analog circuit to compute sine of voltage

[Jim Thompson]

Sure, I like collecting schematics of antique solid state electronics...

Looks like it was "only" 22 years ago...

Newsgroups: alt.binaries.schematics.electronic
Subject: Function Generator - FunctionGen.pdf
Message-ID: <[email protected]>

...Jim Thompson

 

[donald]

Hello -- Can anyone tell me how to make an analog circuit that takes
in a signed voltage v and produces a voltage
a * sin (b * v), where a and b are constants. It doesn't matter to me
what a and b are, since I can always scale the input/output voltages
appropriately.

The application is to drive a large D.C. motor, where the argument to
the sin function is given by a potentiometer connected to the shaft,
in such a way that the motor simulates a swinging pendulum. However,
even a low-power circuit involving op-amps, say, would be okay, since
I could always feed the output into some large power transistors etc.

Thanks, Robert

Just for fun.....

Is this a one off project ??

If its to re-store or re-build a pendulum clock, it would be kind of
cool that you use tubes for the control of the motor.

This would make a cool SteamPunk project.

donald

 

[Phil Hobbs]

I found a data sheet at http://www.analog.com/UploadedFiles/Obsolete_Data_Sheets/55889558331308887AD639.pdf
.. That's EXACTLY what I need.
Darn, it's discontinued!

I sort of like the suggestion to use a sine wave oscillator and a
sample-and-hold. If you use a divide-by-two counter and a 4046 to phase
lock a square wave to half the sine wave frequency, you can put your
control voltage in to shift the phase of the square wave over more than
a whole cycle of the sine, with ~0.1% linearity. Sample the sine wave
using the edge, and clean up the glitches with an RC lowpass.

For motor control, that should be lots fast enough.

Cheers,

Phil Hobbs

 

[Jim Thompson]

I sort of like the suggestion to use a sine wave oscillator and a
sample-and-hold. If you use a divide-by-two counter and a 4046 to phase
lock a square wave to half the sine wave frequency, you can put your
control voltage in to shift the phase of the square wave over more than
a whole cycle of the sine, with ~0.1% linearity. Sample the sine wave
using the edge, and clean up the glitches with an RC lowpass.

For motor control, that should be lots fast enough.

Cheers,

Phil Hobbs

From Last October...

Easiest way to make an adjustable phase shifter without amplitude
variation is push-pull drive....

+E
o
|
\
/ Variable R
\
|
o----> Output
|
|
|
--- C
---
|
|
|
o
-E

( -E is 180° out-of-phase from +E )

I first used this in 1960 to adjust the phase on a two-phase smear
camera (mirror rotator) motor doing ~20,000 RPM (at MIT's Building
20).

...Jim Thompson

 

[Phil Hobbs]

From Last October...

Easiest way to make an adjustable phase shifter without amplitude
variation is push-pull drive....

+E
o
|
\
/ Variable R
\
|
o----> Output
|
|
|
--- C
---
|
|
|
o
-E

( -E is 180° out-of-phase from +E )

I first used this in 1960 to adjust the phase on a two-phase smear
camera (mirror rotator) motor doing ~20,000 RPM (at MIT's Building
20).

...Jim Thompson

Right, but the OP wants something like an AD639, which is more or less
independent of frequency.

Cheers,

Phil Hobbs

 

[Robert]

This sounds very interesting, especially so because you exhibit the
fortitude ( or maybe the cluelessness, dunno ) to attempt a purely
analog solution....

I think I've been making this unnecessarily complicated. Instead of a
linear potentiometer attached to the shaft, fed into a sin function
circuit, I could simply get a sin potentiometer (I just found out
there are such things, eg. http://www.p3america.com/sine_cosine_potentiometers.htm).

As for the control loop, I think the torque due to gravity of a rigid
unbalanced body acting as a pendulum is solely dependent on the sin of
the position angle (the mass, rotational inertia, and radius of the
center of gravity being fixed). And if the torque produced by a D.C.
motor is reasonably proportional to the current, I'd just have to
arrange things so the the sin potentiometer proportionally controls
the current drawn by the motor.

 

[Spehro Pefhany]

If you don't mind the function having knees in it, you can do a fairly
good job with a couple of packages quad of rail to rail op-amps. Each
op-amp amplifies the input voltage with a different gain. The outputs
of the op-amps are summed together with resistors.

Near zero volts all of the op-amps are providing gain so the slope of
F(X) is large. As you move away from zero, the op-amps start hitting
the rails (each in turn) and reducing the slope of F(X). When the
lowest gain op-amp hits the rail, you are at the peak.

With two packages of quad op-amps, you can do a 16 point curve.

But this thing's not monotonic- he wants +/- pi range, not the rather
easier +/- pi/2. I guess you could still do it, but it might be easier
to put some analog switches and comparators in there and just
implement the 0~pi/2 curve, just as you'd likely do it digitally.



Best regards,
Spehro Pefhany

 

[whit3rd]

Hello -- Can anyone tell me how to make an analog circuit that takes
in a signed voltage v and produces a voltage
a * sin (b * v),...
The application is to drive a large D.C. motor, where the argument to
the sin function is given by a potentiometer connected to the shaft,
in such a way that the motor simulates a swinging pendulum.  

The restore force on a pendulum is only approximately
sinusoidal, so you might be able to use a crude approximation
of a sine. A simple resistors-and-diodes attenuator
with the right breakpoints can handle it for B*V in the
range of (-1, +1) though it will fail, of course, at (-pi/2, +pi/2).

 

[MooseFET]

But this thing's not monotonic- he wants +/- pi range, not the rather
easier +/- pi/2. I guess you could still do it, but it might be easier
to put some analog switches and comparators in there and just
implement the 0~pi/2 curve, just as you'd likely do it digitally.

Pendulums rarely swing more than +/- PI/4. It would be a strange one
indeed where the weight goes above the center of rotation.

If I needed to go a full circle, I think I could still do it woth the
R-R op-amp trick. It would take one more op-amp to make a signal that
does this:


! A ! B ! C !
...*........
..*.*.......
.*...*......
*.....*...*.
.......*.*..
........*...
!
0V

In A and C the extra op-amp is clipping. This extra op-amp would be
inverting so that when combined with the input by resistors, you get
the needed shape.

 

[Jamie]

Actually, I don't want the circuit itself to be an oscillator, just
to produce the sine of a given input voltage. That is, if the input
voltage is fixed over time, then so would the output be. I imagine
analog computers must have had circuits like this in the old days...

you mean a simple RC circuit? (resistor and capacitor network) for a
charged time constant?


http://webpages.charter.net/jamie_5"

 

[JosephKK]

Hello -- Can anyone tell me how to make an analog circuit that takes
in a signed voltage v and produces a voltage
a * sin (b * v), where a and b are constants. It doesn't matter to me
what a and b are, since I can always scale the input/output voltages
appropriately.

The application is to drive a large D.C. motor, where the argument to
the sin function is given by a potentiometer connected to the shaft,
in such a way that the motor simulates a swinging pendulum. However,
even a low-power circuit involving op-amps, say, would be okay, since
I could always feed the output into some large power transistors etc.

Thanks, Robert

Google triangle to sine conversions - diode wave shaping. Same technique.

 

[JosephKK]

Oh, and by the way, an input voltage range corresponding to -180
degrees to +180 degrees is sufficient.

Diode wave shaping will get you -90 to +90 degrees and no more.

Only digital will get the range you ask for.

 

[Phil Hobbs]

Diode wave shaping will get you -90 to +90 degrees and no more.

Only digital will get the range you ask for.

Not true. If you use the PLL method I suggested earlier, you can get as
many cycles range as you like by increasing the division ratio. (And if
you want to quibble about the digital divider, you can multiply the sine
frequency instead of dividing the square wave.)

Cheers,

Phil Hobbs

 

[Phil Hobbs]

How are you going to synch your oscillator with the motor speed?

You don't need to. Just make it much faster. It's only the relative
phase of the sampling clock and the sine oscillator that matters.

Cheers,

Phil Hobbs

 

[Nico Coesel]

This means the delay and frequency of the input voltage changes are low.
One idea would be to build a sine oscillator and a sample-and-hold circuit,
which samples the sine output after some time n, starting from zero
crossing from low to high, where n is proportional to the input voltage. I
don't know how to build this, but should be straight forward to design for
an analog expert, maybe with two quad op-amps.

If you have a triangular wave generator and run it's output through an
integrator, you'll have a sine wave. A comparator which compares the
triangular wave with the input signal can be used as a control for a
sample & hold switch.

 

[Tim Williams]

If you have a triangular wave generator and run it's output through an
integrator, you'll have a sine wave.

Not quite. Integeral(x) = 1/2 x^2 (plus C ;-) ). Alternate proof: the
derivative of a sine wave is not a triangle wave, so the integral of a
triangle wave isn't a sine wave.

It would look mighty close though. The O.P. didn't spec error at all.

Tim

 

[Fred Bloggs]

Looks like it was "only" 22 years ago...

Newsgroups: alt.binaries.schematics.electronic
Subject: Function Generator - FunctionGen.pdf
Message-ID: <[email protected]>

...Jim Thompson

Just got it, thanks, great schematic...

 

[Robert Baer]

If you have a triangular wave generator and run it's output through an
integrator, you'll have a sine wave. A comparator which compares the
triangular wave with the input signal can be used as a control for a
sample & hold switch.

Incorrect.
An integrator will not ermove all of those harmonics...

 

[sycochkn]

Isn't that a VCO?
XR2206 came to mind... Dunno if that old chip is still around.


D from BC
British Columbia
Canada.

Synchro?

Bob

 

[D from BC]

Synchro?

Bob

Synchro?
I didn't read much into the problem..(Was tired.)

I later realized the OP wanted the analog version of a pocket
calculator not an oscillator.

Speaking of...
Can't recall who..somebody posted using multipliers..
Considering the Taylor series:
sin(x) = x - (x^3)/3! + (x^5)/5! - (x^7)/7! + ...

Taps off a chain of multipliers???

x-----+----------+--------+-------+---------+
| | | | |
x >multi1--->multi2--->mulit3---multi4---multi5 ..and so on
| | |
| -9.5dB -15dB
| | |
+--------> sum/sub amp <--------+
(mixer)
|
Sin(x)

Perhaps using a bunch of 4 quadrant multiplier chips.
Ex.
http://www.analog.com/UploadedFiles/Data_Sheets/66167428AD734_c.pdf

I'm not saying this will work...Just making ideas..

D from BC
British Columbia
Canada.