On 21 Sep 2023, at 09:38, Nick G4IKZ via groups.io <nds12321@...> wrote:

?Multi-band filters involve an interesting design issues when using the existing QMX filter topology. So it's good to see people looking at the alternatives and posting their results. I'm not sure if you know about 'NanoVNA saver' but it's an easy way of controlling your NanoVNA from a PC and downloading the sweeps from it.

The issue with the existing topology is this.

At first glance it might seem that tri-band filters such as for [20m/17m/15m] would be just as easy to implement as dual-band filters for [80m/60m] or [40m/30m]. After all the fractional bandwidth is very similar for all of them. But there's an important difference. In the dual-band designs the frequencies of the upper and lower bands are in a ratio that fits nicely on to the second and third insertion loss ripples of a fifth-order Chebyshev response. And for the same reason they also fit nicely with the regions of good return loss. So with a little patience the filter responses can be optimised to 'fit' nicely on the two bands. That's basically what I did in my designs for [80m/60m] and [40m/30m]. And it would also be possible when designing a dual-band filter for [20m/15m].

But in such a dual-band filter the 17m band is then at a frequency where the opposite situation is true. 17m is then in the region where insertion loss of a fifth-order Chebyshev is greatest and return loss is worst. The [20m/15m] filter can of course be modified to improve the performance in the 17m band. But that then limits the second harmonic rejection. So the performance of multi-band filters with the existing topology will involve more design compromise - and that will have an impact on the VSWR seen by the PA.

I guess another option is to have three single-band LPFs and accept that it's then just a three-band radio. That would allow excellent harmonic suppression with very good filter match. But I doubt if many builders would want to do that.

Please keep posting your results!

- Nick.