[time-nuts] Temperature stability for Thunderbolt: results

[Mark Sims]

At one time I took some data on a 5370A with a FTS4060 cesium as the reference.  Those numbers tracked the self-reported numbers.  I can't find the raw data anymore.

Warren also pointed out that with the temperature stabilized you can increase the filter time constant 2-3 times over what a non-stabilized unit can run at.  This makes an even bigger improvement in performance.  I usually keep my filter set at 500 seconds.  I need to see what 1000+ seconds will do.

The Tbolt uses a DS1620 temp sensor in its high-resolution readout mode that gives around 0.01C raw resolution.  The Tbolt filters this and spits out micro-degree resolution numbers.  The absolute temperature accuracy is under 1 degree F,  but for thermal management we are really only interested in resolution,  not absolute values.

I did check the consistency of the thermal environment in my box with a micro-degree resolution sensor and it was quite good.  The difference in stability between where the D1620 sensor is and the oscillator was negligible.


----------------
Do you have any measurements of the actual 1PPS output
with and without temperature stabilization? The LH numbers
being quoted recently are internal loop measurements rather
than real 10 MHz or 1PPS BNC outputs of the TBolt, yes?