Dirt Cheap 2% Full Temperature Accurate Clock Needed

By all means, go for it.

Cheers, James Arthur

I'm trying to teach, but you ignore what I offer :-( ...Jim Thompson

That f(out) is a function of Vdd, Vth, and t(pd) is self-evident (and discussed in the references I linked, too).

If I needed such an oscillator I'd simply investigate and solve any problems, as we always do whenever we do new designs.

But I don't need such an oscillator.

Personally I think a crystal would be better in most situations; it's more accurate, and lower-power.

I thought it might be useful for the group's sake, though, and posterity, to quantify the advantages of bootstrapping the cold end of Ct, as done in the classic CMOS oscillator topology. So I spent a few hours investigating, quantified, and posted it.

I also suggested a dual-path topology to eliminate d(Vth)/dT drift.

If you have something to share, why not just share it?

Cheers, James Arthur

"Self evident"? Bwahahahahahaha!

You opined... strenuously... that 2% was attainable with the ICAN schematic. I'm going Missouri on you... show me.

I "shared" that simulation of the real world device does +/- 3.5%.

Show me wrong. ...Jim Thompson

I invite you to point out where I ever said any such thing, much less strenuously.

You posted a bare assertion with no detail and no backup:

"Simulating a real CMOS circuit (inverters, I've done this exercise many times before :-), I get +/- 3.5% (1MHz).

I don't have a breakdown of what it's due to, Vth, versus stage delays, just the whole enchilada."

That's a bare assertion. But the +/- 3.5% figure was at least a useful marker, indicating the performance you achieved for an unknown oscillator using unknown components under unspecified conditions.

If you want to make a better RC oscillator, post your effort here and we'll help you.

Cheers, James Arthur

Discussion ended... you've become as obstinate as Larkin when challenged :-( ...Jim Thompson

I don't mind opinions, if people are right. But "challenges" aren't the best way to start a constructive discourse.

The NXP ARM chips that we use (LPC17xx family) have an internal 4 MHz oscillator that's accurate to 1% over temperature. That's good enough for most applications. We phase-lock the CPU core frequency up to about 100 MHz.

A few years ago I designed several RC oscillators for Microchip. I can assure you that the Microchip (or NXP) schematic is NOT like that of the ICAN note. ...Jim Thompson

;)

Cheers

Phil Hobbs