VCO hacks, tweaks

[Magnus Danielson]

>>>>> "JV" == Joachim Verghese <jocke at netcontrol.fi> writes:

 >> I'm glad you caught the 3300ppm/K vs. 3500ppm/K problem, I missed that one
 >> completely.

 JV> I fear this issue is even more complicated if you opt for really
 JV> accurate compensation.

 JV> The problem is that the temp coefficient of the scale factor is a
 JV> non-linear function of temperature. 3300ppm/K is accurate only around
 JV> 30 deg(C) (303K), since

 JV> (303 + 0.5) / (303 - 0.5)  =  1.003306

 JV> At 40 deg(C) (313K), which perhaps is closer to the operating temperature
 JV> of an enlosed circuit, the required compensation coefficient is

 JV> (313 + 0.5) / (313 - 0.5)  =  1.003200, i.e. 3200ppm/K

 JV> In other words, using a (linear) tempco resistor isn't the most accurate
 JV> compensation method, although it works well enough for most purposes.

I agree with Joachim, I came up with similar results when I dug into
the issue a long time ago. The linear tempco is a quite rude hack
after all and some study and thought might need to go into a good
design. However, many of the corrections will be sufficient enought to
within a range of a few tens of degrees be reduced. If one wants to
learn something about temperature compensatio, go and read about the
bandgap reference cursuits, these bandgap people have spent quite a
lot of time on large range temperature compensations. They talk about
temperature gradients on the chip (it is assumed to be integrated into
the same chip) etc as well as many other things. Bob Pease has some
articles and papers on the web on this.

Cheers,
Magnus