Frank:
It is solved much more easily than that. All you need is to divide into
bands and then analyze what those bands should be. Then the filtering
rejects the high order harmonics.
After we analyzed the QSD, we realized that the filters were not
terminated properly and changed the termination resistors. This
correctly terminated the filters and the odd harmonic response is down
so low that when a 0 dBm signal from an 8640B was put on the front end
on the harmonics of the LO, no response could be seen. So yes, you are
correct this is needed. The correct filters have been designed and how
that they have the proper operating conditions, they work beautifully.
We knew we had a winner when we saw the noise floor drop just by
changing these resistors to good values.
Bob
Frank Eory wrote:
I am just wondering if this topic was addressed. At the RF level, the
hardware uses switches, which is equivalent to multiplying by two
square waves, one for I and one for Q. A square wave has harmonics at
odd integer multiples of the fundamental, so this architecture allows
not only a desired signal at fc, but integer multiples too. Imagine
sampling an AM radio station at 1000kHz where WWV at 5MHz is getting it.
Has there been any effort at making variable center frequency bandpass
filters before the switches? I'm new at this Tayloe detector
architecture, so if it was thorougly addressed and solved, pardon this
noob and point me to some discussion posts or articles.
Frank
------------------------------------------------------------------------
_______________________________________________
FlexRadio mailing list
FlexRadio@flex-radio.biz
http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz