The issues for TRMS, no matter how you do it, is bandwidth and dynamic range. 50/60 Hz power is only part of the need. Audio, SMPS, telecom, and RF push the limits. I have implemented TRMS calculations using FFT front end processes on microcontrollers several times so that I could implement weighting filters for noise and other calculations in software. Using a reasonbly high performance sampling can cover TRMS for SMPS, impedance TDR, and plain old digital oscilloscope functions as well. Less complex TRMS calculations can be done for broadband measurements on wimpier microcontrollers. Until recently, these processes could not compete well with analog TRMS chips such as those from Analog Devices. The analog chips suffer from dynamic range limits verses bandwidth partly because a voltage is squared in some way. Whatever analog scheme you use, this places a hard limit on dynamic range that numerical calculations do not suffer. I have been issued patents for all three approaches but the analog one never generated any sales.