Revisiting the RTLSDR CW Skimmer with a dedicated mini pc required some rethinking of the input stages/filtering when it came to thinking about multi-band support.
Other multi band skimmers appear to have either gone with completely wide open front ends , or multiple antennas/cascaded combiners. With 8bit sampling and a single antenna (active loop) I needed to take a more selective approach to protect the front end.
Inspiration came from looking at papers on multi band front end filtering for 4g networks. It turns out you can easily parallel BPFs with relatively minimal downsides, the main one being reduced stop band attenuation between BPF center frequencies, but it wins in simplicity. I had never contemplated this approach, and have yet to find anything in the amateur literature on it.
First up, while the above is calibrated, so that 0dB is at 0, the two little bumps at 3.5 and 5MHz indicates that the output amp has minor but noticeable second harmonic output. Adjusting for that, I’m guessing the actual insertion loss is less than 1dB and probably 2-3dB for 10MHz. Know your test equipment and it’s limitations.
The second learning, was that if you’re looking at 100-200kHz bandwidth filters at these frequencies, toroid size is actually critical. FT37 will be marginal at best, FT50 will be easily sufficient. charts are also available online plotting Q vs frequency, turns and core type. So one improvement will be to swap the FT37-6 one of the filters for larger cores.
The other, again related to Q, is that capacitor selection is also important. NP0 or silver mica. Trying generic ceramics resulted in 20dB of insertion loss! Probably not anything new for anyone experienced, but nice to see this play out visually, and impressive the difference.
The other test was to attempt to measure the Q of the inductors with the SNA. Tried two approaches, the working one being lightly coupling (4.7pf) to the LC circuit and measuring 3 points of the frequency response curve. Calculated a Loaded Q of 175 for the FT37-6 at 11Mhz which matches closely with a unloaded Q of 190 as expected from other sources. It may be crude, but the SNA is a workhorse.
So time to rebuild this in the new front-end enclosure for the skimmer…