[time-nuts] OT: xtal osc PN
jimlux at earthlink.net
Sun Sep 19 18:56:27 UTC 2010
Magnus Danielson wrote:
> On 09/19/2010 08:23 PM, francesco messineo wrote:
>> On 9/19/10, jimlux<jimlux at earthlink.net> wrote:
>>> francesco messineo wrote:
>>>> Hi Mike,
>>>> as I said, current plans are for a few frequencies in the 20-50 MHz
>>>> range. The current project needs 20, 22 and 42 MHz oscillators.
>>> But you're multiplying that up, it will be 20log(N) worse...
>> no, I'm using these as LO for frequency downconversion. 48, 50, 70 MHz
>> to 28 MHz in the first prototype.
> If you use a 10 MHz OCXO as reference, you will have a 20*log10(50/10)
> and 20*log10(70/10) worse phasenoise shift, i.e. 13,98 dB and 16,90 dB
> higher phasenoise than the reference.
which is why some effort at finding a high quality fundamental mode
crystal at the LO frequency might be worthwhile. If you're running into
a software receiver, then the precise LO frequency might not be
important as long as you know what it is. That lets you design an
oscillator with no tuning input, so you can use a higher Q crystal, and
you'd just worry about the temperature not changing too fast. (a TCXO is
probably a non-starter from noise properties here, the fact that it
*can* be compensated implies that the Q is low enough to be moved enough
You could ovenize it, of course.
Just for a "what can you get from state of the art", you can look at
Greg Weaver's paper
on the USOs for spaceflight..
We like to use about 75 MHz, and typical performance is -120dBc/Hz at
100 Hz offset. I don't recall if these USOs are a lower frequency
crystal multiplied up or a fundamental mode XO at the 75 MHz.
(scaling the 10811 performance.. -140dBc/Hz at 100Hz, at 10MHz, scaled
up to 75 MHz is about the same: -122dBc/Hz)
(also, the USO's real thing is long term performance for Allan deviation
and drift, not so much the "phase noise"... it's being used in a
measurement system with a <1Hz bandwidth..)
More information about the time-nuts