An Improved Front-End Filter for the Pixie Transceiver

February 12, 2021 by Rick Andersen, KE3IJ

It's a 'given' that the output filter [or front-end, depending on whether you're transmitting or receiving] on a Pixie is as "broad as a barn door", which is one reason Pixie haters hate the Pixie.

A while back, I came up with a modification to a Pi-output filter, which filter gives a low-pass response, and is used in the Pixie transceiver. The original mod was to add identical input and output, series capacitors to each end of the Pi network... giving a rather nice bandpass response, but one that doesn't seem to be plagued by 'ringing' like the simpler parallel L-C tank that I once tried on a Pixie. What I like about the Pi-filter is that the L [inductance] value is low-- typically requiring between 10 and 20 turns of wire on a small coil form or toroid, for the Ham bands.

While simulating the filter, including the 22uH choke that ties the output transistor's collector to +Vcc, I realized that its value might interact with the bandpass filter I had just concocted. What I discovered was that, if its value is lowered to match the inductance in the pi-network, a nice resonant peak appears in the simulated output plot. The ratio of cap values is about 2:1, shunt caps vs. series caps [roughly 20 and 44 ohms of reactance.] So, using the 80 meter band as a starting point, I came up with the following:

1) Change the shunt caps to 2200 pF [.0022uF]
2) Change the .01uF collector coupling cap to 1000pF [.001uF]
3) ADD a series cap at the antenna end [1000 pF; .001uF]
4) Change the original pi-network inductor to about 1.1uH
5) Change the collector choke to match [1.1uH]

The capacitors I used are the small green mylar or polyester caps, which I like to call "chiclet caps" because they resemble the square green 'Chiclets' gum I used to chew as a younger guy with better teeth. You can find a whole box of these, in assorted values, for about $18 on Amazon.

I wound my coils on Amidon T50-2 red-gray cores. 15 or 16 turns of #28 enameled copper wire on each.

If receiving, adjust the turns spacing for a peak in the signals or band noise you're listening to. If you can't hear an audible peak, add or subtract a turn if necessary.

If transmitting into a 50 ohm resistive load, adjust the spacing on the coils for maximum power / minimum SWR; drip candle wax onto the coils to hold the turns in place.

The output of this bandpass filter peaks up sharply at the design frequency [I used 3.56 MHz for 80 meters]. The accompanying schematic shows theoretical values for 160, 80, 40, 30, and 20 meters. [I have built only the 80 meter version at this point.]