Hi
Yes, generally it is a good idea to make the mixer see a constant load vs frequency. So a 6 dB pad will provide a higher return loss than 3 dB will. But given that the output power from a single CLKx pin is ~11 dBm a 3 dB to 4 dB is all the headroom there is for a 7 dBm DBM.
For us radio amateurs spurious matters. To many a simple low pass filter following the Si5351A is fine ¨C ¡°Spectrum, you ask? I have added a LPF so everything is honky dory¡±. Harmonics are easy to do something about. But what about the receivers, that may be affected by bad phase noise and spurious? Karen, RA3APW, has documented his Si5351A spectrum findings here you can try to compare them with a different implementation like the RFzero: where cross-talk doesn¡¯t see to be an issue either.
So just mounting a Si5351A on a piece of PCB doesn¡¯t necessarily mean that it will be a good solution RF-wise. The Si5351A is a device designed for the digital domain where spurious is close to being irrelevant as long as it doesn¡¯t affect the ability to lock/trigger on a stable signal, so jitter, i.e. phase noise, matters.
Bo
Bo, OZ2M