[time-nuts] -hp- 10811 repair (Thermistor)

[Bruce Griffiths]

Dan Rae wrote:
> Bruce Griffiths wrote:
>   
>> Meanwhile, before posting a correction I was trying to find data on the
>> oven setpoint temperature tolerance.
>>
>> The relevant HP Journal with an article on the HP10811A is:
>> http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1981-03.pdf
>>
>> One concern is:
>> How accurate does one' thermometer which one uses to calibrate the
>> thermistor bridge setpoint have to be?
>>
>> Bruce
>>
>>
>>   
>>     
> Bruce, I'm guessing that since they quote values of R20 in the manual in 
> intervals of 0.1 Degrees, that is how close they got.  Except oddly 
> enough for 83.9 and 84.0 degrees, which are the same, but shouldn't be, 
> IMHO.  I don't claim to have got that close in my thermistor tests, 
> maybe 0.5 degrees using a lab type mercury thermometer.  But my original 
> point is still valid: I now have a working oven.  In my case I will use 
> an external reference for the 5370B so it is kind of not that important 
> anyway...
>
> Dan
>
>
>   
Dan

Achieving 0.1C accuracy with a mercury in glass thermometer can be
tricky even with a calibrated one reading to 0.1C.

Some of the problems are:

1) Thermometer calibration accuracy

2) Immersion length correction - the indicated temperature depends on
the temperature distribution over the entire length of the glass column.
The thermometer may or may not have been calibrated fully immersed

3) Glass column hystereisis - correction depends on thermal history and
is smaller for some glass types than others.

4) Ensuring that the thermistor and the thermometer are at the same
temperature - perhaps the most important as internal dissipation in the
thermistor will raise its temperature above its ambient. Using a metal
block immersed in a liquid with both the thermometer and the thermistor
installed in wells within the metal block is one of the better ways.

In many ways its easier to use a themometer with an electrical output
such as a calibrated RTD, PRT or SPRT.
Commonly available RTD s are capable of maintaining calibration to much
better than 0.1C for very long periods if they aren't thermally abused.

Another possibility is t use a noise thermometer as all one needs to do
is measure the resistance and the resultant Johnson noise.
No calibration is required other than determining the equivalent noise
bandwidth of the measurement system.

Bruce