Reading on receiver LO phase noise impact on sensitivity

[Steve Kavanagh]

Receiver local oscillator phase noise sidebands at offsets equal to the
receiver's IF frequency can mix in the receiver's mixer with the LO
carrier to produce noise in the IF which reduces the sensitivity of the
receiver. This can be particularly critical for receivers using low IF
or zero-IF (direct conversion architectures). Can anyone direct me to
a good discussion of this particular effect ? Books, papers or web
links would all be of interest, as are both theoretical treatments and
practical experience.

Steve

 

[rick H]

Receiver local oscillator phase noise sidebands at offsets equal to the
receiver's IF frequency can mix in the receiver's mixer with the LO
carrier to produce noise in the IF which reduces the sensitivity of the
receiver. This can be particularly critical for receivers using low IF
or zero-IF (direct conversion architectures). Can anyone direct me to
a good discussion of this particular effect ? Books, papers or web
links would all be of interest, as are both theoretical treatments and
practical experience.

There are several mechanisms by which LO sideband noise can degrade the
receiver performance. In this case, it sounds like you're describing an
effect known as receiver self-quieting.

Remember that the difference between the wanted RF signal and the Local
Oscillator signal is the IF frequency (if it's not, then your receiver
isn't tuned in properly!). That means that noise on one of the LO's
sidebands at plus-or-minus the IF frequency will be at the same
frequency as the wanted RF signal. If that noise component leaks from
the LO port of the mixer to the RF port of the mixer, then that noise is
co-channel with (i.e. on top of) the signal you wish to receive. This
won't be a problem if the signal you wish to receive is much stronger
than the leaked noise, but if it's not, then the receiver's performance
will be degraded by the fact that the LO's sideband noise has been
plonked squarely on top of the signal you wish to receive. The noise at
one-IF-away-from-the-LO will mix with the "proper" LO to produce noise
at the IF, thereby degrading degrading the signal-to-noise ratio, and
degrading the sensitivity of the receiver.

The reason why it's potentially problematic at low-IF or zero-IF
is that the sideband noise of the LO is much higher close-in to the LO
frequency that it is far away. If your IF is several MHz, then the
sideband noise several MHz away from the intended LO frequency is
usually very small, so the noise leakage at the frequency of interest to
the RF port is small. If, however, you've opted for an IF of a few 100
kHz or even 0 Hz (zero-IF), then the noise on the LO one-IF-frequency
away from the LO frequency is much higher, and hence the leaked noise
is much higher, hence higher noise co-channel with the RF signal you're
trying to detect.

 

[Steve Kavanagh]

Remember that the difference between the wanted RF signal and the Local
Oscillator signal is the IF frequency (if it's not, then your receiver
isn't tuned in properly!). That means that noise on one of the LO's
sidebands at plus-or-minus the IF frequency will be at the same
frequency as the wanted RF signal. If that noise component leaks from
the LO port of the mixer to the RF port of the mixer, then that noise is
co-channel with (i.e. on top of) the signal you wish to receive.

Yes, that is the interference mechanism I am interested in. And
described in a similar way to the way I tend to think about it. That
thinking led me to wonder if the specified mixer LO-to-RF isolation
would be a guide guide to the level of the self-interference, but a
former colleague indicated that it was more complex than that (but did
not elaborate). Any comments ?

Steve

 

[rick H]

Yes, that is the interference mechanism I am interested in. And
described in a similar way to the way I tend to think about it. That
thinking led me to wonder if the specified mixer LO-to-RF isolation
would be a guide guide to the level of the self-interference, but a
former colleague indicated that it was more complex than that (but did
not elaborate). Any comments ?

Well, the LO-RF isolation _will_ provide a measure of how much the LO's
noise sidebands at +/- IF will mix with the "wanted" LO to produce extra
noise at the IF port at the IF frequency.
When your ex-colleague said "it's more complex than that", he could have
meant one of a few things:

1. Signals other than LO sideband noise, generated elsewhere on the pcb,
ca also couple in to the receiver front-end and desensitize it in the
same way as the sideband noise. Digital clock harmonics and the like can
do this, but they often remain at the same frequency, and so they
deafen the receiver only on particular channels. Radio hams call these
birdies.

2. There are a zillion other mechanisms for degrading the performance of
a receiver (IP3 for in-band blockers, IP2 for zero-IF receivers,
reciprocal-mixing for adjacent channel, 1/2 IF, group-delay, etc.)

3. There's the other "more complex than that" whereby errant receivers
still don't behave the way they should, despite you thinking that you've
accounted for all the mechanisms for receiver degradation. This is when
you start bluffing to your boss that receiver design is "An Art"!

 

[Phil Hobbs]

Well, the LO-RF isolation _will_ provide a measure of how much the LO's
noise sidebands at +/- IF will mix with the "wanted" LO to produce extra
noise at the IF port at the IF frequency.

LO-RF isolation is all about how well balanced the mixer is. But LO
sidebands mixing with incoming RF to generate crap at the IF doesn't
altogether depend on the balance. There are two mechanisms that will
cause this problem even with a perfectly balanced mixer:

1. Close-in LO phase noise. The instantaneous phase of the _desired_
IF signal is the sum or difference of the instantaneous phases of the LO
and RF signals. A mixer has no memory, so there's nothing there to
smooth out the LO phase variations--they come through unattenuated.

2. LO amplitude noise. Real mixers don't switch exactly at the zero
crossings of the LO, but at some offset, which depends somewhat on the
RF signal amplitude. This causes a variation in switching duty cycle
with LO amplitude, which translates into IF amplitude noise. There is
also typically a variation in the on resistance of the switches (e.g.
Schottky diodes) with LO drive, which does much the same thing. Driving
the mixer with lots of LO power minimizes both of these, but they never
go away altogether. Using square wave LO drive can be a big help here.

Cheers,

Phil Hobbs

 

[vasile]

LO-RF isolation is all about how well balanced the mixer is. But LO
sidebands mixing with incoming RF to generate crap at the IF doesn't
altogether depend on the balance. There are two mechanisms that will
cause this problem even with a perfectly balanced mixer:

1. Close-in LO phase noise. The instantaneous phase of the _desired_
IF signal is the sum or difference of the instantaneous phases of the LO
and RF signals. A mixer has no memory, so there's nothing there to
smooth out the LO phase variations--they come through unattenuated.

2. LO amplitude noise. Real mixers don't switch exactly at the zero
crossings of the LO, but at some offset, which depends somewhat on the
RF signal amplitude. This causes a variation in switching duty cycle
with LO amplitude, which translates into IF amplitude noise. There is
also typically a variation in the on resistance of the switches (e.g.
Schottky diodes) with LO drive, which does much the same thing. Driving
the mixer with lots of LO power minimizes both of these, but they never
go away altogether.

Using square wave LO drive can be a big help here.

This is arguable as long the mixer is mixing a sinusoidal signal with a
square one.
Phase differences between those two will create a hilarious shape
signal at the mixer output.
Also the huge level harmonics of the square wave used for the LO drive
will produce
a bunch of (unwanted) images in the RF side which need to be supressed.
So then, where is the big help please ? I've seen this ideea published
somewhere on the net but didn't understood it.

thx,
Vasile

 

[Phil Hobbs]

vasile wrote:

This is arguable as long the mixer is mixing a sinusoidal signal with a
square one.
Phase differences between those two will create a hilarious shape
signal at the mixer output.
Also the huge level harmonics of the square wave used for the LO drive
will produce
a bunch of (unwanted) images in the RF side which need to be supressed.
So then, where is the big help please ? I've seen this ideea published
somewhere on the net but didn't understood it.

That's actually not true--the phase detector output is sinusoidal if the
RF input is.

If you look at the spur specs of your average mixer (Mini Circuits,
bless them, publish gobs of this sort of detail), you'll find that the
suppression of LO harmonics is extremely poor anyway, so square wave
drive doesn't cost you much there.

The big help is that the linearity of the mixer is greatly improved.
Third-order intermodulation of different components of the IF is
suppressed, for example--and since they often wind up inside the IF
passband, this really matters.

Cheers,

Phil Hobbs

 

[Steve Kavanagh]

1. Close-in LO phase noise. The instantaneous phase of the _desired_
IF signal is the sum or difference of the instantaneous phases of the LO
and RF signals. <snip>

Fundamentally this is a separate issue. In the applications I have in
mind there is no trouble generating a clean enough LO so that the phase
noise on the desired IF signal is not a problem for demodulation.

But it does get me thinking that the balance of the mixer may be upset
a bit by the fact that the mixer switching behaviour will not have a
50% duty cycle in the short term, because of the noise on the LO. I
suppose this might provide some additional LO-to-RF leakage but I will
have to think a bit harder to understand it than I have so far !

2. LO amplitude noise. <snip> Using square wave LO drive can be a big help here.

Yes, certainly any limiting scheme to eliminate amplitude noise should
reduce the effect by up to 3 dB, I suppose (assuming, as a rough
approximation, equal phase noise and amplitude noise powers on the LO).
However, I am thinking at the moment primarily of vhf through
microwave LO frequencies where square wave drive will probably be hard
to come by. Besides oscillators, mixers and frequency multipliers all
normally have some limiting action built in, so the improvement by
using square wave drive is likely less than the 3 dB maximum, unless
there are quite noisy amplifiers in the LO chain.

And Rick...my former colleague's "more complicated" remark (this was a
few years ago in another context than my current one) was specifically
addressed to this particular effect...numerous other receiver
degradations were also considered at that time.

73,
Steve

 

[vasile]

vasile wrote:




That's actually not true--the phase detector output is sinusoidal if the
RF input is.

Are you vowing that's true ?


Because I've believed the same and that cost me a second PCB
revision.
I can tell you for sure that on some mixers I've used (like TDA6190)
this is just theory.

Maybe we are talking about different things like almost always is
happening on these emails...


greetings,
Vasile