Thanks Spehro.
I've already spotted those and also the TC4017, but neither will do.
Unfortunately I realise I wasn't clear enough about the voltage compliance:
they have to be mounted inside a constant current supplied 2 wires sensor
that settles its average bias point from 6V to maybe 16-17V on the whole
temperature range.
And the 0degreeC->0V is mandatory too, since we have to align on
competition, which is a defacto standard.
The LM35 is the ideal device, except the packaging. I was hoping of maybe a
SOT23 packaged second source.
The '35 and '45 are (I think) the same chip, with the LM35 having a
higher max voltage rating based on self-heating effects.
But be warned: both have a lot of personality. Small spikes can make
either latch up, especially if the supply voltage is above 5. And the
datasheet suggests you can pull the output below ground to measure
negative temps, but that can make it latch, too. And they oscillate
with capacitive loads.
Bob Pease promised me he'd fix it, 10 years ago or so.
One remaining advantage of the Fahrenheit temp scale is that you can
get LM34's, which output zero volts at zero F!
John