Coilcraft recently announced a new version of their 'Filter
Designer' software, which is a customized version of the Nuhertz Technologies' 'FilterSolutions'
filter synthesis and analysis program. Filter Designer will generate schematics and response curves for 3rd-, 5th-,
and 7th-order lowpass Elliptic filters that incorporate Coilcraft inductors. The functionality is therefore somewhat
limited and, honestly, is probably more of an advertisement for the full-blown commercial version of the software,
but I did find one aspect very instructive. If you have built a filter using real-world components, then you know
how sensitive inductors are to package configuration, wire winding characteristics, and placement on the circuit
board. A simulation can be utterly worthless in predicting the performance of a prototype filter if the modeling
is not done precisely to accommodate particular component parameters. That's where having Coilcraft incorporate
its own well-characterized inductor models into the program is a great advantage.
To illustrate the effects of package parameters on filter performance, compare the two screen shots of the default
circuit when the program is fired up. The top plot uses Coilcraft's 0603 surface mount package style while the bottom
plot uses the 0302 package. Note the much higher insertion loss with a major slope going into the stopband, and
the significantly degraded return loss throughout the passband. The problem might be partially caused by the series
inductors that were automatically selected being limited to 34 nH in value because that is the largest value
available in the 0302 package. All the other capacitor and inductors remain the same as with the 0603 package implementation.
To test that theory, I changed the passband to 300 MHz so that smaller inductors would be chosen, and discovered
that both the 0603 and the 0302 package versions use the same value inductors within 1 nH. That resulted in
a much smaller difference between the two plots, although the larger package size still resulted in lower insertion
loss and better insertion loss as might be expected due to a larger diameter wire.
The lesson here is when using software that prefers one manufacturer's parts over another's, to be sure to note
component values when comparing presented implementation options. Space restrictions could be forcing you to use
small components that might be available from another supplier but are not included in the database of vendor-customized
software.
One solution is to buy the vendor-agnostic full version of Nuhertz Technologies' 'FilterSolutions,'
or one of the other programs like those offered by NI (formerly
AWR),
Keysight Technologies (formerly Agilent, formerly HP),
AADE Filter Design
(now free), and others. Many of the older free filter synthesis programs that did a
great job for most cases were compiled as 16-bit and therefore will not run on 64-bit microprocessors. There might
be some good smartphone apps out there today, but I have not researched them. Please send me an e-mail if you know
of any worth mentioning.
Posted August 31, 2015
