Phase detection of RF carriers

You

is

Whatever the idea is, it is just plumb eluding me.

?-/

rs

of

In nature, there are three common reasons why the phase of a signal might vary. One is if it's an AC signal at a certain frequency, in which case the frequency is defined as the rate of change of phase over time. Another is because the distance between the source and receiver is changing, stretching or compressing the wavefronts. That changes the frequency, in a familiar phenomenon called the Doppler effect. Finally, the observed phase of the signal will be altered if more than one copy of the signal is received via reflection. This could happen in an incredibly wide variety of situations, from a kinked coax cable to a UFO flying overhead.

'George' already knows the frequency of the two signals, so he can (and will want to) remove that source of time-dependent phase shift. He can do that trivially by mixing them down to DC. Countless ways to do that exist, from a DDC core on an FPGA to a Gunnplexer in a police radar gun. Regardless of how he downconverts to baseband, any other influences on the signals' phase will be left intact. Neglecting transmitter or receiver phase noise which he presumably doesn't care about, those influences will arise from the other two causes.

Even with only one remotely-transmitted signal, you can infer all sorts of interesting things about environmental reflection and/or Doppler shift by looking at what happens to the phase of the signal. This is how bistatic radar works. Things get very interesting (in a potentially ITAR-related sense) if you can look at the phase from several transmitters at once.

-- john, KE5FX

No, for two different frequencies the 'phase shift' is a linear ramp. If you insist on 0-to-360-degrees measure, that becomes a ramp-like sawtooth function (the number of full cycles of phase difference being an arbitrary constant).

I was thinking the original poster had some idea of fitting that strong linear function against time to a model with perturbations from the linear, and analyzing the perturbation part as a signal.

or

OK that concept i can follow, (delta t)modulo 2*pi versus sine(delta t).

I can even make sense of that. Thanks.

?-)

No.

You just do a quadrature mix of the two signals and get a quadrature sine wave take atan2 of that for the phase difference and you get a ramp that just keeps going....

One way to think of it is to start with two signals that are very close in frequency. when you measure phase difference you get result that shows a slow drift which eventially hits 360 degrees (which looks just like 0 degrees) and keeps on going and going.

I couldn't figure out what the OP was trying to do either. The details must have been in the text out beyond where I'd expect a line wrap. :-/

So, what does that mean when averaged over 100 mS? (I cheated and scrolled to the right a few miles)

That's where the curve fitting comes in. You do the averaging after extracting the phase slope. It's done in interferometry all the time--it doesn't make any sense to average sin and cos, and almost none to try averaging the wrapped (-pi

nonsynchronous or

perturbations

scrolled

Said another way, i think, we can get a clean (average or mean) phase velocity (slope) and second (and higher) order moments of the phase velocity.

??-)

sine

=20

Though it took a while, i finally get that one reasonably well. Now i need to figure out what else OP wants to measure.

??-)

Sure, that's one way to do it. Reconstructing from moments isn't that easy in general, due to the horrible non-orthogonality of the basis set consisting of powers of t. (The normal equations for finding a polynomial representation from the moments give rise to the Hilbert matrix, which is famously ill-conditioned.)

In this case, we'd have the raw data and the fit functions, so provided that there's a tight enough bound on df/dt, the unwrapped phase vs frequency can be reconstructed using something like a spline fit.

Cheers

Phil Hobbs

=20

Glad you asked ... :o)

All I want is to measure the two signals' time difference for the purpose o= f determining their distance displacement. I know their precise frequencie= s. =20

I'm not clear on how to handle phase ambiguity when the two signals are at = different RF frequencies plus displaced in time relative to each other by a= large amount (multiple wavelengths).

Analyzing them at their RF frequencies leads to ambiguous results because t= heir RF periods can be small relative to the distance displacement between = them.=20

So if I do direct conversion of the signals down to DC, I get it that their= displacement in time appears as phase difference that is linearly increasi= ng forever. But that can go to an unlimited value, so a t=3D0 point has to= be defined. But the fuzziness of that point itself can constrain measurem= ent accuracy.

Or, I can down-convert the signals to a low enough frequency (instead of DC= ) that there will be no phase ambiguity since the IF period is guaranteed t= o be long enough. Then I can use a zero-crossing as a t=3D0 mark. I'm thi= nking this is the way to go.

Am I on the right road?

George

determining their distance displacement. I know their precise frequencies.

different RF frequencies plus displaced in time relative to each other by a large amount (multiple wavelengths).

their RF periods can be small relative to the distance displacement between them.

displacement in time appears as phase difference that is linearly increasing forever. But that can go to an unlimited value, so a t=0 point has to be defined. But the fuzziness of that point itself can constrain measurement accuracy.

that there will be no phase ambiguity since the IF period is guaranteed to be long enough. Then I can use a zero-crossing as a t=0 mark. I'm thinking this is the way to go.

The phase ambiguity is inherent unless you have some way of setting (or measuring) their relative phases at some time t=0. Just knowing their frequencies isn't enough for your measurement.

Downconversion doesn't help, frequency division doesn't help. Unless you can set or measure the relative phase at some known time epoch, you haven't got a measurement.

Cheers

Phil Hobbs

sine

that

ks

se of determining their distance displacement. I know their precise freque= ncies.

at different RF frequencies plus displaced in time relative to each other = by a large amount (multiple wavelengths).

se their RF periods can be small relative to the distance displacement betw= een them.

heir displacement in time appears as phase difference that is linearly incr= easing forever. But that can go to an unlimited value, so a t=3D0 point ha= s to be defined. But the fuzziness of that point itself can constrain meas= urement accuracy.

f DC) that there will be no phase ambiguity since the IF period is guarante= ed to be long enough. Then I can use a zero-crossing as a t=3D0 mark. I'm= thinking this is the way to go.

Agreed. I'm thinking that time epoch can be a zero-crossing if I can selec= t the specific zero-crossing unambiguously, which should be possible with a= low enough IF.

George

sine

of determining their distance displacement. I know their precise frequencies.

different RF frequencies plus displaced in time relative to each other by a large amount (multiple wavelengths).

their RF periods can be small relative to the distance displacement between them.

their displacement in time appears as phase difference that is linearly increasing forever. But that can go to an unlimited value, so a t=0 point has to be defined. But the fuzziness of that point itself can constrain measurement accuracy.

DC) that there will be no phase ambiguity since the IF period is guaranteed to be long enough. Then I can use a zero-crossing as a t=0 mark. I'm thinking this is the way to go.

the specific zero-crossing unambiguously, which should be possible with a low enough IF.

Unless I'm mistaking your situation, that wouldn't help either. The frequencies are low and accurately known, and the SNR is high. Frequency conversion will add exactly nothing to your knowledge of the phase datum.

If you can walk up to each of the transmitters and do a measurement at a known distance from its antenna, you can figure this out. Alternatively, if you can reset the phase of the TX oscillators at some known time, that would work too.

Otherwise you have one completely unconstrained parameter in your measurement.

Cheers

Phil Hobbs

Glad you asked ... :o)

All I want is to measure the two signals' time difference for the purpose of determining their distance displacement. I know their precise frequencies.

I'm not clear on how to handle phase ambiguity when the two signals are at different RF frequencies plus displaced in time relative to each other by a large amount (multiple wavelengths).

Analyzing them at their RF frequencies leads to ambiguous results because their RF periods can be small relative to the distance displacement between them.

So if I do direct conversion of the signals down to DC, I get it that their displacement in time appears as phase difference that is linearly increasing forever. But that can go to an unlimited value, so a t=0 point has to be defined. But the fuzziness of that point itself can constrain measurement accuracy.

Or, I can down-convert the signals to a low enough frequency (instead of DC) that there will be no phase ambiguity since the IF period is guaranteed to be long enough. Then I can use a zero-crossing as a t=0 mark. I'm thinking this is the way to go.

Am I on the right road?

George ________________________________________________________________

What would you get if you put each signal into divide by N counters to get a common frequency? Then put the two common signals into an xor to get their phase relationship.

tm

nonsynchronous or

ramp.

perturbations

scrolled

none

We are still talking different languages; i was thinking rather ordinary statistics like variance, skew, and kurtosis. Not even anything as interesting as Gaussian quadrature.

Sure, but i have no idea if that is where OP is going with this.