2
\$\begingroup\$

What's the general design procedure for trying to do a wideband impedance match to a device (in this case an LNA) which has very frequency-dependent impedance?

As a personal project, I'm trying to design an LNA for the RTL-SDR receiver which operates on the full 24 MHz - 1800 MHz range (it's very wideband, I know).

As I'm a student I've got the full versions of Agilent ADS, Genesys, and EMPro. Designing the amplifier itself in ADS is pretty easy - plop down an RF transistor, pick a bias point that's stable or near stable, and add neutralization if necessary to bring the transistor into stability.

My question is this: how do I match it to 50 ohms in and 75 ohms out? I know that I should match to the optimum impedance for low noise which I can extract from ADS (or Genesys if I export noise data with my S-paramters).

If the impedance was flat over frequency I could just do something like a Chebyshev match, but the impedance for optimum noise figure is frequency dependent. How do I handle something like this?

\$\endgroup\$
4
  • 1
    \$\begingroup\$ Do you have the input impedances of the RTL-SDR? If the input impedance is well behaved, then it's not so hard. If it's highly reactive, then you might be better off by putting a 3 or 4 dB attenuator in front of it, and using that to give you a nice solid 50 ohm impedance to match into from your LNA. Also, why 75 Ohms in? \$\endgroup\$
    – rfdave
    Jul 23 '15 at 1:32
  • \$\begingroup\$ RTL-SDR is 75 ohms... should read 50 ohms in and 75 ohms out. I'll fix that. \$\endgroup\$ Jul 23 '15 at 1:51
  • \$\begingroup\$ After the LNA, you can match with resistors without much loss of system performance. But why match at all? A small mismatch in the middle won't matter much, just some gain ripple over freq. Rather spend the effort at the front of the LNA where it will make a big difference. \$\endgroup\$
    – tomnexus
    Jul 23 '15 at 5:44
  • \$\begingroup\$ That is some bandwidth. You could almost consider that DC to 2ghz. You'll have to put some additional thought into how bias will be supplied to avoid resistors. Two methods pop into my head. The first would be a distributed LNA. Plently of companies sell these, along with papers explaining how to do it. Your idea about using a filter match might work too. I would de-embed the transistors input capacitance from Sopt and use it in a low pass filter. \$\endgroup\$
    – curtis
    Jul 27 '15 at 19:49
1
\$\begingroup\$

If you want to match the impedances at the input and at the output you could use pieces of transmission line and open circuit stubs between the transistor and the "IN"/"OUT" you are talking about.I always prefer open circuit stubs because I find short circuit difficult to create.You must be given the optimum reflection coefficient of your transistor.Use that as the source reflection coefficient (Since you want to minimize the noise figure) and proceed with TL and stub design.At the output you can maximize the gain.Output does not have much to do with noise.Just use conjugate matching.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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