[sdiy] discrete SSM2018?

[Scott Bernardi]

Well, I used to be a bipolar designer a long time ago.....
Actually, I'm getting a bit puzzled by the loose absolute gm spec also.
Loose specs usually mean its a parameter the manufacturer doesn't want
to test for, and therefore can't guarantee.  It would be a simple enough
thing to test for gm at a fixed current, since it is set externally.
The temperature is another factor: specs usually say "25 deg C", but the
actually temperature is whatever the building is at.  And since
temperature has such a direct effect on the gm....

One thing I can answer, is about the structure of the bipolar
transistors in a planar process.  I drew some crude pictures:


http://members.home.net/sbernardi/img/bipolar_process.html



jhaible at t-online.de wrote:

> Czech Martin schrieb:
> > > Tracking of gm between two OTAs of one chip: typ: 0.3dB
> > >
> > > So we have a variation of 2:1 of absolute gm, but a
> > typical gm
> > > tracking of only 3.5%.
> > >
> >
> > This  is the beauty of integration. ;-> Absolute values
> > are (almost)
> > impossible,
> > but matching or tracking is very good. I can see no
> > inconsistance here.
>
> The inconsistance is, like we discussed before, that apparently
> the matching of transistors within one function block (mirror)
> is worse than the matching between two blocks.
>
> You don't need exact absolute values for *any* of this:
> diff amp, one mirror, several mirrors. It's always the
> matching / tracking.
>
> The only possibility I can imagine at the moment (other than
> too loose specs in the data sheet): Maybe they can produce
> mirrors with a block where transistors share some areas
> (emitter for instance), and they have difficulties to get a
> tight current ratio within such a block, but no difficulties
> to reproduce the same block with tight matching to the first
> one. Does this make sense ? No bipolar chip designers on the
> list ??
>
> JH.

--
Scott Bernardi
sbernardi at home.net