[time-nuts] Strange temperature peak

[Magnus Danielson]

On 05/28/2011 04:12 AM, Mark Sims wrote:
>
> The spikes are due to an artifact in the way the Tbolt firmware reads the temperature sensor chip.  It reads two registers and combine the values to get a high-res temp reading.  But if the firmware accesses the registers in-between the time that the temp sensor chip updates them it can produce a bogus value.  Typically this seems to happen on average of once every couple of hours.

Aha... when the temperature is falling around the LSB of the coarse 
register say that it was 10010.0000000 and flips to 10001.1111111 but 
reads the coarse register as 10010 and fine register as 1111111 it will 
incorrectly interpret it as 10010.1111111 and this looks like a jump 
upwards. There is a number of algorithms to overcome such jump, but they 
need to occur in the Tbolt.

Cheers,
Magnus