Low cost phase noise measurements. How to
Not having access to expensive equipment, I would like to find a method to measure the phase noise of an HF oscillator.
I have seen two procedures
http://wenzel.com/documents/measuringphasenoise.htm
http://www.noding.com/la8ak/m23.htm
The second one seems more attractive as it does not require a pll, but I am not confident how it works.
I have a tektronix 491 SA and a 100MHz Oscilloscope available.
Any other thoughts/projects/documents?
Firstly you should have an idea of the phase noise order of magnitude you want to measure. Obviously, the linked measurement methods are limited by the phase noise of the involved reference oscillator and other system components. (Basically the same problem as if measuring phase noise with a spectrum analyzer).
I am building a very low phase noise crystal oscillator, similar to the wenzel one. I am going to test it on a professional phase noise measurement machine once, then write down the measurements to have it as a reference.
In the notch filter method link, I cannot understand how this works. Is it clear to you?
Maybe the width of the notch at the top of the filter curve is used to give relative measurements?
Or maybe the notch is used to deeply attenuate the center of the carrier and leaving only the edges (noise) pass through to the analyzer so they can be measured?
The first circuit is mixing down the test signal to zero IF, the second to 10 MHz. The down converted carrier is suppressed to increase measurement dynamic.
All right, so a similar pattern like the one shown in the second schematic should be seen on the analyzer.
I think one could avoid mixing and the need for a reference oscillator, if he is interested only in one frequency (10MHz in the current schematic). What do you think?
Or even better one could mix the signal to be measured with a low phase noise crystal oscillator say 3KHz apart and produce an audio signal ready to be analyzed by audio DSP techniques on the PC?
What is the dynamic range of your PC's audio input?
I would probably suggest you make a frequency discriminator using one of the crystal resonators. The output of that discriminator would drive a low noise video amp, and be displayed on a low frequency spectrum analyzer. The discriminator center frequency would, of course, have to coincide with the crystal oscillator frequency.
You can do a slightly better noise floor, especially closer to the carrier, if you did two crystal oscillators of identical design, and phase locked one to the other, and then downconverted to zero frequency, as shown in the wenzel paper
As is mentioned in the Wenzel link, a crystal oscillator can exhibits very low phase noise even using common components.
Discriminator method is fine, but degrades the measurement sensitivity close to the carrier.
Perhaps the best method for very low phase noise measurements (with the widest dynamic range) is using a clean reference oscillator and a phase detector.
This method have also the advantage removing any AM noise, but is complicated in terms of calibration because the gain (and losses) of all stages are part of the final results.
I have no idea, it is an ibm x60s, you say the dynamic range would not be enough to measure such things?
Yes, by many orders of magnitude.
The A/D chip in your computer's sound card is specified with 90dB dynamic range.
http://www.hardwaresecrets.com/datasheets/AD1981HD.pdf
Amendment: it seems that high quality external soundcards might be an option:
http://www.sm5bsz.com/linuxdsp/hware/soundeval.htm
As I said, I have no access to expensive equipment other than a tektronix 491 SA and a 100MHz Oscilloscope. Do you think the second link, (notch filter method) could give satisfactory relative measurements for the 10KHz and up (20KHz notch carrier) measurements, or isn't worth trying it at all?
I am afraid it's REALLY difficult. Measuring the phase noise of low noise crystal oscillators is not trivial even if you have the latest and greatest phase noise measurement instruments.
What is the Frequency of the signal you are trying to test, and how close in do you
want to measure?
If I were you I would see about renting a E5052A. That is what big Companies
use to test their products .
It is for 1-30mhz. The closest to the carrier, using homebrew methods, the better. I think the notch filter presented here is only suitable for 10KHz apart from the carrier?
I wish I had access to one of these, but imagine a homemade poor man's lab here :D