[time-nuts] Re: Characterizing a 100M TCXO

[Bob Camp]

Hi

Yes, this is picky, but there’s a point to it. The CVHD-950 is a VCXO and not 
a TCXO. TCXO implies it has some level of temperature compensation. VCXO 
simply says there’s a crystal in it. At the price point, it’s a good bet it’s uncompensated.
That means you will see some pretty significant temperature / draft sort of effects
on the output.

VCXO’s typically have fairly low loop Q and thus are not going to win many races
in the “who’s best at ADEV” category. Of all the crystal oscillator categories they
will lag behind in this category. Just how far behind depends on a bunch of things.
Swing range is one of the many. 

Best guess ADEV at one second …. 1x10^-9 to 1x10^-8 in a stable environment. 
Open the lab window while you are running the test and who knows. (Yes, I’ve 
actually run into folks who have done that while running ADEV …..).

Testing wise, a single mixer setup running against a 100 MHz OCXO would be
a cheap and quick way to go.

Bob

> On Mar 24, 2023, at 5:00 PM, Christophe Huygens via time-nuts <time-nuts at lists.febo.com> wrote:
> 
> Hi,
> 
> I am back on time-nuts after ... 15 years (kids etc...). I need some help in
> characterizing a 100MHz TCXO (Crystek CVHD-950) which we used before
> as a reference for a microwave PLL (ham project, see DUBUS 1/21). It worked
> good from a PN perspective but uses multiple steps to get to the GPS-locked
> 100MHz used as the final reference.
> 
> I am trying to see if there is merit in locking the Crystek directly and therefore
> would like to assess its short term time stability. We intend to make a bunch
> of the above and simple is better here.
> 
> I dusted off and started my modest time lab and I *think* I have sensible results
> with my setup. It may not be by the book: I am using a 5370B using a Symmetricom
> TS3100 GPS-locked 10MHz as a the reference XO (on-time > 20 years), not the
> internal 10811.
> 
> 1. I guess I am seeing TS/HP5370B noise floor when measuring Z3801A 10MHz in
> frequency mode, since the results are not as good as those on leapsecond.com.
> But still a lot better than the frequency results of 5370B on febo.com
> (which I found confusing).
> I am likely better of using the  Z3801A as a ref. for the 5370B but it
> complicates my setup, one does get smarter with age and the lab is what it is.
> I measured a NEO8M ublox 1pps using TI versus TB 1 pps  and that was as expected.
> 
> 2. How do I go about for 100MHz measurements? Is frequency ok - it probably
> will be for the Crystek? Or do I have to divide (MSI? prescaler? what s today's
> simplest approach...) and then use  "pulsepuppy" next so I can get to TI?
> Division has been asked before but a  definite answer has not come up imho.
> 
> My initial Crystek measurements certainly don t look great. 1PPS is already
> better at tau way less than 10s and that is not the performance we want.
> 
> Any hints practical hints appreciated. Theory welcome too :-)
> 
> Xtof.
> 
> 
> 
> 
> 
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe send an email to time-nuts-leave at lists.febo.com