[time-nuts] Tbolt temperature sensor

[Bruce Griffiths]

Jerry

No one has seriously suggested that the temperature sensor is used to
control the oven temperature.
It is likely to merely an input to the Kalman filter to allow for
compensation of the EFC DAC gain and offset drift as well as the effects
of finite OCXO oven thermal gain.
Thermal hysteresis will limit the stability of the external EFC DAC
after such correction.

The question is really is what effect (if any) does the apparent 0.5C
quantisation error have on the performance compared to the earlier
versions that have greater resolution?

If the external DAC has a gain and offset tempco of 10ppm/C and the OCXO
EFC range is 1ppm, then 0.5C temperature resolution can potentially
cause 5E-12 steps in the fractional OCXO frequency after correction for
DAC gain drift.
If the DAC experiences wide temperature swings then thermal hysteresis
may cause variations of several ppm in the DAC gain and/or offset, in
which case there may be little point in using a sensor with a resolution
of better than 0.5C.

The sensor itself may have significant thermal hysteresis.

Bruce

gsteinba52 at aol.com wrote:
> Why the doubt? 
> Couldn't this just be one input to the Kalman filter? Does the external 
> electronics package require the same temperature regulation as the OCXO?
> Why not put a cold pack on the DS chip and see if the OXCO output swings as 
> wildly as you expect (I think it would be irresponsibly poor design to control 
> the OXCO outside of the crystal component).
> Perhaps the early units had microdegree resolution for ambient air, but that 
> doesn't mean the application *requires* microdegree resolution.
>
> Jerry
>
>
>
>
>
> Message: 6
> Date: Thu, 5 Feb 2009 18:44:08 +0000
> From: Mark Sims <holrum at hotmail.com>
> Subject: [time-nuts] Tbolt temperature sensor
> To: <time-nuts at febo.com>
> Message-ID: <BLU125-W41A5B9F9F57936E392BF0ECEC00 at phx.gbl>
> Content-Type: text/plain; charset="Windows-1252"
>
>
> I doubt the temperature sensor is just used for environmental monitoring.  If it 
> was,  they would most likely have just used the basic 0.5C resolution reading 
> and have been done with it.    Instead,  they do the high res read routine and 
> then feed that into a filter to get very smooth high resolution (microdegree 
> scale) values.
>
> The whole purpose of a GPSDO with an expensive double oven OCXO is to provide 
> extremely high quality holdover performance when GPS signals go away.  To 
> achieve this level of performance one needs to compensate for the affects of 
> temperature on the parts of the system outside the oven.  0.5C resolution is not 
> up to the task of maintaining parts per trillion accuracy.   There does not 
> appear to be any other temperature sensor in the Tbolt.
>
>
>
> ------------
> Isn't it likely that a temperature sensor adjacent to the RS232 connector  is 
> just going to monitor unit temperature for environmental  purposes, perhaps, 
> for example, to give the option for flagging up an  overheating situation, in 
> which case surely 0.5C resolution is more than  adequate and the "clunkiness" 
> isn't really an issue?
>
>