I need to convert a PWM to analog signal. But the issue is, frequency of PWM may change. So, RC-filter option is not preferred. What kind of circuitry would be appropriate for this?
There exists an IC named LTC2644, which operates exactly as I want, but not meeting my frequency requirements.
What you want is not possible. You now say that the PWM frequency can vary from 5 Hz to 500 Hz (200 to 2 ms), but that you want the filtered result to settle within 5% in 5 ms.
Think about it. How is the output supposed to settle to the average value within 2.5% of the pulse period? Until the pulse is over, how can you possibly tell what it's on-fraction is. Unless all your pulses are on for less than 5 ms, there is no way to extract the information from each pulse that fast.
For a 5 - 500 Hz PWM frequency range, with a PWM input signal having clean edge transitions, a simple microcontroller seems a reasonable solution. It should contain a timer and a digital-to-analog converter for output. Not many simple microcontrollers have built-in DAC - perhaps one of PSOC-4 types.
The timer is used as an incrementing counter, initialized to zero on one PWM edge, and incrementing until the corresponding alternate edge, whereupon count value is saved (C1), and re-zeroed. The next input edge completes one PWM cycle...the counter's value is saved once again (C2). The ratio of C1/C2 is scaled and written to the DAC.The PSOC-4 series is targeted for simple analog/digital crossover applications like this.
It is inevitable that one PWM period delay must be accepted as the minimum settling time for PWM period-chages.
simulate this circuit – Schematic created using CircuitLab
A DAC can be avoided with some extra effort. Many microcontrollers contain analog-to-digital converters. One general-purpose I/O pin can be used to charge a capacitor, whose voltage is monitored by the microcontroller's ADC. The I/O pin must be capable of normal High/Low as well as tri-state digital states. An op-amp (low bias current) buffer is required to keep the capacitor from discharging through the load.
Although a current source charging and discharging (charge pump) would be ideal, a resistor charging the capacitor is possible too with a little extra calculation of appropriate charge times.
edit: a much better (but rather more mathematical) approach is to use a resistance and a capacitor in series. By using either the general equation for the capacitance (1/jwc) or (1/2*pifC) you can find the frequency. However it might be slightly inaccurate as its not a sinusoidal signal. If you need an accurate representation you can use signal analysis for the simple circuit (look for circuit response)
Finally you can use ways to measure or convert capacitance to resistance. You probably can find either an op amp circuit for that or some other method
