[time-nuts] Hydrogen maser spin exchange

[Ole Petter Ronningen]

Hi. Apologies for a long post.

I'm trying to read up on the "care and feeding of hydrogen masers". While
they are conceptually simple from a distance, there's quite a bit going on
in the quantum mechanics department when looked at up close. Somewhat
frustratingly, I am not mentally equipped to really grasp the finer (or
even coarser) points of that particular department. The topic of this post
is the concept of spin exchange, and it's relation to cavity (auto) tuning.
I've read papers on the subject, but I am having difficulties building a
"workable intuition", so I turn to the group.

Here's what I think I understand, and I respectfully ask for corrections if
I am way off base here..

Spin-exchange in a hydrogen maser happens when two atoms collide, and
exchange spin, as it were.. (Hazy on the details here..) The number of
spin-exchange collisions is directly proportional to the density of atoms
in the cavity. These collisions *will* happen, but is a problem in hydrogen
masers for two reasons: 1) it takes away energy from the cavity, resulting
in lower signal output power, which degrades stability, and, 2) more
significantly, it results in a frequency shift.

The frequency shift, as far as I can gather, is directly related to the
cavity resonant frequency - there is no way to *stop* spin exchange taking
place (apart from reducing the hydrogen density to a level where collisions
are rare, in which case the density will be too low for oscillation to take
place), but it is possible to reduce the impact the spin exchange has on
the output frequency.

While the resonant frequency obviously influences the output power of the
maser cavity, the "mistuning" of the cavity also increases the effect spin
exchange has. In other words, in a perfectly tuned cavity, spin exchange
does not result in a frequency shift. In a badly tuned cavity, increasing
or decreasing the hydrogen flux (thereby increasing or decreasing the
number of collisions taking place) results in a corresponding
increase/decrease of the output frequency. Since the cavity ages, and the
cavity resonant frequency follows that aging, the long term stability of
the maser is degraded unless the aging can be compensated for. Which is
what cavity auto-tuning is all about.

>From my understanding, there are a few ways to implement cavity auto-tuning:
1. From the above, it follows that a modulation of the hydrogen flux into a
mis-tuned cavity will result in a frequency shift following the modulation
frequency. Using a stable reference, this shift can be measured, and
corrections can be made to the cavity varactor voltage. Once the output
frequency no longer shifts in response to the changes in hydrogen flux, the
cavity is correctly tuned.

2. It is also possible to modulate the cavity varactor voltage. By
measuring the output power of the cavity, an error signal can be obtained
and used to correct the average varactor voltage. A square wave of i.e.
100hz, centered on the approximate correct varactor voltate is put in the
varactor, and cavity output power is measured. If the output power measured
on the "low" of the square wave is lower than the signal measured when the
"high", lower the offset by some mV, and vice versa. Suitable filtering
would of course be required.

The idea is that this method should not result in appreciable degradation
of the short/medium term stability of the maser, because the frequency of
the atoms interacting with the electromagnetic field in the maser cavity
takes time to respond to the changes in the resonant frequency, but the
output power responds "instantly". (Hazy on those details as well..) By
modulating the cavity varactor voltage (much) faster that the time constant
of the maser cavity, the modulation is effectively filtered out.

I am very interested in this method, as it seems to me that it would be
easy (feasible) to retrofit this to older masers never equipped with cavity
auto tuning.

There is at least one more way, which involves injecting a signal into the
maser cavity through a second coupling loop. At least one vendor I know of
does this in their newest design. I do not understand even the basics of
this method.

Any insights and/or corrections of my understanding is most welcome.

Thanks,
Ole