Negative values of V_fundamental in calculating Fundamental Voltage in MSP430F6779 Three Phase Watt-Hour meter

I found the reason for the problem and I fixed it:

Problem in calculating fundamental voltage in the background interrupt.
-------------------------------------------------------------------------------------------------
Calculating the fundamental voltage uses the V_pure voltage that is returned
from the dds_lookup function. This is a pure sinus voltage value that must
be synchronized with the phase of the sampled voltage. The dds_lookup function
gets the value from a table that holds only the first quarter of the full sinus
cycle. The other 3 quarters are either a mirror value or negative same value.

I found that the incorrect values are due to the IAR optimizer mistranslates
of the C code. The optimizer at the emeter-toolkit project configuration is set
to maximum optimization for speed.

I added here the TI's Energy Library original code and my modified code that
solved the problem (in ...\Device\emeter-toolkit\dds.c):

#define SLENK 8
#define DDS_STEPS (1 << SLENK)
#define DDS_SHIFT (32 - 2 - SLENK)

   Original code
==============

  1  int16_t dds_lookup(uint32_t phase) 
  2  {
  3      int16_t amp;
  4      uint32_t step;
  5
  6      phase >>= DDS_SHIFT;
  7      step = phase & (DDS_STEPS - 1);
  8      if ((phase & DDS_STEPS))
  9           step = DDS_STEPS - step;
10      amp = sine_table[step];
11      if ((phase & (2*DDS_STEPS)))
12           amp = -amp;
13      return amp;
14  }

Optimizer problems:
1. After the right shift in line 6, the values should be in the range of
     0 - 1023 (10 bits, because DDS_SHIFT = 22). Breaking after this shift
     shows much larger values.
2. All values returned from the function are positive. Trying to place a
    break point at lines 11 or 12, failed. It means that the optimizer discards
    these lines.

Solution:

1. Copy and type cast the input 32-bit value "phase" into an internal 16-bit
    variable "ph" and use it instead of "phase". This change caused the
    optimiser to perform correct shift.
2. Change the type of "step" to 16-bit.

I also added comments to explain the calculations

   Modified code
=============

  1 int16_t dds_lookup(uint32_t phase)
  2 {
  3     int16_t amp;
  4     uint16_t step;
  5     uint16_t ph;
  6
  7     ph = (uint16_t)(phase >> DDS_SHIFT);
  8     step = ph & (DDS_STEPS - 1);     // get place in quarter
  9     if ((ph & DDS_STEPS))                  // check if in 2nd or 4th quarters
10          step = DDS_STEPS - step;      // yes, take a mirror value
11     amp = sine_table[step];                 // Get the Sinus value from the table
12     if ((ph & (2*DDS_STEPS)))           // check if in 3rd or 4th quarters
13     {
14          return (-amp);                             // yes, invert sign
15     }
16     return amp;
17 }