobie with LDMOS, so far I’m just reading from the experts and then I’ll jump on it. Keep up the good work, nice shots by the way.

    1. N4GA - Rob Avatar
      N4GA – Rob

      Awesome Tony! Keep studying…that’s all it takes!

  11. Juergen Avatar
    Juergen

    Hi Rob,
    i like your insights on these LDMOS PAs.
    One comment from me, the first TLT the should be better with two 12Ohm coax cables in parallel to match the impedance of the FETs and the 2nd TLT with the 25Ohm cables. Even though I have to admit, that the difference will be noticed on higher frequencies only. At 50Mhz this will have an influence of approx 0.4 dB. (10% loss)
    The downside is that the two coax lines use more space and you may come to fewer turns through the toroids.

    best regards,

    Jürgen DL4JY

    1. N4GA - Rob Avatar
      N4GA – Rob

      Jürgen,

      Your math is correct. However, all this is very approximate. As a SSB signal swings from 0W to say 1kW, the output imedance swings right with it, as well as the output capacitance!

      Another thing, the shorter the piece of coax, the less critical the impedance is. So one loop of 10 ohm would not be much difference than parallel 12 ohm (net 6 ohm).

      And you pointed this out nicely!

      73

      Rob

  12. mervin dulce Avatar
    mervin dulce

    hi rob,
    mervin /va6md,how did you connect the end of the balun ,i dont see it on the picture,its 4;1 to 4:1 in series then the 1:1 balun….

    Thanks

    1. N4GA - Rob Avatar
      N4GA – Rob

      Mervin,

      Yes back to back 4:1, then balun. A row of caps series between the braid and ground, and the same between the center and output.

  13. Gan Uesli Starling Avatar

    As I am a CW-only op since 1989, I can’t help but long for a Class B, C or even D amp. Less wasted energy, much less heat that needs to be shed. Could a more efficient (purely CW) amp be accomplished on an LDMOS pallet?

    1. N4GA - Rob Avatar
      N4GA – Rob

      Yes, but not broadbanded.

  14. ROSU VASILE Avatar
    ROSU VASILE

    Hi
    what is the price of this pcb without electronic components?

    1. N4GA - Rob Avatar
      N4GA – Rob

      I’m sorry Rosu, this was a one-off and is long gone.

  15. Victor Avatar
    Victor

    Do you sell complete units, reference materials for educational purposes, parts, LPF included, etc? All info you can provide. Thanks.

    1. N4GA - Rob Avatar
      N4GA – Rob

      Sorry Victor, I do not sell anything right now. I probably will in the future.

  16. Rick AI7JN Avatar
    Rick AI7JN

    I want to use an MRF300-B to replace an MRF300-A in my 600W dual LDMOS HF amp (TO-247-3 package) because it’s what I have. Since it’s a mirror configuration I would have to use 2″ wire jumpers to connect to the circuit board. AWG 12 will handle 8+ amps of current but the spec sheet says that it can only be used up to a few KHz. Can you recommend some kind of wire or will I have to just spend the cash and order an ‘A’ so I can solder the flat leads directly to the board? Further confusing me, the center lead (source) is presently only connected uising a wire of about 22 AWG. How can that be handling the current? Thanks for your help. (A little knowledge is a dangerous thing.)

    1. Rick AI7JN Avatar
      Rick AI7JN

      As to the last part, I just realized that the back side also serves as a source terminal and that is how it handles the current.

    2. N4GA - Rob Avatar
      N4GA – Rob

      Correct, the bakside is the source.

      I would not advise using a wire like you talking about. 2 inches of wire will present way too much impedance there and make your push pull very lopsided causing a lot of distortion, and likely destroying the device.. You want those traces to be as short as physically possible, that’s the reason NXP went through all the trouble of making these identical but mirrored parts.

      It might “work” quite compromised on 160 and 80, but definitely not above that.

  17. Ron Toffen Avatar
    Ron Toffen

    Do you build and sell complete
    1kw
    2kw

    1. N4GA - Rob Avatar
      N4GA – Rob

      Not yet. Maybe after I retire. Sorry!

  18. Peter Avatar
    Peter

    Very nice design, congratulation!!!!. I am convinced this amplifier works well on upper HF bands but I doubt about its performance on the lower HF bands (so on 160m, 80m, 60m and even 40m) given my experience playing with the MRFE6VP1K25 (which is a similar transistor to the one you used). The point is that the very high transconductance below 10 Mhz makes it much harder to get a good efficiency on lower HF bands than on upper HF bands and in VHF. (1) If my observations of the pictures are correct, the output of the common mode choke is NOT decoupled to ground by an adequate capacitor. Without that decoupling, the common mode choke has little effects on the waveform of the drains in relation to limiting the negative effects of even harmonics on the efficiency on lower HF bands. (2) There is no feedback to improve linearity and reduce the gain of the amplifier on the lower HF bands. The way I have implemented the feedback is by adding one turn to the output TLT connected to gates via two resistors in the order of 30-70 ohms ( exact value depending on the design of the input and output circuits and depending on the level of feedback desired). This magnetic coupling approach limits considerably the power of the resistor that is needed compared to an approach where the coupling is done directly on the drains using capacitors. The feedback is of course playing principally on lower HF bands where its effect on the linearity, efficiency and stability of the amplifier is just spectacular (so this is no needed for operating in the 10m or 6m band but it recommended in a wide band design to operate with good efficiency at high power between 1.8 and 7.2 Mhz).

    1. N4GA - Rob Avatar
      N4GA – Rob

      Thank you for your comments Peter.

Leave a Reply

Your email address will not be published. Required fields are marked *