multiplier: was Re: 3280 and AD533 availability, EN frequency shifter

[Grant Richter]

I recently designed and I/Q detector that ran at 300Khz
for the front end of a SPARC DSP. I had to evaluate all
the available analog multipliers for cost vs. performance.
Here's my 2 cents worth.

> RC4200 (exp/log, expensive, and one quadrant, can be extended to 4)

What I finally used for the project due to it's low cost and
excellent DC temperature stability. Linearity is 0.1% (best of all types)
and cost in 1000 pcs. was $0.63 (63 cents). DC drift is
0.005% per Deg C. Manufactured by Fairchild/Raytheon and
second sourced by JRC. You simply bias the one quadrant into
four to handle +/- 10 volt inputs on X and Y. Though
this one takes the most outside support circuitry.
The application engineer from Analog Devices told me
"they really made a coup there, nobody can beat that chip
on price/performance".

> MC1495 (linear inputs, why isn't this available in Europe??)

Very old school, high distortion. PAIA 47xx module used this.

> MC1496 (nonlinear input for modulator, ie. expects square modulator)
> AD633  (crap, if you ask me)

My second choice with 1% distortions, but very few support components.
Jameco has them for $6.25 each. Expensive, thats why I picked the
RC4200.

> I have tried AD633 and found it too noisy.

Use the full 20V p-p input range.

>
> I've bought 4200, but expect problems due to the unsymmetry of the
> circuit. It is basically one quadrant.  We'll see.

All the sum of logs circuits have the same problem.
That includes National LF0094 and Burr-Brown LOG100 (?)
general purpose math chips. All use the sum/difference of logs
trick to do math which limits you to one quadrant.