[time-nuts] Re: Low Phase Noise 10 MHz bench signal source sought

[ghf at hoffmann-hochfrequenz.de]

Am 2022-04-08 14:06, schrieb Mike Monett:
> Dear Fans,
> 
> While I was researching cross-correlation, I came across an interesting
> paper that described different methods of reducing OCXO phase noise. 
> They
> involved putting a 10 MHz crystal between the buffer amplifier and the
> output load. One method promised "sub-thermal" phase noise. The idea is 
> the
> crystal impedance rose sharply off resonance, so the energy at those
> frequencies was heavily attenuated.

> I don't know how much additional noise the crystal ESR would introduce, 
> but
> it is an intriguing idea and might be worth pursuing.

Using a crystal that is resonant on the carrier frequency is not a very 
good idea.
At 13 dBm+ it will not enjoy a long life.

It's a different thing to notch away the sidebands, esp. at frequencies 
where
the noise really hurts, or when you want at least a peephole to see who 
is the
culprit when there are several possibilities.

The sideband power is always miniscule, even when you have 23 dBm. And 
since
only 10% or even less of your crystals are really top notch (pun 
intended)
you soon have a drawer full of "free" crystals. Somewhere I have a home-
etched board that supports 12 crystals, i.e. 6 notches, still unused.

The notch crystals can also be cheap; they do not interact with the 
carrier
frequency, all they need to have is low serial resistance at some  +- 
offset.

I've tried this with 2 crystals at 10 MHz +- 1.5 KHz on my SMIQ-04 
signal
generator. That's an easy victim, more a modulation wonder than
excellent phase noise.

With only 1 crystal per offset frequency you cannot get better than ~3 
dB.
Far off, LC filters may do.

Cheers, Gerhard

-------------- next part --------------
A non-text attachment was scrubbed...
Name: Bildschirmfotovom2020-05-1322-05-10.png
Type: image/png
Size: 37179 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20220408/c36fbe22/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smiqnotchfilter.png
Type: image/png
Size: 234112 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20220408/c36fbe22/attachment-0001.png>