[sdiy] Temperature Compensated Exponential ConverterUsingSSM2164

[Magnus Danielson]

David G. Dixon wrote:
> I think that the temperature dependence of the voltage source to the tempco
> voltage divider is basically a non-issue.  Even if it is 3300ppm/K, by the
> time the voltage is divided down (particular if it comes from a 15V rail),
> this temperature effect will be very small.  For example, if we divide 10V
> down to, say, 300mV, then the temperature coefficient is divided from
> 3300ppm/K to 100ppm/K.  This is on par with the temperature coefficient of
> the metal film resistors.

No. You still have 3300ppm/K after division. For a voltage of 10 V with 
a tempco 3300ppm/K the formula for V = 10 (1 + 3,3E-3*(T-T0)) where T is 
the temperature at hand and T0 is what ever is the base temperature for 
the base voltage.

Division only works on the base voltage V0, so dividing down (assuming 
no or canceling tempco) to 2 volts gives  V = 2 (1 + 3,3E-3*(T-T0)) 
which has changed the base voltage but not the tempco.

Besides, 3300 ppm/C is just an approximation as the actual tempco is not 
very linear.

With some care can the division rate of a pair of resistors can have 
less tempco than the resistors themselves, but since the tempco isn't a 
very stable manufacturing property and can be expect it to change, you 
can't cancel it very effectively.

Cheers,
Magnus