[time-nuts] Different Thunderbolt versions
bruce.griffiths at xtra.co.nz
Sat Feb 21 21:22:09 UTC 2009
Markus Kern wrote:
> Hello everyone,
> I am new to the list and currently looking for a timing standard for a
> VLBI application. After some searching I have decided to use two
> thunderbolts for a first attempt and now have some questions regarding
> the different available versions.
> There seem to be two types of Thunderbolts available on eBay right now.
> One is the "flat" board mount version without PSU and which requires
> a (+12V, -12V, +5V) power supply. The other is the "boxed" version
> with integrated PSU and which requires 24V. These are all old
> Thunderbolts with 8 channel GPS. The new Thunderbolt E with 12
> channels obviously doesn't show up on eBay yet. (See  for
> My first question concerns the power supply.  shows that the output
> phase noise is significantly dependent on the PSU. How does the
> integrated PSU of the boxed version compare to the PSUs tested in ?
> Is it worth getting the boxed (somewhat more expensive on eBay) version
> just for the PSU?
> My second question is about the Thunderbolt's OCXO.  shows how to
> replace the built in OCXO with a more stable one and suggest that not
> all Thunderbolts have the same OCXOs. The bad OCXO in  appears to
> be from PIEZO and has S/N 1384. The better Thunderbolts apparently
> have "10811-class" OCXOs (like the replacement). Neither,
> unfortunately, match the labeling of the OCXOs seen on eBay (e.g. ).
> Is there an easy way to distinguish these different Thunderbolts?
>  http://www.febo.com/pipermail/time-nuts/2008-May/031370.html
>  http://www.leapsecond.com/pages/tbolt/noise.htm
>  http://www.thegleam.com/ke5fx/tbolt.htm
>  http://cgi.ebay.de/ws/eBayISAPI.dll?ViewItem&ssPageName=STRK:MEWAX:IT&item=170300146215
The Thunderbolt GPSDO is too noisy to be particularly useful for
generating the sampling clock for VLBI, unless your observation
frequency is relatively low.
To achieve useful results the carrier phase error between any pair of
VLBI sites needs to be less than 50 degrees or so over the signal
For observations at 1420MHz, and short integration times (~ 100sec) the
phase error needs to be 100ps or less.
For longer integration times (10,000 sec is desirable) and higher
frequencies the timing constraints are far more stringent.
Whilst clock modeling can be used to improve performance it can only
correct systematic effects.
More information about the time-nuts