After having concentrated on the H-Mode mixer circuit for quite some time and looking at the results obtained with the combination of FSA3157 switches and T1-6T transformers, it was clear that at this point the mixer was not the weakest chain in the receiver's frontend anymore.
Especially the T50 toroid based input band pass filters I used were a problem. They would not allow the frontends intercept point to exceed +40dBm. But also the quartz roofing filter was limiting the overall IP3 at +42dBm. The overall result of this combo was that the IP3 was in the upper thirties, which is good by all means, but the mixer could do much better, potentially over +50dBm on 40M...
The first breakthrough came with the roofing filter. The roofing filter suffered from IMD not following 3rd order law, causing the dBc values of the IMD products to stay more or less constant with decreasing levels of the 2-tones. This gave IP3 values between +41dBm and +50dBm depending on the level of the 2-tones. Thanks to Helmut Dick, EA5GNI, who organized a batch of truly wonderful 9MHz crystals manufactured by QT in Daun Germany, a roofing filter was possible with very low insertion loss and an IIP3 of +53dBm that followed 3rd order law perfectly. A that point, the remaining bottleneck was the input BPF.
Now that the roofing filter was fixed, it seemed only a matter of time to get the BPF's into the "green zone". It should be possible to build a passive LC filter that contributes less IMD than the active mixer that follows it. However if that filter would be practical to construct and not require for instance 2 inch air coils was the big question!
The focus went for the 40M BPF initially as this is the band were the H-Mode mixer peaked above +52dBm in my lab. Also 40M is an important band because high dynamic range is most needed there, especially in Europe given the band conditions and the strong nearby broadcast band. Like with the H-Mode mixer switches and transformers a more or less systematic approach was followed after trying to get some grip through the help of Colin on how IMD was formed in the BPF's in the first place.
Many iron powder toroids ranging from T50 up to T130 in yellow and T50 up to T94 in black material and also some Lodestone shielded coil forms were tested in order to find the best and most practical solution for a +50dBm IIP3 input filter on 40M. Rather big, but still manageable T94-10 toriods proved to be the best solution and allowed for a BPF that easily fits an a half height euro-card sized PCB.
The T94-10 based BPF with the FSA3157/T1-6T mixer and the QT crystals hybridized roofing filter together produced a frontend on 40M with a (2.2KHz) SSB MDS of -133dBm and an IIP3 of +50.1dBm. This equals a dynamic range (IMD3DR) of 121.7dB!
The experience obtained with the work on the initial 40M BPF prototype filter board has been used to design and construct a band pass filter box with relay switched filters for all HF amateur bands plus 6M. Loaded Q has been kept below 10 to keep the insertion loss low and the IP3 as high as possible. With these BPF's, 5 HF bands obtain an IMD3DR > 120dB in SSB bandwidth (80M, 40M, 30M, 20M and 17M), and the other 5 or not far behind too. Also stop band performance is designed-in with the goal to attenuate the mixer image spur better than 100dB and the mixer IF spur better than 130dB.
The next logical step in the frontend design was to capitalize on all the work done so far and design a complete frontend euro-card sized board that contains the FSA3157 mixer and two roofing filters, . One for SSB and one for CW. The QT quartz turned out to be good enough to build a low insertion loss 500Hz CW roofing filter with excellent IMD behavior as well!
Back to Receiver Overview
Back to the TOC