Re: How to measure LC oscillator phase noise.

From: Don Pearce (donald_at_pearce.uk.com)
Date: 06/15/04 

Date: Tue, 15 Jun 2004 10:57:43 +0100

On Tue, 15 Jun 2004 19:35:06 +1000, Allan Herriman
<allan.herriman.hates.spam@ctam.com.au.invalid> wrote:

>On 15 Jun 2004 01:30:30 -0700, ablight@alphalink.com.au (Andrew
>VK3BFA) wrote:
>
>>I am in the process of building a 20m SSB transceiver using a LC VFO -
>>how can you measure phase noise of the device without a roomfull of
>>complex expensive test equipment - is this possible with simple
>>techniques even if only in a relative good/bad way? - I have a decent
>>CRO and a tempermental HP141 which works if thumped in the right
>>place. I am not an engineer, and the only references I have here are
>>in the ARRL hanbook for 2002 - they talk about pahes noise with
>>respect to DDS synthesisers but give no practical help. Or am I asking
>>an impossible question - can it be done by "good" oscillator design
>>and it flows from this - the groups comments appreciated.
>
>
>You need both good design and good measurements.
>There are many simple design mistakes that can have an adverse affect
>on phase noise, and it's easy to make those mistakes.
>
>
>The simplest way to measure phase noise is with a spectrum analyser.
>They might be good down to about -90dBc/Hz to -100dBc/Hz or so,
>depending on offset freq.
>Better than that requires a phase noise test set.
>
>I'm not sure how good the HP141 is, but I suspect it will have a noise
>floor higher than your oscillator under test at certain offset
>frequencies.
>
>
>
>How to measure the phase noise floor of your spectrum analyser:
>
>1. Get a clean source, e.g. a crystal oscillator. No, you can't use
>that old sig gen you have, it's probably noisier than the spec an.
>
>2. Plug source into spec an. Adjust so the peak is in the middle of
>the display.
>
>3. Set span (or start and stop frequency limits on the display) to
>about 3 times the offset frequency at which you are trying to measure
>the noise. (E.g. if the spec is so many dBc/Hz @ 10kHz, set the span
>to 50kHz.)
>
>4. Reduce the resolution bandwidth to minimum. This will reduce the
>video bandwidth and increase the sweep time, if they are coupled. If
>not coupled, make the appropriate video bandwidth and sweep time
>adjustment manually.
>
>5. Adjust the reference level so that the peak is at the top of the
>display.
>
>6. Wait for a sweep (slow).
>
>7. Use the cursors (in delta mode if available) to measure the
>difference in level between the peak and a point at the offset
>frequency (e.g. 10kHz) away from the peak.
>
>8. If your instrument has a noise mode for its cursors, use that,
>otherwise, normalise the level at the cursor to a 1Hz bandwidth. This
>can be done (roughly) by subtracting 10.log10(rbw) dB from the level
>at the cursor, where rbw is the noise bandwidth of the resolution
>bandwidth filter. (If you don't know the noise bandwidth, just use
>the resolution bandwidth.)
>
>(Sorry, I don't think the '141 has cursors. You'll have to try to
>read levels against the graticule.)
>
>10. Take this value away from the reference level, and you have your
>phase noise at a 10kHz offset.
>
>
>E.g. 100MHz +10dBm signal. Cursor measures -75dBm at 10kHz away from
>the carrier. Resolution bandwidth is 10Hz.
>Phase noise is (+10) - (-75 -10.log(10)) = -95 dBc/Hz @ 10kHz
>
>Regards,
>Allan.

This is all very well, but unfortunately the noise floor of the
analyser is unlikely to be the limiting factor here. The phase noise
displayed on the analyser is the sum of the phase noises of the device
under test and all the various local oscillators in the spectrum
analyser. Analyser local oscillators (the first anyway) are designed
for broad pulling range, not good phase noise. They tend to be
mutually exclusive.

A free-running oscillator at 30MHz or so should be capable of very
good phase noise at the odd few kHz offset - far better than that of
the analyser oscillator, so the measurement is not likely to be very
convincing.

The best way to get the measurement is to build a crystal oscillator
within the range of the VCO, and phase lock them together with a very
narrow loop (a few Hz maximum). Feed both outputs into the LO and RF
ports of a mixer, and examine the output. It will be baseband noise
with the characteristics of the noise of the oscillator. You can
examine this very nicely with a sound card on your PC. A low noise
op-amp will make a good pre-amplifier to get a decent sized signal.

d
Pearce Consulting
http://www.pearce.uk.com

Relevant Pages

  • Re: How to measure LC oscillator phase noise.
    ... on phase noise, and it's easy to make those mistakes. ... depending on offset freq. ... Reduce the resolution bandwidth to minimum. ... If your instrument has a noise mode for its cursors, use that, ...
    (sci.electronics.design)
  • Re: Undersampling: Specifying Clock Jitter?
    ... I'm basically trying to design a power spectrum analyser that can ... The ADC has a full power bandwidth of 1 GHz and has 12 bits. ... I think that for my purpose, I should specify the phase noise from 100 ...
    (comp.dsp)
  • Re: Undersampling: Specifying Clock Jitter?
    ... The ADC has a full power bandwidth of 1 GHz and has 12 bits. ... another signal that is 500 kHz away. ... I think that for my purpose, I should specify the phase noise from 100 ...
    (comp.dsp)
  • Re: A question for Mike Monett.....
    ... > Vdoes not clip and this is good for the phase noise. ... oscillator is badly overdriven, as most are. ... tank without affecting the phase. ... to force it to the phase of the incoming data. ...
    (sci.electronics.design)
  • Re: How to measure LC oscillator phase noise.
    ... You allow for the mixer loss ... Assuming the crystal oscillator is much better than the LC oscillator, ... the phase noise of the oscillator will not be attenuated outside the PLL ... With the 8553 plugin for the 141 you could get crude measurements, ...
    (sci.electronics.design)