AVR-Chat

>>> yahoo@barello.net 04/11/2004 1:52:09 am >>>
>Please re-read my original post on this subject.  I never claimed that
the
>tiny26 band-gap had high accuracy.  I claimed it had high stability. 
Those
>are two very different parameters.

Almost any zener, let alone one designed as a reference will have "high
stability" at a single temperature. But of course in most practical
applications single temperature stability on it's own is of no real
value.

>Recently, somewhere that I cannot seem to put my finger on...,
someone
>actually measured the drift of the tiny26 reference and found it to be
about
>1.1% or so over temperature (-40, 80c) and VCC (3-5v).  

For what sample size? 

>For room temperature
>applications with regulated supply voltage the stability would be
much
>better - probably good enough for many applications when calibrated
(i.e. a
>scale factor stored in eeprom).  In Dave VanHorn's case, even the full
temp
>range drift would be acceptable for his battery charger, had he
calibrated
>each unit.

Sure, the internal reference is useful for many practical applications,
I've used it myself, uncalibrated.

>I would like to see an Atmel data sheet characterizing the band-gap
>reference, however, Before committing any design to it.  That I cannot
seem
>to find.

In most cases with these simple internal references the only thing you
need to know is that provided in the datasheet. i.e, the worst case
tolerance over the max temperature range. In the case of the Tiny26
those figures (or lack thereof of characteristic data) tells you it's a
crap reference, so if your application can't work with that order of
error, you go for something better. As someone else said, most internal
references in micros are similar.

Dave :)