10 MHz Rubidium Clock

[echen1024]

Well, I have decided to take on my "first" precision project by building a 10MHz, Rubidium controlled atomic clock. Has anyone attempted this before, and have any advice, and are there any pitfalls when building this project. I'm currently looking at EFRATOM M-100 or FRS-A series atomic standards, with a eurocard style plugin chassis.

Edit: I need a schematic for a PRECISION sine-square modifier, with a TTL type output. All designs I have seen are quite imprecise.

[codeboy2k]

Good luck, and let us know how you come along with that.

Start here:
http://www.ka7oei.com/10_MHz_Rubidium_FE-5680A.html

I find that this author has done a great job of it. I have not built one myself, but if I did, I would start at his page.
He discovered that the rubidium oscillator he was using had harmonic content that made it unsuitable for his purpose (which was as a reference for a 10Ghz microwave transverter)

So he added a crystal filter at the output get a narrow band filter to try to remove the harmonic content. Even after that, it was not narrow enough to remove the low level spurs and was still unsuitable as a reference for his microwave transverter, so he built a another circuit that uses the output of the Rubidium as a reference to phase lock a separate 10Mhz Butler oscillator; this is the output he uses which is then buffered and amplified for distribution in his lab.

Butlers aren't common these days, you are more likely too see a Colpitts or a Pierce oscillator used instead.  However, the Butler is the best choice for this purpose because it operates the crystal in its series resonant mode, instead of parallel resonant. The crystal will have a higher Q in series resonant mode.  The Butler also places the crystal in a low impedance feedback path, which helps to maintain the high Q of the crystal. Furthermore, operating in series resonant mode means it is less susceptible to stray capacitances in, on or around the board, and thus is able to maintain a higher frequency stability than either the Colpitts or the Pierce oscillators. If you go this route with your own frequency standard, then make sure you actually use a crystal that is made to operate in series resonant mode.

Dave did a video here of the FE-5680A:
http://www.eevblog.com/2012/01/14/eevblog-235-rubidium-frequency-standard

FE-5680A Teardown here:
http://www.eevblog.com/2012/01/14/eevblog-236-fe-5680a-rubidium-standard-teardown/

You can find other rubidium stuff on the EEVBLOG using a google search too, with site:eevblog.com
http://www.google.com/search?q=rubidium%20site%3Aeevblog.com

Gerry Sweeney had a blog entry about his rubidium standard here:
http://gerrysweeney.com/tag/rubidium/

Precision:  You need to specify what that means to you. Does it mean low jitter? low phase delay? What can you accept for these? This determines the design and choice of parts, and what hoops you have to go through (if at all) to bring it under control.   For example, a sine to square converter with detection accuracy to within ±100mv of 0 volts, and with a propagation delay of 50ns, on a 10Mhz, 5V p-p input wave, will introduce jitter:

  sin^-1(100mV/5V) / (10Mhz*2*pi)  = 18ps phase jitter

and add 180 degrees of phase shift.  Will that work for you? If not, you need to specify what you want and find component choices and a circuit design that will meet your needs.

So, having said that, I'd suggest a precision comparator that meets your needs for precision, as described above.  If you want minimal phase delay between input and output then you also need a fast comparator.  Only you know what you want and can accept here.

Peruse the following linear app note for some applications of a fast comparator:
A Seven-Nanosecond Comparator for Single Supply Operation, Linear Technology, Application Note 72

There is no zero crossing detector in there, but zero crossing detection with a comparator is dead simple.