Crushing Spurs with Better Bandpass Filters

PA3AKE's passband (on screen) an mine (on NanoVNA) 

While I was away in the Dominican Republic (3-9 August 2023),  I was thinking about spurs.  While there I watched Nick M0NTV's video about mixers.  The video was all great, but I was especially taken by the way he used a spectrum analyser to evaluate the output of various mixers.  This made me think that I should do the same thing with the output of each of my dual-band BITX rigs. 

I was especially worried about the output from my 17-12 rig.  The IF is at 21.4 MHz.  The VFO runs around 3.5 MHz.   So if you add the IF and the carrier oscillator signal you get to 12 meters.  If you subtract them you get to 17 meters.  But you need some good bandpass filtering to sufficiently knock down the unwanted output from the mixer.  And the BP filter should be sufficiently narrow to take out any remnants of the carrier oscillator signal. I had taken the easy way out and had used simple dual-tuned-circuit (DTC) filters.  I started to wonder if these simple BP filters would be enough to knock down the 12 meter signal while on 17 and the 17 meter signal while on 12.  I pulled out my NanoVNA to look at the passbands: 

Here is what the 17 meter DTC filter passband looked like. The cursor is at 29.6 MHz and you can see that near the 12 meter band it is only providing about 21 db of attenuation.  That is not enough. 

And here is what the 12 meter filter looked like. Here the cursor is at 18.150 Mhz and shows about 25 db of attenuation at this frequency.  Again, not enough.  

The results are as you would expect: I could see 24.9 MHz signals in the output when I was on 17 meters, and I could see 18 MHz signals in the output when I was on 12 meters.  The spurs weren't strong, but they were there.  I knew that more robust BP filters would help.  

At first I used the circuits prescribed by Martein PA3AKE.  He used larger toroidal cores, I used smaller T-50-6 (yellow) cores.  The results were very similar.  See the first picture on this blog post. 

The results were really good.  See pictures below.  I was using a TinySA with the signal fed through a 50 ohm 30db attenuator.  I was putting a 1 kHz signal in to the mic input.  

This picture shows the 12 meter output and the now non-existent 17 meter spur.  The cursor is at 18.142 MHz:    

This picture shows the 17 meter signal and the now non-existent 12 meter spur. The cursor is at 24.993 MHz.   We can see some second harmonic signal getting past the LP filter -- I will fix this.  

Then Farhan commented on Martien's filters, noting that they are all in the "LSB" configuration.   You can see from the charts below how they would be really good when you are trying to use the "difference" output from your mixer while knocking down the sum output,  but not vice versa.     So I built new USB filters for 12 meters,  and for 10 meters in my nee 15-10 rig.  I got better results on the two "sum" bands in my rigs (10 and 12 meters)