[time-nuts] Tbolt temperature sensor

[Mark Sims]

My latest Thunderbolt is one from March 2005 (like Asa's).  The others are much earlier production. 

The later production units seem to handle the temperature sensor differently.  Tbolts use a DS1620 sensor chip.  This chip reports the temperature in 0.5C increments,  but has support for getting higher resolution.  The earlier production tbolts generated a smooth temperature curve that tracked temperature changes gracefully.On the later production tbolts you usually see a flat temperature plot with the temperature quantized to 0.25C/0.75C values.  

This indicates that the Tbolt firmware is doing at least part of the high resolution temperature read cycle shown in  the DS1620 data sheet.  The first part of this is to take the temperature reading (0.5C steps),  knock off the lower bit,  and subtract 0.25C.  You should then read some counter registers and do some arithmetic to generate the high res temp reading.  It appears that the later production units are not doing this (or later DS1620 chips do not support the counter registers) because the temperature plot is always quantized to 0.25C/0.75C (except for some apparent software filtering that is performed when the value changes).


One side effect of  the coarse temperature steps in the later Tbolts is that cause Lady Heather to log lots of temperature spikes.  The eralier tbolts produced nice smooth curves with an occasional temperature spike (that I now suspect is due to boundary conditions in the DS1620 count registers) that quickly faded away.  The newer units can show lots of high frequency oscillations in the temperature reading as the DS1620 readings jitter around  the 0.5 degree temperature steps.  
_________________________________________________________________
Windows Live™ Hotmail®:…more than just e-mail. 
http://windowslive.com/explore?ocid=TXT_TAGLM_WL_t2_hm_justgotbetter_explore_012009