I scored a nifty YIG Tuned Filter (YTF) on ebay the other day. I bought it on account of the 1.85mm connectors, as they suggested a frequency range that was relevant to my interests. Of course, it was still a gamble, so the next step was to characterize the filter and see if the gamble paid off. This repo tells the story of said characterization.
Spoiler alert: it paid off :)
- Programmable Power Supply: GPP-4323
- VNA to 50GHz: HP 8510C
- SNA above 50GHz: Marki quadruplers & mixers, Centellax amplifier, SignalHound spectrum analyzer
The YTF's "deliverable" is a current-controlled bandpass characteristic. Put in a current I, get a bandpass centered at frequency f. I(f) is the key control characteristic. Up to 50GHz, I characterized this with my repaired 8510C network analyzer. Results:
Click for Live
- F3560R Data Gathering.ipynb: controls instruments, conducts measurement
- F3560R YIG Data Fit.ipynb: fits a highly empirical model
- F3560R YIG Polynomial Fit.ipynb: attempts to compress the model
Above 50GHz, I had to use an "improvised" scalar analysis setup. The noise floor is rather high, but it demonstrates an important point: the energy demand to drive the YIG coil goes parabolic at 60GHz, where the high-permeability alloy starts to saturate. However, the YIG becomes leaky between 65GHz and 75GHz, and this could be used to extend the useful range of a spectrum analyzer slightly beyond the nominal upper frequency limit of the YIG! Cool!
The frequency is controlled by current, so the key engineering question is: how much current (and power) is required for a given frequency? A quantitative answer to this question allows for frequency planning and controller design.
How good is the polynomial spline model at capturing the slow frequency response of the YIG? The residual plot tells us.
