Ping J. Larkin chip scale atomic clock

That's cool. The idea of using it against roadside bombs is fanciful at best.

The cost of this is similar to rubudium clocks, but of course this is smaller and uses less power. Phase noise is mediocre, probably because the package is so small and they can't include a good OCXO inside. The physics package in these things just applies a slow discipline to a crystal oscillator, so the XO dominates phase noise.

I have the schematic of a rubidium clock around here somewhere...

John

That "chip scale" part isn't cheap, either.

-- Bill Sloman, Nijmegen

That Symmetricom clock *is* rubidium. There is a reason it is called an "atomic" clock. Rubidium does have poorer phase noise than crystal, and not just for chip scale.

The PhysicsWorld article says it's caesium.

John

...

I guess it could be (I didn't read the phywo article). When I worked for Symmetricom (not for too long, 2006), I remember it being rubidium, or so I was told. It was developed in a different branch than I worked at.

-s...

se

he

I think Symmetricom makes Rb atomic clocks too. I was wondering why Cesium in this one? (I went and downloaded the white paper from symmetricom. It says Cesium there too, but no explaination as to why.)

George H.

Probably some lucky coincidence between cs spectral lines and available vcsel lasers.

Rubidiums use a rubidium discharge lamp to make the light that the rb absorption cell uses, which solves the wavelength problem automagically. But the laser is nice and small.

The Efratom we have here seems to rely on some lucky numerical coincidences, too.

John

hip-s...

ful

s is

cause

. The

Yeah but I think you can get VCSEL at 780 or 795 line of Rb. (I think the NIST guys who published this used Rb.) The cool thing about a laser is that you can tune it right to the transistion you want. In the discharge bulb things you've got extra filters to get only one spectral line. And with a laser you don't need any microwave cavity you modulate the laser directly. It'd be fun to try and build one of these.

SRS sent us one of their broken Rb atomic clocks to take a part. (It's in my bosses hands, so I haven't been able to rip in to it.) I'd like to see how they do the RF part. It's hidden under a mu-metal shield.

George H.

Do they RF modulate the laser? That would be in the 4 GHz range.

The optical absorption is small in a conventional rubidium, like a tenth of a per cent or some such, so it takes a slow lockin loop to find the resonance.

The old monster HP atomic clocks didn't use the optical interaction. They vaporized caesium into a flat beam, hit it with RF, and found a resonance. Eventually the tubes would run out of caesium.

And they really *did* have a clock, with hands, on the front panel. We have one of them, too. Beast.

Some day I should scan and post the Efratom manual.

John

s/chip-s...

nciful

this is

because

ide. The

to a

...

lled

Yeah, that's what I read.

Hmm I assume they heat up the cell to get more absorption. I've been doing Rb type things for the past ~10 years, it's now like an old friend.

The pumping times (time from when you turn on the light till the cell reaches steady state) can be fairly long. Several milliseconds, but I don't know if that is part of the locking loop. The part of the loop that I'm least clear about is how the GHz resonance is feedback to the

The atomic beam type, I've only read the manuals. (And heard a nice talk by one of the inventors, can't recall his name.)

I'd like to know how they get ~6.7GHz from the 10 MHz.

George H.

fanciful

is

because

The

OK, I'll scan/post the thing. The microwave generation is remarkably simple, a little TTL do do some lucky math and a step-recovery diode banging a cavity, something like that. They generate a horrible mess of spectral lines, one of which just happens to resonate the rubidium.

John

tors/chip-s...

fanciful

se this is

bly because

inside. The

ine to a

ere...

called

rked

ch

why

rb

ink

of

tal

Remarkable simpe sounds promising, there'e some chance I'll understand it. Any amplification after the step recovery diode?

George H.

fanciful

this is

because

inside. The

to a

somewhere...

called

I don't think so. I'd wild-guess that they generate a microwatt or less at the desired frequency. It's an x114 SRD multiplier!

John

llators/chip-s...

is fanciful

ourse this is

obably because

XO inside. The

ipline to a

ewhere...

is called

han

worked

ranch

ing why

rom

s to

he rb

think

=A0In

vity

ne of

t I

op

We

nice

-metal

Yeah that's what the schematic shows! I assume the cavity Q is high enough that only the 114th harmonic gets big.

I'll have to take apart the SRS 'physics package' and see how the cavity is done.

There's this hugh amount of 'energy' gain in an optical pumping signal. Each RF 'photon' changes the absorption of one optical photon. With 6GHz RF that's ~10^5 or so. (lamda of 795nm, ~3.8 x10^14 Hz.) But you can do optical pumping with 1 kHz 'RF' photons too. Energy gains of >10^10! It's like knocking over a train with a ping pong ball. (If a ping pong ball weights a gram, is a train close to 10^7 kg?)

George H.

fanciful

course this is

probably because

inside. The

discipline to a

somewhere...

called

worked

Only if the medium has an atomic resonance at 1KHz, and if it interacts with the optical resonance enough to detect. The trick in a rubidium clock is the interaction.

John

scillators/chip-s...

mbs is fanciful

f course this is

probably because

OCXO inside. The

iscipline to a

somewhere...

it is called

e than

n I worked

ing

t branch

ing why

rom

s to

t the rb

think

. In

one

vity

ne of

to

en

ell

t I

op

the

ion.

nd a

el. We

t.)

-metal

um.

Oh, it's a Zeeman resonance. So you've got to add a bit of a B-field to see this.

Figure 2 in the brochure shows the signal at 100kHz.

By the time you get to 5 MHz, it gets even more complicated. (figure

Optical pumping can be made into a very sensitive magnetometer using these Zeeman signals.

George H.