[sdiy] OTA diodes

[Ian Fritz]

At 11:08 PM 6/14/2014, Donald Tillman wrote:
>On Jun 14, 2014, at 12:01 PM, Justin Owen <juzowen at gmail.com> wrote:
>
> > In Electronotes EN62 pg12 (one of the pages that is publicly available 
> AFAIK...) there is a schematic of a basic OTA where the top current 
> mirror hangs off a 680R resistor with 3 diodes to GND. I remember seeing 
> this in another discrete OTA schem but never in the simplified schematics 
> of the major production OTA's.
> >
> > Recently I noticed that the simplified 2164 schematic uses a similar 
> config - but in a different way (I think...).
> >
> > What are these doing, what purpose do they serve, what are they 
> improving? I'd be particularly interested in answers from an OTA POV.
>
>Circuits that rely on matched transistors work much better when those 
>transistors are operating in identical, or near identical, electrical 
>environments.  In this case the issue is Vcb, and you'll lose the effects 
>of matching if the transistors are operating at different Vcb's.
>
>An example would be the current mirror at the top of EN-62 p12.  The 3 
>transistor Wilson current mirror is superior to the 2 transistor vanilla 
>current mirror because the third transistor assures that the Vcb of the 
>two matched transistors are within one diode drop.  So the Wilson current 
>mirror is a win.
>
>For the diff amp here, the second transistor's collector is at 0 volts as 
>it's a virtual ground.  A current mirror operating at a high voltage would 
>mean that the collector of first diff amp transistor would be way up 
>high.  So running the current mirror at a low voltage assures that the 
>diff amp collectors are within one diode drop of each other.
>
>The SSM2164 circuit works exactly the same way.
>
>--------
>
>Ironically... with the EN62 circuit running the current mirror at this low 
>voltage... the Wilson current mirror is no longer an advantage over the 
>plain vanilla current mirror.  Heh-heh.
>
>   -- Don

A few years ago I put some effort into making the best discrete OTA I 
could.  My goal was to  have performance as good as the CA3280.  It took 
some care, but it worked fine.  However, it was necessary to use a fully 
symmetric structure, with a four transistor current mirror -- using MAT 
transistors -- and a fully differential I/V converter.  Without all this, 
the offset variation with Iabc was excessive.  Keeping the C-B voltages 
small reduces problems with the Early effect.  I believe this may be what 
Don is referring to, although there may be other benefits.

But a two or three transistor mirror did not work well enough for what I 
was trying to achieve, even with the reduced voltage at the top of the mirror.

Ian