Waveform via analog multiplier

Hi all, just looking for opinion or anecdotes.
I want to modify a low frequency analog signal. Most multipliers use SIN^2(X) for frequency doubling. I am hoping, using the MLT04 and/or AD633 to run the function y=sin(x) +cos^n(x).
the AD539 has an additional exponential input, up to 5th order, but my function has different results with odd-vs-even “n” exponents.
So, does anyone have nightmare stories about upper harmonics coming from multipliers?


Do note that the MLT04 is obsolete, though still available through Rochester.

I’d expect transfer function nonlinearities to introduce harmonic distortions of some degree, but it’s not apparent whether such would be of concern in the present case. The overall context of what you’re trying to do here isn’t entirely obvious, nor is the intended meaning of the AD539 being described as having “an additional exponential input.”

If you could provide a bit more context on your application and the underlying thought processes, that would be helpful in terms of offering information responsive to your interests.

I really have 2 goals, both in the audio spectrum.
But first, I apologize for mentioning the AD539, when the IC with powers/roots is the AD538.
The first goal is a guitar effect that follows the fundamental (most use LC/RC tanks with a fixed frequency) . I want to be able to add and effect the input waveform while keeping the fidelity of the source. I want to do this as an audio output because hearing sometimes tells more than a scope can.
Next I want to use it as a compressor that works in real time. I believe that first and second differentiators and an error amplifier will allow me to ‘read’ changes in slew rate as the wave approaches the upper rail, and reduce gain accordingly in real time. I have heard similar companders and know the distortion that can occur, but the test is to do it without an integrator feeding back.
I am sure it would be very easy in the digital domain. But Bob Pease said it wouldn’t work, so I feel I at least should build the thing.
So I tried the double angle formula for overtones. These multipliers use SIN^2, which is not the same, and does not provide 90° of lead, which is what I need to analyze and effect the waveform “ahead of time”. And d^2y/dx^2 is negative SIN(x), so as long as my first differentiator is clean [COS(X)], I will be able to extract the data from the change in slew rate that I need.
I found that in y=SIN(X) + COS(X)^n , as ‘n’ increases, the added pulse narrows, and flips up/down as ‘n’ goes even/odd.
The MLT04 and AD539/633/7xx are obsolete, i think, because newer chips have a 450V/uS slew rate, and these old chips are under 100, with unity gain in the low MHz range.
And again, sorry to confuse, but the chip I was looking at was the AD538, not 539. The AD538 can produce x^5 by adding an external resistor, but varying that resistor changes the exponent from odd to even, and I am interested in mapping that effect.

In terms of “what does this squiggle sound like?” I’d agree, though this this not the sorta room where one wants to talk about being able to hear things that an APx555 can’t :wink: No reason a person can’t 'scope a speaker waveform for multi-sensory analysis though…

I’m not quite sure I understand how you’re thinking of putting the mentioned pieces together; pictures worth a thousand words and all. Maybe 2K with inflation… Regardless, differentiation tends to make some pretty wild outputs from small amounts of noise, and talking about doing that twice kinda pulls upward on the eyebrow.

If I’m pondering what you’re pondering, it’s not apparent to me how the end result would be all that different from a soft clipping behavior.

So a peasey question to ask then would be whether or not a “why” was offered along with the binary prognostication, and the contents of if so. That’s usually where the good stuff is…

Indeed, it’s an interesting thing to play with on the spreadsheet and nobody said n has to be an integer either; it certainly won’t be in practice if one’s doing analog math. What happens though if instead of sin(x) you model your signal source more realistically as a semi-arbitrary f(t)? It’s not sine waves that’re coming out of that 1/4" jack…

They charge half a liver for those AD538’s, but if that’s not really a concern I’d say go for it; maybe the results won’t be as expected, but you’ll be better off spending your time experimenting with that sort of thing than ingesting the latest netflix programming…