8

I have read about matching networks, authors usually will describe a target impedance to be presented, to say, the output of an amplifier, and then they design the matching network. To widen the bandwith they usually create a matching network with more sections, thus, lowering the general Q and presenting the same impedance to a wider range of frequencies.

But my question is this: What's the point on presenting the same impedance over a wide range of frequencies if the S-parameters (and thus the required load/source to be presented to the amp) change with frequency? For example the optimum source impedance for LNA design is not the same at different frequencies, in a power amplifier the optimum load varies with frequency, so whats the point on designing a matching network that presents the same load over a wide range of frequencies?, instead, shouldnt a network be designed to "track" those impedance changes, thus presenting the ideal source/load required for each frequency?, how is a network like that designed?

S.s.
  • 2,700
  • 6
  • 26
  • 50
  • 1
    I'm not familiar with the issue you're asking about, but I suspect a matter of scale. Say you have a system operating around 100 MHz. A simple matching circuit might give a good match from 98 to 102 MHz. A better matching circuit might give a good match from 90 to 110 MHz. If you tried to operate at 50 or 150 MHz, then the amplifier characteristics might change enough that you'd need a different matching circuit. – The Photon Nov 07 '17 at 03:31
  • If you gave some concrete examples of the amplifiers you're thinking of, and the matching circuits you've seen proposed to use with them, it might make it more clear. – The Photon Nov 07 '17 at 03:32
  • Thats actually my question, if I want to match over a wide range say between 100MHz to 800MHz, how does one design a matching network since the S parameters are different over frequency? – S.s. Nov 07 '17 at 17:53
  • Can you give an example of the amplifiers you're looking at? – The Photon Nov 07 '17 at 18:00
  • FWIW, my field is high speed digital rather than classical rf, but we often want to amplify very broadband signals (like 100 kHz to 10 GHz). Our solution is "choose an amplifier that's well-enough-matched over the required band" rather than "design a matching network". – The Photon Nov 07 '17 at 18:42
  • Yes, im actually refering to discrete amplifier design, for example one using this transistor http://www.cel.com/pdf/datasheets/NE3210S01.pdf say you want to match from 2GHz to 3.5 GHz, the S parameters are similar but not the same, so how does one design a matching network which tracks the S parameter changes? – S.s. Nov 07 '17 at 18:46
  • We'll see what the real RF guys say, but I suspect the answer is "don't do that". Most classic RF applications span much less than an octave, so it may be unreasonable to expect to use a single-transistor amplifier in an application like yours that spans 3 octaves. – The Photon Nov 07 '17 at 18:48
  • I've worked in RF PA design, in which bandwidths of say 1GHz or more are highly desireable and achievable using single transistor amplifiers. – S.s. Nov 07 '17 at 18:51
  • Does a typical customer application use all that bandwidth or does each customer just use a small band within the amplifier bandwidth? – The Photon Nov 07 '17 at 19:39
  • Usually the wide bandwith is to allocate several users, so each customer uses a small band within the amp bandwidth. – S.s. Nov 07 '17 at 23:06

1 Answers1

4

I've developed a tool (based on the algorithms described in [2]) for matching a frequency dependent complex source to a frequency dependent complex load impedance. It is still WIP, but I think it works a treat. You'll have to compile that from the source and provide the impedances to match as s1p files.

https://github.com/Qucs/qucs/pull/555

On the other hand, Keysight ADS comes with a similar tool based on the RFT and the analytic methods.

Some references:

[1] Wideband circuit design. H.J. Carlin. 1998. CRC Press

[2] http://ethw.org/images/a/a6/RF_Networks_OCR.pdf

Hope it helps

clabacchio
  • 13,481
  • 4
  • 40
  • 80
  • I've read the work of Carlin and Yarman aswell both are incredibly complex texts, if im not mistaken they use the real frequency technique, I always though making a code using their algorithms was too laborious and eventually I gave up. AFAIK ADS doesnt have a tool which allows you to match a frequency dependent load. I appreciate your tool, but I would like to know more about the nuts and bolts on how to make them by myself. – S.s. Nov 09 '17 at 01:47