[sdiy] Temperature stable lin-exp converter with a CA3086or CA3046

[Ian Fritz]

At 04:50 AM 5/28/2002, Neil Johnson wrote:

> > All free-electron-like metals have the correct tempco.
>
>Indeed, but don't forget that this also depends on the purity of the
>metal, and can vary over a wide range.  For example, standard Pt100 RTDs
>are specified (DIN) as having a temp.co of 3850 ppm/'C, but higher purity
>Pt (eg. 99.999% pure) can go as high as 3923ppm/'C (from info on the web).

Sure. To clarify, pure metals have a linear T dependence above the Debye 
temperature. But this is not linear-through-the-origin -- there is an 
offset due to the T^5 dependence at low temperature. But a small series 
resistance will buck the offset to give the correct behavior at high 
temperature (and over hundreds of degrees). I believe the impurity effects 
you refer to mainly come into play at low temperature.


>I guess the old Q81s were wire-wound copper because its easier to make
>that vapourizing platinum onto a ceramic substrate??  And standard copper
>has a temp.co of 3930ppm/'C.
>
>Indeed, in Tim Orr's Transcendent 2000 design he describes making a copper
>temp.co from handwound copper around a high-R resistor substrate, then
>bulking out the temp.co to the desired value with a low-value metal-film
>resistor (from memory something like 820R copper, 180R metal-film).
>
>Certainly for op-amps, inductors in the feedback path is bad news (unless
>you specifically design for it) as it can lead to loop instability, such
>as overshoot.

Unless the resistor is wound non-inductively. I can't imagine why this 
would not be done. Has anyone actually seen this inductive effect? I've had 
no problems with the old Q81's, but it is certainly something to look at 
when ordering a tempco.


>I tend to use platinum surface-mount temp.cos, as (a) they're quite cheap
>and available, and (b) easier to glue to the side of tranny-pairs.  As
>they are not wire-wound I could put them in either of the two places
>(feedback or divider) but convenience suggests the divider, directly
>across the base-emitter pair on one of the transistors (with maybe a
>padding resistor to adjust the temp.coeff).  Even though, as Rene pointed
>out, its not *quite* as perfect is being in the feedback path :-)

Platinum is excellent, but a bit pricey. I figured out the correct series 
resistance to add for pure Pt at one time -- as I recall it was quite small.

><flame>
>Heck, if you want "perfect", go build a digital oscillator!!
></flame>
>
>Confuscious say: "A bank of analogue oscillators should be like bamboo --
>pure, simple, drifting in the breeze"

Sure. But I got tired of retuning every 20 minutes during my recording 
sessions. If you want beating it can be added.

   Ian