When I read about the frequency generation in the Rb or CS there are
normally many numbers associated with the actual frequency. Down to at
least the 1 Hz level. Many of these PLLs are intended for multi-KHz steps.
I speculate you might need 2 PLLs one thats very fine in I hz increments
that gets added to something like these PLLs that step in 200 KHz
increments.
Regards
Paul
WB8TSL

On Tue, Apr 11, 2017 at 2:32 PM, Alex Pummer alex@pcscons.com wrote:

You always want two frequency sources.  One generates a carrier
frequency offset many MHz from 6.834 GHz and the other frequency
source modulates the carrier with a sideband that is at the
exact ~6.834 GHz frequency that finds the atomic line.  The
sideband is in turn modulated with audio to find the exact
center of the line.  This is now the easiest section to
design.

This architecture automatically gives you gobs of resolution.
Of course you always have the C field for infinite resolution.
The other thing it does is prevent RF leaks from exciting the
atoms, since the strong LO is safely offset from 6.834.

Rick N6RK

On 4/11/2017 2:59 PM, paul swed wrote:

Hi

If you make the cells in the basement (or even in most factories) the ability to have a wide range
synthesizer will come in handy. The whole “6.834xxx GHz” thing is dependent on a number of variables.
It is not at all uncommon to produce cells that come out 10’s or 100’s of KHz off of the “right number”.
Wide range synthesizer = higher yield.

Bob

But it is as Richard said and after he sent it I remembered reading about
it in the "Quantum beat".
The trick is to get that accurate frequencies sideband to land on the RB or
CS frequency. Just 1 sideband. So using a large step synthesizer and then
another to get the finer detail is a reasonable way to go.
Regards

On Tue, Apr 11, 2017 at 10:13 PM, Bob kb8tq kb8tq@n1k.org wrote:

An update about this thread...

It turns out that OP (Andre) would like the readings of his vintage Black Star (UK) Nova 2400[X] 8-digit 2.4 GHz bench frequency counter to be accurate and stable to the 8th digit. That's all. The internal XO or TCXO is not good enough for that last digit or two. Not to mention the cal date is 1996.

So the good news is that he is not embarking on a life-long project to build his own atomic clock. This thread is just a newbie looking for a simple answer to an accuracy / stability / calibration question. His counter has a BNC for external timebase, so ...

I think it would help him if any UK/EU time-nuts:

• have a 1e-9 or 1e-10 level OCXO to spare -- which he would have to calibrate, or
• can help him locate a surplus telecom Rb -- inherently good enough that calibration is not even required, or
• point him to a cheap newbie-friendly 10 MHz GPSDO -- if his environment would make that possible and reliable.

I would guess that a mere 8-digits of accuracy/stability should cost no more than 100 these days. If you can help, please send him email, off-list.

Again, OP is not looking for laboratory-grade accuracy or to build his own atomic clock from scratch. It's just an 8 digit counter.

/tvb

----- Original Message -----
From: "Andre" Andre@Lanoe.net
To: time-nuts@febo.com
Sent: Tuesday, April 11, 2017 12:31 AM
Subject: [time-nuts] Re. DIY atomic "resonator"

Hi

According to the data sheet the X version of the 2400 has a TCXO in it. The “not an X” has an XO. “Setability” of the XO is <0.5 ppm and
the TCXO is rated at 0.2 ppm. Maximum resolution on the device occurs with a 20 MHz input. That gives you +/- 0.05 ppm. The 200 MHz
range is scaled by 10 and the 2.4 GHz input is scaled by 128. Neither one gives higher resolution than you get at 20 MHz. A source good
to 0.05 ppm would only degrade the device by 1 LSB. A 0.01 ppm source would be adequate to ensure the final LSB is good. Any of the
normal eBay telecom Rb’s would likely be good enough forever and ever, even without calibration. I have yet to see one that was off by

Bob