**GENERAL DISCUSSION ON PRACTICAL FILTER REALISATION PROBLEMS**
During this month of January I have built and tested several prototypes. The outcome of of the results is as follows.

1) Practical realization of filters with in/out 50 Ohm impedances is difficult, because capacitor values near or above 1000 pF are needed and temperature stable high value precision capacitors are difficult to find. At this cutoff frequency (2.200 KHz) it is advisable to go for higher internal impedances (such as 500 Ohm) which require more easily available components.

2) My choice of

**Series First topology **(see previous posts) in order to minimise the number of coils and hence the insertion losses,

** was not ideal**, because then one would be obliged to use transformers (additional components) for impedance trasformation, so losses would go up again, apart from making the filter more complex.

- NEW HP filter topologies .jpg (74.51 KiB) Viewed 6271 times

3) To conclude, the best topology, after all, is the

**Shunt First** as shown above.With this configuration the filter has a shunt inductor at the beginning and one at the end. If we change these inductors into transformers by adding an additional link (winding) we obtain a big simplification, hence smaller size, and lower losses in practice. With this topology we have the advantage we can use the impedance level as a parameter for filter component optimisation, because the turns ratio between main winding and link permits to accommodate any impedance shift. This configuration makes the use of toroids mandatory, at least for the two end inductors, in order to have strong coupling between the windings.

4) Another advantage is that any filter, whatever the level of the internal design impedance, can be easily measured with standard 50 Ohm instrumentation by just adding a link to the two external coils. Furthermore the links can even provide double ground isolation if desired, resulting in exceptional ground balance, useful for local noise rejection.

Following the above criteria, I calculated and built the experimental proptotype whose excellent performance was illustrated in the previous posts. In the next post complete data for this prototype will be uploaded.