[sdiy] ca3280e in germany?

[jhaible]

> Serge suggested using larger input resistors connected to the  - supply
> (instead of ground). If I remember, somebody here tried that once but
> didn't see an improvement.

In fact it makes the situation *worse*.

Short outline of what has happened, from my point of view:

Serge suggested this, probably never tried this (?), at least didn't correct
it.

I read Serge's article, thought it was a great idea, and passed it on via
our mailing list. I also built it (in my Morphing Programmer), but never
really made measurements.

Joergen Bergfors made a comparison of VCAs, and noticed that the
linearisation over-compensated, i.e. what was a convex shape without
diodes, and which is nicely linearized with current sources, and still
fairly linearized with National Semiconductor's resistor network, becomes
concave with Serge's suggestion.

I measured my circuit and confirmed Joergen's findings. I also made
simulations and found that using current sinks for the lower path (or
even larger resistors, like Serge suggested), but leaving the upper
path in the 10k range results in the convex shape.

So much for numerical verification. An analytical explanation would
be nice, but I'm lazy, and slightly irritated by National's misleading
explanation on their data sheet (where they speak of approximated
current sources (;->) ), so I thought I'd contact Bob Pease about this.

His reaction was quick, but not exactly helpful (basically suggesting
I should do the maths myself.) I'm still to lazy to do this, because
I don't consider the problem *that* urgent. Any volunteers?

JH.