[sdiy] OTA's, Iabc, and the magic of V to I

[Magnus Danielson]

From: "Nils Pipenbrinck" <np at inverse-entertainment.de>
Subject: [sdiy] OTA's, Iabc, and the magic of V to I
Date: Sun, 16 Dec 2001 14:42:29 +0100

> Hiho list!

Hej Nils!

> I'm on a personal quest to design and build my own sine-waveshaper (and get
> some more expericences with OTA's).
> 
> Therefore I fooled around with a 3080 OTA last week. One thing I found out
> was (to my surprise) that the resistance of the amplifier bias input pin
> varies widely and nonlinear with the OTA output current.

According to my datasheet shall the amplifier bias voltage vary from
about 500 mV (at 100 nA and +25 degree C) to about 700 mV (at 1 mA and
+25 degree C).

> I built a simple test-circuit to plot some curves. Just some resistors and
> the OTA itself. Iabc was provided via a 100K resistor from V+. Vdiff was
> setup with a voltage divider powered by a low impendance voltage source, and
> the output was loaded with a 10K resistor (where I measured the voltage drop
> across).
> 
> As soon as the output current went up (by changing the differential voltage)
> Iabc went down (in my setup by factor 3 from 300µA to 100µA). This gave a
> very nice atan shaped curves at the output. For my sine-waveshaper that's
> brilliant, initially planed to overdrive the differential input, not to load
> down the current source. I'm very satisfied with the resulting shape, but:
> 
> Sometimes you don't want distortion. How to go linear and provide a voltage
> controlled current to this pin? Will a current mirror do the job for me?

You may not realize it, but the pin is allready connected to a current
mirror. In the schematic diagram of the chip D1 and Q3 forms a current
mirror. Maybe another current mirror provides sufficient isolation.

Also note that the atanh curves from signal-input to output is due to
the diffrential pair (Q1 and Q2) and linearisation of that would
require additional diodes to some positive voltage and then feed with
current in order to logarithmize the input voltages. To use the input
diffamp current to vary the gain is a gross oversimplification.

> I'm curious, because I see the following problem:
> 
> Say I have a mirror built around two trannies (T1 and T2), T1 is the
> "master", collector current is generated via a resistor from CV. T2 is the
> mirror which powers the Iabc of the OTA.
> 
> When the resistance of T2's collector path changes (because the OTA does so)
> wouldn't this affect the base-emitter resistance of T1, which in turn
> changes the "master" current? If so I would have gained nothing to make the
> current to voltage converter more linear than the resistor I used in the
> first place. (it's a mirror, no current copy machine :)

A current mirror achieves two things: It changes the direction of the
current and it separates voltage differances. However, different
setups are more or less good at this tentative goal.

> Gee, I'm on thin ice here, and I'm meditating over this for quite a time
> now. Trannies make me scratch head all the time because any current and
> resistance is somewhat related. The only possible way I can think of is a
> sledgehammer method: Adding an opamp, measuring the voltage drop across a
> resistor in the Iabc current path and do some kind of feedback compensation
> to get rid of the drift.

There ougth to be a better way indeed.

> There must be a simpler way - Could someone please enlighten me?

Let me ponder some over this.

> 	Nils 'getting headaches' Pipenbrinck
> 
> Btw. I'm on a business trip for 3 days. I'm not sure if it's possible for me
> to follow the list for this time, so if you answer me, and I stay quited
> just wait a day or two.

Just don't scratch your head too much on OTAs and current mirrors
then...

Cheers,
Magnus