Quartz Oscillator low phase noise

[nenea dani]

Hello colleagues,
    I want to build a fixed frequency oscillator of 240Mhz and the main requirement is the smallest possible phase noise. It's a good idea to use a quartz crystal overton no. 9 on 240Mhz manufactured on order with 10ppm tolerance?

[dmills]

The accuracy has very little to do with the PN performance (Which is at least partially down to the surface cleaning of the quartz).

What oscillator topology are you considering? Whatever you use MUST not operate in saturation (to avoid AM->PM conversion), and you should carefully consider the crystal drive level.

Driscoll did some good work on low noise sources for multiplication into the microwave bands, and has at least one publication that would be worth tracking down.

There is also some good stuff in Wenzels on line technical library.

Regards, Dan. 

[Wolfgang]

... most of the solutions I know of use multipliers and a low-PN oscillator like the Wenzel ones.
I am just building one for 5MHz using an SC cut crystal in an OCXO configuration.

https://electronicprojectsforfun.wordpress.com/a-5mhz-low-phase-noise-crystal-oscillator/

[nenea dani]

    Thanks Wolfgang and Dan for attention.
 I come with some clarifications. The oscillator will have to generate 240 Mhz, followed by an  GALI-74+  amplifier, a 3dB attenuator, LPF, 3dB attenuator and a mixer 310Mhz -240Mhz = 70Mhz  for the IF. For the oscillator I want to use the BFU910FX. I do not know what configuration to be. The Wenzel documents do not explicitly state what is the best solution for low noise. There the job is interpretable and commercial. Do I understand that I'd better start from a crystal to say 24Mhz and use a Schottky multiplier? Actually I found a company that makes crystals on the desired frequency.

[Wolfgang]

...just a few thoughts about crystal choice:

- the higher the frequency, the thinner (and more fragile) the crystals get. Manufacturing tolerances get incredibly thights as frequency increases.
- All super-precision oscillators use SC cut at or below 10MHz in 3rd OT. Reasons:
  the crystals are relatively thick (and mechanically robust, plus makeable with regards to tolerances)
  3rd OT has a smaller pulling range compared to the fundamental, which is good for a reference
  phase noise is lowest

There are a lot of ultra low phase noise oscillators published, also from Ulrich L. Rohe N1UL. If you think you have found an ideal circuit and components, just try it.

Do you have an FSWP to measure the phase noise of your circuit ?

[CJay]

...just a few thoughts about crystal choice:

- the higher the frequency, the thinner (and more fragile) the crystals get. Manufacturing tolerances get incredibly thights as frequency increases.
- All super-precision oscillators use SC cut at or below 10MHz in 3rd OT. Reasons:
  the crystals are relatively thick (and mechanically robust, plus makeable with regards to tolerances)
  3rd OT has a smaller pulling range compared to the fundamental, which is good for a reference
  phase noise is lowest

There are a lot of ultra low phase noise oscillators published, also from Ulrich L. Rohe N1UL. If you think you have found an ideal circuit and components, just try it.

Do you have an FSWP to measure the phase noise of your circuit ?

Kind of freaked me out to find that Ulrich Rohde is active on the timenuts mailing list (which is a great source of information about low phase noise sources.)
 

[nenea dani]

Do you have an FSWP to measure the phase noise of your circuit ?

I have no noise analyzer. I read what the N1UL says, but it overcomes me terribly. Basically I understand that I need an  BJT/silicon with Ft as small as possible and Ic=15% Ic.max  . I really do not research.
I found a quartz of 24Mhz and another of 48 Mhz. I will try to make an oscillator and then watch with the spectrum analyzer to see what appears at 240Mhz. I also have a factory oscillator on 80Mhz. In all cases  harmonics worry me up to 240Mhz. I'm thinking about a block with  SI5351A that I'm familiar with, but I do not know how to stand with the noise  vs quartz  on 240Mhz, vs quartz on  48Mhz then x 5. In fact, this is the capital question and especially what do I choose between the two in the end if I'm not mistaken?

[Wolfgang]

... what difference does it make when you have no possibility to measure anything ? In your case I would recommend the solution with an Si chip.
Its not super brilliant, but it definitively works.

[dmills]

You can measure PN fairly easily by building TWO of your oscillators (it helps if one can be electronically pulled a bit) and just locking them in quadrature, then a simple passive mixer will give you phase noise at baseband that you can just stuff into an ADC or SA (depending on where you want to look).

Doing it directly with a notch filter, preamp and spectrum analyser is perfectly workable as long as you don't want to see nearer then a few kHz from the carrier. 

PA3AKE did some good work on this as part of his investigation into what turned out to bandgap AM noise on the AD9910 DDS chip.

Regards, Dan.

[Wolfgang]

... thats exactly the point. If you want a high precision measurements (plus near the carrier), this is the end of improvisation, unfortunately.
Measurements down to ca. 110 to 120dBc are makeable with "normal" equipment. If you want to see -180dBc, you have to
- kill *all* RFI around (otherwise you will be flooded with spurs)
- pay a lot of attention to power supplies (every little signal will be mixed up into the oscillator output)
- care for thermals
- care for components (no microphonic multilayer caps, e.g.)
- also important, mechanical stability is a must. Vibrations will also get mixed into the output at these low levels
- ... and a lot of other vodoo.

I often read some articles about amateur phase noise measurements down to -170dBc. Too good to be true, IMHO.

[dmills]

But all of that stuff is just what you need to do to build a low noise source!

Also, amplifier linearity matters (AM->PM conversion), and careful selection of crystal drive power level makes a difference, Leesons eqn. becomes very real when playing with that stuff.

Metal biscuit tins with battery packs and bulkhead BNC connectors are very much your friends for this sort of work.

-120dB sufficiently far from the carrier is **EASY** with a notch filter and a spectrum analyser, for all that seeing that close in calls for quadrature mixing method with careful attention to the control voltage supplied to the pullable osc (Noise pickup here can really screw you up).

-170dB, yea, that's going to be tough.

Regards, Dan.

[Wolfgang]

Yeah, thats about it. My Keysight N9000A CXA has a specified phase noise limit of 110dBc, so on a good day it could be 120.
Far from the carrier everything is easy, clear enough. Just notch the carrier out and off we go.
The freaks of the freaks (Holzworth) claim 190dBc. Vodoo ? Probably. 

[Gerhard_dk4xp]

<    https://www.researchgate.net/publication/260332569_How_Low_Can_They_Go_Oscillator_Phase_Noise_Model_Theoretical_Experimental_Validation_and_Phase_Noise_Measurements     >


<  https://synergymwave.com/articles/2013/04/full_article.pdf   >

regards,
Gerhard

[nenea dani]

    As I said, I'm not doing research, just building a low-noise oscillator or at least that's what I'm thinking. Suppose in the case of SA the oscillators 2, 3 or 4 must be low noise or just the first? Is it point to knock my head with my last or penultimate phase noise oscillator or focus on the first oscillator & mixer? I have heard that 2-3 pieces of the SI5351A somehow with parallel outputs would generate less noise than single one.
Generally speaking, a very wide base signal is noisier in phase than a very narrow one (something like a needle)? I understand this after the definition of phase noise or just confuse a signal with prominent side bands and another without?