[time-nuts] Injection locking via common power supply

[Matt Huszagh]

To what extent is oscillator injection locking a risk when powering
multiple oscillators from the same power supply?

I measured a couple OCXOs to see how much of the output signal made its
way back into the power supply input. A Datum 1000B showed a 10 MHz
signal of -53 dBm for a +13 dBm output. A Wenzel 10 MHz ULN gave a
similar result. Curiously, the Wenzel showed much better isolation
(-74 dBm) with the output unterminated.

If we compare that to the reverse isolation of buffer/isolation
amplifiers, that seems to put the power supply isolation in the "ok"
territory, but possibly still a risk, depending on other characteristics
of the OCXO. Since the PSRR of many linear regulators is quite poor at
10 MHz (and even worse at higher typical OCXO frequencies), I would
expect using separate final regulators for each OCXO is probably only
slightly helpful. The same is probably true for the regulators internal
to the OCXOs themselves. There are exceptions to this, though. For
example, the LT3045 provides 50 dB PSRR at 10 MHz, assuming the layout
recommendations are followed.

I found Bruce's analysis page on injection locking
(http://www.ko4bb.com/~bruce/InjectionLocking.html) quite helpful in
this process. Unfortunately, I don't often know the Q factor of a given
OCXO or the amplitude of the tank circuit, so I just tried to guess.

Any thoughts on ways to minimize the risk of injection locking via power
supply connections? Am I right in thinking this is a genuine concern? I
was hoping to place a couple Datum 1000Bs in an enclosure and power them
all from the same AC/DC supply. Each would use separate LT3045 final
regulators. Given the good PSRR of that regulator, I should be ok, I
think. But I'm also curious about cases where the PSRR of the linear
regulator cannot be relied upon. I suppose ferrite beads in the form of
pi filters could be used, but ferrite beads generally aren't as
effective at 10 MHz as they are at 100 MHz. There are also DC power line
filters, sold by Schaffner and others. These can be quite effective at
10 MHz. I suppose these could be used if really needed.

Anyway, curious to hear thoughts and ideas. Thanks!

Matt