Howdy,


There are a of couple newer devices that you might like
even more than the UC3906. Check out the UC3909
and BQ2031.

All zeners are not created equal! The 1N825 is a common
temperature compensated diode and it's uncommonly good.
Transfer standards built in the 1960's and 70's often
used zeners... and they were good enough by golly. The
venerated Fluke 335D used an ion implanted or burried
zener diode and it's still one of the most stable units
available.

Here's the deal, the battery voltage varies with temperature.
About minus 3 to minus 5 milivolts per degree C. Not very
stable! (However for temperatures between 5C to 35C it can be
safely ignored.)

Now if the circuit is exposed to the same ambient temperature
as the battery and your reference drifts about the same amount
and in the same direction as the battery, it can be only a good
thing.

A true zener has a negative temperature coefficient.
An avalanche diode has a positive temperature coefficient.
The cross over point between the two effects is around 5.6V
A common silicon rectifier diode has a negative 2 milivolts
per degree C coefficient.

It's not difficult to select a couple of parts that when
combined in series will produce the minus 5 milivolts per
degree C desired. A 6.8V zener in series with three 1N4001
running 1 milliamp of bias current is in the ball park.

Honest to god temperature compensated charging requires a
temperature sensor in the battery (thermistor usually)
but it's only needed where you're charging at a rate that
heats the battery or when charging over a very wide range
of ambient.


The above is just my opinion and all that jazz...


73,
Grumpy