Design of general filtersGLM.JPG


Summary
-There are two adjustable filters on General Learning Machine
-Filters are designed with AADE -filter design and QUCS -circuit simulator programs.
- Both have same topology and are based with two-gang adjustable capacitor and one tapped/changeable air-core inductor. Air core is used because easy estimation of inductance.
-Filters are not band-pass but sharp filters. "Sharpness" depends of input/output capacitor.
-I ended up to this topology because:
1) It is less critical for component tolerances than other simulated. Tunable capacitors unbalance will not ruin the filter performance.
2) There is only one inductor to chance, therefore not problem of tolerances in inductors also.
3) I wanted continously lowering graph toward higher frequencies, because filters are meant to use also for cleaning square wawe signal from Raspberry Pi.

Picture: On first pic, bandscope of Si4735-receiver is used with filter tuned at middle of scope, around 25Mhz.. On next pic, filter is bypassed.
scopeFilterscopeNoFilter

I made main design work with AADE, and QUCS was used more like confirming results. There is capture from QUCS first, because filter topology and parts are more clearly seen on it. Resistor is added only for simulation purpose, it is not a real component.

Qucs2.png

Chancing inpu/output capasitors


In these two simulations on AADE impact of chancing input/output capacitor is clearly visible. Chancing capacitor value from 30 to 10 pf makes filter much more sharp, but also increase losses. Also it moves filters "tuning point", but this is acceptable on tunable filter.
General rule is: when frequency increase, input/output capacitors should decrease. Low frequency and little capacitors causes high losses. In lowest freqs, 50-100 pf can be used. On high side, somewhere 10 meter, even 5pf can be useful. On simulations I used same value for both capacitors, but it is not absolute rule. You can use what you have.

17.5mhz10pf.png

Again, resistors are for simulations only, not real components.

AADE16mhz30pf.png

I played with filter simulations couple of days and ended to this topology. It turns out that I had reinvented "Chebychew"-type filter.
By the way when I made this filter "from scratch" on AADE, it gave a bit different simulation results as when I taked chebyshew-topology readily available on AADE, even with exactly same components and design. Therefore I made some confirmation work with QUCS.
On these two picture readily available topology is simulated.

I did not tried to make the-best-specs-ever -filter. I morely tried to create flexible, simple and reliable design.

This is first time I have used AADE and QUCS, so do not take it too serious.