Phase differences in direct conversion receivers
 

I'm curious... with the current popularity of simple (e.g., QRP usage)
direct conversion receivers, whatever happened to the problem of having to
synchronize the cariier phases? I'm looking at Experimental Methods in RF
Design, and they just use an LC oscillator for the input to the mixer. If
input carrier is cos(f*t) and the LC oscillator is generating cos(f*t+phi),
where phi is the phase offset between them, you end up with a cos(phi) term
coming out of the mixer. If the frequencies are ever-so-slightly off, phi
essentially varies slowly and cos(phi) should slowly cause the signal to
fade in and out.

Why isn't this a problem in practice?

Thanks,
---Joel Kolstad

This is indeed what happens only if the VFO and an incoming single are
at almost the same frequency ("zero beat"). However, in practice, if
the signal is a cw signal, we listen to a signal that is 600 Hz or so
away from the VFO so that we hear the 600 Hz tone difference.

For a SSB signal, we listen to the audio content contained in the
sideband, which is 300 Hz to 3 KHz away from the VFO signal when it is
tuned in correctly.

- Dan, N7VE

Joel Kolstad wrote:

I'm curious... with the current popularity of simple (e.g., QRP usage)
direct conversion receivers, whatever happened to the problem of having to
synchronize the cariier phases? I'm looking at Experimental Methods in RF
Design, and they just use an LC oscillator for the input to the mixer. If
input carrier is cos(f*t) and the LC oscillator is generating cos(f*t+phi),
where phi is the phase offset between them, you end up with a cos(phi) term
coming out of the mixer. If the frequencies are ever-so-slightly off, phi
essentially varies slowly and cos(phi) should slowly cause the signal to
fade in and out.

Why isn't this a problem in practice?

Thanks,
---Joel Kolstad

This is indeed what happens only if the VFO and an incoming single are
at almost the same frequency ("zero beat"). However, in practice, if
the signal is a cw signal, we listen to a signal that is 600 Hz or so
away from the VFO so that we hear the 600 Hz tone difference.

For a SSB signal, we listen to the audio content contained in the
sideband, which is 300 Hz to 3 KHz away from the VFO signal when it is
tuned in correctly.

- Dan, N7VE

Joel Kolstad wrote:

I'm curious... with the current popularity of simple (e.g., QRP usage)
direct conversion receivers, whatever happened to the problem of having to
synchronize the cariier phases? I'm looking at Experimental Methods in RF
Design, and they just use an LC oscillator for the input to the mixer. If
input carrier is cos(f*t) and the LC oscillator is generating cos(f*t+phi),
where phi is the phase offset between them, you end up with a cos(phi) term
coming out of the mixer. If the frequencies are ever-so-slightly off, phi
essentially varies slowly and cos(phi) should slowly cause the signal to
fade in and out.

Why isn't this a problem in practice?

Thanks,
---Joel Kolstad

Dan Tayloe wrote:
This is indeed what happens only if the VFO and an incoming single are
at almost the same frequency ("zero beat"). However, in practice, if
the signal is a cw signal, we listen to a signal that is 600 Hz or so
away from the VFO so that we hear the 600 Hz tone difference.

....or at least, say, 595-605Hz is the local oscillator tends to drift +/-5Hz
over time, eh? Good enough.

With SSB, presumably you have the same 'problem' -- the entire voice signal
is shifted in pitch by the difference between the LO and the real carrier.
In fact, with SSB and direct conversion, how do you even decide you have the
correct LO frequency? Just when people sound 'most natural?'

Thanks,
---Joel Kolstad

Dan Tayloe wrote:
This is indeed what happens only if the VFO and an incoming single are
at almost the same frequency ("zero beat"). However, in practice, if
the signal is a cw signal, we listen to a signal that is 600 Hz or so
away from the VFO so that we hear the 600 Hz tone difference.

....or at least, say, 595-605Hz is the local oscillator tends to drift +/-5Hz
over time, eh? Good enough.

With SSB, presumably you have the same 'problem' -- the entire voice signal
is shifted in pitch by the difference between the LO and the real carrier.
In fact, with SSB and direct conversion, how do you even decide you have the
correct LO frequency? Just when people sound 'most natural?'

Thanks,
---Joel Kolstad

This is a problem in general in direct conversion schemes: and this is the
reason that "quadrature" detectors are made, with two mixing channels 90
degrees apart, so that the phasing is no longer a problem. (the sqrt of sum
of squares of the signal out of the two channels (or "magnitude") is not
sensitive to phase.)

Cliff

"Joel Kolstad" wrote in message
...
I'm curious... with the current popularity of simple (e.g., QRP usage)
direct conversion receivers, whatever happened to the problem of having to
synchronize the cariier phases? I'm looking at Experimental Methods in RF
Design, and they just use an LC oscillator for the input to the mixer.
If
input carrier is cos(f*t) and the LC oscillator is generating
cos(f*t+phi),
where phi is the phase offset between them, you end up with a cos(phi)
term
coming out of the mixer. If the frequencies are ever-so-slightly off, phi
essentially varies slowly and cos(phi) should slowly cause the signal to
fade in and out.

Why isn't this a problem in practice?

Thanks,
---Joel Kolstad

This is a problem in general in direct conversion schemes: and this is the
reason that "quadrature" detectors are made, with two mixing channels 90
degrees apart, so that the phasing is no longer a problem. (the sqrt of sum
of squares of the signal out of the two channels (or "magnitude") is not
sensitive to phase.)

Cliff

"Joel Kolstad" wrote in message
...
I'm curious... with the current popularity of simple (e.g., QRP usage)
direct conversion receivers, whatever happened to the problem of having to
synchronize the cariier phases? I'm looking at Experimental Methods in RF
Design, and they just use an LC oscillator for the input to the mixer.
If
input carrier is cos(f*t) and the LC oscillator is generating
cos(f*t+phi),
where phi is the phase offset between them, you end up with a cos(phi)
term
coming out of the mixer. If the frequencies are ever-so-slightly off, phi
essentially varies slowly and cos(phi) should slowly cause the signal to
fade in and out.

Why isn't this a problem in practice?

Thanks,
---Joel Kolstad

Cliff Curry wrote:
This is a problem in general in direct conversion schemes: and this is the
reason that "quadrature" detectors are made, with two mixing channels 90
degrees apart, so that the phasing is no longer a problem. (the sqrt of
sum of squares of the signal out of the two channels (or "magnitude") is
not sensitive to phase.)

Hmm... I went through the math, and indeed, this is the case!

But this then begs the question: Since the quadrature detector obtains the
correct magnitude of the transmitted signal for ANY phase difference between
the carrier and the LO, and if we model the phase difference as a function
of time that slowly changes due to the fact that, in actuality, our LO isn't
_quite_ the same frequency as the carrier, will the system still work? This
almost seems too good to be true...

Thanks,
---Joel Kolstad

Cliff Curry wrote:
This is a problem in general in direct conversion schemes: and this is the
reason that "quadrature" detectors are made, with two mixing channels 90
degrees apart, so that the phasing is no longer a problem. (the sqrt of
sum of squares of the signal out of the two channels (or "magnitude") is
not sensitive to phase.)

Hmm... I went through the math, and indeed, this is the case!

But this then begs the question: Since the quadrature detector obtains the
correct magnitude of the transmitted signal for ANY phase difference between
the carrier and the LO, and if we model the phase difference as a function
of time that slowly changes due to the fact that, in actuality, our LO isn't
_quite_ the same frequency as the carrier, will the system still work? This
almost seems too good to be true...

Thanks,
---Joel Kolstad

Of course, if you were demodulating DSB suppressed carrier and you
injected the carrier at the wrong phase, you indeed would get the two
sidebands going through constructive and destructive phases. If
you're 90 degrees out with your LO, it looks a lot like narrowband FM,
though very slightly different as I posted in the thread on SSB-FM.
If you do the quadrature detector thing with DSB-suppressed carrier,
then when one of the two is just the wrong phase (and you get no
output from that one), the other will be just the right phase, and
vice-versa. When it's in between, does it work out right to just sum
the two? I suppose so, though it's worth going through the math to
make sure. And of course, with quadrature mixers, you can combine the
outputs with audio phase shifting to select just one of the two
sidebands (or just CW signals on one side of the LO). In fact, the
mixer LO inputs don't have to be exactly in quadratu it's possible
to apply a calibration to account for a phase error (and also an
amplitude error, where the gain through one mixer path is slightly
different from the gain through the other). That's all practical to
do digitally...we do that sort of thing at 100 megasamples per second
with some custom chips.

Cheers,
Tom


"Joel Kolstad" wrote in message ...
Dan Tayloe wrote:
This is indeed what happens only if the VFO and an incoming single are
at almost the same frequency ("zero beat"). However, in practice, if
the signal is a cw signal, we listen to a signal that is 600 Hz or so
away from the VFO so that we hear the 600 Hz tone difference.

...or at least, say, 595-605Hz is the local oscillator tends to drift +/-5Hz
over time, eh? Good enough.

With SSB, presumably you have the same 'problem' -- the entire voice signal
is shifted in pitch by the difference between the LO and the real carrier.
In fact, with SSB and direct conversion, how do you even decide you have the
correct LO frequency? Just when people sound 'most natural?'

Thanks,
---Joel Kolstad