Chip Scale Atomic Clocks

It has recently come to my attention that these devices are being manufactured!

A cursory glance at the documentation seems to indicate that they have resolution and accuracy better than 1 part in 10e10.

No other comment than just Wow!

CSAC - SA.45s - Chip Scale Atomic Clock - Symmetricom

Symmetricom is one of the major players in the timing business.

They provide timing equipment for cellphone base stations and similar systems that must maintain synchronization over a large geographic area.

These devices are all based on the GPS satellite system, which is extremely accurate. Each satellite has three Cesium standards.

A part in 10E10 is really pretty sloppy for timing. Good systems can surpass that by one to two orders of magnitude.

- Leigh

Symmetricom bought the cesium standard business from HP. I did repair and cal on HP's standards before that. The accuracy of the 5061B was 3 parts in 10^12. Never understood why we sold that business. Even the old 105B's were money makers. We usually had to replace the tubes at a price of $8500 for a rebuild. I will say that new product is pretty impressive. Beats running an ovenized crystal oscillator if the phase noise is good.

Forestgnome said:

I did repair and cal on HP's standards before that.

Click to expand...

That's interesting. You wouldn't be named Corby, would you?

Forestgnome said:

Beats running an ovenized crystal oscillator if the phase noise is good.

Click to expand...

I was running a Cesium-disciplined Rubidium to get around the phase noise problem.

I also have a Sulzer 5B, which is amazingly accurate. Considering the inner oven holds the crystal at ±1/600th of a degree C, that's not surprising.


- Leigh

Just looked it up. This new oscillator's phase noise doesn't come close to our latest ovenized oscillator. -55dbc vs. -100dbc at 1hz, -78dbc vs. -137dbc at 10hz. That thing is pretty noisy. My job is safe!

The real Leigh said:

That's interesting. You wouldn't be named Corby, would you?

Click to expand...

No, Eric. Maybe before my time? Seems to me Symmetricom acquired the business sometime around 1998-1999 if memory serves me.

The real Leigh said:

...
These devices are all based on the GPS satellite system, which is extremely accurate.
...

Click to expand...

This device is a self contained reference.

Forestgnome said:

Just looked it up. This new oscillator's phase noise doesn't come close to our latest ovenized oscillator. -55dbc vs. -100dbc at 1hz, -78dbc vs. -137dbc at 10hz. That thing is pretty noisy. My job is safe!

Click to expand...

How large is that unit. I was aware of rack contained systems provilding 13 digits accuracy.

But that this unit fits in the palm of your hand and is easily battery powered just seems amazing to me. Opens up whole new fields of application for reference timekeeping.

35 grams!

Looks like it might make a reasonable back-end to an ntp server. A lot more stable than a non-temperature-controlled quartz crystal and with a good enough rate of drift change that the software could manage it easily. It still wouldn't actually count as a Stratum 1 device though


I wonder what the cost is in one-off quantities...

Mark (been running an ntp server for 17 years now)

<jbc> said:

This device is a self contained reference.


How large is that unit. I was aware of rack contained systems provilding 13 digits accuracy.

But that this unit fits in the palm of your hand and is easily battery powered just seems amazing to me. Opens up whole new fields of application for reference timekeeping.

35 grams!

Click to expand...

I'm talking our most recent version of the 10811 oscillator. Definitely heavier and more power hungry than the Symmetricon unit, but fits in the palm of your hand. Everything has it's applications, and I'm sure this new one will have some good ones. The phase noise performance of our stuff is needed for testing and development of new 802.11 comm stuff, but not for end-use products, which is where this little oscillator will be.

As pointed out above - these are not GPS based devices - that's what need they fill. In comparison to ovenized or microcontroller comped TCXO, the long term drift is greatly improved (given its size). Cost is a little over a grand if I recall correctly. Lead time has been problematic.

As far as using it in an NTP server - not sure why you would. A GPS disciplined oscillator would provide the best long term drift performance at a fraction of the cost. If you're doing this ad hoc - you can pump a trimble thunderbolt into a linux server for probably $100.

The primary use is the need for accurate time in a GPS denied environment - jammed or subsea. The folks I work with put them in Ocean Bottom Seismometers, where the timing performance improvements over oscillators running on the order of a year is greatly improved.

As far as phase noise goes, I guess if you wanted excellent phase noise in a GPS denied environment, you could discipline a TCXO with this clock. Still not as good as a GPSDO, but better than nothing.

Brent

Phase noise is generally not a concern in a timing environment because it gets averaged out.

It becomes significant if you're trying to synchronize events over a short period of time.

Neat item.

- Leigh