Re: temperature jumps

[brian whatcott <inet@INTELLISYS.NET>]

At 00:49 2/6/99 +0100, Mark Sylvester wrote:

> The increments are very small so that the curve looks smooth. Then it jumps
> up 4 degrees and continues to rise smoothly for a few minutes before
> jumping down again.

> Every now and then a section of the curve is displaced upwards.
> For a period
> of about 5 minutes the temperatures recorded are about 4 degrees higher,
> then they drop back down to be once again consistent with earlier
> measurements. Obviously a measurement problem. Does anyone know what
> causes it?

> Mark

Lacking insight into this arrangement I will speculate:

One might arrange a temperature probe from a diode junction, from a
thermocouple or from a thermistor.

One would wish to provide a measurement range of say -50 to plus 120 degC
and arrange the measurement to be converted in either 11 or 12 bits of an
AD converter.

Let us suppose that 11 bits are used to convert a 120 degree range:
then there are 2024 discrete positive values available , so a bit change
might represent 0.063 degC

If bit zero represets 0.063 degC
bit 1 = 0.125
bit 2 = 0.25
bit 3 = 0.5
bit 4 = 1
bit 5 = 2
and bit 6 = 4 ...
It certainly looks as though the sixth or seventh bit from the least
significant end could be sticking.

How would one test this hypothesis without a 'scope or DVM?

Perhaps a beaker of water with a regular thermometer could be set to
heat slowly over a bunsen to calibrate the probe.
 If this hypothesis had merit, one might well see a pattern of
curve displacement repeated every eight degrees up the scale towards
boiling point:
e.g
Actual  Indicated

25       29
30       30
35       39
40       44
45       45
50       54
55       55
60       64   and so on...

Brian
brian whatcott <inet@intellisys.net>
Altus OK