snipped-for-privacy@b36g2000pri.googlegroups.com...
y y=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 !
=A0 =A0 =A0 =A0 =A0 =A0 =A0!
+---- tk,
oo
If you do enough knees in the curve, it would be within a few percent of the smooth curve. Unfortunately, I haven't found many smooth curve circuits that don't involve matched diodes or transistors etc.
Some ideas:
(1)
If the user doesn't mind a thing that takes a long time to settle, he could do something like this:
RT1 -------+----------/\/\------+------Vout ! ! \ \ R1 / .................../ RT2 \ \ ! ! GND GND
R1 heats RT2 making it a fixed number of degrees above RT1 for a given input voltage.
Since thermistors follow an exp() curve, a fixed temperature difference will make a fixed resistor ratio.
(2) Using something like a lm339 you can make the input into a variation in pulse width or frequency. Consider this circuit:
Vcc ! Vcc [R1] ! Vin-------!+\ ! [R2] ! >--+-+------------!+\ ! -!-/ ! ! >---+-+-- To low pass ! =3D=3D=3DC1 Vcc/2--!-/ ! ! ! ! ! GND ! -------------------------------+ ! =3D=3D=3DC2 ! GND
C1 and R1 are just enough to ensure that the second stage completely discharges C2.
The voltage on C2 is a bunch of Ramps up to Vin. For small inputs, the gain is
1/2. When the Vin starts to become a significant fraction of the Vcc, the gain increases. It is too early in the morning to do the math. !