> > > hey guys,
> > > 
> > > I am currently dealing with sinusoidal signals ranging between 200k and 30MHz and have to measure a phase difference of a few millidegrees between the two signals. I have read up on a few stratergies like using a PLL or IQ demodulation, but since we are working with such high frequencies it is best to obtain advice from experts here. thanks very much.
> 

sounds similar to an instrument that  RF guys call a "RF NETWORK ANALYZER"

try those keywords and see how those instruments work

Mark

On Saturday, 20 December 2014 02:21:31 UTC, gyans...@gmail.com  wrote:
> On Wednesday, December 17, 2014 12:59:57 AM UTC+13, zoulz...@googlemail.com wrote:
> > hey guys,
> > 
> > I am currently dealing with sinusoidal signals ranging between 200k and 30MHz and have to measure a phase difference of a few millidegrees between the two signals. I have read up on a few stratergies like using a PLL or IQ demodulation, but since we are working with such high frequencies it is best to obtain advice from experts here. thanks very much.
> > 
> > Best wishes
> > Zoul
> 
> Is that phase you are after or rate of change of phase?

Thanks everyone,

gyans, its the phase difference between the excitation signal and the induced signal that we are after. The following is an open access paper on the topic we are working on and the following paragraph (from the paper) summarizes the measurement protocol 
'http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0097179'

'In magnetic inductive measurement, a signal source is used to output a certain frequency current into a transmitting coil to induce eddy currents in the measured object. The eddy currents will, in turn, produce a new magnetic field known as the perturbation field. These two fields, constituting a superimposed field, are sensed by a receiving coil, and the phase shift between the exciting field and the superimposed field is proportional to the conductivity of the object and the signal frequency. This phase shift, also called magnetic inductive phase shift (MIPS), is usually measured by phase measurement hardware.'

It has been established that the phase drift and the phase noise overtime is a major limiting factor. The system we envisage should have a comparable performance and output conductivity images at 50 frames a second. Obviously a lot depends on the inverse problem solver (used to create conductivity images from measured phase), but speeding up the phase measurement close to real time would be ideal. Hope this clears things up.

I am running simulations on the ideas provided so far, any new ones are welcome. Grotzel algorithm on a DSP chip or a FPGA is another method I am speculating on. Thanks.

Zoul

On Wednesday, December 17, 2014 12:59:57 AM UTC+13, zoulz...@googlemail.com wrote:
> hey guys,
> 
> I am currently dealing with sinusoidal signals ranging between 200k and 30MHz and have to measure a phase difference of a few millidegrees between the two signals. I have read up on a few stratergies like using a PLL or IQ demodulation, but since we are working with such high frequencies it is best to obtain advice from experts here. thanks very much.
> 
> Best wishes
> Zoul

Is that phase you are after or rate of change of phase?

thanks mnentwig, i will try getting it set up and report back. much appreciated.

>> it can be very simple
at least conceptually. Whether the accuracy can be achieved is another
question, but I don't see any obvious "showstoppers" against averaging over
a long time. 
0.001/360 would be around -110 dB at the detector output, doesn't sound
impossible to me with a single reference clock. If needed, I could use a
ballast wire to balance the reference clock delay, so that phase noise
cancels out...	 

_____________________________		
Posted through www.DSPRelated.com

Hi,

accessing the article requires a subscription. But this

>> it is possible to use the phase shifts between inductor and detector
signals for image reconstruction

seems fairly straightforward.
The achievable accuracy depends on signal-to-noise ratio, the spectrum of
the signal (high frequencies are more "valuable" than low frequencies) and
how rapidly the phase shift changes. If it is constant, accuracy can be
improved (almost) arbitrarily by increasing the measurement duration.
 
Without having read the article, my first question would be whether you can
get the inductor signal in quadrature (sine and cosine). In that case and
for narrow-band signals it can be very simple, multiply / filter /
integrate / arctan.

	 

_____________________________		
Posted through www.DSPRelated.com

thanks mnentwig. 
TV, yes apologies, it was the phase drift over time which needed to be a few millidegrees, the actual phase shift is about 15 degrees. (reference below shows the sort of system i am after, page 91 paragraph 3 reports the measured phase difference.

http://iopscience.iop.org/0967-3334/21/1/311/pdf/0967-3334_21_1_311.pdf

do you still think this is feasible? any input is welcome.

Thanks
Zoul

On 16.12.14 13:59, zoulzubazz@googlemail.com wrote:
> hey guys,
>
> I am currently dealing with sinusoidal signals ranging between 200k and 30MHz and have to measure a phase difference of a few millidegrees between the two signals. I have read up on a few stratergies like using a PLL or IQ demodulation, but since we are working with such high frequencies it is best to obtain advice from experts here. thanks very much.
>
> Best wishes
> Zoul


One degree at 30 MHz is about 100 ps in time.  This time corresponds to 
3 cm at the speed of light. (1 millidegree is 0.03 mm at speed of 
light). You are specifying a very difficult situation in the analog signals.

How do you think to get the signals into the comparison with the 
required precision?

-- 

-TV

Hi,

without trying to give an exhaustive answer: The frequency response of your
design is one important design decision. Ideally, it should match the
spectrum of your signal, including the transient response.

For an implementation, have a look here:
http://en.wikipedia.org/wiki/Goertzel_algorithm at "Phase detection"

For a quick-and-dirty solution, I'd use quadrature multiplication with
sine-/cosine, lowpass filter matched to predicted signal spectrum,
integrator, integration over an integer number of periods. Done correctly
and making just the right assumptions, this could even be optimal.	 

_____________________________		
Posted through www.DSPRelated.com

hey guys,

I am currently dealing with sinusoidal signals ranging between 200k and 30MHz and have to measure a phase difference of a few millidegrees between the two signals. I have read up on a few stratergies like using a PLL or IQ demodulation, but since we are working with such high frequencies it is best to obtain advice from experts here. thanks very much.

Best wishes
Zoul