[sdiy] OTA nonlinearities

[Magnus Danielson]

From: "Nils Pipenbrinck" <np at inverse-entertainment.de>
Subject: [sdiy] OTA nonlinearities
Date: Mon, 7 May 2001 18:23:25 +0200

> Hi there.

Hi Nils,

> Just wondering where the non-linearities of the CA3080 OTA are.. I mean.. as
> long as I keep the current on the current-input pin constant, can I expect
> the amplification of the OTA in "hifi"-quality, or does it distort the
> signal?
> 
> Or put it different: Under which circumstances will the OTA work nonlinar?

The OTA core is the input transistor pair (bases to pin 2 and 3) with
a common emitter current current mirror sourced by pin 5. The
collectors of the input transistor pairs is sent through two equalent
current mirrors where one is hooked to the output and the other is sent
through another current mirror before hitting the output.
So, a CA3080 is nothing else than 4 current mirrors and one input
transistor pair.

Now, the input transistor pair will operate just as the transistors in
an Moog-ladder, that is, the exponential/logarithmic curve. In effect
the endresult will be an arctan function, which has the behaviour of
approximating linear as the signal diffrence on the input
reduces. There's a separate graph showing the actual curve. Naturally
will other imperfections also make themselfs known, like those of the
4 current mirrors.

There are means to compensate for this arctan function by sending the
insignal through voltage-to-current converters and then pull current
through diodes hooked to some voltage. The midpoint of current source
and diode is hooked to the input transistor bases. These diodes then
act as "linearizing" diodes by effectively use the logarithmic
function. This is a wellknown trick for those that have digged into
how Gilbert multipliers work.

I've never actually tried to do the compensation, but it should be
possible, but thermic connections should be tight.

> And now back to work...

Why? ;)

Cheers,
Magnus