replicating FB architecture?

Hal Murray hmurray at
Mon Mar 30 06:07:44 UTC 2020

>> It depends on what you mean by "stratum 1".
> I meant the FB architecture stratum 1 = atomic clock.
> Nowadays we can apparently get chip-scale atomic clocks, which are on the
> order of an Apple laptop in cost.

 You haven't said anything about how good a clock you want.  Dig out the spec 
sheets and compare the drift rate of a chip scale clock with its high end 

If all you want is "atomic clock" so your marketing guys can brag about it, 
the chip scale is probably the best you can do.

If you want good time but don't like the cost of high end Cesium clocks, the 
telco industry uses Rubidium clocks at roughly the same cost as the chip scale 
clocks.  They are not as good as the high end Cesium clocks, but much longer 
lifetime which turns into much much lower cost per year.

The target market for the chip scale clocks is low power, light weight and 
fast warmup, not the ultimate in accuracy or low drift.  That's what you want 
for portable radios that need a good clock to sync up crypto channels.


If you are using GPS, one interesting parameter is what happens when your GPS 
stops working for whatever reason.  The usual technology is GPSDO - GPS 
disciplined oscillator.  The oscillator part is typically a good ovenized 
crystal.  While GPS is working, you adjust the crystal to have the desired 
frequency.  That corrects for crystal aging.  When GPS stops, you switch to 
"holdover" mode.  The aging on the crystal (and everything else like 
temperature) is low enough that you can repair it before it drifts out of spec 
relative to your requirements.

If you look at the pile of gear below cell phone antennas, you can often spot 
the GPS antenna.  They are typically a white plastic cone, 6 inches high.

A good atomic clock gives you much more time to repair.

If you are in the data center business, the cost of a high end atomic clock is 
not a major part of your budget.

These are my opinions.  I hate spam.

More information about the users mailing list