- Hence I will probably fall back to using a schmitt trigger (or a NAND gate to have one input as an enable pin) oscillator. The 555 also works, but has more pins and require more components.
- - The single NAND SN74AHC1G00 seems to be a good choice:
Symmetric input levels for close to 50 % duty cycle and low-leakage which enables me to use large resistors (instead of a large capacitor)
That's very much not a good idea. The frequency of a single gate Schmitt inverter oscillator is highly dependent on the width of its hysteresis band, which isn't specified for the SN74AHC1G00, and you will certainly violate its max recommended input transition time (100ns @3.3V). If you use a dual NAND you can implement a two gate oscillator with predictable period, but that needs three passives (see
Fairchild AN-118), and you are back up to an 8 pin package.
You could use a CMOS 555, with Threshold and Trigger tied together (which gives you an inverter with a guaranteed 1/3 Vcc hysterisis band), and to the timing capacitor, and a single feedback resistor from Out to the Threshold + Trigger + TC node.
However that still leaves you needing a fairly large, extremely low leakage timing cap, and a high value feedback resistor, so it will be highly sensitive to board surface contamination and enviromental conditions.
The best option if you want to avoid programmable chips, is an higher frequency oscillator + a divider chain. 74HC4060 if you can afford the board space, or bite the bullet and pay the premium for the LTC6995 timerblox chip if you need something in a smaller package.