Measuring phase noise without a spectrum analyzer

[luky315]

Is it possible to measure the single sideband phase noise of an oscillator (8MHz) without a spectrum analyzer? Or at least compare various oscillators to see which has the lowest phase noise (no absolute value required)?.
The specified (typical) value for the phase noise (100 kHz, 1Hz BW) is -107dBc and for 1MHz -123dBc.
All ideas are appreciated.

[PA4TIM]

Even with a SA this is not easy. My Tek goes deep enough but not at 100 kHz or 1 MHz bandwidth. If I measure phase noise of a oscillator I use  300 Hz  RBW to get around -125dBm.  You can use a very good receiver, a very steep narrow filter and a trms meter like you make noisefloor measurement:  http://www.pa4tim.nl/?p=252 it is in Dutch but there are pictures (and a translate button) My racal was better as -154dBm/Hz a FT101e measured -166dBm/Hz



This is a Tek 2712 spectrum analyser

You can get old selective voltmeters rather sheep. I have a HP312 bought for 30 euro, almost DC to 17MHz , these are a sort of non-sweeping spectrum analysers http://www.pa4tim.nl/?p=680

[G0HZU]

The specified (typical) value for the phase noise (100 kHz, 1Hz BW) is -107dBc and for 1MHz -123dBc.

At 8MHz that phase noise is very high at those offsets! A decent spectrum analyser can measure that noise spec very easily.

However, you asked for a cheap and cheerful alternative and I would suggest the following:

Make a 'decent' 8MHz oscillator that gives about 5mW output and use this as the (ultra clean) local oscillator to a regular double balanced mixer. eg SBL-1 from Minicircuits or even a home made mixer.

You could use a decent commercial signal generator as an alternative for the clean LO but even with a modern lab grade $$$ sig gen the phase noise won't be anywhere as good as a decent homemade 8MHz oscillator at 100kHz and 1MHz offsets. But a decent sig gen should still be easily clean enough for your very 'noisy' requirements above.

Then feed your original oscillator under test into the RF port of the mixer after attenuating it down to be about -5dBm. Make sure the difference in frequency between the two oscillators is initially about 10kHz.

Then feed the IF port of the mixer into a PC soundcard via a 6dB attenuator and view the response on some freebie spectrum analyser software for a PC soundcard. You should be able to see the noise on the signal and this will be dominated by your noisy test signal at 8MHz.

Obviously, if your signals drift apart by more than the frequency span of your soundcard then you won't see the main signal appear on the PC spectrum analyser display. But you can then tune one of the oscillators away by 100kHz or 1MHz and just look at the noise at the relevant offset.

[luky315]

Is there some sort of circuit that is able to generate a DC voltage proportional to the noise? I am not really interested in the exact value for the 100kHz / 1MHz phase noise. An output that is somehow correlatet to the phase noise will do.

[G0HZU]

If you don't want to use the soundcard method then you could still mix down to an IF of maybe 455kHz and feed this into a log amp via some amplification and a narrow filter at 455kHz. eg something about 1 or 2 kHz wide. The log amp would give a DC voltage with something like 25mV/dB slope.

The log amp isn't the ideal part to measure noise but it would be able to crudely measure the noise although you would ideally have to calibrate the system.

The soundcard method is far better though...

I'm still surprised at how poor your target phase noise is? i.e. -107dBc/Hz at 100kHz offset. For an 8MHz oscillator it should be easy to get close to -170dBc/Hz at 100kHz offset even with a simple design. What type of oscillator is it? (apart from the 'very noisy' type)

[luky315]

The oscillator is ultra low power, experimental and made of discrete components. And it needs improvment.
What type of filter can I use? I have to block the "carrier" (100Khz distance from the wanted signal) SAW is not suitable / obtainable in small quantities in this frequency range (or is it?).

[KJDS]

There are many methods of measuring phase noise, however just how low is this low power oscillator?

Depending on cricumstances, a frequency discriminator, a PLL or an FLL can be used effectively.

[JackOfVA]

It's also possible to use the oscillator under test as its own reference for PN measurements with a delay line discriminator.  The concept is that phase noise is not highly correlated so you compare the current oscillator output with a delayed copy of the oscillator output. The difference is proportional to the PN.

There are some practical problems with this, as one might expect, but if one is looking for a minimalist approach and if the oscillator has relatively poor PN performance, a delay line discriminator may work.

There's a useful overview of PN measurement techniques at
http://www.home.agilent.com/upload/cmc_upload/All/PhaseNoise_webcast_19Jul12.pdf?&cc=US&lc=eng