diff --git a/cli/imgui_example.cpp b/cli/imgui_example.cpp
index 8f6ba22..64a5e77 100644
--- a/cli/imgui_example.cpp
+++ b/cli/imgui_example.cpp
@@ -13,6 +13,7 @@ extern "C" {
 #include "cpu.h"
 #include "rom.h"
 #include "lcd.h"
+#include "instructions.h"
 }
 
 static const char *A_FORMAT = "A: 0x%02x";
@@ -24,8 +25,10 @@ static const char *H_FORMAT = "H: 0x%02x";
 static const char *L_FORMAT = "L: 0x%02x";
 static const char *SP_FORMAT = "SP: 0x%02x";
 static const char *PC_FORMAT = "PC: 0x%02x";
+static const char *INSN_FORMAT = "Next instruction: %s";
 
 unsigned char output_image[256 * 256 * 4];
+unsigned char visible_pixels[160 * 144 * 4];
 unsigned char vram_tiles[256 * 96 * 4];
 
 struct key_input {
@@ -75,6 +78,17 @@ void convert_output(struct lcd *lcd) {
             output_image[out_index++] = 255;
         }
     }
+    out_index = 0;
+    for (y = 0; y < 144; y++) {
+        for (x = 0; x < 160; x++) {
+            int val = lcd->pixels[y * 160 + x];
+            int fill = default_palette[val];
+            visible_pixels[out_index++] = fill;
+            visible_pixels[out_index++] = fill;
+            visible_pixels[out_index++] = fill;
+            visible_pixels[out_index++] = 255;
+        }
+    }
 }
 
 void convert_vram(struct dmg *dmg) {
@@ -194,6 +208,7 @@ int main(int argc, char *argv[])
 
     // setup output
     GLuint texture = make_output_texture();
+    GLuint vis_texture = make_output_texture();
     GLuint vram_texture = make_output_texture();
 
     editor.ReadFn = read_mem;
@@ -276,6 +291,9 @@ int main(int argc, char *argv[])
                 ImGui::SameLine();
                 ImGui::Text(PC_FORMAT, dmg.cpu->pc);
 
+                u8 opc = dmg_read(&dmg, dmg.cpu->pc);
+                ImGui::Text(INSN_FORMAT, instructions[opc].format);
+
                 ImGui::Checkbox("Z", &z_flag);
                 ImGui::SameLine();
                 ImGui::Checkbox("N", &n_flag);
@@ -302,10 +320,10 @@ int main(int argc, char *argv[])
                 ImGui::End();
             }
 
+            convert_output(dmg.lcd);
             {
                 ImGui::Begin("Output");
 
-                convert_output(dmg.lcd);
                 glBindTexture(GL_TEXTURE_2D, texture);
                 glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, 256, 256, 0, GL_RGBA, GL_UNSIGNED_BYTE, output_image);
                 ImGui::Image((void*)(intptr_t) texture, ImVec2(512, 512));
@@ -313,6 +331,16 @@ int main(int argc, char *argv[])
                 ImGui::End();
             }
 
+            {
+                ImGui::Begin("LCD");
+
+                glBindTexture(GL_TEXTURE_2D, vis_texture);
+                glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, 160, 144, 0, GL_RGBA, GL_UNSIGNED_BYTE, visible_pixels);
+                ImGui::Image((void*)(intptr_t) vis_texture, ImVec2(320, 288));
+
+                ImGui::End();
+            }
+
             {
                 ImGui::Begin("VRAM");
 
diff --git a/src/cpu.c b/src/cpu.c
index fff4b21..4e07831 100644
--- a/src/cpu.c
+++ b/src/cpu.c
@@ -269,7 +269,8 @@ static void add(struct cpu *cpu, u8 value, int with_carry)
 {
     u8 sum_trunc;
     int sum_full = cpu->a + value;
-    if (with_carry && flag_isset(cpu, FLAG_CARRY)) {
+    int carry = (with_carry && flag_isset(cpu, FLAG_CARRY)) ? 1 : 0;
+    if (carry) {
         sum_full++;
     }
     sum_trunc = (u8) sum_full;
@@ -279,12 +280,12 @@ static void add(struct cpu *cpu, u8 value, int with_carry)
         clear_flag(cpu, FLAG_ZERO);
     }
     clear_flag(cpu, FLAG_SIGN);
-    if (sum_full > sum_trunc) {
+    if (sum_full > 0xff) {
         set_flag(cpu, FLAG_CARRY);
     } else {
         clear_flag(cpu, FLAG_CARRY);
     }
-    if (((cpu->a & 0xf) + (value & 0xf)) & 0x10) {
+    if (((cpu->a & 0xf) + (value & 0xf) + carry) & 0x10) {
         set_flag(cpu, FLAG_HALF_CARRY);
     } else {
         clear_flag(cpu, FLAG_HALF_CARRY);
@@ -296,10 +297,11 @@ static void subtract(struct cpu *cpu, u8 value, int with_carry, int just_compare
 {
     u8 sum_trunc;
     int sum_full = cpu->a - value;
-    if (with_carry && flag_isset(cpu, FLAG_CARRY)) {
+    int carry = (with_carry && flag_isset(cpu, FLAG_CARRY)) ? 1 : 0;
+    if (carry) {
         sum_full--;
     }
-    sum_trunc = (u8) (sum_full & 0xff);
+    sum_trunc = (u8) sum_full;
     if (!sum_trunc) {
         set_flag(cpu, FLAG_ZERO);
     } else {
@@ -508,9 +510,6 @@ static u16 check_interrupts(struct cpu *cpu)
     return 0;
 }
 
-/*
-*/
-
 void cpu_step(struct cpu *cpu)
 {
     u8 temp;
@@ -570,6 +569,13 @@ void cpu_step(struct cpu *cpu)
             cpu->a = rotate_right(cpu, cpu->a);
             break;
 
+        case 0x37: // SCF
+            set_flag(cpu, FLAG_CARRY);
+            break;
+        case 0x3f: // CCF
+            clear_flag(cpu, FLAG_CARRY);
+            break;
+
         // incs and decs
         case 0x03: write_bc(cpu, read_bc(cpu) + 1); break;
         case 0x04: inc_with_carry(cpu, &cpu->b); break;
@@ -864,7 +870,7 @@ void cpu_step(struct cpu *cpu)
             break;
 
         case 0xde: // SBC A, u8
-            subtract(cpu, read8(cpu, cpu->pc), 0, 1);
+            subtract(cpu, read8(cpu, cpu->pc), 1, 0);
             cpu->pc++;
             break;
 
@@ -992,6 +998,7 @@ void cpu_step(struct cpu *cpu)
             break;
         case 0xf1: // POP AF
             write_af(cpu, pop(cpu));
+            cpu->f &= 0xf0;
             break;
         case 0xf2: // LD A,(C)
             cpu->a = read8(cpu, 0xff00 + cpu->c);
diff --git a/src/dmg.c b/src/dmg.c
index 0fde300..8bca6f6 100644
--- a/src/dmg.c
+++ b/src/dmg.c
@@ -67,13 +67,12 @@ u8 dmg_read(void *_dmg, u16 address)
     if (address < 0x4000) {
         return dmg->rom->data[address];
     } else if (address < 0x8000) {
-        // TODO switchable rom bank
         return dmg->rom->data[address];
     } else if (address < 0xa000) {
         return dmg->video_ram[address - 0x8000];
     } else if (address < 0xc000) {
-        // TODO switchable ram bank
-        return 0;
+        printf("RAM bank not handled by MBC\n");
+        return 0xff;
     } else if (address < 0xe000) {
         return dmg->main_ram[address - 0xc000];
     } else if (lcd_is_valid_addr(address)) {
@@ -182,6 +181,23 @@ static void render_background(struct dmg *dmg, int lcdc)
     }
 }
 
+static void scroll(struct dmg *dmg)
+{
+    int scroll_y = lcd_read(dmg->lcd, REG_SCY);
+    int scroll_x = lcd_read(dmg->lcd, REG_SCX);
+
+    int lines;
+    for (lines = 0; lines < 144; lines++) {
+        int src_y = (scroll_y + lines) & 0xff;
+        int cols;
+        for (cols = 0; cols < 160; cols++) {
+            int src_off = (src_y << 8) + ((scroll_x + cols) & 0xff);
+
+            dmg->lcd->pixels[lines * 160 + cols] = dmg->lcd->buf[src_off];
+        }
+    }
+}
+
 struct oam_entry {
     u8 pos_y;
     u8 pos_x;
@@ -272,6 +288,8 @@ void dmg_step(void *_dmg)
                 render_objs(dmg);
             }
 
+            scroll(dmg);
+
             // now copy 256x256 buf to 160x144 based on window registers
             lcd_copy(dmg->lcd);
             lcd_draw(dmg->lcd);
diff --git a/src/mbc.c b/src/mbc.c
index 2601820..f556152 100644
--- a/src/mbc.c
+++ b/src/mbc.c
@@ -19,7 +19,10 @@ static int mbc_noop_write(struct mbc *mbc, struct dmg *dmg, u16 addr, u8 data)
 static int mbc1_read(struct mbc *mbc, struct dmg *dmg, u16 addr, u8 *out_data)
 {
   if (addr >= 0x4000 && addr <= 0x7fff) {
-    int use_bank = mbc->rom_bank || 1;
+    int use_bank = mbc->rom_bank;
+    if (!use_bank) {
+      use_bank = 1;
+    }
     *out_data = dmg->rom->data[0x4000 * use_bank + (addr & 0x3fff)];
     return 1;
   } else if (addr >= 0xa000 && addr <= 0xbfff) {