In article , "Joel Kolstad"
writes:

With all this discussion of phasing fun... could someone answer the
following question for me?

Say I'm transmitting binaural audio, with I being L and Q being R. I
receive this signal and generate my own I' and Q' outputs. However, if the
RF carrier and my LO have a phase difference, the entire IQ (phasor) diagram
is rotated by that difference and, e.g., a 90 degree difference will result
in the left and right channels I receive being swapped.

How do IQ-binaural receivers recover a phase lock to present this?

You are using an example of separation of conventional AM
sidebands. Phase synchronization to the carrier can be done
separately or by using parts of the multi-mixer I-Q circuitry.
Phase synchronization is not absolutely necessary for listening
and still hearing separate sidebands.

Relative phase in mixers is NOT disturbed. (basic fact)

Single-sideband phasing systems use at least two mixers, the
LO of one in quadrature (90 degrees) phase with the other LO.
Since the LO frequency is the same, the two mixers' output will
have a relative phase difference of 90 degrees.

In addition to that, the mixer outputs are put through an audio-
frequency-range wideband phasing network. The Gingell 4-phase
network is ideal for this (it works fine with just two phase inputs).
With the Yoshida value optimization, the Gingell network can be
made with excellent constant-relative-phase-quadrature over a
broad audio frequency range without using precision tolerance parts.

The "trick" now is to linearly combine two of the four audio phases
such that the TOTAL relative phase shift is 0 degrees (or very nearly
so). With the LO having a relative phase differential of 90 degrees
the audio output of the mixers will have a differential phase of 90
degrees. Since the audio polyphase network provides additional
90 degrees relative phase difference, the total is 180 degrees...or 0
degrees if an inverting unity gain amplifier is used.

So, what happens if the LO isn't "locked" to the incoming carrier?
Actually, very little. If the 2 LOs remain in quadrature phase realation-
ship, the two mixer output relative phase relationships are STILL in
quadrature. The only thing that has changed is the slight frequency
difference in the mixer outputs relative to the original modulation
frequency. This has no effect on any broadband audio phasing network
following the mixer outputs...those maintain the additional quadrature
relative phase and linear addition and subtraction will be the same.
Unwanted sideband AND carrier suppression in demodulation will be
essentially unaffected.

In going back through messages after a short absence, I detect some
worry about a slow "beat" effect if the LO isn't synchronized. That's not
a real worry if you've gone through the full expansion of the basic AM
equation and shifted the whole series in phase by 90 degrees, then did
a linear comparison with the same series unshifted in phase, then taking
the TWO audio frequency components from the series and did a linear
addition or subtraction with an additional 90 degree relative difference.

Synchronization-to-the-carrier-frequency-and-phase is necessary only
with conventional AM and the audio circuitry being broadband all the
way down to DC. There are several ways to make the DC component
represented by the carrier mixed down to baseband either disappear or
reduce greatly in value. Hint: Using the full series expansion, use a
small phase shift error and note the comparisons of all series
components, including the carrier.

With SSB, there's no real worry since the transmitting end carriers are
already reduced reduced in amplitude. If the LO isn't quite in sync or
even not on-frequency, all that will be noticed is the slight change in
demodulated audio frequency relative to original frequency. The
amount of rejection of RF in the unwanted side of the carrier will
vary by the error of exact LO quadrature and the error of the phasing
network quadrature. On conventional AM, those errors are the same
as the isolation between sideband modulation content. If the AM has
left ear content on lower sideband and right ear content on upper, that
isolation is the same as "left-ear v. right-ear separation of stereo."


Have any suggestions for a nice simple mixer (ala the NE602) that retains
both the I and Q signals at the output?

"A" mixer, no. You must use at least two to get an In-phase and
Quadrature output. The Tayloe detector has several advantages.
First, the CMOS switch can be driven with 4 phases, not just 2, and
the equivalent conversion transconductance is far higher than any
passive mixer; mixer noise is also reduced relative to active mixers
due to the nature of the CMOS switch structure. With four phases
in the output, all at 90 degree multiples, it fits the Gingell-Yoshida
polyphase network just dandy such that quadrature errors from
network components are greatly diminished.

The original Gingell polyphase network as described by Peter
Martinez* in RSGB's Radio Communication magazine in 1973 had
only 0 and 180 degree audio phase differences at the input. The
network outputs were still at 90 degree multiples over a wide audio
bandwidth. That bandwidth will increase and with less error when
inputs are already at four phases of relative quadrature.

The only disadvantage of the Tayloe mixer is the need to use a 4x
frequency master LO if the four phases are derived digitally for
broad tuning range.

*G3PLX, the same that inovated PSK31 some years later.

Len Anderson
retired (from regular hours) electronic engineer person