A quality GPSDO is good to 1e-13. I'm seeing 1e-10 over several days in a very unstable temperature environment from a single OXCO with a lab supply and loads of EMI.
I plan 4 OXCOs which will be filtered to the fundamentals and fed to 6 mixers. Each mixer output will be the difference frequency of a pair of OXCOs. These low frequency signals will then be fed to zero crossing detectors and counters to measure the relative frequency differences.
As needed an MCU will update the DAC outputs controlling the OXCO frequencies based on measured environmental and aging history to attempt to maintain 1e-11. A factor of 10 better than current generic Chinese PCB version.
That doesn't seem that big a stretch. We'll just have to see.
Have Fun!
Reg
On Sunday, September 10, 2023 at 05:12:34 PM CDT, David Kirkby <drkirkby@...> wrote:
On Tue, 8 Aug 2023 at 02:42, Reginald Beardsley via <pulaskite=
[email protected]> wrote:
After?some?study,?I've?concluded?that?the?main?limitation?on?the?stability?of?these?is?power?supply?noise?and?temperature?stability.??I'd?like?to?get?to?0.01?ppb/yr?just?to?see?if?I?can.??The?10x?rule?of?thumb?dictates?that?I?need?noise?and?voltage?drift?in?the?single?uV?range?which?will?not?be?easy?or?quick.
Reg
0.001 ppb/year sounds unrealistic to me - by a long way. Looking at the specifications of a high-end frequency counter from Keysight, with an optional ultra high stability oven, its 50 ppb/year
Keysight specs are pretty conservative, so we might assume they actually achieve 5 ppb/year, but what you would like to get is 5 000 times better than that.
The datasheet for the Stanford Research PRS10 rubidium oscillator gives (5 x 10^-11 / year), which is 50 times worst than you would like to get.
I know you laughed at my earlier suggestion of using an LTZ1000A, but even with that, you are not going to achieve anywhere near what you want.
