[time-nuts] Double balanced mixer question

[ed breya]

Is the 1 Hz difference signal the one you want, or do you want the sum? 
The sum will include 10.000001 MHz but also 9.999999 MHz, and 10.000002 
and 9.999998 and 10.000000, and so forth, and a whole lot of other 
products - basically the sums and differences of every fundamental and 
harmonic of every frequency involved, at various levels (and phase). For 
investigating this, it's best to use a spectrum analyzer too.  The scope 
can see the envelope of the modulation, but not the frequency content.

The 1 Hz difference signal is easy to LPF out, but the sums have lots of 
stuff close together, nearly impossible to separate with filters. If you 
start with a quadrature mixing scheme, you can get much better rejection 
of what you don't want, but there will still be plenty of spurious 
stuff, because it can't be done perfectly - as always, it depends on how 
good it needs to be.

I'm guessing the objective is to make a 10 MHz +/- 1 Hz offset reference 
frequency. Unfortunately, it's very difficult with simple or even fancy 
up-converting analog mixing and filtering. Better to use a fundamental 
XO at that frequency, or a DDS, or start with frequencies way higher 
than the 10 MHz, that can be mixed down to the 10 MHz range, and readily 
filtered out. Most of the garbage you don't want will be much higher up, 
and a LPF can be made as good as needed to reject it.

Depending on your 5 MHz sources, a good possibility, presuming they can 
be tuned or PLLed, is instead of adding them, just set one for say 1/2 
Hz away from 5 MHz, then double it to 10 MHz +/- 1 Hz, which will be way 
easier to separate from the fundamental and higher harmonics with a BPF.

Ed