Matt, Jeff

The TrueRMS part used in the new violin mode BLRMS appears to be having trouble... its output may be positive with no input, or zero with a large input.  It appears that the trouble comes from an uninitialized variable in the RCG part (PART_n).  The following code is an approximation to what the RCG generates:

// TrueRMS:  PART
if (PART_first_time_through) {
    PART_first_time_through = 0;
    PART = in0;
    PART_sqrsum = in0 * in0;
    PART_indatsqrd[0] = PART_sqrsum;
} else {
    if (PART_n < PART_WINSZ) {
        PART_index = PART_n++;
    } else {
        PART_index = (1+PART_index) % PART_WINSZ;
        PART_sqrsum -= PART_indatsqrd[PART_index];
    }
    PART_indatsqrd[PART_index] = in0 * in0;
    PART_sqrsum += PART_indatsqrd[PART_index];
    PART_sqrval = PART_sqrsum / (double) PART_n;
    if (PART_sqrval > 0.0)  {
        PART = lsqrt(PART_sqrval);
    }else{
        PART = 0.0;
    }
}
 

With PART_n uninitialized, the PART_sqrsum is polluted by uninitialized values in the PART_indatsqrd array, resulting in a persistent offset.

We'll need to get this fixed and rebuild the SUS FEs.