Hallo Vivek,

100RPM...4000RPM gives a frequency of 1.67Hz...66.7Hz. When you count pulses you always have a uncertainty of +-1pulse. Referring to RPM this gives an uncertainty of 60RPM.

I would solve problem like it is done in several frequency meters for the low frequency range: Use a PLL circuit, producing a frequency of 100 times of input frequency. PLL circuit divides this again by 100 and compares to input signal which rersults in a lock-in. If you now count the pulses of this 100 times higher frequency, counting uncertainty is decreased by the same factor.
Use CD4518 in combination with CD4046. From pin 11 of CD4046 to ground use 5k6, from pin 12 to ground 4M7, between pin6 and pin7 10n, between pin 9 and pin 13 1M0, from pin 9 to ground a series combination of 100k and 2.2µF.
At CD4518 connect pin 6 to pin 9, pin 1 to pin 4 of CD4046, pin 14 to pin 3 of CD4046. Input signal goes to pin 14 of CD4046, output is at pin 4 of CD4046.

You can, of course, change time constants of PLL loop in order to match with your requirements. Also, there is some settling time for locking to the correct frequency, but this settling time you will always have, even if you make some pure digtal signal condition.

Good luck,
Kai