diff --git a/plugins/channel/lora/lorabits.h b/plugins/channel/lora/lorabits.h
index 9698a5618..0ef527279 100644
--- a/plugins/channel/lora/lorabits.h
+++ b/plugins/channel/lora/lorabits.h
@@ -3,12 +3,15 @@
  Chosen mode "spreading 8, low rate" has 6 bits per symbol, so use 4:6 FEC
 
  More spreading needs higher frequency resolution and longer time on air, increasing drift errors.
- Want higher bandwidth when using more spreading, which needs more CPU.
+ Want higher bandwidth when using more spreading, which needs more CPU and a better FFT.
 
- Six bit Hamming can only correct drift errors. Want 7 or 8 bit FEC for QRM
+ Six bit Hamming can only correct drift errors. Need 7 or 8 bit FEC for QRM
+ Easier to use 5:4 FEC with Reed-Soloman messages.
+
+ Implicit Mode: explicit starts with a 4:8 block and has different whitening
 */
 
-// Needs adjusting for different sizes
+// Six bits per symbol, six chars per block
 void LoRaDemod::interleave6(char* inout, int size)
 {
 	int i, j;
@@ -46,11 +49,11 @@ void LoRaDemod::hamming6(char* c, int size)
 		i++;
 		c[i] = ((c[i] & 1)<<2) | ((c[i] & 2)<<2) | ((c[i] & 4)>>1) | ((c[i] & 8)>>3);
 		i++;
-		c[i] = ((c[i] & 32)>>2) | ((c[i] & 2)<<1) | ((c[i] & 4)>>1) | ((c[i] & 8)>>3);
+		c[i] = ((c[i] &32)>>2) | ((c[i] & 2)<<1) | ((c[i] & 4)>>1) | ((c[i] & 8)>>3);
 		i++;
 		c[i] = ((c[i] & 1)<<3) | ((c[i] & 2)<<1) | ((c[i] & 4)>>1) | ((c[i] & 8)>>3);
 		i++;
-		c[i] = ((c[i] & 1)<<3) | ((c[i] & 2)<<1) | ((c[i] & 4)>>1) | ((c[i] & 16)>>4);
+		c[i] = ((c[i] & 1)<<3) | ((c[i] & 2)<<1) | ((c[i] & 4)>>1) | ((c[i] &16)>>4);
 		i++;
 		c[i] = ((c[i] & 1)<<3) | ((c[i] & 2)<<1) | ((c[i] & 4)>>2) | ((c[i] & 8)>>2);
 	}
@@ -61,7 +64,12 @@ void LoRaDemod::hamming6(char* c, int size)
 void LoRaDemod::prng6(char* inout, int size)
 {
 	const char otp[] = {
-	"5^ZSm0=cOGMgUB=bNcb<@a^T;_f=6DEB]2ImPIKg:j]RlYT4YZ<`9hZ\\PPb;@8X8i]Zmc_6B52\\8oUPHIcBOc>dY?d9[n5Lg]b]R8hR<0`T008h9c9QJm[c?a:lQEGa;nU=b_UbTW3=W5Aa<9i;F;ondS[LBA;[4S9]kkh]Vc2j>kX"
+	//explicit mode
+	"cOGGg7CM2=b5a?<`i;T2of5jDAB=2DoQ9ko?h_RLQR4@Z\\`9jY\\PX89lHX8h_R]c_^@OB<0`W08ik?Mg>dQZf3kn5Je5R=R4h[<Ph90HHh9j;h:mS^?f:lQ:GG;nU:b?WFU20Lf4@A?`hYJMnW\\QZ\\AMIZ<h:jQk[PP<`6[Z"
+#if 0
+	// implicit mode (offset 2 symbols)
+	"5^ZSm0=cOGMgUB=bNcb<@a^T;_f=6DEB]2ImPIKg:j]RlYT4YZ<`9hZ\\PPb;@8X8i]Zmc_6B52\\8oUPHIcBOc>dY?d9[n5Lg]b]R8hR<0`T008h9c9QJm[c?a:lQEGa;nU=b_WfUV2?V4@c=8h9B9njlQZDC@9Z<Q8\\iiX\\Rb6k:iY"
+#endif
 	};
 	int i, maxchars;
 
diff --git a/plugins/channel/lora/lorademod.cpp b/plugins/channel/lora/lorademod.cpp
index cea99b88a..ceb827368 100644
--- a/plugins/channel/lora/lorademod.cpp
+++ b/plugins/channel/lora/lorademod.cpp
@@ -77,25 +77,30 @@ void LoRaDemod::dumpRaw()
 	char text[256];
 
 	max = m_time / 4 - 3;
-	if (max > 100)
-		max = 100; // about 2 symbols to each char
+	if (max > 140)
+		max = 140; // about 2 symbols to each char
 
 	for ( j=0; j < max; j++) {
-		bin = (history[j * 4 + 12] + m_tune ) & (LORA_SFFT_LEN - 1);
+		bin = (history[(j + 1)  * 4] + m_tune ) & (LORA_SFFT_LEN - 1);
 		text[j] = toGray(bin >> 1);
 	}
-
 	prng6(text, max);
-	interleave6(text, max);
-	hamming6(text, max);
+	// First block is always 8 symbols
+	interleave6(text, 6);
+	interleave6(&text[8], max);
+	hamming6(text, 6);
+	hamming6(&text[8], max);
 
 	for ( j=0; j < max / 2; j++) {
-		text[j] = (text[j * 2 + 2] << 4) | (0xf & text[j * 2 + 1]);
+		text[j] = (text[j * 2 + 1] << 4) | (0xf & text[j * 2 + 0]);
 		if ((text[j] < 32 )||( text[j] > 126))
 			text[j] = 0x5f;
 	}
+	text[3] = text[2];
+	text[2] = text[1];
+	text[1] = text[0];
 	text[j] = 0;
-	printf("%s\n", text);
+	printf("%s\n", &text[1]);
 }
 
 short LoRaDemod::synch(short bin)