[time-nuts] Stable32 Input (Multiplier)

Magnus Danielson magnus at rubidium.dyndns.org
Sat Aug 11 12:05:15 EDT 2018


On 08/11/2018 05:32 PM, Tom Van Baak wrote:
>> Should the multiplier value be +1.0000000000000e-09 or +1.0000000000000e-10?
> I worked with Chris off-list on this one. It turns out he was measuring his new PRS10 and trying to match his ADEV plots against the published plots from SRS. Hence the request about 1e-9 (which is ppb and should be correct) and 1e-10 (which is not ppb, but seems to give a much closer match of the plots). It's an interesting mystery and I'm glad he posted his question. He's new to the list, but clearly has gone well down the time nuts path already.
> He was using Heather to monitor the PRS10, extracted the "FC" column from the log file, and used Stable32 to scale and plot the ADEV. Can you spot the problem?
> The integer FC (Frequency Control) values from the PRS10 is a pair of DAC settings; the DAC used to steer the internal OCXO to the rubidium resonance. Heather scales these two numbers by 1.5e-9 and 1.5e-12 (per the manual) [1] and then scales them by 1e9 (so they logged as ppb). Stable32 (or TimeLab) then has to unscale them back by 1e-9. So far so good.
> But the actual problem is that the FC value is *not* a measurement of PRS10 performance. It's simply telemetry that shows how the DAC is operating; how much steering the OCXO needs to generate a Rb-locked 10 MHz output. In order to actually measure the output of a PRS10 one needs a rather high-res counter as well as a better-than-PRS10 frequency standard as a reference. A PRS10 can't measure itself, and Heather is a reporting tool not a measurement tool.

The FC values would be how the OCXO long term performance compares to
the rubidium, and we expect it to be worse than the rubidium, or else
the rubidium reference isn't very helpful in the long-term.

Actually, you could use the PPS lock mechanism and read out the
time-errors from that.

> I'm guessing that the reason scaling by 1e-10 looked closer to spec is that by tau 100 s the Rb is about an order of magnitude better than the OCXO. But let's find out for sure.
> Two questions for time-nuts:
> 1) Do any of you have log files of FC values from your PRS10? I'd like to compare yours with Chris's to see if they are in the same ballpark. A couple of hours or a day is all that's needed. Either raw ascii from the PRS10 itself or processed values via Heather is ok.
> 2) Have any of you ever logged the FC values while at the same time taken precision measurements of PRS10 output phase? I ask because that would allow one to sort of back out the DAC adjustments and make ADEV plots of both the OCXO and the Rb. It would then nicely show the effect of the Rb disciplining the OCXO, similar to how we make plots of GPS disciplining a OCXO. [2].

Now, that would be a worthy lab-setup to do. Sounds like a fun exercise.
I'll see if I can swing something together as I'm down there anyway.


More information about the time-nuts mailing list