Apologies if the answer to this is already in the encyclopaedia, I looked but I couldn't find it.
    I'm just wondering how >20GHz sources work. I know they use crystal oscillators as reference, which oscillate at ~6GHz, but how is that reference then used to create other frequencies? and in particular, how do they make them go to >20GHz?
    And I guess this will become clear from the previous answers, but why does the cost of the sources increase exponentially with bandwidth?
  • UnknownEditor

    This topic has not been dealt with very well on Microwaves101, our apologies! People tend to write textbooks about what they know, and I never really had to work on a source.

    Modern sources use voltage controlled oscillators and phase-locked loops, with the output usually below 6 GHz. Then, mixers, multipliers or other nonlinear devices are used to achieve higher frequencies.

    Bandwidth in a VCO is limited, maybe an octave. So for wideband, you will need more than one VCO, or more than one multipler, and then somehow mux it all together. Here's an example from Analog Devices


    I am hopeful that someone smarter than I will add to this answer!

    Steve (the Unknown Editor)
  • JPD
    Thanks Steve! That was very useful.
    I have a follow-up question that you might also be able to help me with. I've been looking at frequency synthesizers and noticed that some of them have phase control with >1 degree resolution. How is this done? I understand that these synthesizers use PLL to get their phase reference, but how does one then change the relative phase at the output?
  • UnknownEditor
    We are going to need some expert help here... I have exhausted my limited knowledge of sources. My guess is that this is done in the digital domain, it would be hard to maintain 1 degree resolution using analog networks.

    Where are all the smart people out there when you need them....

  • JPD
    Hahaha thanks
    Indeed I also thought it needed to be done digitally, but even then, if you naively think about digitally shifting a 10 GHz waveform by 1 degree, it would imply that you need 3.6 TSa/s sampling rate!
    I spoke to Yu Gao, a tech rep from Anapico about this, and he explained that the trick is not to shift the signal directly in time, but to just change the amplitudes of your digitised waveform, so that you get the shifted signals. With ~10 bits of resolution in the amplitude, then it is more straightforward to get the tiny phase resolution.
  • Gugeta
    PLL to get their phase reference, but how does one then change.
Add a Comment


Join the international conversation on a broad range of microwave and RF topics. Learn about the latest developments in our industry, post questions for your peers to answer, and weigh in with some answers if you can!