When making analog op-amp or 555 timers, many times RC time constant is employed.
Lets say we want 1 sec time constant:
We could chose for instance the combination R=100k and C=10uF.
Or we could choose R=100Meg and C=10nF.
We have zillions of possibilities here.
If possible I was always choosing the combination such that I can avoid using
electrolytic capacitor type.
For example in my last circuit, to obtain a 10 sec RC time constant I used R=1Meg C=10uF. I would like to use nF level capacitors to avoid using electrolytic capacitor. But in this case then I have to increase R to very high resistances.
Here my question is:
1-) Is keeping the cap value low in nF level and increasing the R to hundreds of mega-ohms problematic?
I thought we might be avoiding the use of electrolytic capacitor but then the very high resistor value could introduce more error?
2-) If I keep the resistor in kOhm level I should use then uF level cap. In this case what is the best alternative to electrolytic caps in uF level to obtain better accuracy and repeatability.
edit: Capacitors I use should not be surface mount.
Best Answer
Electrolytics have quite high leakage currents so they don't make a good choice and they also have a poor tolerance that gives significant variability in the design of a time delay circuit.
A commonly available non-polarized, 5% tolerance, surface mount type, I would consider is the X7R dielectric ceramic capacitor. In 1206 size you can get 10 uF 16V. If you are prepared to accept a 10% tolerance you can get a 47 uF in a 1210 case but it might only be 10 volts rated.
Worse dielectrics than X7R (such as X5R) yield higher values with wider tolerances and some change in capacitance with applied voltage thus making them unsuitable for decent timing applications.
So, depending on how accurate your timing needs are, I would make the assumption that 10 uF is the maximum (for half decent surface mount ceramic capacitors) and calculate the resistor based on this. If the resistance value is above 1 Mohm, I would start to worry about leakage currents from the pins on the chip and go another way entirely such as using a small MCU.