|
I think all three versions are functionally
similar, and the cancelling of the input sideband is something that falls out of
the maths. If you consider the 90 degree phase shift to operate only at
the freqeuncies you are interested in, then look at the signs of the two image
responses ie. IF = RF1- LO and IF = LO - RF2 then the technique
works just the same as visualising it backwards. By then reversing teh
signal flow you end up with my first solution.
After I posted the conundrum (spot my silly
spelling mistake in the subject!) we had a long session on the whiteboard at
(work with a mathematician along as well), and there is a reasonably
straightforward answer as to why the maths appears to fail. There is one
fundamental property of the sin and cos functions which have to be
remembered. Wolf nearly got there. I'll post our
solution on Monday when I get back into work as it is still up on the board and
will be easier to copy..
----- Original Message -----
Sent: Friday, April 25, 2003 11:45
PM
Subject: Re: LF: Conundrum - image
cancelling mixer
Dear Andy and LF group,
we had that very same
discussion at work about a year ago... can the multi-octave IF
quadrature network not be replaced by a simple narrowband RF phase-shifter?
No, it can't.
As you say, the first circuit with quadrature RF and LO
drives will actually cancel one of the IF output frequencies, eg. the sum
frequency, and deliver only the difference. But that is not what you normally
want, because it does not differentiate between the two possible RF input
frequencies, signal and image. With low IF, these are closely spaced and
difficult to separate by a preselector filter.
Example: Direct
conversion receiver for 14.2 MHz USB voice. Circuit Nr. 1 is not a true SSB
RX, it will convert both 14199 and 14201 kHz to 1 kHz audio. It would however
cancel the IF sums near 28.4 MHz, but this is much more easily done with even
the most simple IF lowpass. - Circuit Nr.2 is right, it will accept 14201 and
produce both 1 and 28401 from it, and it will reject any input at 14199
kHz.
For a transmitter, the same holds true. Circuit 1 would be
insensitive at 28.4 MHz "audio", but not be able to split between RF outputs
at USB and LSB.
There is a third variant with quadrature LO and IF
which will also provide a true SSB mixer. This is advantageous if you need a
large tuning range, as the LO is only a single frequency at a time, and
digital LO-divider schemes can easily provide matched 90-degree outputs. Also,
thanks to mixer saturation, amplitude balance is less critical.
73 de
Markus, DF6NM
|
