[sdiy] Analog Computational Unit

[Ryan Williams]

> Yes, those are fun analog computational blocks.  I have used them for
> computing ARCTAN and Square Root Sum of Squares.  However, there is one
> caveat .... the input voltages can only be positive (because they are log
> circuits).  It is possible to get around these limitation.  Also,
bandwidth
> is a function of signal level.  At the lower input voltages (lower input
> current really), the bandwidth is significantly lower.  These are just
> things you need to be aware of.  It might make an interesting way of
> processing control voltages.

right now I'm justing thinking about how this could be done, and the main
problem I'm having is the part about only positive voltages being allowed. I
guess I'll get that "non-linear handbook" that Paul mentioned.

> You can pretty easily build one of these by just using an LM3046
> transistor array for the log transistors and your favorite opamp.  You
> should be able to make one that will perform adequately for about $2....

I"m *trying* to analyze the circuit shown in the functional diagram of AD538
and I think I understand pretty much what it's doing. to lazy to do all the
math tonight (I ran out of space on the single sheet of paper I allocated
for this task) but I think I'll just build a version from a 3046 and some
opamps (like you said) and see how that works first. that is after I figure
out the part about inputing negative voltages.

I'm assuming you want the 3046 because it's important that the temperature
and matching of the transistors is close? It looks to me like if they are
perfect then the temperature effects will be canceled out?

--Ryan (apparently starting another project when I have about a million
other things I want to build)