[time-nuts] Rb short-term noise (was RE: Is this a cesium...)

[John Miles]

> I tried converting the "negative saw-tooth" and injecting it into
> the PRS10 via the serial port, but the results were at best very
> mixed.  In the end I used the PRS10 to timestamp the 1PPS, pulled
> the measurement out via the serial port, and implemented my own
> neg-saw corrected PLL in software, which injected the correction
> back into the PRS10 via the serial port.
> 
> When you do that, you get much better results, at 2Hz and elsewhere.

If so, that's probably a different issue.  The hump in Tom's PRS10 ADEV plot
is just a (normal) artifact of the PLL used to lock the internal crystal
oscillator to the 6.8 GHz signal.   It is the same effect seen in the plots
at www.ke5fx.com/rb.htm .

Unlike most small rubidiums, the PRS10 uses a decent ovenized crystal, so it
can use a much longer disciplining time constant than usual.  That is
generally a Good Thing, because a longer time constant means that the noisy
Rb control loop does less damage to the unit's short-term stability.

Basically, the only reason you're noticing the ADEV hump on plots of the
PRS10 is because the loop time constant is long enough to show up at the
taus that are usually portrayed in an ADEV plot.  So we need to be careful
not to blame SRS for something that they are actually doing right.

If you look at the ADEV of an LPRO-101 down to t=1 ms (attached) you can see
this effect in action.   The LPRO exhibits a similar hump, except it's
present near t=0.02 seconds instead of t=2 seconds... and it's 4x as high!


The phase noise plot gives you another view of the same loop
characteristics.  The area under the LPRO's PN curve is quite a bit larger
than the area under the PRS10's.  The LPRO's relatively short loop TC causes
the noise from the physics package to appear at offsets past 100 Hz, while
the PRS10's crystal oscillator has started to dominate by 3 Hz.

JohnA also measured the ADEV of an LPRO at t0=1 ms -- see green trace at
http://febo.com/pages/oscillators/rubes/ -- and he got a similar result.
Note how much worse everything else was.  The short-term performance of the
PRS10 would have blown the others away if he had tested one as part of that
batch.

-- john, KE5FX
www.miles.io

-------------- next part --------------
A non-text attachment was scrubbed...
Name: adev.png
Type: image/png
Size: 32782 bytes
Desc: not available
URL: <http://www.febo.com/pipermail/time-nuts/attachments/20120823/310a5e39/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pn.png
Type: image/png
Size: 39583 bytes
Desc: not available
URL: <http://www.febo.com/pipermail/time-nuts/attachments/20120823/310a5e39/attachment-0003.png>