KT81 in feedback loop

[Martin Czech]

:::The relative error (compared with an ideal tempco) is approx. given,
:::depending on temperature and input voltage. The point of taylor
:::development is 273+35K, ie.  the error is 0 for this temperature.
:::
:::
:::
:::Temperature [K]  rel error  voltage [V]
:::---------------------------------------
:::273+35 K         0          -3 , +3
:::273+40 K         -0.03 %    -3 
:::273+40 K          0.03 %    +3 
:::273+30 K         -0.03 %    -3 
:::273+30 K          0.03 %    +3 
:::273+45 K         -0.15 %    -3
:::273+45 K          0.15 %    +3
:::273+25 K         -0.15 %    -3
:::273+25 K          0.15 %    +3
:::

If should have mentioned that without tempco the error would be 13%
in the worst case (more then 2 half tones!), compared to 0.15% with
compensation this is quite a difference .  Without compensation the tilt
direction depends on temperature, it is positive for temperatures lower
than the 273+35K, and vice versa.

I think it is a reasonable goal to have 0.1% error on 6 octaves.

With regard to this result, the other case (tempco in lower leg)
seems not to be so interesting, it could give better results,
but parameter dependency make life harder for this case.

m.c.