CA3019 Sine Shaper

[Steben]

Looks fun.
Anyone mentioned this lately or built it?

[ElectricDruid]

Looks like a variation on the typical tri-to-sine shaper. Soft overdrive on the triangle to remove the peaks, and what comes out is "sine" in 1980's parlance. Usually close enough for rock'n'roll, within a few percent distortion. Since the output is totally dependent on the input level, it works nicely for fixed-volume-output triangle oscillators, but won't help you for variable level signals.

[Rob Strand]

From what I see the idea is to allow some flexibility in
setting the output level.

I'm a little skeptical that the extra complexity follows from the
non-inverting configuration. Perhaps there's no need for this
if an inverting scheme is used?  I haven't simulated it.

The need for the extra "input/output" diodes comes from
the non-inverting clipper.

The non-inverting clipper doesn't hard clip it always adds
the signal on the opamp +input to the clipper voltage. In other
words the input signal leaks though even when clipping.

If the input was a triangle wave the triangle wave, in particular
the tips, would leak through to the output.

In order to clamp +input signal a hard clipper is required at the input.
The set of input/output diodes and input resistors forms an adjustable
soft clipper.  It's soft in the sense the Thevenin impedance
of the output divider (430 ohm and R2=1k) and the 5k1 input
resitor form a divider which prevents the diodes hard clipping.
=>  That means the triangle tips cannot be flattened and still
    leaks through to the output but probably below the level of
    causing visible glitches.
    (same happens on the diff pair shaper with emitter resistors.)

Let Vi is the input, Vi' the voltage of the opamp +input, Vo the output.
    A amplifier gain Vo/Vi'
    k is the division ratio of the divider (430 ohm + R2=1k)

The voltage across the input/output diodes is,

Vd_clamp = Vi' - kVo
        = Vi' - k*A.Vi'
        = (1 - k*A) Vi'

If k*A > 1 the input/output diodes don't clip but provide positive feedback
This will *stretch* the peaks not compress them - we don't want that.
The 430R is chosen to prevent this when R1 is maximum.
However, when R1 is minimum the gain is higher and R2 needs to be set to a
lower setting to make k lower.

k is essentially making the input clip point adjustable.
The clip point is Vi' = Vd / (1-kA).

R1_min
- gain about A=2
- to clip both feedback diodes
  need output > 4*Vd
  then input > 2*Vd

R1_max
- gain about A=3/2
- to clip both feedback diodes
  need output > 6*Vd
  then input > 4*Vd

I suspect we need to set k to the resulting Vi' when Vd_clamp=Vd to just over
these upper input knee points in order that the input clipper
starts clipping when the feedback clipper cannot (due to the
non-inverting configuration "leakage" issue).

[You might want to check those numbers - there's a bug here somewhere.

Will fix later.  however, idea is the same.
.. well, it seems OK. ]

[ElectricDruid]

Hi Rob,
There are some variations of this approach that use some "second order" cancellation of the little "nibs" that you tend to get at the peaks that might be closer to what this is trying to do, or might be exactly what this is trying to do. I could try and dig up some references if you're interested? Or maybe you'Ve already seen this stuff?

Tom

[Rob Strand]

There are some variations of this approach that use some "second order" cancellation of the little "nibs" that you tend to get at the peaks that might be closer to what this is trying to do, or might be exactly what this is trying to do. I could try and dig up some references if you're interested? Or maybe you'Ve already seen this stuff?

If they are easy to find I'd be interested.  (I've probably clocked up too many hours on sine shapers but it was some years ago.)

Just throwing some stuff into spice it seems "easier" for the non-inverting stage to flatten-off the tops.   That follows from the front-end "hard" clipper.  So perhaps spike flattening one of the motivations behind the circuit.  Based on that maybe the circuit does offer something different.

I haven't made much effort in trying to come-up with an equivalent inverting clipper (if that is even possible in general).  Superficially have two clippers in cascade would be expected to produce a harder clipping.  However after juggling these things around in the past is its surprising how things unravel themselves into equivalents circuits - perhaps not obvious from the start.

[iainpunk]

i build it in 2019, this exact schematic with the prescribed CA3019 and CA3140.but with +/- 9v. i boosted the input signal with another CA3140 with variable gain (up to 11x).

it sounded like an overdrive/distortion cross. a bit to 'farty', undefined chords, mushy sound, probably needs some filtering upfront to sound great, but i could hear the potential.
the 10k control is also a nice feature, it changes the clipping feel in a subtle way, hard to explain. it also feels like going to the middle it gets softer and going away from the middle gets harder, while it doesnt matter if you go up or down from the middle.
the 1k control goes from hard to soft clipping, but is very interactive with the other control. i dont find this circuit particularly worth it due to its cost and complexity for a marginally more interesting set of clipping controls.

cheers

[Steben]

i build it in 2019, this exact schematic with the prescribed CA3019 and CA3140.but with +/- 9v. i boosted the input signal with another CA3140 with variable gain (up to 11x).

it sounded like an overdrive/distortion cross. a bit to 'farty', undefined chords, mushy sound, probably needs some filtering upfront to sound great, but i could hear the potential.
the 10k control is also a nice feature, it changes the clipping feel in a subtle way, hard to explain. it also feels like going to the middle it gets softer and going away from the middle gets harder, while it doesnt matter if you go up or down from the middle.
the 1k control goes from hard to soft clipping, but is very interactive with the other control. i dont find this circuit particularly worth it due to its cost and complexity for a marginally more interesting set of clipping controls.

cheers

I understand.
I think the CA3019 package "might" be interesting for any diode ladder in any other circuit  to mimick a certain transfer curve.
But the use of an inverting stage is interesting as well.

[iainpunk]

im contemplating using a CA3019 for a transistor based wave folder, maybe ill throw in another metal can transistor array to achieve full MOJO status.

cheers