Analog Correlator Circuit.

It isn't just the sign of the top comparator. My first thought was that the

To make it work for both positive and negative, I think you need to double it and use an AND-OR function. Alternatively you could use 1 circuit (with an AND) and reference the dividers to -V.

There are too many adjustments as well.

Then there's the issue JL raised of both inputs at 0V, which is solved with the dividers referenced to -V. Another solution would require, I think, that instead of resistive dividers you use an adjustable offset, like the TL431, in which case it should be possible to have 1 adjustment.

_ /| / A ------+-/\/\/--+--/\/\/-+ | / | _|_ | | - | | |\ | `-------------|-\ __ | | >-----| \ .---------------------------|+/ |AND)---. | | |/ .--|__/ | | | | | | | | | | | |\ | | | `----------------------|+\ | | | _ | >--' | | /| .------------------|-/ | | / | |/ | B -+-/\/\/--+--/\/\/-+ | / _|_ | - | | __ `-----\ \ )OR>--- out .-----/__/ _ | /| | / | A ------+-/\/\/--+--/\/\/-+ | | / | _|_ | | | - | | | |\ | | `-------------|+\ __ | | | >-----| \ | .---------------------------|-/ |AND)---' | | |/ .--|__/ | | | | | | | | |\ | | `----------------------|-\ | | _ | >--' | /| .------------------|+/ | / | |/ B -+-/\/\/--+--/\/\/-+ / _|_ -

Then again, JL's "subtractor with gain followed by a window comparator" would be a simpler way to avoid all those problems.

But that won't do a "percentage" compare as Owen (Steve) Roberts requested in his original post.

I'm puzzled by an "additive" fix. Why not see what the servo driver output looks like as it converges? ...Jim Thompson

I assumed he meant percent of full scale. For percent of signal you would need to take my doubled version of Jasen's circuit and find a way to fix the

Or you could use the subtractor idea, and make that window comparator such that it has a smaller window for smaller input signals. That would be simpler overall, and only 1 adjustment.

OK. That's good, I like it :-) Window proportional to signal, and adjustable. Nice!

Of course we have no response time specifications. It would probably suffice to wait x amount of time after a command change and call it good.

Though I still fall back to watching the driver output... seems like all the info you need should be right there. ...Jim Thompson

I can't take credit. It's only a small change to Larkin's idea. :)

Since feedback is partly mechanical and there might be gear backlash, maybe the top and bottom of the window should be separately adjustable.

LASER on with delay and off immediately?

Yeah, I guess you'd want to see both control and servo signals before designing it.

Well, if you really want an analog multiplier cheap, look at the LM13700 data sheet; figure 6 is a four-quadrant multiplier.

These chips are about a dollar, and there's two sections (so, two multipliers) in each.

Ok, I need to do some scoping this weekend.

Servo has a effective bandwidth of DC-2.5 Khz for small steps. As actual time step specs are a trade secret, I'll do some settling time tests.

Usually it damps with a decaying sine wave within 2-3 cycles on a input step. This is at the first resonance of the shaft/mirror combination.

Data sheet is here.

450 uS is a more typical number.

I'm jammed up this week, I'll have some scope shots by the weekend.

Steve