Ceramic Filter Load Optimization

[k6sti]

This demonstrates how to optimize load resistance for a ceramic filter. The plots are for a Murata SFE10.7MA5-A. It was measured using a 50 ohm VNA. I renormalized the response to 330 ohms with the plotting program. The white curve is the group delay response had the filter been loaded with 330 ohms. I made a copy of the file and varied the load resistance. 220 ohms yielded lowest group delay variation. At 330 ohms, it is just over a microsecond in the 10.6-10.8 MHz passband. It's about 75% as much at 220 ohms.

This neglects the effect of other filters and other circuitry. Still, it's quite easy to optimize the load resistance for an isolated filter for lowest group delay variation. This will reduce audio distortion.



This is the reverse direction. 230 ohms was slightly better than 220 ohms. The filter was measured in both directions with a VNA that measures S11 and S21. A program included with the plotter merged the two files into one that contained S11, S21, S12, and S22. All four S-parameters are needed for the most accurate reference impedance renormalization.

The plotting program is at the top of this page:

Measurement Utilities

ham-radio.com

Downloading instructions are at the bottom of the page.

Brian

 

[k6sti]

I neglected to show the magnitude response. It flattened and shifted about 15 kHz lower in frequency after minimizing the group delay variation.

Brian

 

[Bob@FM]

That program appears quite useful, nice job.
Not sure if you are using an especially bad filter that deviates from normal MA5 280s, or there is a glitch in the calibration or elsewhere. Most the MA5 280's I measure have a more normal GDT curve with 330 ohm loading than shown above. Ditto for the Murata data books, but realize there is a large variation in units out there. It seems Murata tested and then selected "out" the better testing 280's as either low loss, low spurs, or low GDT 280 variants.
so it may have almost left the rest to be somewhat flawed one way or the other. It seems the newer SMD version are a little more consistent and have better quality for shape and center frequency.
On the other hand, maybe it's my eyes expect to see the less blown up Y axis GDT uS scale, and having it re-drawn like that makes it appear more odd than it really is.

Below is a typical MA5 280 re-labeled with a later Murata model number. I also attached the conversion chart / part marking guide if anyone needs. This lets you IF filters, and convert back and forth from old to new model numbers