Low-power crystal oscillator in 0.18um?

I don't know squat about 0.18um processes, but what you really want is just an unbuffered complementary pair at a rather slow speed biased into linear operation.

I don't think speed can possibly be a concern; even with the CD4007 the outputs of the so-called gates (really you should think of them as complementary pairs) are generally slowed down with capacitors for oscillator use.

If there's any gate array equivalent of tiny logic unbuffered inverters eg. NC7SU04 etc. that's what you want.

Tim.

Hi Steve,

What is it exactly that you need?

A crystal oscillator, in CMOS, is nothing more than an unbuffered inverter with a big-ass feedback resistor for the DC Q-point, and an R-2C pi network to hang around the crystal.

The PAD's are usually dictatorial by TSMC rules... all you can eliminate is the series R.

I don't have Vittoz' paper. Could you send me a copy? Thanks! ...Jim Thompson

The oscillators in the library available to me (might be Artisan) are basically gate jobs. But they're not very power-efficient. I need something, perhaps 16MHz, that's well under the 100uA/MHz that the more basic cells provide. I know there are much niftier circuits that include various amplitude stabilization tricks, I'd just prefer not to have to design one just now (though it would be a great learning experience).

I'll post Vittoz' paper tomorrow from work. He's published quite a bit, but this one seems to be the most often referenced. I've had a photocopy for years. Frerking's and Parzen's books, both of which I have, get a lot of play too. Less so Matthys.

If there's anything like an unbuffered inverter in the library, that's what you want.

An inverter that can make nice sharp 16 MHz square waves is *not* what you want. You want something far mushier. I suspect if there are any cells in the low MHz range (do these even exist in the library?) they could oscillate just fine but maybe still not below your power goal.

I think many libraries do have a low-power 32kHz oscillator cell but don't know about your specific process and that exact part probably wouldn't work at all at 16 MHz.

Don't completely ignore the possibility of a tinylogic unbuffered inverter living outside the gate array. It'll be smaller than the crystal.

Tim.

Or, use a transistor. CMOS is intrinsically a class AB amplifier, it isn't as power-efficient as a class C amplifier in oscillator applications.