diff --git a/src/capture_line_sixbits_4bpp_8bpp.S b/src/capture_line_sixbits_4bpp_8bpp.S
index 620c355e..6d2b8aa0 100644
--- a/src/capture_line_sixbits_4bpp_8bpp.S
+++ b/src/capture_line_sixbits_4bpp_8bpp.S
@@ -29,15 +29,15 @@ capture_line_sixbits_4bpp:
         tst     r3, #BIT_DEBUG
         eorne   r6, r6, #0x50         //magenta in leftmost
         eorne   r6, r6, #0x02000000   //green in rightmost
-        SKIP_PSYNC
+        SKIP_PSYNC_FAST
 loop:
-        WAIT_FOR_PSYNC_EDGE              // expects GPLEV0 in r4, result in r8
+        WAIT_FOR_PSYNC_EDGE_FAST         // expects GPLEV0 in r4, result in r8
         CAPTURE_0_BITS_WIDE              // input in r8, result in r10, corrupts r9
-        WAIT_FOR_PSYNC_EDGE              // expects GPLEV0 in r4, result in r8
+        WAIT_FOR_PSYNC_EDGE_FAST         // expects GPLEV0 in r4, result in r8
         CAPTURE_1_BITS_WIDE              // input in r8, result in r10, corrupts r9
-        WAIT_FOR_PSYNC_EDGE              // expects GPLEV0 in r4, result in r8
+        WAIT_FOR_PSYNC_EDGE_FAST         // expects GPLEV0 in r4, result in r8
         CAPTURE_2_BITS_WIDE              // input in r8, result in r10, corrupts r9
-        WAIT_FOR_PSYNC_EDGE              // expects GPLEV0 in r4, result in r8
+        WAIT_FOR_PSYNC_EDGE_FAST         // expects GPLEV0 in r4, result in r8
         CAPTURE_3_BITS_WIDE              // input in r8, result in r10, corrupts r9
         WRITE_WORD_FAST
         subs    r1, r1, #1
@@ -64,14 +64,21 @@ capture_line_sixbits_8bpp:
         tst     r3, #BIT_DEBUG
         eorne   r5, r5, #0x05           //magenta in leftmost
         eorne   r6, r6, #0x02000000     //green in rightmost
-        mov     r1, r1, lsl #1
-        SKIP_PSYNC
+        //mov     r1, r1, lsl #1
+        SKIP_PSYNC_FAST
 loop_8bpp:
-        WAIT_FOR_PSYNC_EDGE                      // expects GPLEV0 in r4, result in r8
+        WAIT_FOR_PSYNC_EDGE_FAST                      // expects GPLEV0 in r4, result in r8
         CAPTURE_LOW_BITS_8BPP_WIDE               // input in r8, result in r10, corrupts r9
-        WAIT_FOR_PSYNC_EDGE                      // expects GPLEV0 in r4, result in r8
+        WAIT_FOR_PSYNC_EDGE_FAST                      // expects GPLEV0 in r4, result in r8
         CAPTURE_HIGH_BITS_8BPP_WIDE               // input in r8, result in r10, corrupts r9
-        WRITE_WORD_FAST
+        mov     r7, r10
+        WAIT_FOR_PSYNC_EDGE_FAST                      // expects GPLEV0 in r4, result in r8
+        CAPTURE_LOW_BITS_8BPP_WIDE               // input in r8, result in r10, corrupts r9
+        WAIT_FOR_PSYNC_EDGE_FAST                      // expects GPLEV0 in r4, result in r8
+        CAPTURE_HIGH_BITS_8BPP_WIDE               // input in r8, result in r10, corrupts r9
+        mov     r9, r7
+     //   stmia  r0!, {r7, r10}
+        WRITE_WORDS_8BPP_FAST
         subs    r1, r1, #1
         bne     loop_8bpp
         pop     {pc}
diff --git a/src/macros.S b/src/macros.S
index 92ccb7a2..38c7137c 100644
--- a/src/macros.S
+++ b/src/macros.S
@@ -19,7 +19,20 @@ wait\@:
         bne    wait\@
 
         // toggle the polarity to look for the opposite edge next time
+        eor    r8, r3    // restore r8 value
+        eor    r3, #PSYNC_MASK
+.endm
+
+.macro WAIT_FOR_PSYNC_EDGE_FAST
+waitPF\@:
+        // Read the GPLEV0
+        ldr    r8, [r4]
         eor    r8, r3
+        tst    r8, #PSYNC_MASK
+        bne    waitPF\@
+
+        // toggle the polarity to look for the opposite edge next time
+        eor    r8, r3    // restore r8 value
         eor    r3, #PSYNC_MASK
 .endm
 
@@ -31,7 +44,8 @@ wait\@:
         WAIT_FOR_CSYNC_1
         READ_CYCLE_COUNTER r9
         // Calculate length of low hsync pulse (in ARM cycles = ns)
-        sub    r10, r9, r10
+        subs   r10, r9, r10
+        rsbmi  r10, r10, #0
         // Start with the configured horizontal offset
         // Implement half character horizontal scrolling:
         // - a "short"  hsync is 3.5us, leave h_offset as-is
@@ -54,6 +68,17 @@ skip_psync_loop\@:
 skip_psync_loop_exit\@:
 .endm
 
+.macro SKIP_PSYNC_FAST
+        // Wait for the start of hsync
+        WAIT_FOR_CSYNC_0
+        // Wait for the end of hsync
+        WAIT_FOR_CSYNC_1
+skip_psync_loop_fast\@:
+        WAIT_FOR_PSYNC_EDGE_FAST           // wait for next edge of psync
+        subs   r7, r7, #1
+        bne    skip_psync_loop_fast\@
+.endm
+
 .macro CAPTURE_0_BITS_WIDE
         // Pixel 0 in GPIO  7.. 2 ->  7.. 4
         // Pixel 1 in GPIO 13.. 8 ->  3.. 0