Voltage Divider

Google on summing junction.

Analog Device's big book, "Basic Linear Design" discusses this a fair amount... the trick is to exploit the exponential relationship of a diode (or transistor) in a feedback path of an op-amp to obtain log(a) and log(b), then compute log(a)-log(b), and finally (using the same trick) exp(log(a)-log(b)) = exp(log(a/b))=a/b.

Actually doing this accurately requires a bit of experience, I suspect -- I'm not surprised those Analog Devices ICs aren't exactly cheap. (Granted, not all applications need *that* much accuracy. What are you doing?)

---Joel

Probably is. While you're looking for it, and assuming discrete parts, how about a log amp driven by each signal, those feeding a subtraction circuit, and that followed by an anti-log amp. Four op amps and a handful of discrete parts depending on input ranges, accuracy, etc.

A nice bit on log/antilog amps:

-- Silvar Beitel

The other (and probably more common) approach is to put an analogue multiplier in the feedback loop of an op amp--the so-called "inverted multiplier". There are always lots of stability vs speed issues with log amps and dividers, and it matters a lot whether one or both voltages can go negative.

If this isn't an analogue computing application (which would be rare and interesting these days), it's frequently enough to use a diff pair in a feedback loop--you get the log ratio instead of the actual ratio, and the simple method requires that the voltages have the same polarity and that one can be made to be always bigger than the other.

Cheers,

Phil Hobbs

BTW all: I bought a book on analogue computers by one Clarence L. Johnson...anybody know if this is the same Clarence L. (Kelly) Johnson who started the Skunk Works?

Cheers

Phil Hobbs

What does that mean?

John

NatSemi was never very big in analog multipliers or dividers.

TI (formerly Burr-Brown) has the MPY634.

Analog devices has the AD633.

Those are nominally multipliers but with a feedback loop they become dividers, see the data sheets for examples. The AD633 needs an external op-amp to help out with division, but the MPY634 with its extra pins doesn't need an external op-amp. Accuracy of X/Y becomes worse and worse as Y gets smaller and smaller due to offset voltages, some manual trimming can help out.

Tim.

This is not the best of their two app notes on the subject, but it is the one I could find.

See:

Steve

If it was published in the US, you can look it up on the 'Library of Congress' website for information about the author.

--

aioe.org, Goggle Groups, and Web TV users must request to be white listed, or I will not see your messages.

If you have broadband, your ISP may have a NNTP news server included in your account:

There are two kinds of people on this earth: The crazy, and the insane. The first sign of insanity is denying that you're crazy.

Dear Mr/Ms Van Horn (guessing from your username, sorry if it's inaccurate),

Before this thread degenerates any further into the usual sci.electronics.design slag-fest, do you suppose you could enlighten us further about what you're after? And whether or not any of the replies you've seen have been of any use to you?

-- Silvar Beitel

Wow.. So much argument over terminology. FWIW I'm no "beginner". I know what I'm after is NOT a resistive divider, and I didn't want to fend off 1000 replies telling me to use two resistors. That's why I included the part illustrating that what I wanted was the division of voltage A by voltage B, and not the division of voltage A by some constant.

The end result is a circuit to do monopulse resolution enhancement of two sensors with overlapping detection fields.

The circuit needs to take two inputs, A and B, and give the result of (A+B)/abs(A-B) I had everything worked out at that point except for the analog divider.

I've found several approaches using log-antilog now, and I'm satisfied with that at this point. I didn't know what to call it when I asked, as I'd not seen such a circuit in quite some time, and looking thru H+H I also came up blank. It's hard to ask for something precisely when you don't yet know what the name of it is..

So thanks for the helpful replies, but I think I'll pass on the "attitude" replies.

(or

then

b)) =3D

Right, I don't have that one handy, but I have found some examples in NS app notes.

I'm

not

Monopulse resolution enhancement.

Think of two sensors, could be phototransistors, CdS cells, etc looking out with their detection fields overlapping. If you plot the response of either, you get a broad hump. If you sum them, you get a broader hump with a dip in the middle, but still indistinct. Difference is similarly unhelpful. But if you sum them, and divide that by the absolute value of the difference, you get a very sharp peak.

Thanks, that covers it nicely..

I wonder, WWBPD? :)

Only in the strictest sense..

I can satisfy those requirements. Thanks for the pointer.

Hope This Helps! Rich

OK. You want a true analog divide, thus the answer is related to the log+antilog solutions already posted. There are IC's that perform (X*Y)/10Z available (-10 V - 0 V - +10 V; full sphere/cube and half sphere/cube). Perhaps one of these would make your design faster/easier.

Neat part.