[time-nuts] Thunderbolt Monitor (Didier Juges)

[Didier Juges]

> -----Original Message-----
> From: time-nuts-bounces at febo.com 
> [mailto:time-nuts-bounces at febo.com] On Behalf Of Bruce Griffiths
> Sent: Saturday, September 13, 2008 10:48 PM
> To: Discussion of precise time and frequency measurement
> Subject: Re: [time-nuts] Thunderbolt Monitor (Didier Juges)
> 
> >      7.  Provide the ability to read several temperature devices.
> >      Something as simple as the DALLAS One-wire devices as 
> >      probes would work.
>
> The capability of using higher accuracy, higher resolution 
> temperature sensors such as RTDs and thermistors may also be useful.

Regarding temperature sensing,

Using easily obtained NTC thermistors and the built-in 12 bit ADC of a
Silabs processor, I was able to obtain stable temperature readings with
about 0.02 degree C resolution at 25 degree without any digital filtering or
analog processing (other than decoupling capacitors), by directly feeding
the voltage from a voltage divider composed of the thermistor and a
precision 10k resistor to the ADC, and using the processor's built-in
voltage reference to drive the divider. A small amount of dithering (or
simply by averaging several samples and using noise as the dither) will
easily give 0.01 degree C resolution if you like to see all the digit values
used. I did not bother to calculate the actual precision, but a rough check
shows the main absolute error to be the thermistor itself, all other errors
can be made negligible compared to it by using a precision resistor and
clean layout. The basic unadjusted accuracy is a fraction of a degree,
sufficient for most applications.

This with a thermistor that costs about $1.00 at Digikey and is as big as a
1/8 Watt resistor...

The software does the voltage/temperature conversion using math functions
(log particularly), so make sure your compiler supports it. Alternately, you
could use lookup tables, but that becomes impractical for anything greater
than 8 bits of resolution.

Due to the non-linearity of the circuit and that of the NTC thermistor
itself, resolution of that particular circuit degrades away from 25 degree C
to about 0.08 degree C at -35 and at +85 degree C. Different choices of
thermistor/resistor values can optimize the resolution at the temperature of
interest, but make sure you take the dissipation in the thermistor itself
into consideration, as it will create a non-linear offset. Any type of
analog linearization (in front of the ADC) over a broader range of
temperatures will cost a lot of money because you will need an exponentially
greater number of precision components. If greater resolution is needed, a
16 bit ADC would be the best solution in terms of cost and hassle.

Didier KO4BB