Fix a numerical overflow in wavegen() which could cause distortion when a voice file increases the height of formant peaks.

git-svn-id: https://espeak.svn.sourceforge.net/svnroot/espeak/trunk@92 d46cf337-b52f-0410-862d-fd96e6ae7743

src/wavegen.cpp

@@ -1341,7 +1341,7 @@ static int Wavegen()
 			z2 = (z2 * mix_wave_amp)/32;
 		}
 
-		z1 = z2 + (((total>>7) * amplitude2) >> 14);
+		z1 = z2 + (((total>>8) * amplitude2) >> 13);
 
 		echo = (echo_buf[echo_tail++] * echo_amp);
 		z1 += echo >> 8;