Diode losses - I think I've zapped them. Ring mod schemo using virtual diodes.

[brett]

Hi.
Over the last few years I've built a few passive ring modulators.
The "gating" and losses due to the losses in the diodes always bugged the cr*p out of me.

The schematic below might overcome these nasty problems, because if I'm not crazy (always a concern) these rectifiers have nil loss.  Anybody like to check the concept and suggest whether it will work?
I think somebody smarter (like Mark or RG) might be able to work out how to ditch the LM311 and use 4 of the six buffers in a CD4050 to drive the mosfets.  Or use dual op-amps.  In the meantime, the 4 comparators should work, and they only cost $1 each.

The idea came from circuits for reverse-polarity protection (which is all these rectifiers do, but with the signal rather than a power supply).
cheers

[R.G.]

Good theory. Synchronous rectifiers are really good if you can get them implemented.

There are some practical problems with what you have shown there. The LM311 comparators have to have floating power supplies so they can follow the MOSFETs around as the signal voltages move them, and the power supplies need high resistance and low capacitance between them to avoid interactions.

The complexity gets out of hand as well. Now you have a passive ring modulator and four 8 pin dips as well as five power supplies. It might be simpler to buy a 1496 modulator, or even simpler to make a high voltage power supply by doubling or tripling the 9V supply of the first stage, amplifying the signal up to near that, driving the modulator, then dividing the signal back down. That would get the diode losses to be insignificant.

There are LED-solarcell MOSFET drivers that do isolated gate drive, but they're not fast enough for the turn-on in this case, I don't think.

The idea is a good one. I just haven't yet come up with a practical way to both sense the voltage across the MOSFETs and drive the gates as they fly around on top of the signal voltages to do a good synchronous rectification scheme.

[toneman]

What about a (simple)diode-connected transistor array?
Sure, it's silli-con, but,
5 trannys in one dip package.
All trannys on same substrate, so they R matched as close as can B.
Could U boost signal into diodes to compensate 4 voltagedrop??

staymatched
tone

[Tim Escobedo]

TI makes the UC3611 chip, which is a four Schottky diode monolithic chip. Lower forward voltages than the transistor array, better matching than four discrete Ge diodes.

In the long run, though, the balanced modulator chips like the 633 or 1496 look like the best solution overall. Simple noise gates can also cover up a multitude of sins.

[R.G.]

Here's a line of reasoning: Diodes have a "dead zone" of about the clipping voltage where they do not respond. To make this dead zone smaller, you can trickle current through it. This is in fact how the Fender Blender and Green Ringer run their rectifier diodes - they trickle a little current through them.

To do this in a flying diode ring modulator, you could use the LED-photovoltaic things I mentioned. These put out up to 5V at a few microamps, and come two to an 8 pin DIP. You could use two of these, and drive each diode through a high value resistor to eat up most of the diode dead zone.

Again, the complexity grows. But it ought to work.

If you think about it, two transistors biased slighly on does a similar thing. The full wave rectifier in the SuperFuzz works like this. If you take two of those, then a diffamp to switch between them, you get something very like the 1496 balanced modulator again.

OK, here's another one. Use transformers with multiple windings. The extra windings are only for turning MOSFETs on and off. That ought to work.

[brett]

Hi.
Thanks for the comments.
I didn't understand the bit about multiple power supplies.  Can't the LM311 run off a single-sided supply? (+9V in this case, but also off +5V.)

I like the simplicity of "Use transformers with multiple windings. The extra windings are only for turning MOSFETs on and off."   I've used an LM386 to drive a 8ohm:1k transformer in a passive ring modulator before, and with reasonable success.  8:1k gives an 11 x boost to voltage.  (But assuming a 3V gate on the mosfet, that's still a 0.3 threshold on the rectifier.  No better than a Ge diode.  Bummer.)

The "boost the signal until the diode voltage is insignificant" doesn't really work well because of the log scale of volume.  Germanium diodes nullify about 0.5 V p-p, which is only 10dB down from the max 5V p-p that a 9V pedal can be expected to handle.  Simple modulators with 20 to 30dB dynamic range would be excellent.

Maybe I've got to accept the inevitable and use that 1496 that's been in the shed for a year...

[Paul Perry (Frostwave)]

It all depends what you want from your ring modulator.
If you trace out a couple of cycles of sine waves of different frequencies, (ior, much better, simulate it mathematically) you would see a very different output for the normal "curved" diode transfer, and the "straight line" transfer of the "ideal" diode.

only a diode with a curve approaching a square law, will give you the textbook AxB multiplication, like you get from an AD633 multiplier chip. But, that isn't to say that something else sounds better!!!
(actually, now I think about it, I'm wondering whether an "ideal" diode isn't going to cancel out completely! Is there a mathematician in the house??)

[A.S.P.]

information lost...