I want to know if it is possible to use a shunt regulator in series with a load. I need to make a 5mA 3.3v supply from a 12v rail, however it needs to be efficient and extremely cheap.

The average load at the 3.3v rail will only be about 200 uA, but the burst current could be 5-10 mA for a second or two at a time. A series regulator such as an LM317 would work just fine, but it is expensive. It's about 7 cents in quantity. A 431 on the other hand can be had for less than 5 cents.

Essentially, during sleep periods there is just enough current coming from the divider and a trickle through the 431 to power to the MCU. Once a burst starts, it will start draining the capacitor. That will cause the 3v3 rail to drop, and the reference voltage to rise above 2.5v, which should open the shunt regulator and increase the current. It will turn off when it has recharged the cap back up to 3.3v.

I can tolerate +10/-20% ripple on the 3.3v rail, but I don't have a feeling for how fast the 431 will respond or if I will have any problems with oscillations. Has anyone ever tried to do this before? Will it even work?

2 cents really makes a difference here. We could be making over a million of these. I don't want to use the more expensive part unless I absolutely have to.

Thoughts?

 

Bite the bullet.....use the more expensive part, and stuff a rag in your beancounters yap.

Beancounters have been the bane of an engineers' existence almost forever.

 

Quite so, and there exists the problem that the cheaper part may become obsolete in the near future. If you think there's a posssibility of making these in a huge volume in future money talks when it comes to volume pricing and you might be able to get the better part at a lower cost anyway.

You're just too early in the prototype stage to worry about volume production costs, try it with both parts and if they interchange without a hitch then just choose which one to use as you build them. The documentation can easily read XXX, alternate part # yyy

 

At a million it probably won't go obsolete before the OP buys them.
My thought is to look at the market. If you can buy something premade you can still negotiate a price break (at your quantity), and it will probably be cheaper than making them yourself.

Is the 12VDC stable? If it is why not use a simple zener in series? 12VDC - 8.7 V zener = 3.3V. You could even use the shunt regulator for precision.

If the 12VDC is stable you could also use a simple emitter follower with 2 resistors. This is the most basic of voltage regulators. I don't know how expensive your transistors are, but I suspect they are much cheaper, and my personal costs are 7¢ for a PN2222 and 2¢ each for resistors. Again, if you need precision you could still use the LM431, but I suspect it is major overkill.

You may want to add a collector resistor to provide a crude form of current limiting in the event of a short on the emitter. Protection is always nice.

My instructor always told us in class that skull sweat is cheaper than components. It may have been over 30 years ago, but it is still truth.

Welcome aboard to AAC!