double full wave rectification

[egasimus]

Here's what I have so far. A circuit that takes a sine wave, full wave rectifies it, and, between two inverting buffer stages, removes its DC bias to get a (non-sine) wave at twice the frequency.

Now, I want to full wave rectify that, but I can't seem to. I've tried adding a two-transistor phase splitter at the output of the second opamp, but that doesn't work. Help (and theory)?

[egasimus]

So far I have made progress - I added an opamp phase splitter.
Simulation.

I would prefer a transistor-based phase splitter (for/against this idea?), but I don't understand it very well and thus have been unsuccessful in simulating it.

[amptramp]

Your simulation circuit shows a gain of -1 for the inverting signal but a gain of 2 for the non-inverting signal.  I hope you have something different from 100 ohm resistors for an op amp stage - 1K to 10K would be more appropriate.  There are also circuits using diodes in the feedback path of the op amps that eliminate the diode drop.  At typical signal levels, this would matter.

I prefer the op amp solution because the transistor phase splitter has a much lower source impedance at the emitter than at the collector and this is variable with the transistor.  This may give you an imbalance that is difficult to tune out even with a variable resistor on one output.

[egasimus]

Good call. I've considered Schottky diodes for their lower voltage drop, but placing the diodes in the feedback path indeed does eliminate this.

So this is my next draft. However, I notice occasional transients in the signal - what are they? Am I doing something wront, or are they artifacts of the simulation?

[earthtonesaudio]

A workalike circuit can be made with fewer components if you use certain types of op-amp:

CA3140: http://freecircuitdiagram.com/2008/09/15/single-supply-precise-full-wave-rectifier-using-ca3140-op-amp/
OP90FZ: http://freecircuitdiagram.com/2010/09/24/op90-single-op-amp-full-wave-rectifier/
Two of the above circuits may be AC coupled to achieve double-FWR.

The key feature is being able to work with input voltages that exceed the negative supply.

These and your third simulation also have the feature/problem of different impedances for the negative- and positive-cycles.  A buffer in front (AC coupled again) would isolate the source from this.
Also a buffer on the output (for any of the circuits) would go a long way to help the circuit "play nice with others".

[R.G.]

Double full wave rectification is a fairly well known technique. The problems are the nitty details, having to remove the DC accurately and fast enough, and making sure your diode drops don't eat up your signal. In practice it has to have signals big enough that a diode drop is negligible or precision electronic diodes.

[egasimus]

Okay, so I was revising this circuit with the intent of building it. Any tips for making this work IRL, besides adding a buffer in front?

[~arph]

You can do it with a NE570/571 don't have the schematic here, but there is two rectifiers in there, just put them in series. Very little external components needed

[egasimus]

Interesting, thanks! I'll see if I can figure that out.

[amptramp]

There is another method of getting a double frequency output and that is to use the identity:

cos(2x) = 2cos²(x)-1 where x is an angle or 2*pi*frequency*time

You can use a multiplier IC like a Motorola 1496 to get a square of the signal and use capacitive coupling to eliminate the "-1" term which is a DC component.  If you use fullwave rectification, you get a lot of harmonics that may not be desirable as shown in:

http://www.falstad.com/fourier/e-fullrect.html

Just drag your cursor over the harmonic magnitude elements in the picture to get the exact magnitudes and over the phase elements to get the phase.  As you can see, Figure 30 in the first data sheet on the following site has already done it, but beware of overloading - analog multipliers like to operate at a few hundred millivolts:

http://www.datasheetcatalog.com/datasheets_pdf/M/C/1/4/MC1496.shtml

[Brossman]

Still a noob on this, so try to bear with me...

Why is it important to get double FWR? Is it good practice? more efficient? safer? more options? or just to sound cool?

[amptramp]

Still a noob on this, so try to bear with me...

Why is it important to get double FWR? Is it good practice? more efficient? safer? more options? or just to sound cool?

Each stage of fullwave rectification doubles the frequency, creating a pitch that is an octave above the input signal.  Cascading two of these stages gives you four times the frequency for a signal that is two octaves up.  Frequency doubling is commonly used in transmitters in the microwave region where fullwave rectification is often used to avoid using active components.  Other harmonics are introduced as well as shown in:

http://www.falstad.com/fourier/e-fullrect.html

These other harmonics are not necessarily that pleasant, which is why i suggested the squaring method to avoid their creation.  What you see in the above site working for a rectified sinewave does not necessarily sound good for a sinewave input, let alone a complex waveform.

[egasimus]

I'm only trying to multiply the frequency, in order to run it through a comparator, then through a PLL, and eventually get a square wave. The MC1496 might be a step in the right direction.