discrete transistor ring mod and wavefolding

[EATyourGuitar]

the possibility for non-linearity seems perfect for a guitar pedal. what can we do with this?

http://en.wikipedia.org/wiki/Gilbert_cell

[Transmogrifox]

In addition to what you have suggested, the following are possible:
Compressor
Expander
VCF
Tremolo
Dynamically controlled distortion
Octaver (frequency doubler via x^2)

[R.G.]

The Gilbert Cell is what's inside nearly every analog multiplier, notably the MC1496 and many others.

Tansistor matching is critical to its function. That's why you don't see many discrete implementations. Inside a monolithic chip, they're all matched by default.

Buy a 1496.

[EATyourGuitar]

check the price on DMMT3904

[R.G.]

check the price on DMMT3904

Yep. Mouser says they're $0.39 each.

A Gilber Cell needs a matched quad and either one or better yet two matched pairs. The matched quad is a real toughie. To do a Gilbert with multiple duals, you'd need to sort/match the duals into a more-or-less matched quad, then use the other pairs. You need three or four pairs, and at $0.39 each, you get a simple Gilbert Cell for $1.17. Mouser sells the MC1496 for $0.90. It's cheaper, better matched, and more thermally coupled to help with thermal offsets.

I once looked into using matched duals for building an OTA. One of the things I ran into is that some "dual" devices are not monolithic devices on the same substrate, just two separate dies in the same package. The datasheet for the DMMT3904 from MCC at mouser does not specify a matching parameter, only the parameters for an SMD 3904. It's possible that they *might* be a matched monolithic dual, but they do not specify that they do, so if you get them and they're not, they haven't lied to you.

Real monolithic duals will almost always have a parameter expressed as a variation in Vbe or gain or something. After all, there has to be a reason to pay more for a dual. In the case of the DMMT3904, that reason may be PCB area. MCC does talk about how tiny they are.

On the other hand, Diodes Incorporated's version of the DMMT3904 *is* matched, and the specify the matching: http://www.mouser.com/ds/2/115/ds30311-19227.pdf
So I would trust the Diodes Incorporated part to match to 2% and the MCC version not at all.

Mouser also lists matched quads of NPNs for as little as $6.25 each, up to about $10 each.

[~arph]

So is the MC1496 like a cheap version of the AD633?

[R.G.]

So is the MC1496 like a cheap version of the AD633?

I'd say it differently. The MC1496 is pretty much a basic Gilbert Cell multiplier which was primarily intended for RF applications. The upscale multipliers are what the Gilbert Cell grew up to be when fed steroids all their lives. They're tweaked, laser-trimmed, and with extra additions to get wider range, lower distortion, more and less yada.

[~arph]

Ah, yes, so depending on your application you might not need all the yada that the 633 has. Saves a couple of bucks then. Nice to know about this IC.