phase noise measurement

[pablohoney]

i have designed a 4.5-5.8 LC VCO. My problem is measuring(on wafer, probe station) phase noise of the oscillator. Since the VCO is free-running, the phase noise option of my spectrum analyser cant catch the exact carrier so i cant make a healthy measurement.

Another way i tried was reading directly from the spectrum analyser output (1 MHz offset from the carrier). but since the resolution bandwidth can not go down to 1 Hz, i still doubt whether my measurement is right.

What do you think the best way of measuring phase noise of a "free-running" oscillator. I need some advice about phase noise measurement.

Thanks.

 

[toonafishy]

It is very difficult to do this unless the VCO can be kept at a constant temperature. I use an external PLL application board with a very narrow loopbandwidth. National and Analog Devices make PLL boards in this frequency range for not too much money.

 

[biff44]

For free running oscillators, you need to use a delay line to measure phase noise.

 

[vfone]

**broken link removed**

 

[biff44]

Here is a paper describing various ways to make the phase noise measurement. Around page 30 it talks about delay line methods. Basically, to measure phase noise you need to downconvert your VCO signal with a local oscillator. One way to do that is to take a copy of your VCO's output, delay it in time long enough so it is no longer coherent with the present VCO output, and use it to downconvert to baseband.

You basically only need an amplifier (to boost the VCO signal in power), a power splitter, a long length of low loss cable (say 100 feet), a manual 0 to 360 degree phase shifter, and a mixer/10 Hz to 40 MHz video amp (part of the phase noise measurement stand if you have a preassembled one from HP, etc), and a low frequency spectrum analyzer. You turn the crank on the manual phase shifter until there is ~ 0 VDC coming out of the mixer, and then measure the AC part on the spectrum anlyzer. You then calibrate the spectrum analyzer output to dBc/Hz (using one of various methods) and have your answer.

Since you are using the VCO as its own local oscillator, you can make measurements even though the VCO is jumping around a lot.

Delay line frequency measurements are not that sensitive close to the carrier, and are useless at very high offsets from the carrier (because the delay line has a null in its transfer function at 1/τ)


tf.nist.gov/general/tn1337/Tn190.pdf

 

[elkimhoc]

VCO phase noise characterization
Xu, D.Q.; Branner, G.R.
Circuits and Systems, 1997. Proceedings of the 40th Midwest Symposium on
Volume 1, Issue , 3-6 Aug 1997 Page(s):619 - 622 vol.1
Digital Object Identifier 10.1109/MWSCAS.1997.666214
Summary:A measurement procedure is developed specifically to characterize the phase noise of low Q, relatively unstable free-running oscillators. The phase noise performance of a VCO is evaluated theoretically using nonlinear computer aided design (CAD) tool and experimentally using phase noise measurement system. General procedure and equipment setup is introduced. Effect of power supply on the phase noise is illustrated

 

[mriss]

Another way i tried was reading directly from the spectrum analyser output (1 MHz offset from the carrier). but since the resolution bandwidth can not go down to 1 Hz, i still doubt whether my measurement is right.

there is a easy way to read the noise level with 1Hz bandwidth by the spectrum analyser.
move the marker to the offset ,and press the "marker"->"mark noise". the spectrum analyser will automatically calculatel the noise level within 1Hz bandwidth and display it on the screen.

you can also calculate the noise level within 1Hz bandwidth by youself, read out the noise level at the offset, and subtract 10*log(RBW) ,RBW is in Hz.

 

[xjbocx]

But there is some difference between maker noise and phase noise measurement,can you please tell me why

 

[mriss]

i don't think there is much difference between mark noise and phase noise measurement. the noise is random, so sometimes there is some difference between the results of this two method. you may get different results between any two times measurement by phase noise measurement method.

If you want to make a precise measurement of the phase noise, you should use Phase Noise Measurement System to measure it, not the specturm analyser.

 

[biff44]

xjbox is correct. When you use a stand alone spectrum analyzer to measure noise, you are measuring both AM and PM noise. So you have to be careful. If you are close to the carrier (say 10 KHz offset) for an oscillator, then one can safely say that the noise measured is mostly PM, since the clipping nature of the oscillator suppresss the AM noise. However, out at say 5 MHz from the carrier, it is possible for the AM noise and PM noise to be ~equal.

If measuring other things, like residual noise in a non-compressed amp, you might guess that the AM and PM components are ~equal, and you subtract 3 dB from the spectrum analyzer to yield one of them.

A real phase noise measurment system uses a phase detector which is set up to null out AM noise.

 

[normaldist]

Another thing about what xbjocx said, he recommended that you use the noise marker function on the spectrum analyzer. This feature is not on all SA. Normally, a SA uses a peak detector. When the noise marker function is turned on, it samples the signal and gives the average. There no reliable way to go from peak to average signal that is why the function is built into some SA.

With that said, an SA with the noisemarker function can typically measure phase noise down to -90dBc @ 10KHz. The SA users manula should tell you it phase noise floor.

C