how to measure phase noise of signal above 100GHz?

[beoward]

I use Agilent E4440A spec analyzer (up to 20GHz) and OML 110-170GHz harmonic mixer for the measurement.

A vco oscillating at 110GHz has been designed, but it's very difficult to accurately measure the phase noise. There are two main reasons:

1, the CW frequency is not stable, randomly vibrating in a range of 20 MHz.

2, 110-170GHz mixer is very noisy, lifting the noise floor too much, and almost emerging the signal.

Could you give me some suggestion on the measurement? Thanks a lot!

 

[biff44]

You should phase lock the vco in a narrow loop bandwidth, like 100 hz.

You may or may not be able to use the harmonic mixer, depending on the phase noise of the vco. If the vco phase noise is good, the harmonic mixer and spectrum analyzer LO phase noise will dominate. If VCO is poor, then you will be ok.

Rich

 

[BigBoss]

If CW is not stable, it's not possible to measure PN.
20MHz random fluctuation is huge!!

 

[beoward]

thank biff44 and BigBoss very much for your answer! let me think of it.

 

[biff44]

Thinking about your problem some more, I think I might form some dedicated test gear for measuring the phase noise. Assuming you have standard test components lying around, I would split the vco power, amplify one signal and drive the LO of a fundamental waveguide mixer (with a dc coupled output). I would then take the other signal and send it thru some sort of time delay (like a resonator, quasioptic resonator, long length of waveguide, etc) and then send this second signal thru a manual phase shifter and then to the RF port of the mixer.

You then could adjust the manual phase shifter so that the mixer has a 0 DC volt output at the IF port. Now you have a delay line frequency discriminator. As the vco frequency dithers, the video output at the IF port of the mixer will dither in unison. You can read the amount of peak to peak frequency deviation readily by just looking at the output on an oscilloscope You can run the IF output into a low frequency spectrum analyzer (or ADC/DSP software), do a little math, and have the phase noise measurment. If the noise is very good, you might need a 30 dB gain low noise video op amp at the IF mixer output.

Rich

 

[volker_muehlhaus]

Here's a basic setup for measurement of a single source using delay line:

http://www.thegleam.com/ke5fx/phase_noise.pdf
Pages 29-31

 

[beoward]

Thank you very much biff44 and volker_muehlhaus !

I think this is a very nice way to measure high frequency phase noise. unfortunately we have no fundamental mixer working in that frequency band.

I'm thinking of another possible way of measurement:

if the vco dithers around 110 GHz for example, I inject a 110-GHz signal into the output port of vco. The 110-GHz signal comes from multiplication of signal source, so it can stabilize the vco oscillating. then I measure the phase noise of the injection locked vco, and estimated the phase noise of free oscillating vco according to some formulas from injection lock oscillator theory.

What do you think of this way? Thanks a lot!

 

[biff44]

when you injection lock the vco to the multiplied up signal source, you do so in a very large "effective" locking bandwidth. In essence, what you end up with is the multiplied up phase noise of the lab source...which does not tell you much about the vco itself.

I guess you are going to have to do the best you can with the harmonic mixer/spectrum analyzer approach. The first thing I would do is to isolate the vco from everything else:

I would use a plastic sheet and plastic screws to connect the vco output to the harmonic mixer waveguide input...thereby isolating the DC ground of the vco from the rest of the system.

I would get a 10 turn potentiometer, a bunch of 10 uF ceramic caps, and a large battery. I would power the vco off of the battery voltage and some 10 uf caps, and derive the varactor tuning voltage from the 10 turn pot with a couple 10 uf caps across it. All the while I would make absolutely sure there is no DC ground connection outside of the vco and battery circuit. Keep all wiring short so as to not pick up any ac or rf stray noise.

I would then put it all under a cardboard box to keep any air circulation to a minimum.

Then rerun your phase noise test with the spectrum analyzer/harmonic mixer, and hope for the best. Drive the mixer RF input with the biggest RF vco voltage you can get.

Rich

 

[byteptr]

Hi,
if your oscillator is very noisy, try injection locking tecnique.

 

[beoward]

Thank you biff44!

I agree with you, buttery solution should be the best one. I think I should try it. So there is one thing cannot be isolated, it is ground of circuit and spectrum analyzer.

Thanks again!

Added after 12 minutes:

Hi,
if your oscillator is very noisy, try injection locking tecnique.

Thank you, byteptr!

Yesterday, I tried injection locking technique to stabilize my vco. The vco was stabilized, and phase noise was measured. But when I try to measure phase noise of 110GHz injection signal multiplied by 12 from Agilent signal source, I cannot measure the phase noise at 110 GHz because the harmonic mixer has too low conversion gain that the 110-GHz signal power is only 50 dB higher than the noise floor, so the phase noise at 1MHz offset is emerged in the white noise. Without the phase noise of injection signal, we cannot estimate vco's phase noise.

Do you have any suggestion? Thanks!

 

[tony_lth]

do link budget. Assume your X12=X3 X4, insert Amp between X3 and X4, focus on signal level.

 

[beoward]

Hi biff44,

After I isolate the VCO by using a piece of paper from probe station and use buttery for power supply, the VCO is very stable. Now, I can measure phase noise easily.

Thanks a lot for your suggestion!

Added after 1 minutes:

do link budget. Assume your X12=X3 X4, insert Amp between X3 and X4, focus on signal level.

Thank you, tony_lth! That's a good idea! we have amplifier on this frequency band. I will try it.