That's a beautiful graph, don't you think? In addition to the very pleasing results, I liked Bob's methods: the "by hand" collection of the data points using an AD9850, a 'scope and a notepad (see below); the filing down of ceramic disc caps; the use of nail polish hardener -- all this adds a definite artisanal element to this project and puts more soul in the new machine.
Like Bob, I too kind of bailed out on the Q calculation when I was doing this. But as I recall there is a variation on the G3UUR method that yields this parameter too, right?
I'm reluctant to share this with you because the results appear too
good. I've attached a graph showing my four crystal, 11Mhz
filter measurements. The graphed points are read values using my
AD9850 DDS VFO feeding to a TEK scope.
A TIA amp was used for input and output of the filter. My development software was the Steder-Hardcastle software as presented in November 2009, QEX.
I am now practiced in the black art of filing off the tops of disc
ceramic caps to "adjust" their values. This black art also involves
the mysterious qualities of Sally Hansen Nail Hardener.
Let me share my method for developing the filter.
I've built CW filters before but this was my first effort at SSB
bandwidth which is less forgiving.
The G3UUR oscillator method (see page 3.19 in EMRFD) is a simple and
effective filter design method. but it does not provide a measure for
crystal Q, a value which impacts filter insertion loss.
So I decided to choose crystals with proven pedigree. These were 11 Mhz crystals from Mouser, part number 20-HCA1100-S. A lot of ten costs $5. These were the crystals selected by Jim Kortge, K8IQY, for use in his 2N2/20 rig.
For software, I use the Dishal package that can be downloaded from the ARRL and other sources. This package was the basis for the
Steder-Hardcastle article in November 2009, QEX. The "Xtal" pull down
menu provides entries for an individual G3UUR oscillator.
Simply put, all critical filter input values are generated by reading
the change in crystal frequency as an additional capacitor is added
into the oscillator circuit.
I suggest starting with a 4 crystal filter. Only two capacitor values
were required for my filter. Five capacitors were required--two series
and three shunt.
Start by reading the "open switch" frequency for each crystal. Sort
the crystals into increasing frequency order and choose the four with
the most narrow frequency span.
Using the pull down menu measure the individual crystal measures for
Lm, Cm, and series frequency. Average these across the four crystals.
The Lm or Cm and series frequency are placed into the Dishal software
main menu. Also enter the average Cp which is the measured capacitance across the crystal leads.
Finish up the main menu entries by entering the number of crystals (4), and the desired bandwidth--generally 2.4 to 2.9 Khz. Finally enter the acceptable ripple, which is often 0.1db.
Let the software calculate the filter values. Expect some odd
capacitance values. By changing the filter bandwidth--say from 2.4 Khz to 2.35 Khz I can move one of the capacitance values to a standard value.
The software also displays the input and output impedance. If the
filter is centered between two TIA amps, this filter impedance must be
transformed to 50 ohms in the amps. This provides the transformer
The Dishal software has always given me good results. But I haven't
compared its results to Ladpac--especially GPLA.